PEER REVIEW SUMMARY REPORT

External Peer Review of EPA’s Draft Document

Approach to Assessing Non-cancer Health Effects from Formaldehyde and
Benefits from Reducing Non-cancer Health Effects as a Results of
Implementing Formaldehyde Emission Limits for Composite Wood Products

Prepared for:

U.S. Environmental Protection Agency

National Program Chemicals Division

Office of Pollution Prevention and Toxics

1200 Pennsylvania Avenue, NW (7404T)

Washington, DC 20460

Prepared by: 

Versar, Inc. 

6850 Versar Center

Springfield, VA 22151

Contract No. EP-C-07-025

Task Order 129

Peer Reviewers:

John R. Balmes, M.D.

Glenn C. Blomquist, Ph.D.

A. Myrick Freeman III, Ph.D.

Shelby Gerking, Ph.D.

David Kriebel, Sc.D.

Frederick J. Miller, Ph.D., Fellow ATS

Rebecca T. Parkin, Ph.D., MPH

TABLE OF CONTENTS

  TOC \o "1-2" \h \z \u    HYPERLINK \l "_Toc299520720"  I.	INTRODUCTION
  PAGEREF _Toc299520720 \h  1  

  HYPERLINK \l "_Toc299520721"  II.	CHARGE TO REVIEWERS	  PAGEREF
_Toc299520721 \h  3  

  HYPERLINK \l "_Toc299520722"  III.	INDIVIDUAL REVIEWER COMMENTS ON
NON-CANCER HEALTH 

EFFECTS – DERIVATION OF CONCENTRATION-RESPONSE FUNCTIONS	  PAGEREF
_Toc299520722 \h  9  

  HYPERLINK \l "_Toc299520723"  John R. Balmes, M.D.	  PAGEREF
_Toc299520723 \h  10  

  HYPERLINK \l "_Toc299520724"  David Kriebel, Sc.D.	  PAGEREF
_Toc299520724 \h  15  

  HYPERLINK \l "_Toc299520725"  Frederick J. Miller, Ph.D., Fellow ATS	 
PAGEREF _Toc299520725 \h  20  

  HYPERLINK \l "_Toc299520726"  Rebecca T. Parkin, Ph.D., MPH	  PAGEREF
_Toc299520726 \h  27  

  HYPERLINK \l "_Toc299520727"  IV.	INDIVIDUAL REVIEWER COMMENTS ON  THE
BENEFITS ASSESSMENT APPROACH	  PAGEREF _Toc299520727 \h  34  

  HYPERLINK \l "_Toc299520728"  Glenn C. Blomquist, Ph.D.	  PAGEREF
_Toc299520728 \h  35  

  HYPERLINK \l "_Toc299520729"  A. Myrick Freeman III, Ph.D.	  PAGEREF
_Toc299520729 \h  44  

  HYPERLINK \l "_Toc299520730"  Shelby Gerking, Ph.D.	  PAGEREF
_Toc299520730 \h  52  

  HYPERLINK \l "_Toc299520731"  V.	PEER REVIEWER COMMENT TABLE ON
NON-CANCER HEALTH  

EFFECTS – DERIVATION OF CONCENTRATION-RESPONSE FUNCTIONS	  PAGEREF
_Toc299520731 \h  61  

  HYPERLINK \l "_Toc299520732"  VI.	PEER REVIEWER COMMENT TABLE ON  THE
BENEFITS ASSESSMENT APPROACH	  PAGEREF _Toc299520732 \h  80  

 

	

I.	INTRODUCTION

In March 2008, the Sierra Club and other organizations and individuals
petitioned EPA under Section 21 of the Toxic Substances Control Act
(TSCA).  The petitioners requested that EPA use TSCA Section 6 to adopt
a newly-promulgated California Air Resources Board (CARB) Airborne Toxic
Control Measure (ATCM) as a national standard for formaldehyde emissions
from hardwood plywood, particleboard, and medium-density fiberboard
products.  The petitioners expressed particular concern over the levels
of formaldehyde found in emergency housing provided for Hurricane
Katrina survivors, but noted that there are no federal regulations on
formaldehyde emissions from composite wood products other than those
applicable to manufactured housing regulated by the U.S. Department of
Housing and Urban Development.  In response to the petition, EPA
announced in the Federal Register on June 27, 2008 (73 FR 36504) that
EPA would initiate a proceeding to investigate whether and what type of
regulatory or other action might be appropriate to protect against risks
that may be posed by formaldehyde emitted from composite wood products. 
In December 2008, EPA published an Advance Notice of Proposed Rulemaking
(ANPR) to publicly initiate the investigation and obtain stakeholder
input (73 FR 73620).   

On July 7, 2010, the Formaldehyde Standards for Composite Wood Products
Act (FSCWPA) was signed into law.  This statute, which adds a Title VI
to TSCA, establishes formaldehyde emission standards for hardwood
plywood, particleboard, and medium-density fiberboard that are identical
to the CARB ATCM Phase II standards.   TSCA Title VI directs EPA to
promulgate implementing regulations by January 1, 2013 that address:
sell-through dates for products; stockpiling; third-party testing and
certification; auditing and reporting of third-party certifiers;
recordkeeping; chain of custody; labeling; enforcement; products made
with no-added formaldehyde (NAF) and ultra-low emitting formaldehyde
(ULEF) resins; laminated products; finished goods; hardboard; and
products containing de minimis amounts of composite wood products.  

In support of the implementing regulations, as with other rulemakings
that are determined to be “significant regulatory action[s]” under
Executive Order 12866—Regulatory Planning and Review, EPA is required
to conduct an economic analysis of the costs and benefits associated
with the rulemaking.  On June 2, 2010, EPA released a draft document
titled Toxicological Review of Formaldehyde-Inhalation Assessment: In
Support of Summary Information on the Integrated Risk Information System
(IRIS). This draft IRIS assessment, which recently underwent independent
scientific peer review by the National Academy of Sciences (NAS),
addresses both non-cancer and cancer human health effects that may
result from chronic inhalation exposure to formaldehyde.  However, this
draft IRIS assessment does not include concentration-response functions
for non-cancer health effects, which are needed for the benefits
assessment.  Therefore, this draft report titled Approach to Assessing
Non-cancer Health Effects from Formaldehyde and Benefits from Reducing
Non-cancer Health Effects as a Result of Implementing Formaldehyde
Emission Limits for Composite Wood Products (referred to hereafter as
draft Approach Document) was developed, which provides discussion of
several non-cancer endpoints as well as a discussion of the derivation
of concentration-response functions, uncertainties, and EPA’s proposed
approach for using the concentration-response functions in the benefits
assessment.  

Peer Reviewers:

Experts in Non-cancer Health Effects:

John R. Balmes, M.D. 

University of California, San Francisco

San Francisco, CA  94110

David Kriebel, Sc.D.

University of Massachusetts Lowell

Lowell, MA 01854

Frederick J. Miller, Ph.D., Fellow ATS

Fred J. Miller & Associates LLC

Cary, NC 27511

Rebecca T. Parkin, Ph.D., MPH

The George Washington University Medical Center

Washington DC, 20037

Experts in Environmental and Health Economics:

Glenn C. Blomquist, Ph.D.

University of Kentucky

Lexington, KY 40506

A. Myrick Freeman III, Ph.D.

Bowdoin College

Georgetown, ME 94548

Shelby Gerking, Ph.D.

University of Central Florida

Orlando, Florida 32816

II.	CHARGE TO REVIEWERS

EPA is requesting comment on the draft document, “Approach to
Assessing Non-cancer Health Effects from Formaldehyde and Benefits from
Reducing Non-cancer Health Effects as a Result of Implementing
Formaldehyde Emission Limits for Composite Wood Products” (draft
Approach Document). The purpose of this draft Approach Document is to
describe how EPA derived concentration-response functions for
formaldehyde effects on sensory irritation, asthma, and female
reproductive toxicity, as well as the approach EPA will take to use
these concentration-response functions to estimate benefits from the
contribution of reduced non-cancer health effects as a result of
implementing formaldehyde emission standards for composite wood
products. The draft document describes in detail how the
concentration-response functions were derived as well as uncertainties.
The draft document also briefly describes the epidemiological studies
that were used to derive the concentration-response functions; however,
a detailed analysis of these studies (as well as others) is included in
the 2010 draft IRIS assessment for formaldehyde and is therefore not
included in the draft Approach Document. The draft Approach Document
also describes how EPA plans to use the concentration-response functions
to estimate benefits. It provides sample calculations but not the actual
benefits assessment. The full benefits assessment will be provided in
the economic analysis that will be publicly available (at  HYPERLINK
"http://www.regulations.gov" www.regulations.gov ) when the draft
proposed rule is published.

The draft IRIS assessment for formaldehyde underwent peer review by the
National Academy of Sciences (NAS), and EPA requests that the peer
reviewers of the draft Approach Document do not duplicate the efforts of
the NAS.  EPA is requesting comment specifically on the derivation of
the concentration-response functions and the approach on how the Agency
plans to use these functions to assess benefits. EPA requests that if
the reviewers have comments on EPA’s proposed approach that they
provide specific recommendations for improvements, including specific
references and how those references are relevant to the Approach. In
addition, if reviewers have information that EPA should have considered,
please provide that information. 

Specific Charge Questions on Derivation of Concentration-Response
Functions

Sensory Irritation

The literature base of residential epidemiology studies of the
irritation effects with quantitative concentration-response data was
limited, but data were available from which a concentration-response
function could be estimated for eye irritation.

Please comment on whether the concentration-response function is
appropriately derived from the studies used in the draft Approach
Document.

Please comment on the validity of extrapolating the
concentration-response function and using the function to estimate
effects at concentrations below 100 ppb.

The studies considered for the concentration-response function are
residential studies because residential exposures are the focus of the
rule and the residential studies avoid some of the limitations of other
types of studies (e.g., controlled exposure chamber studies used high,
acute exposures, and occupational studies reflect high exposure levels
and potential dose-rate effects from peak exposures).  Are there other
studies (e.g., chamber or non-residential studies) that should be used
as supporting evidence for the currently derived concentration-response
function for eye irritation?  If so, please provide information about
the studies and discuss how they should be used to support the
concentration-response function.

Are there other sensory irritation endpoints besides eye irritation for
which a concentration-response function can and should be developed for
assessing effects at concentrations below 100 ppb?  If so, please
provide references of appropriate studies and a description of how the
studies could be used to develop a concentration-response function.

Asthma

Several studies of the association between formaldehyde and asthma have
been published and these were quantitatively summarized in a
meta-analysis by McGwin et al. (2010).

Please comment on whether the McGwin et al. (2010) meta-analysis is an
appropriate summary of the association between formaldehyde inhalation
exposures and childhood asthma.

Please comment on whether EPA chose the appropriate function from the
McGwin et al. paper to use in the benefits assessment.

Female Reproductive Toxicity

The literature base describing reproductive effects of formaldehyde
exposure in humans is limited; however, there are supportive findings
from the animal toxicological literature.  The epidemiologic literature
suggests an association between formaldehyde and increased risk of
spontaneous abortion and a related delay in time to pregnancy (i.e.,
reduced fertility).  The only study with quantitative
concentration-response data was Taskinen et al. (1999).

 glove use yielded effect estimates whose confidence intervals
overlapped.  The FDR for the 17 women who were classified as not wearing
gloves was 0.51 [95% CI: 0.28−0.92].  The FDR for women who were
classified as always or sometimes wearing gloves was 0.79 [95% CI:
0.47−1.23].  

In light of the results regarding glove use, in what way(s) can the
Taskinen et al. results be used to support calculating a concentration
response for formaldehyde inhalation risk to time to pregnancy?

Please comment on whether the concentration-response function is
appropriately derived from the Taskinen et al. (1999) study.

Specific Charge Questions on the Benefits Assessment Approach

Sensory Irritation

The concentration-response function for the prevalence of eye irritation
is based on Hanrahan et al. (1984). While Hanrahan does not identify the
relevant time period for this relationship, Liu et al. (1991) produced
qualitatively similar results for a two-week time period. We assumed
that the concentration-response relationship for eye irritation is for a
two-week period and multiplied the concentration-response function times
26 to obtain the number of cases of eye irritation experienced in one
year. 

The concentration-response curve used is non-linear, which means that
the annual approach may not produce an exact measure of the benefits,
but it is believed that the bias is probably relatively small given the
small change in exposure.

Please comment on whether the concentration-response function is used
appropriately in the benefits assessment approach.

The concentration-response curve is for an acute effect occurring at a
time less than the one-year time step of the exposure model. The
willingness to pay to avoid cases of eye irritation is calculated for
each year of analysis by multiplying the number of cases in a year by
the marginal willingness to pay value to avoid one case. This may
produce biased results, but the direction of that bias is unknown.

On one hand, some studies have suggested that the willingness to pay per
symptom-day declines as the number of symptom-days avoided increases. In
other words, the value to reduce two symptom-days of some effect is less
than two times the value for one symptom-day. Therefore, this approach
may overstate the total value. 

On the other hand, this analysis only applies values to one symptom, eye
irritation, largely because this is the only symptom for which an
effective concentration-response curve was derived. However, there is
evidence that there are other forms of minor eye, nose, and throat
irritation associated with formaldehyde exposure. Since eye irritation
is the largest of the median values from the Tolley et al. (1986) study,
the values used here likely represent a lower bound on the true effect
of multiple symptoms. A reasonable upper bound of the total effect for
one symptom-day would be sum of the willingness to pay for one
symptom-day of each effect individually.  

Please comment on whether the annualized approach as opposed to the
lifetime approach is appropriate in this context.

The concentration-response function for the prevalence of eye irritation
is for a single two-week period.  The annualized approach used an
adjusted concentration-response relationship that consisted in
multiplying the concentration-response function obtained from Hanrahan
et al. (1984) by 26 two-week periods.

Please comment on whether the adjustment of the concentration-response
function is appropriate in the annualized approach.

Asthma

The concentration-response relationship for asthma incidence is given as
an odds ratio for children exposed to formaldehyde. An odds ratio is a
way of comparing the difference in the probability of an event, in this
case asthma diagnosis, occurring in groups exposed at different levels.
The odds ratio is used to determine the willingness to pay for any
change in emissions exposure over a 16 year period. We assume that the
probability of being diagnosed with asthma is uniformly distributed
across these 16 years. 

The use of an odds ratio implies that the concentration-response
relationship is non-linear. However, given that the change in exposure
is relatively small, the results from using this implied non-linear
function are anticipated to be similar to the results if we were to use
a linear function.

Please comment on whether the concentration-response function is used
appropriately in the benefits assessment approach.

The valuation of asthma effects from formaldehyde exposure is based on
valuing the incidence of asthma rather than the more standard valuation
based on the exacerbation of asthma symptoms. The reason for this is
that a statistical relationship between formaldehyde exposure and asthma
exacerbation has not been sufficiently well established, but there is a
statistical relationship between formaldehyde exposure and asthma
occurrence or incidence. However, the only time that EPA has valued
asthma exacerbation is for the first prospective study of the benefits
and costs of the Clean Air Act (EPA 1999). This is probably due to a
lack of an appropriate concentration-response relationship. The EPA’s
Office of Air does include recommended values for the monetization of
chronic asthma in its BenMAP documentation (EPA 2011b), which is based
on an average of the results from Blumenschein and Johannesson (1998)
and O’Conor and Blomquist (1997).

For this analysis, EPA treats prevalence as a measure of cumulative
incidence by assuming that the disease duration is the rest of the
child’s life. However, the EPA (2000a) has previously assumed that a
portion of asthmatic children will become asymptomatic as they move into
adulthood.

Please comment on whether the annualized approach as opposed to the
lifetime approach is appropriate in this context.

The concentration-response relationship is based on an odds ratio for a
period of time longer than the one year. The exposure model produces
results for one-year of exposure to formaldehyde for each age, housing
type, and climate cohort. As a consequence, an adjustment to the simple
model needed to be made to model the effect on individuals who are
currently alive when the regulation takes effect.

As an upper bound estimate, we assume that the odds ratio can be applied
to the one year reduction in the same fashion as it was applied for the
16 year time period. The problem with this approach is that it
effectively treats the exposure as an acute effect. If the odds ratio
indicates a particular percentage reduction in asthma cases from a
reduction in formaldehyde exposure over 16 years, the upper bound
approach would suggest the same percentage reduction in asthma case in
one single year if we were to witness the same reduction in formaldehyde
exposure in one year. The lower bound approach assumes that a particular
exposure reduction for one year produces the same effect as if that
exposure reduction were spread out over the entire 16-year time period. 

Please comment on the upper and lower bound methodology used.

The prevalence data suggest that 7.2% of children age 0-4 have been told
that they have asthma. 16.4% of children aged 5-14 have been told that
they have asthma. We assume that the probability of being diagnosed is
uniformly distributed across the 0 to 16 age bracket, so that the
incidence rate would be 1.44% for the first five years and 1.49% for the
next eleven years. Allocating cases uniformly across the age brackets
provides us with a way to calculate the annual benefits of any exposure
reduction that occurs for the entire time period, but may not be
appropriate if the diagnosis of asthma for very young children (e.g.,
age 0-2) differs from older children.

Is the assumption about applying the concentration-response function to
children aged 0 to 16 appropriate, given the asthma onset studies’
focus on school-aged children and adolescents, and given the potential
uncertainty of the Rumchev study involving younger children?

Female Reproductive Toxicity

The linear concentration-response function accounting for background
exposures based on Taskinen et al. (1999) is used for the annual
approach.

Please comment on whether the concentration-response function is used
appropriately in the benefits assessment approach.

Monetization of delayed fertility has been discussed in the literature
but has never been included in primary analyses of EPA rules because it
was believed that the scientific knowledge on reproductive and
developmental health effects was not strong enough to quantify risk in
the primary benefits analysis.

While there appears to be good data on the cost of treatment, there are
few studies on the willingness to pay to reduce delayed conception. Most
economic valuation studies regarding fertility focus on an
individual’s or a couple’s willingness to pay for infertility
treatment, particularly in-vitro fertilization (IVF).  Conceptually, the
willingness to pay approach is preferable and could be deduced from the
valuation studies. However, given that there has been less research on
this willingness to pay values than other values and the fact that the
values have not been used in previous economic analyses, we used the
more straight-forward cost of illness approach for this analysis. As an
upper bound, we assume that any woman who has difficulty conceiving
(i.e., time to pregnancy exceeds 12 months) will obtain some type of
fertility service. This may be a reasonable upper bound, but there are
certainly women who have trouble becoming pregnant and do not seek
fertility treatment. 

The willingness to pay for fertility treatment is treated as an expected
value. The average cost of various cycle-based fertility treatments is
multiplied times the probability of obtaining that treatment at various
ages. There is probably a small opposite effect associated with an
increase in pregnancy from reduced formaldehyde exposure by women who do
not want to become pregnant, but we do not consider this effect

Please comment on whether the annualized approach as opposed to the
lifetime approach is appropriate in this context.

The time period associated with the exposure-response function is longer
than the time step of the exposure model. The Taskinen (1999) study
examined the fecundity of women age 20-40. This causes a problem if we
have an individual who does not experience the exposure reduction for
the full time period. We can form an upper and lower bound of this
willingness to pay as we did for asthma prevalence. The upper bound is
based on the assumption that the linear concentration-response function
holds for an acute exposure of one year. The lower bound assumes that an
exposure reduction for one year represents 1/20th of an exposure
reduction for the entire 20-year time period. 

 Please comment on the upper and lower bound methodology used.

The ages used in this analysis reflect the ages reported in the Taskinen
(1999) study, ages 20-40. However, women older than age 40 obtain
fertility treatment. The value of reduced infertility effects from
reduced formaldehyde exposure is not reflected in this analysis in order
to remain consistent with Taskinen (1999).

 Please comment on whether the concentration-response function was
applied to appropriate age groups given the age groups in the studies. 

III.	INDIVIDUAL REVIEWER COMMENTS ON NON-CANCER HEALTH EFFECTS –
DERIVATION OF CONCENTRATION-RESPONSE FUNCTIONS

Review By:

John R. Balmes, M.D.

Peer Review Comments on EPA’s Draft Document Approach to Assessing
Non-cancer Health Effects from Formaldehyde and Benefits from Reducing
Non-cancer Health Effects as a Results of Implementing Formaldehyde
Emission Limits for Composite Wood Products

John R. Balmes, M.D. 

University of California, San Francisco 

July 25, 2011

I.  GENERAL IMPRESSIONS 

The document is clearly and concisely written.  The background
subsections of Section 2: Non-cancer Health Effects – Derivation of
Concentration-Response Functions for sensory irritation (2.1.1), asthma
(2.2.1), and female reproductive toxicity (2.3.1) are extremely brief,
especially the latter two.  While it is understandable that the EPA does
not wish to be duplicative of the material in the draft IRIS document on
formaldehyde or the NAS review of the IRIS document, some readers of the
“Approach” document will not have read the IRIS and NAS documents
and thus will benefit from a bit more background context on formaldehyde
and asthma or formaldehyde and female reproductive toxicity.

II. RESPONSE TO CHARGE QUESTIONS

 

Sensory Irritation

The literature base of residential epidemiology studies of the
irritation effects with quantitative concentration-response data was
limited, but data were available from which a concentration-response
function could be estimated for eye irritation.

1. Please comment on whether the concentration-response function is
appropriately derived from the studies used in the draft Approach
Document.

The concentration-response function (CRF) for sensory irritation is
derived from one small field study, Hanrahan et al. (1984).  Because
most individuals exposed to formaldehyde from the products of interest
will be exposed at levels below 100 ppb, the EPA is appropriately
concerned about the form of the CRF in this range.  Unfortunately, the
data for exposures below 100 ppb are not given in the original Hanrahan
paper and the regression of prevalence of eye irritation and
formaldehyde concentration in the paper covers the range of 100-800 ppb.
 Another issue regarding CRF derivation from the Hanrahan paper is the
relatively small sample size (n=61).  The Liu et al. (1991) study, which
involved a much larger population, is supportive of the concept that
there is a CRF for formaldehyde and eye irritation at relatively low
levels of exposure, but was not used in the actual derivation of such a
CRF because of the way the authors aggregated their data into low,
medium, and high levels of exposure.

		

2. Please comment on the validity of extrapolating the
concentration-response function and using the function to estimate
effects at concentrations below 100 ppb.

Given the limitations of the data available from the Hanrahan paper, the
method of derivation of the CRF including the portion of the curve below
100 ppb is appropriate, but the extrapolation is pushing the envelope of
what reasonably can be extracted from the data.  The assumption that the
reported standard deviation of the outdoor formaldehyde measurements
represents the lower bound on the standard deviation of the indoor
concentrations appears reasonable, but nonetheless the distribution of
exposure data below 100 ppb in the Hanrahan study was assumed rather
than based on the actual data.  The EPA used the graphical display of
the regression data over the range of 100-800 ppb in the Hanrahan paper
to derive a CRF and then to infer the shape of the CRF below 100 ppb. 
Given that the Hanrahan study only included 61 subjects and that the
formaldehyde measurements in the mobile homes were only for 30-60
minutes, the validity of the extrapolated CRF can be questioned even if
an appropriate method was used.

	

3. The studies considered for the concentration-response function are
residential studies because residential exposures are the focus of the
rule and the residential studies avoid some of the limitations of other
types of studies (e.g., controlled exposure chamber studies used high,
acute exposures, and occupational studies reflect high exposure levels
and potential dose-rate effects from peak exposures).  Are there other
studies (e.g., chamber or non-residential studies) that should be used
as supporting evidence for the currently derived concentration-response
function for eye irritation?  If so, please provide information about
the studies and discuss how they should be used to support the
concentration-response function.

I do not know of other studies that would be better to use for
derivation of a CRF for formaldehyde exposure and sensory irritation
given the EPA’s interest in chronic, low-level, non-peak exposures.

4. Are there other sensory irritation endpoints besides eye irritation
for which a concentration-response function can and should be developed
for assessing effects at concentrations below 100 ppb?  If so, please
provide references of appropriate studies and a description of how the
studies could be used to develop a concentration-response function.

I do not know whether a CRF can be developed for other sensory
irritation endpoints.  

Asthma

Several studies of the association between formaldehyde and asthma have
been published and these were quantitatively summarized in a
meta-analysis by McGwin et al. (2010).

5. Please comment on whether the McGwin et al. (2010) meta-analysis is
an appropriate summary of the association between formaldehyde
inhalation exposures and childhood asthma.

The McGwin et al. (2010) meta-analysis of the association between
formaldehyde exposure and risk of asthma in childhood appears to have
been well conducted, using state-of-the-art methods, including both
fixed and random effects models and various sensitivity analyses. 
Because the meta-analysis was based on a systematic review of the
literature that appears to have been comprehensive, the results can be
considered an appropriate summary of what is known about the association
from epidemiological studies.

6. Please comment on whether EPA chose the appropriate function from the
McGwin et al. paper to use in the benefits assessment.

Exponentiating the linear slope of the fixed effects model from the
McGwin et al. paper is appropriate, especially given that the random
effects model gave essentially the same pooled odds ratio.

Female Reproductive Toxicity

The literature base describing reproductive effects of formaldehyde
exposure in humans is limited; however, there are supportive findings
from the animal toxicological literature.  The epidemiologic literature
suggests an association between formaldehyde and increased risk of
spontaneous abortion and a related delay in time to pregnancy (i.e.,
reduced fertility).  The only study with quantitative
concentration-response data was Taskinen et al. (1999).

d.  The FDR for the 17 women who were classified as not wearing gloves
was 0.51 [95% CI: 0.28−0.92].  The FDR for women who were classified
as always or sometimes wearing gloves was 0.79 [95% CI: 0.47−1.23].  

7. In light of the results regarding glove use, in what way(s) can the
Taskinen et al. results be used to support calculating a concentration
response for formaldehyde inhalation risk to time to pregnancy?

The results of the stratified analysis regarding glove use are
suggestive of the possibility that the dermal route may contribute to
total exposure, but are not adequate to consider modification of a CRF
for decreased fertility based on inhalational exposures, especially
given the small size of the two strata and the wide confidence intervals
of the FDR point estimates.

8. Please comment on whether the concentration-response function is
appropriately derived from the Taskinen et al. (1999) study.

The CRF for formaldehyde exposure and decreased fertility appears to be
appropriately derived from the Taskinen et al. (1999) study.  The
adjustment for probable exposure to formaldehyde outside of the
workplace is appropriate given the interest of the EPA in potentially
controlling indoor exposures due to off-gassing from certain products. 
The use of reasonably contemporaneous background exposure data from
Finland (Jurvelin et al., 2001) for this adjustment is also appropriate.

III.	SPECIFIC OBSERVATIONS

Page 10, 1st full paragraph; page 14, figure 4a legend; page 15, 2nd
full paragraph.  The use of “inhalation” exposure in the context of
eye irritation seems inappropriate.  I would eliminate the use of
“inhalation” in all three sentences.

Page 16, paragraph. 2.2.1.  Asthma is defined by the Global Initiative
for Asthma (GINA) as a disease characterized by inflammation of the
airways, not lungs.  I would eliminate “and lungs” in the first
sentence of this paragraph.  To avoid confusion, I would not use
“sensitization” when discussing non-allergic, neurogenic mechanisms
of irritant chemical-induced asthma in the third sentence.

Page 20, paragraph 2.3.1.  “Increased relative risk estimates (RRs)
were in the range of 1.7 to more than 3.0.” To avoid confusion, I
would cite the specific papers that reported these increased relative
risks.

Page 22, last sentence of text.  The text here refers to “estimated
concentration-response functions” but does not explain why there is
more than one CRF in Figure 6.  It is not until page 24 that one learns
that “Regression equations for linear and exponential fits are also
provided.”  If the two lines in Figures 6 and 7 are for both types of
fits then this should be clearly indicated.

Page 26, last full sentence.  The use of a contraction here
(“didn’t”) is unprofessional.

Review By:

David Kriebel, Sc.D.Peer Review Comments on EPA’s Draft Document
Approach to Assessing Non-cancer Health Effects from Formaldehyde and
Benefits from Reducing Non-cancer Health Effects as a Results of
Implementing Formaldehyde Emission Limits for Composite Wood Products

David Kriebel, Sc.D.

University of Massachusetts Lowell

July 21, 2011

I.  GENERAL IMPRESSIONS 

The overall impression is of a thoughtful and well-written document. The
authors have provided a generally clear and logical explanation of their
reasoning throughout. There are a few exceptions, noted below, where
additional explanation or documentation could improve the presentation.

II. RESPONSE TO CHARGE QUESTIONS

 

Sensory Irritation

The literature base of residential epidemiology studies of the
irritation effects with quantitative concentration-response data was
limited, but data were available from which a concentration-response
function could be estimated for eye irritation.

1. Please comment on whether the concentration-response function is
appropriately derived from the studies used in the draft Approach
Document.

I see no problems with the derivation of the concentration-response
function itself. In Section 2.1.3, I think it would strengthen the
argument to at least acknowledge the problem of acclimatization. There
is evidence from a number of studies – notably the studies of studies
conducting autopsies – that there is a change over weeks in the
short-term irritant effects. One way that this will impact the
calculations is that cumulative exposure is probably not the appropriate
exposure metric. The authors may have little choice to use the
simplified approach they have taken which assumes a constant
exposure-response relationship over time, but they could acknowledge
that there is a source of uncertainty from the changes in potency of
formaldehyde over time for various endpoints. I suspect that the
irritant effects diminish with time exposed, but the other two chosen
endpoints probably do not.

				

2. Please comment on the validity of extrapolating the
concentration-response function and using the function to estimate
effects at concentrations below 100 ppb.

I am comfortable with this extrapolation. However, I think that it would
be appropriate to acknowledge that errors in exposure estimation in the
studies on which the function is based have not been included in the
uncertainty calculations. This is a general concern – for all 3
endpoints, there were certainly errors in the exposure estimation, and
the impact of these on the uncertainty in the exposure-response function
has not been considered.

It might be noted that the extrapolation downwards assumes a logistic
shape, which is reasonable but not subject to validation. It might be
noted that the excellent R2 for Figure 3b is a bit misleading since it
is fitting a curve to categorical data, while the actual individual
exposure data were of course continuous – the fit of this curve to the
original data would not have been nearly as good. And again, see my
comment just above about failure to point out that the exposure data had
error, which was not taken into consideration in the model error
calculations.

3. The studies considered for the concentration-response function are
residential studies because residential exposures are the focus of the
rule and the residential studies avoid some of the limitations of other
types of studies (e.g., controlled exposure chamber studies used high,
acute exposures, and occupational studies reflect high exposure levels
and potential dose-rate effects from peak exposures).  Are there other
studies (e.g., chamber or non-residential studies) that should be used
as supporting evidence for the currently derived concentration-response
function for eye irritation?  If so, please provide information about
the studies and discuss how they should be used to support the
concentration-response function.

I think you should use the other studies – particularly the chamber
studies and the autopsy studies – to show the consistency of the
findings across many settings. I am comfortable with the argument that
the function should be derived from the residential studies. But I think
that the logic of extrapolating downwards is supported by some of the
other studies. Kriebel et al. Arch Env Hlth 2001; 56:11- 18 contains
fairly detailed exposure data linked to eye irritation scores. While
much of the data are above the range you are concerned with, there is
evidence here that the relationship observed was valid across the full
range of exposures.

4. Are there other sensory irritation endpoints besides eye irritation
for which a concentration-response function can and should be developed
for assessing effects at concentrations below 100 ppb?  If so, please
provide references of appropriate studies and a description of how the
studies could be used to develop a concentration-response function.

Both nose and throat irritation can be linked to increased risk of lower
respiratory tract infection including bronchitis. So you could argue
that these would have greater economic impacts. 

  

Asthma

Several studies of the association between formaldehyde and asthma have
been published and these were quantitatively summarized in a
meta-analysis by McGwin et al. (2010).

5. Please comment on whether the McGwin et al. (2010) meta-analysis is
an appropriate summary of the association between formaldehyde
inhalation exposures and childhood asthma.

I agree that it is appropriate.

6. Please comment on whether EPA chose the appropriate function from the
McGwin et al. paper to use in the benefits assessment.

I am comfortable with the choice to exclude Rumchev, and agree with the
derivation of the function as described.

Female Reproductive Toxicity

The literature base describing reproductive effects of formaldehyde
exposure in humans is limited; however, there are supportive findings
from the animal toxicological literature.  The epidemiologic literature
suggests an association between formaldehyde and increased risk of
spontaneous abortion and a related delay in time to pregnancy (i.e.,
reduced fertility).  The only study with quantitative
concentration-response data was Taskinen et al. (1999).

The a priori hypothesis in Taskinen et al. (1999) was that formaldehyde
would be associated with decreased fertility.  The primary analyses
showed that there was a statistically significant reduction of fertility
among the women in the highest exposure group that appeared to be
concentration dependent.  Post-hoc subgroup analyses showed that
stratification of the 39 women in the highest exposure group by reported
glove use yielded effect estimates whose confidence intervals
overlapped.  The FDR for the 17 women who were classified as not wearing
gloves was 0.51 [95% CI: 0.28−0.92].  The FDR for women who were
classified as always or sometimes wearing gloves was 0.79 [95% CI:
0.47−1.23].  

7. In light of the results regarding glove use, in what way(s) can the
Taskinen et al. results be used to support calculating a concentration
response for formaldehyde inhalation risk to time to pregnancy?

Yes, but you could strengthen your argument. First, there is a trend in
both FDR and spontaneous abortion across levels of formaldehyde, so you
should not only focus on the effect modification by glove use in the
high exposure group. While the high exposure is the only group for which
the authors provided the results stratified by glove use, the fact that
there is a trend across air concentrations suggests that glove use –
not constant across exposure levels – could not be a very important
source of protection or you would not see the trend. Second, you have
not correctly dealt with the question of whether or not there is effect
modification by glove use. The overlap of confidence intervals is not
the correct way to do this. Construct the null hypothesis that glove
users and non-users have the same FDR and then calculate a p-value
testing this null. You can do it with the data provided – back
calculate the standard errors from the confidence intervals, or ask the
authors for the standard errors. I am pretty sure that this p-value will
be large, and then you will be in a strong position to say that the
appropriate conclusion is to act as if glove use has not modified the
FDR. 

8. Please comment on whether the concentration-response function is
appropriately derived from the Taskinen et al. (1999) study.

Can’t you ask Taskinen to confirm your (and my) assumption that age
was the conditioning time variable in their FDR calculations?

The process of adjusting for background exposure is OK, but I don’t
think it strengthens the presentation. First, I doubt very much that the
relevant background formaldehyde exposure for the women in this plant
could have been so much higher than the levels inside the plant. My
guess is that the mean (one mean for an entire country over a long
period of time? How useful is that?) had a wide range around it. Second,
any of my graduate students could have told you that adding a constant
to all the x values in a regression model will not change the slope.

III.	SPECIFIC OBSERVATIONS

Page 20, section 2.3.1, 6th and 7th lines. The statement “Increased
relative risk estimates (RRs) were in the range of 1.7 to more than
3.0” is not helpful – without explaining whose risk is being
compared to whose, the absolute value of an RR is not useful. 

Section 2.0 Background on the Identification of Non-Cancer Health
Effects, last sentence. It would be helpful to summarize at least
briefly the logic behind focusing on these 3 endpoints. If I have
understood the NAS comments correctly, that Committee also raised
questions about these choices, and in particular on the decision not to
use pulmonary function evidence. Given this concern from the NAS, I
think a bit more explanation of the perceived inadequacies in the data
for the other endpoints would strengthen this document. 

Review By:

Frederick J. Miller, Ph.D., Fellow ATSPeer Review Comments on EPA’s
Draft Document Approach to Assessing Non-cancer Health Effects from
Formaldehyde and Benefits from Reducing Non-cancer Health Effects as a
Results of Implementing Formaldehyde Emission Limits for Composite Wood
Products

Frederick J. Miller, Ph.D., Fellow ATS

Fred J. Miller & Associates LLC

July 18, 2011

I.  GENERAL IMPRESSIONS 

The comments provided are related to the health assessment and
concentration-response evaluation portion of the above cited document as
this reviewer was not asked to comment on the economic benefits
assessment approach. For the most part, the sections on sensory
irritation, asthma, and female reproductive toxicity are clearly
presented and the material is developed in a logical manner. Some
exceptions are noted in this reviewer’s responses to the specific
charge questions. The decision on the concentration-response functions
to take forward to the economic benefits analyses is best documented for
sensory irritation. This reviewer provides a suggestion for how far
below 100 ppb of formaldehyde the modeling results developed by the
Agency based on the Hanrahan et al. (1984) study can be used. The
material included in support of using the McGwin et al. (2010) meta
analysis for asthma in children does not adequately describe the short
comings of the meta analysis. The uncertainties and conflicting results
of the studies make this reviewer reticent to endorse the
concentration-response function for the odds ratio based on the McGwin
et al. (2010) meta analysis. The case to use background adjusted
exposure data for the female reproductive toxicity
concentration-response curve is adequately defended. However, the Agency
does not indicate if the linear risk or the exponential risk curve will
be used.

Overall, the information presented is accurate. To their credit, in a
number of places, the EPA authors have incorporated criticisms included
in the NAS report of the Agency’s draft IRIS assessment document. By
addressing the responses to the charge questions, the Agency will be in
a good position to move forward with the cost benefit analysis required
by Executive Order 12866 related to the Formaldehyde Standards for
Composite Wood Products Act. 

II. RESPONSE TO CHARGE QUESTIONS

Sensory Irritation

The literature base of residential epidemiology studies of the
irritation effects with quantitative concentration-response data was
limited, but data were available from which a concentration-response
function could be estimated for eye irritation.

1. Please comment on whether the concentration-response function is
appropriately derived from the studies used in the draft Approach
Document.

The draft Approach Document adequately discussed the type of data
available in the literature from various studies and presented their
strengths and weaknesses. The Agency wisely stayed away from using the
Liu et al. (1991) study as the primary driver for the
concentration-response function for eye irritation. The categories from
that study were formed on using a ppm-hr per week metric. The eye
irritation studies discussed in Section 2.1.2 for a 10 % additional risk
for 560 ppb for 3 hours (i.e., 1,680 ppm-hrs) compared to 240 ppb for 5
hours (i.e., 1,200 ppm-hrs) clearly show that Haber’s Law does not
apply for the effects of formaldehyde exposure on eye irritation, which
agrees with Miller et al. (Haber’s rule: a special case in a family of
curves relating concentration and duration of exposure to a fixed level
of response for a given endpoint. Toxicology 149:21–34, 2000). Thus,
the Agency would have been hard pressed to defend a translation of the
ppm-hrs metric of Liu et al. (1991) as the basis for developing a
concentration-response function for eye irritation.

Rather, the Hanrahan et al. (1984) study that reported eye irritation
was associated with in-home formaldehyde exposures (p < 0.05) (both as
“burning eyes” and “eye irritation”) was used as the data for
developing a concentration-response relationship for eye irritation.
Moreover, the logistic regression model developed by Hanrahan et al.
(1984) gave insight as to the needed nature of the curve if one were to
extrapolate below the range of exposures reported by those authors
(i.e., from 100 to 800 ppb). In the opinion of this reviewer, the
concentration-response function was appropriately derived by the Agency.

		

2. Please comment on the validity of extrapolating the
concentration-response function and using the function to estimate
effects at concentrations below 100 ppb.

The Agency appropriately focused on the data from the Hanrahan et al.
(1984) study to develop the concentration-response curve for eye
irritation. Importantly, the Agency used a power law curve to fit the
prevalence odds data. This ensures that the fitted curve will
asymptotically approach zero for a zero formaldehyde exposure level and
makes it reasonable to use Equation 2.2 below 100 ppb. Given (1) the
limit of detection (LOD) was 10 ppb at the time the Hanrahan et al.
(1984) study was conducted, (2) the Agency’s assumption about the
outdoor standard deviation of measurements being 30 ppb around a
log-normal distribution of measurements, and (3) the known overall shape
of the concentration-response curve below 100 ppb, this reviewer
believes it would be reasonable for the Agency to use the fitted
equation down to a level of about 40 ppb, which would correspond to the
LOD plus one standard deviation.

3. The studies considered for the concentration-response function are
residential studies because residential exposures are the focus of the
rule and the residential studies avoid some of the limitations of other
types of studies (e.g., controlled exposure chamber studies used high,
acute exposures, and occupational studies reflect high exposure levels
and potential dose-rate effects from peak exposures).  Are there other
studies (e.g., chamber or non-residential studies) that should be used
as supporting evidence for the currently derived concentration-response
function for eye irritation?  If so, please provide information about
the studies and discuss how they should be used to support the
concentration-response function.

Beyond the studies considered by EPA, this reviewer knows of no chamber
or epidemiology studies that would be useful in developing the
concentration-response function for eye irritation. Given the definition
of an inhalation RfC and the focus on extended periods of exposure, the
residential studies were conducted in a manner that best fits the
criteria for developing an RfC and using it in the benefits analyses
that the Agency is being required to conduct.

4. Are there other sensory irritation endpoints besides eye irritation
for which a concentration-response function can and should be developed
for assessing effects at concentrations below 100 ppb?  If so, please
provide references of appropriate studies and a description of how the
studies could be used to develop a concentration-response function.

Sensory irritation may be indirectly reflected in pulmonary function
responses measured after exposure to formaldehyde. However, these
studies were not conducted at levels that would be relevant to what the
Agency is being charged with investigating.

Asthma

Several studies of the association between formaldehyde and asthma have
been published and these were quantitatively summarized in a
meta-analysis by McGwin et al. (2010).

5. Please comment on whether the McGwin et al. (2010) meta-analysis is
an appropriate summary of the association between formaldehyde
inhalation exposures and childhood asthma.

There are many problems with the studies used in the meta analysis
conducted by McGwin et al. (2010). Relative to the description of the
metal analysis, the Agency did a reasonable job of pointing out many of
the strengths and weakness that McGwin and colleagues noted in their
paper, but this description could be improved. 

There are various inconsistencies in the studies selected for the meta
analysis that would make this reviewer reticent to put much trust in the
findings for applicability to the U.S. population. First, only one study
was conducted in the U. S. and that study did not find a statistically
significant odds ratio (OR) with exposures being in the range of 30 to
80 µg/m3 of formaldehyde. Moreover, the sample size of the U.S. study
was greater than that of 5 of the other studies contained in Table 1 of
McGwin et al. (2010).

A Swedish study with exposures no greater than 10 µg/m3 had an OR more
than twice that of any other study even though several of those studies
involved formaldehyde exposures in the 5 to 70 µg/m3 range, including
another Swedish study. The Zhao et al. (2008) study results are
inconceivable to this reviewer for inclusion in a reliable meta
analysis. With school exposures (1 to 5 µg/m3) and outdoor exposures (5
to 7 µg/m3), mean OR estimates were found to represent decreased risk,
yet the upper bound of the 95 % confidence interval was 17 for the
school OR and more than 2.3 million for the outdoor OR for the same set
of 1,993 children having a mean age of 12.8 years. 

McGowan et al. (2010) state that to evaluate whether the results they
found were unduly influenced by any individual study and to determine if
there was any publication bias, an influence plot and a funnel plot,
respectively were used. However, they did not present these plots in
their paper. The computed values of the Q and I2 statistics were stated
to support the presence of moderate between-study heterogeneity. The EPA
document cites values for these statistics but does not explain how they
are calculated, but should do so. Q (better known as Cochran’s Q) is a
classical measure of heterogeneity and is calculated as the weighted sum
of squared differences between individual study effects and the pooled
effect across studies, with the weights being those used in the pooling
method. Q is distributed as a chi-square statistic with k (number of
studies) minus 1 degrees of freedom (d.f.). The I2 statistic describes
the percentage of variation across studies that is due to heterogeneity
rather than chance and is computed as 100* (Q – d.f.)/Q.

On page 18 of the EPA document, the overall OR for a fixed effects model
is given as 1.026 but on page 20 the same result is stated as an OR of
1.03. The Agency needs to be consistent. If the data are only sufficient
to give an OR with 2 digits after the decimal point, say so and change
the value on page 18 and vice versa. For the reader to better judge the
reasonableness of the meta analysis results, the EPA document should
include Table 1 of the McGwin et al. (2010) paper in addition to Figure
1 from that paper.

6. Please comment on whether EPA chose the appropriate function from the
McGwin et al. paper to use in the benefits assessment.

The EPA document does not specifically state which model result (i.e.
fixed effects or random effects) was selected. Rather they state that
the mean OR was 1.24 per 10 ug formaldehyde per m3 for both models. One
only determines that the Agency chose to use the random effects model if
one calculates the transformed 95 % confidence interval for an OR for an
increase of 1 ppb and sees that what is stated on page 18 corresponds to
the confidence interval for the random effects model. 

The lack of significant between-study heterogeneity in the meta analysis
of McGwin et al. (2010) supports using a fixed effects model; however,
the extent of the uncertainties among the studies and the fact that the
random effects confidence interval will always be larger than the fixed
effects confidence interval supports the conservative selection by the
Agency to use the random effects model results.

In view of the additional problems this reviewer discussed earlier
concerning the choice of studies for inclusion in the meta analysis,
this reviewer would be reticent to have the benefits assessment use only
one function (i.e., the OR of 1.03 per one ppb increase in formaldehyde
exposure with a 95 % confidence interval from 1.008 to 1.047). The
single U.S. study should also be evaluated wherein the OR is 1.0083 per
1 ppb increase in formaldehyde exposure with a 95 % confidence interval
of 0.97 to 1.045.  

Female Reproductive Toxicity

The literature base describing reproductive effects of formaldehyde
exposure in humans is limited; however, there are supportive findings
from the animal toxicological literature.  The epidemiologic literature
suggests an association between formaldehyde and increased risk of
spontaneous abortion and a related delay in time to pregnancy (i.e.,
reduced fertility).  The only study with quantitative
concentration-response data was Taskinen et al. (1999).

The a priori hypothesis in Taskinen et al. (1999) was that formaldehyde
would be associated with decreased fertility.  The primary analyses
showed that there was a statistically significant reduction of fertility
among the women in the highest exposure group that appeared to be
concentration dependent.  Post-hoc subgroup analyses showed that
stratification of the 39 women in the highest exposure group by reported
glove use yielded effect estimates whose confidence intervals
overlapped.  The FDR for the 17 women who were classified as not wearing
gloves was 0.51 [95% CI: 0.28−0.92].  The FDR for women who were
classified as always or sometimes wearing gloves was 0.79 [95% CI:
0.47−1.23].  

7. In light of the results regarding glove use, in what way(s) can the
Taskinen et al. results be used to support calculating a concentration
response for formaldehyde inhalation risk to time to pregnancy?

The results after stratifying for glove use are indirectly supportive
for an inhalation effect of formaldehyde on the fecundity density ratio
(FDR). Those women who did not wear gloves had a statistically
significantly decreased FDR of 0.51 (i.e., 95 % confidence interval did
not include 1). Women classified as always or sometimes wearing gloves
had a FDR closer to 1 and their confidence interval included 1. Thus,
the inclusion of women wearing gloves would tend to move the overall FDR
closer to 1, where FDR values lower than 1 can be interpreted as an
adverse effect in that the time to achieve conception is delayed. This
was indeed the case since a FDR of 0.64 was found for all women in the
highest exposure group. This provides evidence in support of the
hypothesis that effects on FDR were most likely due to inhalation
exposure rather than to dermal exposure to formaldehyde.

8. Please comment on whether the concentration-response function is
appropriately derived from the Taskinen et al. (1999) study.

A concentration-response function can be derived from the Taskinen et
al. (1999) study, and the selected function must be one for which there
was an adjustment for background exposure levels; however, the EPA
document does not actually state whether they will use the linear
function of risk or the exponential function of risk equations that are
discussed in the document (i.e. will the Agency use Eq. 2-5 or 2-6?).

The two curves presented in Figures 6 and 7 of the Agency document are
not explained either in the figure legend or in the text. While they
obviously relate to the regression of risk on formaldehyde
concentration, their difference is not explained. The Agency clearly
picked up on most of the criticisms provided by the NAS review of the
draft IRIS document related to the Taskinen et al. (1999) study.
However, the Agency’s section on the discussion of uncertainties does
not comment at all on the clear possibility of confounding due to xylene
exposures in the Finnish worker study. Also, there is no discussion of
the possible magnitude or direction of any exposure misclassification
bias in the Taskinen et al. (1999) study. If the Agency is going to use
the Taskinen et al. (1999) study in their economic analysis, more
attention to a description and discussion of the study’s strengths and
weaknesses is needed. 

III.	SPECIFIC OBSERVATIONS

None.Review By:

Rebecca T. Parkin, Ph.D., MPH

Peer Review Comments on EPA’s Draft Document Approach to Assessing
Non-cancer Health Effects from Formaldehyde and Benefits from Reducing
Non-cancer Health Effects as a Results of Implementing Formaldehyde
Emission Limits for Composite Wood Products

Rebecca T. Parkin, Ph.D., MPH

The George Washington University Medical Center 

July 13, 2011

I.  GENERAL IMPRESSIONS 

The draft is clearly organized and, in large part, well-written.  Minor
inaccuracies are noted below.  Most sections were clear, although some
key issues were not mentioned or are not clear; these concerns are noted
below.  Except in a few instances as noted below, the conclusions are
effectively supported by the text provided.

Uncertainties are important to discuss.  EPA has captured many of the
important ones, but may need to add more to these sections based on this
and other reviewers’ comments.  Being thorough in all uncertainty
sections is crucial to the foundations of the benefit assessments.

II. RESPONSE TO CHARGE QUESTIONS

 

Sensory Irritation

The literature base of residential epidemiology studies of the
irritation effects with quantitative concentration-response data was
limited, but data were available from which a concentration-response
function could be estimated for eye irritation.

1. Please comment on whether the concentration-response function is
appropriately derived from the studies used in the draft Approach
Document.

First, I will comment on each study and then on the derivation process.

Liu et al (1991): This paper adequately describes how homes were
randomly selected, but omits some key information related to monitoring
results for the homes included.  The study was conducted throughout
California (p. 93); a very large state with a wide range of climatic
conditions.  “Summer” and “winter” sampling periods were used,
but these seasons are quite different in different parts of the state. 
It is well-known that temperature and relative humidity affect
formaldehyde (HCHO) levels, but there is no mention of how these factors
varied in the study homes or how they may have affected measured HCHO
levels.  Further, the Quality Assurance/Quality Control methods are not
described, and the number of laboratories used to analyze the passive
area monitoring samples is not stated.  These omissions leave this
reviewer without key information to assess the accuracy or validity of
the HCHO levels reported.  

Residents received instructions on how to uncap, place and mail back the
samplers.  The use of the kitchen and master bedroom as monitoring sites
may be most relevant for the homeowner and less relevant for other
occupants.  Although averaging the monitored levels for each home may
have lost some useful variation, it is not likely that in fairly small
homes (such as mobile homes) that this loss – except for peak
exposures – would be particularly important.  

The study appropriately controls the impact of smoking indoors, and does
not discuss gas cooking stoves or appliances, opening of windows and
doors, or other factors that have been reported by other authors as
having non-significant impacts on HCHO levels.  

More importantly, if I am reading the article correctly (p. 92), some
indoor samples in this cross-sectional study were taken after symptoms
were recorded.  This conflicts with determining the appropriate temporal
sequence of cause preceding effect.  I believe my reading is different
from what EPA states on p. 10 of their draft, which indicates that
exposures were recorded in a period preceding the documentation of
symptoms.

Hanrahan et al (1984):  This cross-sectional study, conducted in
Wisconsin during July 1979, sampled a smaller number of mobile homes
(65) than did Liu et al.; it also had a lower percentage of participants
among those contacted (31% vs. 44%).  In this study, personal samplers
were used and 33 outdoor samples were taken (at least 1 in each mobile
home park).  The authors noted that prior to and during sampling periods
home windows were closed, gas appliances were turned off and smoking was
prohibited.  

The Quality Control and specific laboratory methods were noted (unlike
in Liu et al).  The indoor samples for each home were averaged, as in
Liu et al, with similar but less specific sampling locations noted.  It
is not clear whether the residents or the researchers collected the air
samples in this study.  

Smoking was appropriately considered, and indoor and outdoor temperature
and humidity, and appliance and construction characteristics were noted.
 Some but not all of these factors’ influences on HCHO levels were
presented in the article.

Derivation:  The seasonal and cumulative averages and ranges of indoor
levels, the concentration-response data are correctly extracted from the
Liu et al. article.  The method to estimate the average number of hours
indoors seems appropriate; however, this reviewer wonders whether EPA
attempted to determine whether the authors still have the raw data that
could be used for a more accurate assessment of hours spent indoors.

EPA had to make many more adjustments when using the Hanrahan et al.
article.  First, EPA estimated the lower bound for the indoor levels
based on the standard deviation for the outdoor levels; this is a
reasonable approach given the missing data for the indoor levels. 
Second, the Agency assumed that the distribution of indoor levels is
log-normal in order to estimate the number of samples with HCHO levels
below 100 and 50 ppb.  However, if this assumption or the estimated
standard deviation is not correct, the numbers of samples could be quite
different.  

It is not clear in Figure 2 whether the data shown are for EPA’s or
the authors’ use of “eye irritation;” e.g., does “eye
irritation” here mean the “burning eyes” AND “eye irritation”
results in the article?  If combined data were used, did EPA assume that
all of these symptoms were distinct and not co-occurring events?  Are we
looking at data for the number of symptoms reported or the number of
people with eye symptoms (meaning “burning eyes” and “eye
irritation”) at each level of exposure?  This question should also be
asked of the data shown in the draft Table 1.  Without knowing whether
EPA combined data in Figure 2 and Table 1, it is difficult to comment
further on this part of the draft.

The conversion of data in the article to prevalence odds seems
appropriate.  The display of estimates below 150 ppb in Figure 4b are
derived from Equation 2-2 and are likely correct.   This reviewer did
not verify the mathematics.

The EPA authors have noted many key uncertainties in the two studies and
how these affect their estimates of the relational function.  However,
the validity of the HCHO monitored levels remains a question, due to
missing information in the articles; exposure misclassification is
possible in both studies but is not discussed in the draft.  

		

2. Please comment on the validity of extrapolating the
concentration-response function and using the function to estimate
effects at concentrations below 100 ppb.

The validity of the extrapolation in the Hanrahan et al. paper depends
in part on what data were used for the outcome measure.  If “burning
eye” AND “eye irritation” data were combined, more information is
needed in the draft to describe what assumptions were made in the
combining process.  (See comments above.)

The validity of the estimates below 100 ppb based on the Hanrahan et al.
data is strengthened by their alignment with the results in the Liu et
al. study.

However, clearer discussion of the Agency’s choice of and
uncertainties in the points of departure would improve this portion of
the draft. 

3. The studies considered for the concentration-response function are
residential studies because residential exposures are the focus of the
rule and the residential studies avoid some of the limitations of other
types of studies (e.g., controlled exposure chamber studies used high,
acute exposures, and occupational studies reflect high exposure levels
and potential dose-rate effects from peak exposures).  Are there other
studies (e.g., chamber or non-residential studies) that should be used
as supporting evidence for the currently derived concentration-response
function for eye irritation?  If so, please provide information about
the studies and discuss how they should be used to support the
concentration-response function.

I am not aware of any other studies that should be considered for this
function.

4. Are there other sensory irritation endpoints besides eye irritation
for which a concentration-response function can and should be developed
for assessing effects at concentrations below 100 ppb?  If so, please
provide references of appropriate studies and a description of how the
studies could be used to develop a concentration-response function.

I agree with EPA and NAS that eye irritation is the most sensitive
sensory irritation endpoint and therefore is the most appropriate
endpoint to use. 

Asthma

Several studies of the association between formaldehyde and asthma have
been published and these were quantitatively summarized in a
meta-analysis by McGwin et al. (2010).

5. Please comment on whether the McGwin et al. (2010) meta-analysis is
an appropriate summary of the association between formaldehyde
inhalation exposures and childhood asthma.

The authors clearly built on well-accepted guidelines for systematic
meta-analysis (Stroup et al., 2000).  The study selection/exclusion
criteria and process used by McGwin et al. is well-documented in their
article.  The variables extracted and utilized are stated.  The
determination to use a consistent metric (odds ratio per 10 µg/m3 with
the related 95% confidence interval) across the 7 studies for which
actual HCHO measurements were available is appropriate.  The methods
used to pool the odds ratios and test for heterogeneity are standard
analytic approaches.  Their evaluation of the potentially undue
influence of any one study, using influence and funnel plots, adds
strength to their analysis.  The results are shown with and without
Rumchev et al., allowing the reader to judge whether this paper should
have been included or not.  The data in Table 2 indicate that the
magnitude of the HCHO effect on asthma prevalence is significant whether
this study is included or not.  The presentation of the Q and I2
statistics guides the reader to the key results for interpreting the
data in this table.  In summary, this reviewer finds that the McGwin et
al. meta-analysis is an appropriate summary and synthesis of the
eligible studies reporting a quantitative relationship between HCHO and
prevalence of childhood asthma.

6. Please comment on whether EPA chose the appropriate function from the
McGwin et al. paper to use in the benefits assessment.

The Agency has chosen the appropriate function from the McGwin et al.
paper.  With the exception of the discussion of the rationale for
excluding Rumchev et al., however, more insightful comments about the
strengths of the meta-analysis would lend greater support to the
Agency’s use of and reliance on the fixed-effects model’s OR=1.24. 
Deepening the discussion would improve the support for the Agency’s
conclusions and recommendation.  For example, McGwin et al. stated that
an I2 value of <50% would point to relying on the fixed-effect results
(p. 314).  Although the Agency uses the fixed-effect modeling results,
correctly stating that the Q and I2 data indicate low heterogeneity, the
draft does not include the key cut-point (I2<50%) for the use of fixed-
vs. random-effect modeling results.   In this reviewer’s opinion,
pointing the reader explicitly to this cut-point and the relevant data
in McGwin et al.’s Table 2 would sharpen the Agency’s rationale for
choosing the fixed-effect results

The discussion of uncertainties in this section includes the key issues
(such as the changes in definition and determination of “asthma”
over time) and is appropriately written.

Female Reproductive Toxicity

The literature base describing reproductive effects of formaldehyde
exposure in humans is limited; however, there are supportive findings
from the animal toxicological literature.  The epidemiologic literature
suggests an association between formaldehyde and increased risk of
spontaneous abortion and a related delay in time to pregnancy (i.e.,
reduced fertility).  The only study with quantitative
concentration-response data was Taskinen et al. (1999).

s 0.51 [95% CI: 0.28−0.92].  The FDR for women who were classified as
always or sometimes wearing gloves was 0.79 [95% CI: 0.47−1.23].  

7. In light of the results regarding glove use, in what way(s) can the
Taskinen et al. results be used to support calculating a concentration
response for formaldehyde inhalation risk to time to pregnancy?

The article (p. 208) is not clear as to whether the post-hoc analysis
was adjusted for the variables noted in the footnote of Table VI; if
not, the post-hoc results would not be directly comparable to the data
presented there.  That said, the reported post-hoc high exposure group
glove/not odds ratios and confidence intervals expand and are consistent
with the trend shown in Table VI.  However, all of the confidence
intervals overlap each other, yielding a set of non-significant odds
ratios.  If the adjusted post-hoc odds ratios are available, it would be
interesting to see the results of a test for trends comparing high
exposure-gloves not used, high exposure-gloves used, medium exposure,
and low exposure.  The trend probably is not significant, but testing
its significance may offer more information for interpreting the
concentration-response relationship.

8. Please comment on whether the concentration-response function is
appropriately derived from the Taskinen et al. (1999) study.

Exposure measurements were available only for the middle of the study
period, but the authors were able to verify “most” women’s
self-reported exposure levels based on data from their workplaces.  The
article does not state whether this verification was completed to the
same extent across the entire study period.  This reviewer also wonders
whether for some women (outside of the undefined “most”) the authors
assumed that self-reported HCHO exposure levels were accurate and did
not change during the study period.  The Agency draft notes general
exposure measurement concerns and agrees with Taskinen et al. regarding
their modest response rate to the mailed questionnaire.  

 

Although Taskinen et al. note that reporting biases may be present in
their study, they did not report any evaluation of their data to address
that concern.  For example, the response rates for women with
pregnancies in the 1985-90 vs. 1991-95 periods were quite different. 
Although the authors note that “only a few women could not report the
length of time to pregnancy,” they do not indicate whether more women
in the earlier vs. later period were non-reporters.  Also, there is no
mention to what extent the authors believe the recall of the number of
months-to-pregnancy was accurately recalled by either group of women. 
Neither Taskinen et al. nor EPA adequately discuss whether self-reported
time-to-pregnancy (or spontaneous abortion) data are reliable for use as
an outcome measure.  If outcome misclassification is an issue as well as
exposure misclassification, the odds ratios could be further compromised
toward the null.  Discussion of the reliability of self-reported outcome
data should be considered for the next draft.

The Agency’s adjustment for non-occupational HCHO exposures, based on
contemporaneous data, shifts the relationship to the right (from Figure
6 to Figure 7) but does not consider the variation in exposure levels
found by Jurvelin et al.  The draft assumes that all of the women in the
study had the same non-occupational HCHO exposures.  Considering the
range, not just the mean, of Jurvelin et al.’s measurements would be
more informative and would allow for the identification of upper and
lower bounds on the concentration-response relationship.  

III.	SPECIFIC OBSERVATIONS

Eye Irritation 

Page 10, last paragraph, line 4 and Figure 2, line 2.  The EPA draft
states that these samples were taken from 30-60 minutes, but the authors
only mention pumps running “for approximately one hour” (p. 1026). 
Does EPA have information in addition to what is in the article or is
the 30-minutes an error?  

Pages 11-12.  Figure 2 in the draft has a more detailed caption, but
otherwise seems to be exactly the same as the one in the Hanrahan et al.
paper (which is data for “burning eyes”); is it the same as in the
paper?  If so, shouldn’t it clearly indicate that it is “burning
eye” data?  

Page 12.  Table 1 may need a clearer caption indicating whether the data
shown are for “burning eyes” AND “eye irritation” or “eye
irritation” alone.  Wherever combined data are used, EPA should
clarify this usage and possibly create a new name for combined eye
symptom data.   Further, the title of Table 1 would be more accurate if
it included “…extracted and calculated from…”  

Asthma

Page 16, last paragraph, last 2 lines and top of page 1.  The
descriptive language in the Agency draft about using the McGwin et
al.’s I2 statistic would be stronger if made more specific; e.g.,
noting that McGwin et al. used a cut-point of <50% for choosing fixed-
vs. random-effects data (page 314, top of middle column).  It appears
that EPA agreed with this cut-point.  

Page 18, first complete paragraph, line 6.  The Agency states that the
I2 statistic is 11.2 when the Rumchev et al. paper is excluded, but
Table 2 in McGwin et al. shows 11.3.

Female Reproductive Toxicity

Page 22.  The title on Table 2 should state which “response” (TTP)
is presented.IV.	INDIVIDUAL REVIEWER COMMENTS ON 

THE BENEFITS ASSESSMENT APPROACH



Review By:

Glenn C. Blomquist, Ph.D.

Peer Review Comments on EPA’s Draft Document Approach to Assessing
Non-cancer Health Effects from Formaldehyde and Benefits from Reducing
Non-cancer Health Effects as a Results of Implementing Formaldehyde
Emission Limits for Composite Wood Products

Glenn C. Blomquist, Ph.D.

University of Kentucky

July 19, 2011

I.  GENERAL IMPRESSIONS 

Overall, the draft Approach Document is well-written.  Refreshingly,
Standard English language is used extensively and technical terms
introduced when necessary.  Examples to explain terms such as odds
ratios should make the document understandable to educated lay readers
with varied backgrounds.  The key concentration-response functions are
clearly presented in the form of equations and graphs or tables.  This
draft is readable.  Information presented appears to be accurate and
representative of studies that are the source of the information.  Brief
descriptions of key studies are given in addition to the findings
relevant to the topic of this draft Approach Document.  The extensive
list of references allows the reader to learn more about the studies
from which crucial estimates are taken.  Conclusions regarding
concentration-response functions and valuation (monetization) as
reflected in their use in the examples follow from the information and
models presented in a logical and reasonable way.  My assessment of
presentation applies to the entire document.  My assessment of accuracy
and soundness applies mostly to Section 3 Benefits Assessment Approach
because of my background in environmental and health economics.

II. RESPONSE TO CHARGE QUESTIONS

 

Sensory Irritation

The concentration-response function for the prevalence of eye irritation
is based on Hanrahan et al. (1984). While Hanrahan does not identify the
relevant time period for this relationship, Liu et al. (1991) produced
qualitatively similar results for a two-week time period. We assumed
that the concentration-response relationship for eye irritation is for a
two-week period and multiplied the concentration-response function times
26 to obtain the number of cases of eye irritation experienced in one
year. 

The concentration-response curve used is non-linear, which means that
the annual approach may not produce an exact measure of the benefits,
but it is believed that the bias is probably relatively small given the
small change in exposure.

1. Please comment on whether the concentration-response function is used
appropriately in the benefits assessment approach.

The change in exposure expected from the policy appears to be small
relative to typical exposure.  It follows that the linear approximation
implied by multiplication by 26 to obtain annual changes in number of
cases of eye irritation should produce only a small bias.  It seems as
reasonable as other approaches.

The concentration-response curve is for an acute effect occurring at a
time less than the one-year time step of the exposure model. The
willingness to pay to avoid cases of eye irritation is calculated for
each year of analysis by multiplying the number of cases in a year by
the marginal willingness to pay value to avoid one case. This may
produce biased results, but the direction of that bias is unknown.

On one hand, some studies have suggested that the willingness to pay per
symptom-day declines as the number of symptom-days avoided increases. In
other words, the value to reduce two symptom-days of some effect is less
than two times the value for one symptom-day. Therefore, this approach
may overstate the total value. 

On the other hand, this analysis only applies values to one symptom, eye
irritation, largely because this is the only symptom for which an
effective concentration-response curve was derived. However, there is
evidence that there are other forms of minor eye, nose, and throat
irritation associated with formaldehyde exposure. Since eye irritation
is the largest of the median values from the Tolley et al. (1986) study,
the values used here likely represent a lower bound on the true effect
of multiple symptoms. A reasonable upper bound of the total effect for
one symptom-day would be sum of the willingness to pay for one
symptom-day of each effect individually.  

2. Please comment on whether the annualized approach as opposed to the
lifetime approach is appropriate in this context.

I admit to being confused by the statement that eye irritation has the
largest of the median values in the Tolley et al. (1986) study.  (For
full disclosure, I am one of et al.)  Much of that study was published
in Tolley, Kenkel, and Fabian Valuing Health for Policy (1994)
particularly in Chapter 4 by Kenkel, Berger, and me.  In Table 4.3A on
page 88, the median values for willingness to pay to avoid an additional
symptom day are $11 for coughing, $12.50 for eye irritation, $13 for
throat irritation, and $14 for sinus congestion.  All three are similar,
but eye irritation is not the largest.  It will not make much
difference, however, if either of the other two is used instead of
$12.50 for eye irritation.

Comparing results reported in Table 4.3A and Table 4.3B suggests that
willingness to pay per symptom-day declines as the number of symptom
days avoided increases.  For example, the median value for avoiding 30
symptom-days of eye irritation is $100 which is less than 30 times
$12.50 (or $375).  This suggests that multiplying the $12.50 times the
number of days of eye irritation avoided will bias the annual estimate
for avoiding eye irritation upwards if there are multiple days for
individuals.  Declining marginal willingness to pay for avoided
illnesses is also found in the recent study,   SEQ CHAPTER \h \r 1
Bosworth, Ryan, Trudy Ann Cameron, and J.R. DeShazo. “Demand for
Environmental Policies to Improve Health: Evaluating Community-Level
Policy Scenarios” Journal of Environmental Economics and Management 57
(2009): 293-308.

Table 4 also shows willingness to pay values for avoiding combinations
of symptoms in a day.  The willingness to pay values for avoiding
combinations of symptoms in a day are less than the sums of the
willingness to pay values for avoiding the symptoms separately.  The
declining marginal value of adding another light symptom suggest that a
reasonable upper bound of the annual value of the total effect for one
symptom-day would be the sum of the willingness to pay for one
symptom-day of each effect individually.

On page 34, the paragraph that begins with Tolley et al. (1986) is
mostly correct.  (I am a coauthor on that Tolley report and the Berger
et al. (1987) article.)  It is correct, as stated, that the mean
willingness to pay to relieve one symptom-day of a light symptom was $25
- $50.  It is mostly correct, as stated, that Berger et al. report
higher willingness to pay values for these same symptoms.  The
difference is due to different samples.  The Tolley et al. values are
based on a sample of 176 respondents, some of whom had experienced the
symptoms and some of whom had not.  The Berger et al. values are based
on a subsample of 119 who actually had experienced a symptom.  The mean
values for the subsample that had experienced the symptoms are greater
than for the full sample for five of the seven light symptoms, nearly
the same for one, and less for the remaining one.  The mean value for
eye irritation was $48.48 for the 16 who reported eye irritation whereas
the mean value was $27.73 for all 176 who responded.  The values
reported are in 1984 dollars in both the Tolley et al. report and the
Berger et al. article.  The median value for eye irritation of $12.50
that Weitzel (1990) found is also reported in Tolley, Kenkel, and Fabian
Valuing Health for Policy (1994) in Table 4.3 on page 88.  It is correct
that it too is in 1984 dollars. 

On page 36, if the BLS Inflation Calculator is used to inflate the
median value of $12.50 for eye irritation from 1984 dollars to 1999
dollars, the value is $20.03 as shown in Table J-10.  If the same is
done for 1984 to 2006, the value is $24.25, which rounds to $24, as
reported below Table J-10.

Two recent studies find that different lifetime risk profiles are valued
differently;  see   SEQ CHAPTER \h \r 1 Nielsen, Jytte Seested, Susan
Chilton, Michael Jones-Lee, and Hugh Metcalf. “How Would You Like Your
Gain in Life Expectancy to Be Provided? An Experimental Approach”
Journal of Risk and Uncertainty 41 (December 2010): 195-218 and Cameron,
T.A., J.R. DeShazo, and P. Stiffler. “Demand for Health Risk
Reductions: A Cross-national Comparison between the U.S. and Canada”
Journal of Risk and Uncertainty 41 (December 2010): 245-273.  Given the
limited information on the symptoms of interest it is not clear how
incorporate these finding into the Approach Document.

The concentration-response function for the prevalence of eye irritation
is for a single two-week period.  The annualized approach used an
adjusted concentration-response relationship that consisted in
multiplying the concentration-response function obtained from Hanrahan
et al. (1984) by 26 two-week periods.

3. Please comment on whether the adjustment of the
concentration-response function is appropriate in the annualized
approach.

It seems as reasonable as other possible approaches.

Asthma

The concentration-response relationship for asthma incidence is given as
an odds ratio for children exposed to formaldehyde. An odds ratio is a
way of comparing the difference in the probability of an event, in this
case asthma diagnosis, occurring in groups exposed at different levels.
The odds ratio is used to determine the willingness to pay for any
change in emissions exposure over a 16 year period. We assume that the
probability of being diagnosed with asthma is uniformly distributed
across these 16 years. 

The use of an odds ratio implies that the concentration-response
relationship is non-linear. However, given that the change in exposure
is relatively small, the results from using this implied non-linear
function are anticipated to be similar to the results if we were to use
a linear function.

4. Please comment on whether the concentration-response function is used
appropriately in the benefits assessment approach.

It seems as reasonable as other possible approaches.

The valuation of asthma effects from formaldehyde exposure is based on
valuing the incidence of asthma rather than the more standard valuation
based on the exacerbation of asthma symptoms. The reason for this is
that a statistical relationship between formaldehyde exposure and asthma
exacerbation has not been sufficiently well established, but there is a
statistical relationship between formaldehyde exposure and asthma
occurrence or incidence. However, the only time that EPA has valued
asthma exacerbation is for the first prospective study of the benefits
and costs of the Clean Air Act (EPA 1999). This is probably due to a
lack of an appropriate concentration-response relationship. The EPA’s
Office of Air does include recommended values for the monetization of
chronic asthma in its BenMAP documentation (EPA 2011b), which is based
on an average of the results from Blumenschein and Johannesson (1998)
and O’Conor and Blomquist (1997).

For this analysis, EPA treats prevalence as a measure of cumulative
incidence by assuming that the disease duration is the rest of the
child’s life. However, the EPA (2000a) has previously assumed that a
portion of asthmatic children will become asymptomatic as they move into
adulthood.

5. Please comment on whether the annualized approach as opposed to the
lifetime approach is appropriate in this context.

The approach used seems reasonable except for ignoring that some of the
asthmatic children are likely to become asymptomatic as adults.  If the
incidence and prevalence data can support outgrowing asthma, then it
should be incorporated in the estimates of valuing the incidence of
asthma.  Otherwise, the estimates will be biased upward.

I agree with the choice of the best estimate from O’Conor and
Blomquist (1997) for the implied annual value of asthma control (page
47).  It is $1,474 and is based on the risk-risk tradeoff and the
assumed value of statistical life of roughly $6 million.  The $1,474 is
close to the nonparametric estimates of $1,500 for the 5 unit change and
$1,440 for the 10 unit change.  It is also close to the parametric
estimates assuming a value of statistical life of $6 million of $1,468
for Model 3 and $1,504 for Model 4.  They are the more complete
specifications of the drug choice logistic regression.  Since surveys
were done in 1995, the values are in 1995 dollars.

The implied annual value of asthma control of $1,474 in 1995 dollars is
also close to the best estimates from our more recent work; see Glenn C.
Blomquist, Mark Dickie, and Richard M. O’Conor. “Willingness to Pay
for Reducing Fatality Risks and Asthma Symptoms:  Values for Children
and Adults of All Ages” Resource and Energy Economics 33 (May 2011):
410-425. In this recent work, we are able to estimate the willingness to
pay without assuming a value of statistical life.  As Table 5 shows, for
the ordered logistic regression for willingness to pay for the more
effective drug we have a larger sample for the preferred specification
(263 in 2011 > 75 in 1997) and the ratio of the coefficient for efficacy
in controlling asthma to its standard error is much larger (3.14 in 2011
> 0.70 in 1997).  This facilitates estimation directly from the
willingness to pay logit rather than indirectly from the risk-risk logit
and an assumed value of statistical life.  We were also able to test for
selection bias related to the choice of drug at the first stage of the
hybrid contingent valuation.  (We found we could not reject the null of
no correlation between errors and so based estimates of willingness to
pay on the logit regression without selection.)  Lastly, we made
estimates based on using only “definitely yes” as a “yes”
instead of both “definitely yes” and “probably yes” as a
“yes” as in O’Conor and Blomquist (1997).  In Table 6 we report
values that turn out to be comparable to the best estimate from our
earlier work.  For the average age adult (45 years) our estimate for the
annual value of asthma control is $1,960 in 2007 dollars (or $1,441 in
1995 dollars).  From Table 6 which reports annual values of asthma
control by age we cautiously summarize saying that the pattern of
results hints that values are higher for children and young adults.  For
the average age child (11 years) our estimate for annual value of asthma
control is $2,842 in 2007 dollars (or $2,089 in 1995 dollars.)  This
pattern is relevant for estimating the asthma related benefits because
the ages considered are 0 – 16 and point estimates for values for
children are greater than the value for the average age adult.  The
pattern suggests that estimates based on the average value for adults
are probably underestimates, at least with respect to this factor.

The concentration-response relationship is based on an odds ratio for a
period of time longer than the one year. The exposure model produces
results for one-year of exposure to formaldehyde for each age, housing
type, and climate cohort. As a consequence, an adjustment to the simple
model needed to be made to model the effect on individuals who are
currently alive when the regulation takes effect.

As an upper bound estimate, we assume that the odds ratio can be applied
to the one year reduction in the same fashion as it was applied for the
16 year time period. The problem with this approach is that it
effectively treats the exposure as an acute effect. If the odds ratio
indicates a particular percentage reduction in asthma cases from a
reduction in formaldehyde exposure over 16 years, the upper bound
approach would suggest the same percentage reduction in asthma case in
one single year if we were to witness the same reduction in formaldehyde
exposure in one year. The lower bound approach assumes that a particular
exposure reduction for one year produces the same effect as if that
exposure reduction were spread out over the entire 16-year time period. 

6. Please comment on the upper and lower bound methodology used.

It seems as reasonable as other possible approaches.

The prevalence data suggest that 7.2% of children age 0-4 have been told
that they have asthma. 16.4% of children aged 5-14 have been told that
they have asthma. We assume that the probability of being diagnosed is
uniformly distributed across the 0 to 16 age bracket, so that the
incidence rate would be 1.44% for the first five years and 1.49% for the
next eleven years. Allocating cases uniformly across the age brackets
provides us with a way to calculate the annual benefits of any exposure
reduction that occurs for the entire time period, but may not be
appropriate if the diagnosis of asthma for very young children (e.g.,
age 0-2) differs from older children.

7. Is the assumption about applying the concentration-response function
to children aged 0 to 16 appropriate, given the asthma onset studies’
focus on school-aged children and adolescents, and given the potential
uncertainty of the Rumchev study involving younger children?

It seems as reasonable as other possible approaches.  A question about
this and the other concentration-response functions and exposure is
whether or not they take into account averting behavior.  Smell leads to
open windows and more time outside of house and less exposure.

Female Reproductive Toxicity

The linear concentration-response function accounting for background
exposures based on Taskinen et al. (1999) is used for the annual
approach.

8. Please comment on whether the concentration-response function is used
appropriately in the benefits assessment approach.

It seems as reasonable as other possible approaches.

Monetization of delayed fertility has been discussed in the literature
but has never been included in primary analyses of EPA rules because it
was believed that the scientific knowledge on reproductive and
developmental health effects was not strong enough to quantify risk in
the primary benefits analysis.

While there appears to be good data on the cost of treatment, there are
few studies on the willingness to pay to reduce delayed conception. Most
economic valuation studies regarding fertility focus on an
individual’s or a couple’s willingness to pay for infertility
treatment, particularly in-vitro fertilization (IVF).  Conceptually, the
willingness to pay approach is preferable and could be deduced from the
valuation studies. However, given that there has been less research on
these willingness to pay values than other values and the fact that the
values have not been used in previous economic analyses, we used the
more straight-forward cost of illness approach for this analysis.

As an upper bound, we assume that any woman who has difficulty
conceiving (i.e., time to pregnancy exceeds 12 months) will obtain some
type of fertility service. This may be a reasonable upper bound, but
there are certainly women who have trouble becoming pregnant and do not
seek fertility treatment. 

The willingness to pay for fertility treatment is treated as an expected
value. The average cost of various cycle-based fertility treatments is
multiplied times the probability of obtaining that treatment at various
ages. There is probably a small opposite effect associated with an
increase in pregnancy from reduced formaldehyde exposure by women who do
not want to become pregnant, but we do not consider this effect

9. Please comment on whether the annualized approach as opposed to the
lifetime approach is appropriate in this context.

It seems as reasonable as other possible approaches.

On page 58, the choice is made to use the cost-of-illness estimates from
Katz et al. (2011) for valuing difficulty in conceiving due to
formaldehyde exposure.  A strength of these estimates is that apparently
most fertility treatment costs are paid for out-of-pocket by
consumer/patients rather than by third parties through private or public
insurance.  As such, consumers reveal that the treatments are worth at
least as much as they paid.  Their maximum willingness to pay would be
expected to be greater than these costs because of consumer surplus. 
(This does not take into account any utility or disutility of
treatments.)  If data were available and demand for fertility treatment
could be estimated, then a better estimate of maximum willingness to pay
could be made.  This suggests using cost-of-illness will produce an
underestimate.  However, if the values are applied to all women who are
exposed to formaldehyde, benefits will tend to be overestimated.  The
reason is that not all the women whose fertility is affected will pay
the market prices for the treatments.  In other words, the price is
known for the various packages of treatments, but the number of women
who are willing to pay the price is not.  The maximum willingness to pay
for treatment for the women whose fertility is affected but choose to
forgo treatment and the number of them are unknowns.  Their values are
likely to be positive, but they are less than the costs.  Katz et al.
report treatment costs per successful outcome of more than $70,000. 
They also report that 72% of the women had college education.  A hunch
is that those women have above average household incomes and spend most
of their time in single family detached homes where exposure reduction
is smallest.  Less educated women with lower household incomes are
probably less likely to pay for treatments at the going prices. 
Assuming that women who have below average incomes and live in mobile
homes will pay the market prices for the same distribution of fertility
treatments is likely to bias the estimate of benefits of reducing
exposure to formaldehyde upward.

The time period associated with the exposure-response function is longer
than the time step of the exposure model. The Taskinen (1999) study
examined the fecundity of women age 20-40. This causes a problem if we
have an individual who does not experience the exposure reduction for
the full time period. We can form an upper and lower bound of this
willingness to pay as we did for asthma prevalence. The upper bound is
based on the assumption that the linear concentration-response function
holds for an acute exposure of one year. The lower bound assumes that an
exposure reduction for one year represents 1/20th of an exposure
reduction for the entire 20-year time period. 

10. Please comment on the upper and lower bound methodology used.

It seems as reasonable as other possible approaches.

The ages used in this analysis reflect the ages reported in the Taskinen
(1999) study, ages 20-40. However, women older than age 40 obtain
fertility treatment. The value of reduced infertility effects from
reduced formaldehyde exposure is not reflected in this analysis in order
to remain consistent with Taskinen (1999).

11. Please comment on whether the concentration-response function was
applied to appropriate age groups given the age groups in the studies. 

It seems as reasonable as other possible approaches.

 

III.	SPECIFIC OBSERVATIONS

Page 53, bottom.  $4.9 instead of $4,9.

Page 56, 3rd paragraph.  Monetization is misspelled.

Page 58.  Van Houtven and Smith (1999) instead of 1997 and $12,500
instead of $12,50.

Page 59.  Is the probability of obtaining cycle-based fertility
treatment shown in Table 8, the best projection for future treatment? 
The utilization percentages given on page 916 in Katz et al. seem
different.

Page 68.  O’Conor instead of O’Connor.

Review By:

A. Myrick Freeman III, Ph.D.Peer Review Comments on EPA’s Draft
Document Approach to Assessing Non-cancer Health Effects from
Formaldehyde and Benefits from Reducing Non-cancer Health Effects as a
Results of Implementing Formaldehyde Emission Limits for Composite Wood
Products

A. Myrick Freeman III, Ph.D.

Bowdoin College

July 15, 2011

I.  GENERAL IMPRESSIONS 

The general model for aggregating over time and across individuals that
is outlined in Section 3.1 is correct.  For other “impressions,” see
my responses to specific charge questions below.

II. RESPONSE TO CHARGE QUESTIONS

 

Sensory Irritation

The concentration-response function for the prevalence of eye irritation
is based on Hanrahan et al. (1984). While Hanrahan does not identify the
relevant time period for this relationship, Liu et al. (1991) produced
qualitatively similar results for a two-week time period. We assumed
that the concentration-response relationship for eye irritation is for a
two-week period and multiplied the concentration-response function times
26 to obtain the number of cases of eye irritation experienced in one
year. 

The concentration-response curve used is non-linear, which means that
the annual approach may not produce an exact measure of the benefits,
but it is believed that the bias is probably relatively small given the
small change in exposure.

1. Please comment on whether the concentration-response function is used
appropriately in the benefits assessment approach.

It was used appropriately.

The concentration-response curve is for an acute effect occurring at a
time less than the one-year time step of the exposure model. The
willingness to pay to avoid cases of eye irritation is calculated for
each year of analysis by multiplying the number of cases in a year by
the marginal willingness to pay value to avoid one case. This may
produce biased results, but the direction of that bias is unknown.

On one hand, some studies have suggested that the willingness to pay per
symptom-day declines as the number of symptom-days avoided increases. In
other words, the value to reduce two symptom-days of some effect is less
than two times the value for one symptom-day. Therefore, this approach
may overstate the total value. 

On the other hand, this analysis only applies values to one symptom, eye
irritation, largely because this is the only symptom for which an
effective concentration-response curve was derived. However, there is
evidence that there are other forms of minor eye, nose, and throat
irritation associated with formaldehyde exposure. Since eye irritation
is the largest of the median values from the Tolley et al. (1986) study,
the values used here likely represent a lower bound on the true effect
of multiple symptoms. A reasonable upper bound of the total effect for
one symptom-day would be sum of the willingness to pay for one
symptom-day of each effect individually.  

2. Please comment on whether the annualized approach as opposed to the
lifetime approach is appropriate in this context.

I don’t understand the question.  What do you mean by the “lifetime
approach”?  If you mean the results shown in Table 3 on p. 42, then
either approach is appropriate.  They are just different ways of
summarizing the data.

The concentration-response function for the prevalence of eye irritation
is for a single two-week period.  The annualized approach used an
adjusted concentration-response relationship that consisted in
multiplying the concentration-response function obtained from Hanrahan
et al. (1984) by 26 two-week periods.

3. Please comment on whether the adjustment of the
concentration-response function is appropriate in the annualized
approach.

Yes, it is appropriate.

Asthma

The concentration-response relationship for asthma incidence is given as
an odds ratio for children exposed to formaldehyde. An odds ratio is a
way of comparing the difference in the probability of an event, in this
case asthma diagnosis, occurring in groups exposed at different levels.
The odds ratio is used to determine the willingness to pay for any
change in emissions exposure over a 16 year period. We assume that the
probability of being diagnosed with asthma is uniformly distributed
across these 16 years. 

The use of an odds ratio implies that the concentration-response
relationship is non-linear. However, given that the change in exposure
is relatively small, the results from using this implied non-linear
function are anticipated to be similar to the results if we were to use
a linear function.

4. Please comment on whether the concentration-response function is used
appropriately in the benefits assessment approach.

It was used appropriately.

The valuation of asthma effects from formaldehyde exposure is based on
valuing the incidence of asthma rather than the more standard valuation
based on the exacerbation of asthma symptoms. The reason for this is
that a statistical relationship between formaldehyde exposure and asthma
exacerbation has not been sufficiently well established, but there is a
statistical relationship between formaldehyde exposure and asthma
occurrence or incidence. However, the only time that EPA has valued
asthma exacerbation is for the first prospective study of the benefits
and costs of the Clean Air Act (EPA 1999). This is probably due to a
lack of an appropriate concentration-response relationship. The EPA’s
Office of Air does include recommended values for the monetization of
chronic asthma in its BenMAP documentation (EPA 2011b), which is based
on an average of the results from Blumenschein and Johannesson (1998)
and O’Conor and Blomquist (1997).

For this analysis, EPA treats prevalence as a measure of cumulative
incidence by assuming that the disease duration is the rest of the
child’s life. However, the EPA (2000a) has previously assumed that a
portion of asthmatic children will become asymptomatic as they move into
adulthood.

5. Please comment on whether the annualized approach as opposed to the
lifetime approach is appropriate in this context.

Again, I don’t understand the question.  What do you mean by the
“lifetime approach”?  The term is not used in the document.

The concentration-response relationship is based on an odds ratio for a
period of time longer than the one year. The exposure model produces
results for one-year of exposure to formaldehyde for each age, housing
type, and climate cohort. As a consequence, an adjustment to the simple
model needed to be made to model the effect on individuals who are
currently alive when the regulation takes effect.

As an upper bound estimate, we assume that the odds ratio can be applied
to the one year reduction in the same fashion as it was applied for the
16 year time period. The problem with this approach is that it
effectively treats the exposure as an acute effect. If the odds ratio
indicates a particular percentage reduction in asthma cases from a
reduction in formaldehyde exposure over 16 years, the upper bound
approach would suggest the same percentage reduction in asthma case in
one single year if we were to witness the same reduction in formaldehyde
exposure in one year. The lower bound approach assumes that a particular
exposure reduction for one year produces the same effect as if that
exposure reduction were spread out over the entire 16-year time period. 

6. Please comment on the upper and lower bound methodology used.

The approach seems reasonable.  The bounds are relatively narrow, as
shown in Table 7.  How will the results be presented in the final
report, as upper and lower bounds or averaged to give a “best”
estimate?

The prevalence data suggest that 7.2% of children age 0-4 have been told
that they have asthma. 16.4% of children aged 5-14 have been told that
they have asthma. We assume that the probability of being diagnosed is
uniformly distributed across the 0 to 16 age bracket, so that the
incidence rate would be 1.44% for the first five years and 1.49% for the
next eleven years. Allocating cases uniformly across the age brackets
provides us with a way to calculate the annual benefits of any exposure
reduction that occurs for the entire time period, but may not be
appropriate if the diagnosis of asthma for very young children (e.g.,
age 0-2) differs from older children.

7. Is the assumption about applying the concentration-response function
to children aged 0 to 16 appropriate, given the asthma onset studies’
focus on school-aged children and adolescents, and given the potential
uncertainty of the Rumchev study involving younger children?

I can’t answer this.  It lies outside my area of expertise.

Female Reproductive Toxicity

The linear concentration-response function accounting for background
exposures based on Taskinen et al. (1999) is used for the annual
approach.

8. Please comment on whether the concentration-response function is used
appropriately in the benefits assessment approach.

It was used appropriately.

Monetization of delayed fertility has been discussed in the literature
but has never been included in primary analyses of EPA rules because it
was believed that the scientific knowledge on reproductive and
developmental health effects was not strong enough to quantify risk in
the primary benefits analysis.

While there appears to be good data on the cost of treatment, there are
few studies on the willingness to pay to reduce delayed conception. Most
economic valuation studies regarding fertility focus on an
individual’s or a couple’s willingness to pay for infertility
treatment, particularly in-vitro fertilization (IVF).  Conceptually, the
willingness to pay approach is preferable and could be deduced from the
valuation studies. However, given that there has been less research on
these willingness to pay values than other values and the fact that the
values have not been used in previous economic analyses, we used the
more straight-forward cost of illness approach for this analysis.

As an upper bound, we assume that any woman who has difficulty
conceiving (i.e., time to pregnancy exceeds 12 months) will obtain some
type of fertility service. This may be a reasonable upper bound, but
there are certainly women who have trouble becoming pregnant and do not
seek fertility treatment. 

The willingness to pay for fertility treatment is treated as an expected
value. The average cost of various cycle-based fertility treatments is
multiplied times the probability of obtaining that treatment at various
ages. There is probably a small opposite effect associated with an
increase in pregnancy from reduced formaldehyde exposure by women who do
not want to become pregnant, but we do not consider this effect

9. Please comment on whether the annualized approach as opposed to the
lifetime approach is appropriate in this context.

Again, I don’t understand the question.  What do you mean by the
“lifetime approach”?  The term is not used in the document. Perhaps
the charge question should be "Please comment on the annual approach to
calculating a total WTP per year as described in Section 3.4.2."  My
answer to that question is that if a valid individual WTP per year can
be obtained, equation (18) is the appropriate way to calculate aggregate
WTP per year.  But for reasons I point out in the Specific Observations
section below (see my comments to question #3), I have serious doubts
about the ability to obtain a valid individual WTP per year.

The time period associated with the exposure-response function is longer
than the time step of the exposure model. The Taskinen (1999) study
examined the fecundity of women age 20-40. This causes a problem if we
have an individual who does not experience the exposure reduction for
the full time period. We can form an upper and lower bound of this
willingness to pay as we did for asthma prevalence. The upper bound is
based on the assumption that the linear concentration-response function
holds for an acute exposure of one year. The lower bound assumes that an
exposure reduction for one year represents 1/20th of an exposure
reduction for the entire 20-year time period. 

10. Please comment on the upper and lower bound methodology used.

The approach seems reasonable.  How will the results be presented in the
final report, as upper and lower bounds or averaged to give a “best”
estimate?

The ages used in this analysis reflect the ages reported in the Taskinen
(1999) study, ages 20-40. However, women older than age 40 obtain
fertility treatment. The value of reduced infertility effects from
reduced formaldehyde exposure is not reflected in this analysis in order
to remain consistent with Taskinen (1999).

11. Please comment on whether the concentration-response function was
applied to appropriate age groups given the age groups in the studies. 

Given my reservations about using the cost-of-fertility treatment to
monetize this effect (see  my comments to question #3 in the Specific
Observations section below), I have no problem in limiting the
quantification of this effect to the age group covered in the Taskinen
study.

 

III.	SPECIFIC OBSERVATIONS

There are a number of typographical and editorial errors.  This is not a
complete list:

Table of Contents.  Page numbers are wrong.

Page 17.  Figure 2 should be labeled figure 5.

	Section 3.  Why do the table and figure numbers start over instead of
continuing throughout the report?  And some of the table numbers seem to
be wrong (e.g., J-10 on page 36).

Page 6.  The report gives the aim of developing concentration-response
relationships as being “ ... to inform the economic benefits
assessment of potential cost savings ...”  This is not about “cost
savings;” it is about welfare gains.

	In some places, coefficients in the concentration-response
relationships are Greek letters.  In other places they are not.

Section 2.1.1, 3rd sentence in the first paragraph.  Needs editing.

Page 53.  Where it says “($93,095 + $58,662 + $90,852= $242,602),”
it should be = $242,609.

Page 58.  It says “willingness to pay of between $3,750 and $12,50 for
fertility treatment.”

Page 59.  The probability of using IVF should be 2.5%, not .25%.

Page 61, last paragraph.  There are two references to Table 10, which
should be Table 12.

Page 63.  The sentence, “This is probably the largest category of
value associated fertility issue from a reduction in formaldehyde
exposure,” needs editing.

I have added the following three charge questions based on my expertise
in non-market valuation:

(1) Please comment on the validity of the monetization of sensory
irritation. The approach to the monetization of this effect is the same
as that used in the First Prospective Section 812 Report.  In my
judgement, this is a valid approach and makes the best use of the
available empirical evidence.  However, it should be noted that the
effect actually valued in the  First Prospective Section 812 Report as
“Upper Respiratory Symptoms” defined as “two or more of the
following symptoms: runny or stuffy nose; coughing; and eye
irritation.”  Since this document states that formaldehyde can cause a
variety of types of sensory irritation in the upper respiratory system
(p. 7), this is appropriate.  But it should be made clear that more than
simply eye irritation is involved.

(2) Please comment on the validity of the monetization of chronic
asthma.  The approach to the monetization of this effect is the same as
that used in the First Prospective Section 812 Report.  In my judgement,
this is a valid approach and makes the best use of the available
empirical evidence. 

(3) Please comment on the validity of the monetization of reduced
fertility. I have serious reservations about the approach taken here to
monetize this effect.  First, “reduced fertility” is not a
well-defined effect.  There is ambiguity as to both the degree of
reduction and its duration.  This makes determining the proper valuation
difficult.  

I agree that conceptually, the willingness to pay (WTP) approach is
preferable to a cost-of-illness or cost-of-treatment approach.  However,
as I read the available WTP studies, they apply to the more serious
infertility rather than to the “subfertility” (p. 21) associated
with exposure to formaldehyde.  Thus, the available WTP studies are not
suitable for benefits transfer.

 

The report takes a cost of treatment approach to valuation.  But it is
not at all clear to me that the “cycle-based treatments” described
here are either appropriate treatments or likely to be effective in the
face of continued formaldehyde exposures.

For these reasons, I recommend that this effect be described and
quantified to the extent that evidence supports quantification, but that
it be acknowledged that there is not an adequate basis for monetization
of this effect.

Review By:

Shelby Gerking, Ph.D.Peer Review Comments on EPA’s Draft Document
Approach to Assessing Non-cancer Health Effects from Formaldehyde and
Benefits from Reducing Non-cancer Health Effects as a Results of
Implementing Formaldehyde Emission Limits for Composite Wood Products

Shelby Gerking, Ph.D.

University of Central Florida

July 23, 2011

I.  GENERAL IMPRESSIONS 

The report is, for the most part, complete and clearly written.  I had
no problem following the narrative, however, an executive summary, a
conclusion and a map of the climate zones would have been helpful.  
Also, in Section 3, there are at least seven issues that warrant
additional thought.  First, the choice of the 3% discount rate, while
not unreasonable, appears from out of nowhere.  Some attention needs to
be given to this choice as well as to the assumption of exponential
discounting.  Other types of discounting considered by behavioral
economists may also be appropriate.   Second, the contingent valuation
estimates of WTP for eye irritation are rather dated—those obtained by
Tolley et al. are 25 years old.  EPA still uses these estimates, but
there have been a number of advances over the years in the application
of stated preference methods aimed at reducing hypothetical bias which
causes stated preference estimates to be too high.  Third, the Dickie et
al. estimates in Table J-10 may not be directly comparable to those of
Tolley et. al. and Loehman et. al. in that they are based on the
averting behavior method (which looks at actual purchases of goods),
rather than on contingent valuation.   In any case, the narrative might
usefully reflect the fact that sound willingness to pay estimates are
elusive and that different methods of estimating WTP often give (very)
different results.  Fourth, both asthma and eye irritation estimates
rely on transferring adult WTP estimates to children.  This procedure,
which may be the only available option, may well lead to substantial
inaccuracies.  Parents, for example, may be WTP more to reduce health
symptoms for their children than they are WTP to reduce those same
symptoms for themselves.  In any case, the issue of WTP for adults vs.
WTP for kids (with citations to recent literature) should be a
discussion point in the report.  Fifth, the WTP estimates presented
implicitly assume that switching to different building materials, or
treating wood products to cut down formaldehyde off-gassing will not
occur for the 30 year horizon analyzed.  Is this assumption reasonable? 
In this same vein, could health effects from formaldehyde off-gassing be
lessened by providing particularly residents of new homes with
information that would lead them to install better ventilation systems? 
If so, the marginal cost of better ventilation (if such a strategy is
appropriate) might then form the basis of a simple WTP estimate along
the lines of the headache example given on p. 31.  Sixth, the population
by housing type and climate zone apparently are assumed not to change
over the next 30 years, even though there have been major changes in
population by housing type and climate zone over the past 30 years. 
This assumption seems hard to defend, although as noted on p. 42 it
would be a difficult task to predict these movements.   Seventh, it
would be helpful to develop the main features of the cohort x housing
type x climate zone model in introductory Section 3.1.  It would be
helpful in this section to explain about the assumptions underlying this
approach (some of these are identified above).  This reorganization will
clarify the narrative at a number of points and avoid misunderstandings
created when readers simply encounter model assumptions in the
explanation of results.

II. RESPONSE TO CHARGE QUESTIONS

 

Sensory Irritation

The concentration-response function for the prevalence of eye irritation
is based on Hanrahan et al. (1984). While Hanrahan does not identify the
relevant time period for this relationship, Liu et al. (1991) produced
qualitatively similar results for a two-week time period. We assumed
that the concentration-response relationship for eye irritation is for a
two-week period and multiplied the concentration-response function times
26 to obtain the number of cases of eye irritation experienced in one
year. 

The concentration-response curve used is non-linear, which means that
the annual approach may not produce an exact measure of the benefits,
but it is believed that the bias is probably relatively small given the
small change in exposure.

1. Please comment on whether the concentration-response function is used
appropriately in the benefits assessment approach.

The non-linear concentration response function is appropriately used in
the benefits assessment approach.  However, the approach used here
appears to take no account of other sources of formaldehyde exposure,
such as workplace exposures.  In any case, the policy-related reductions
in formaldehyde exposure should be treated as reductions in total
exposure, not simply reductions in home exposure.  This refinement may
alter the benefits calculation given the use of a non-linear
concentration response function.

The concentration-response curve is for an acute effect occurring at a
time less than the one-year time step of the exposure model. The
willingness to pay to avoid cases of eye irritation is calculated for
each year of analysis by multiplying the number of cases in a year by
the marginal willingness to pay value to avoid one case. This may
produce biased results, but the direction of that bias is unknown.

On one hand, some studies have suggested that the willingness to pay per
symptom-day declines as the number of symptom-days avoided increases. In
other words, the value to reduce two symptom-days of some effect is less
than two times the value for one symptom-day. Therefore, this approach
may overstate the total value. 

On the other hand, this analysis only applies values to one symptom, eye
irritation, largely because this is the only symptom for which an
effective concentration-response curve was derived. However, there is
evidence that there are other forms of minor eye, nose, and throat
irritation associated with formaldehyde exposure. Since eye irritation
is the largest of the median values from the Tolley et al. (1986) study,
the values used here likely represent a lower bound on the true effect
of multiple symptoms. A reasonable upper bound of the total effect for
one symptom-day would be sum of the willingness to pay for one
symptom-day of each effect individually.  

2. Please comment on whether the annualized approach as opposed to the
lifetime approach is appropriate in this context.

A search of the review document turned up no references to “lifetime
approach”, so the issue to be addressed here is not entirely clear. 
Nonetheless, it is likely that WTP per symptom day of eye irritation
falls as the number of symptom days increase.  This outcome would be a
general implication of utility maximization under a fixed budget.  As
additional resources are spent to relieve eye irritation, the
opportunity cost of foregone consumption will increase.  If eye
irritation is coupled with other symptomatic discomforts, then we may
have a different problem.  On the one hand, suppose that there exists a
good that will reduce eye irritation together with all of the other
symptoms.  Then, the marginal cost (inclusive of both time and money) of
using the good to reduce these symptoms will be the WTP for all
combined.  On the other hand, if more than one good needs to be
purchased to reduce all of the symptoms, then the WTP would be the
marginal cost of the combination of goods that will do the job.  In any
case, it is problematic to simply add up CVM estimates of WTP for
individual symptoms as this is likely to yield a total WTP value that
really is upward biased.  Note that the CVM estimates themselves are
probably upward biased.  

The concentration-response function for the prevalence of eye irritation
is for a single two-week period.  The annualized approach used an
adjusted concentration-response relationship that consisted in
multiplying the concentration-response function obtained from Hanrahan
et al. (1984) by 26 two-week periods.

3. Please comment on whether the adjustment of the
concentration-response function is appropriate in the annualized
approach.

An issue here is that there appear to be different ways to interpret the
Total WTP expression on p. 39.  Should we assume that the same people
experience eye irritation repeatedly for each 26 two-week periods in the
year?  If this (Groundhog Day) assumption is correct, then the WTP value
applied may well be on the high side.  On the other hand, if we assume
that a different set of people experience these symptoms in each of the
two week periods, then the calculations presented may be more
defensible.  

Asthma

The concentration-response relationship for asthma incidence is given as
an odds ratio for children exposed to formaldehyde. An odds ratio is a
way of comparing the difference in the probability of an event, in this
case asthma diagnosis, occurring in groups exposed at different levels.
The odds ratio is used to determine the willingness to pay for any
change in emissions exposure over a 16 year period. We assume that the
probability of being diagnosed with asthma is uniformly distributed
across these 16 years. 

The use of an odds ratio implies that the concentration-response
relationship is non-linear. However, given that the change in exposure
is relatively small, the results from using this implied non-linear
function are anticipated to be similar to the results if we were to use
a linear function.

4. Please comment on whether the concentration-response function is used
appropriately in the benefits assessment approach.

I agree that the bias from using the implied non-linear function is
likely to be small, but please see #1 in the eye irritation section for
a possible qualification.

The valuation of asthma effects from formaldehyde exposure is based on
valuing the incidence of asthma rather than the more standard valuation
based on the exacerbation of asthma symptoms. The reason for this is
that a statistical relationship between formaldehyde exposure and asthma
exacerbation has not been sufficiently well established, but there is a
statistical relationship between formaldehyde exposure and asthma
occurrence or incidence. However, the only time that EPA has valued
asthma exacerbation is for the first prospective study of the benefits
and costs of the Clean Air Act (EPA 1999). This is probably due to a
lack of an appropriate concentration-response relationship. The EPA’s
Office of Air does include recommended values for the monetization of
chronic asthma in its BenMAP documentation (EPA 2011b), which is based
on an average of the results from Blumenschein and Johannesson (1998)
and O’Conor and Blomquist (1997).

For this analysis, EPA treats prevalence as a measure of cumulative
incidence by assuming that the disease duration is the rest of the
child’s life. However, the EPA (2000a) has previously assumed that a
portion of asthmatic children will become asymptomatic as they move into
adulthood.

5. Please comment on whether the annualized approach as opposed to the
lifetime approach is appropriate in this context.

A search of the review document turned up no references to “lifetime
approach” so the issue to be addressed here is not entirely clear. 
Nonetheless, the report makes clear that valuation of asthma effects is
based on valuing the incidence of asthma, rather than symptom
exacerbation.   This approach is crude for two reasons.  First, there is
substantial heterogeneity in cases of asthma.  For some people, asthma
is little more than an annoyance, while for others this illness is
debilitating.  Willingness to pay to avoid asthma would be expected to
vary according to how serious a case is contemplated.  Second,
willingness to pay to reduce additional cases of asthma will not include
the willingness to pay to reduce exacerbated symptoms of those who
already have asthma.   Two other issues: (1) While estimates of
increased incidence of asthma pertain to children (see p. 19), estimates
of WTP to reduce asthma incidence are based on two studies of adults. 
Parents may well be willing to pay more to reduce asthma incidence among
their children than they are WTP to protect themselves from the same
illness.  (2) It might be useful to perform a sensitivity analysis on
the estimates provided to determine how much difference it would make to
assume that a percentage of childhood asthmatics become asymptomatic
when they become adults, rather than assume that childhood asthmatics
carry their illness forward to adulthood.  

The concentration-response relationship is based on an odds ratio for a
period of time longer than the one year. The exposure model produces
results for one-year of exposure to formaldehyde for each age, housing
type, and climate cohort. As a consequence, an adjustment to the simple
model needed to be made to model the effect on individuals who are
currently alive when the regulation takes effect.

As an upper bound estimate, we assume that the odds ratio can be applied
to the one year reduction in the same fashion as it was applied for the
16 year time period. The problem with this approach is that it
effectively treats the exposure as an acute effect. If the odds ratio
indicates a particular percentage reduction in asthma cases from a
reduction in formaldehyde exposure over 16 years, the upper bound
approach would suggest the same percentage reduction in asthma case in
one single year if we were to witness the same reduction in formaldehyde
exposure in one year. The lower bound approach assumes that a particular
exposure reduction for one year produces the same effect as if that
exposure reduction were spread out over the entire 16-year time period. 

6. Please comment on the upper and lower bound methodology used.

The upper bound estimates are obtained by assuming that asthma onset is
an acute effect of formaldehyde exposure particularly among older
teenagers.  I am uncertain as to whether this assumption is at variance
with scientific information regarding the development of this disease
and would defer to another panel member with greater expertise on this
point.  If asthma onset from formaldehyde exposure cannot be considered
an acute effect, then the upper bound estimates would be less credible. 
Nonetheless, the upper bound estimates and the lower bound estimates do
not differ greatly.  Table 7 reports that for the three housing profiles
total discounted upper bound WTP is $4,939,166 and the corresponding
lower bound figure is $4,340,740.  So, the upper bound estimate is about
14% larger than the lower bound estimate.  An alternative to the lower
bound estimate might be to assume that currently living children are
unaffected by reduced formaldehyde exposure.  Instead, we might assume
that the reduced exposure applies only as “new” children as they
enter the model.  These “new” children are then assigned cases of
asthma with the incidence rates described on p. 48 of the review
document.  

The prevalence data suggest that 7.2% of children age 0-4 have been told
that they have asthma. 16.4% of children aged 5-14 have been told that
they have asthma. We assume that the probability of being diagnosed is
uniformly distributed across the 0 to 16 age bracket, so that the
incidence rate would be 1.44% for the first five years and 1.49% for the
next eleven years. Allocating cases uniformly across the age brackets
provides us with a way to calculate the annual benefits of any exposure
reduction that occurs for the entire time period, but may not be
appropriate if the diagnosis of asthma for very young children (e.g.,
age 0-2) differs from older children.

7. Is the assumption about applying the concentration-response function
to children aged 0 to 16 appropriate, given the asthma onset studies’
focus on school-aged children and adolescents, and given the potential
uncertainty of the Rumchev study involving younger children?

I defer to another panel member here as I have no expertise in this
area.  

Female Reproductive Toxicity

The linear concentration-response function accounting for background
exposures based on Taskinen et al. (1999) is used for the annual
approach.

8. Please comment on whether the concentration-response function is used
appropriately in the benefits assessment approach.

 ?  If so, the notation should be changed.  If not, the equation needs
additional explanation.  Second, in these equations, population refers
to women who have difficulty conceiving.  The population of women who
are actively attempting to become pregnant would instead be more
appropriate.  Page 59 indicates that for the proposed rule this latter
definition would be incorporated.  I agree that this change would be
appropriate.  

Monetization of delayed fertility has been discussed in the literature
but has never been included in primary analyses of EPA rules because it
was believed that the scientific knowledge on reproductive and
developmental health effects was not strong enough to quantify risk in
the primary benefits analysis.

While there appears to be good data on the cost of treatment, there are
few studies on the willingness to pay to reduce delayed conception. Most
economic valuation studies regarding fertility focus on an
individual’s or a couple’s willingness to pay for infertility
treatment, particularly in-vitro fertilization (IVF).  Conceptually, the
willingness to pay approach is preferable and could be deduced from the
valuation studies. However, given that there has been less research on
these willingness to pay values than other values and the fact that the
values have not been used in previous economic analyses, we used the
more straight-forward cost of illness approach for this analysis.

As an upper bound, we assume that any woman who has difficulty
conceiving (i.e., time to pregnancy exceeds 12 months) will obtain some
type of fertility service. This may be a reasonable upper bound, but
there are certainly women who have trouble becoming pregnant and do not
seek fertility treatment. 

The willingness to pay for fertility treatment is treated as an expected
value. The average cost of various cycle-based fertility treatments is
multiplied times the probability of obtaining that treatment at various
ages. There is probably a small opposite effect associated with an
increase in pregnancy from reduced formaldehyde exposure by women who do
not want to become pregnant, but we do not consider this effect

9. Please comment on whether the annualized approach as opposed to the
lifetime approach is appropriate in this context.

Cost-of-illness estimates are conceptually inappropriate in this context
for two reasons.  First, cost-of-illness is not a conceptually sound
approach for estimating willingness to pay.  This point is noted in the
document on p. 58.  In consequence, I would suggest that cost-of-illness
estimates might be used only as a last resort.  The WTP estimates
obtained by Smith and van Houtven may well be a more defensible choice. 
Careful thought should be given to whether this alteration would be an
improvement.  Second, the COI estimates utilized (see equation (20) and
Table 10) do not really pertain to the delay in conception addressed in
the Taskinen study.   Women (or couples) may be willing to pay to
reduced time to conceive, but may at the same time be reluctant to seek
fertility treatment.  At a minimum, this potential misalignment should
be addressed.  

The time period associated with the exposure-response function is longer
than the time step of the exposure model. The Taskinen (1999) study
examined the fecundity of women age 20-40. This causes a problem if we
have an individual who does not experience the exposure reduction for
the full time period. We can form an upper and lower bound of this
willingness to pay as we did for asthma prevalence. The upper bound is
based on the assumption that the linear concentration-response function
holds for an acute exposure of one year. The lower bound assumes that an
exposure reduction for one year represents 1/20th of an exposure
reduction for the entire 20-year time period. 

10. Please comment on the upper and lower bound methodology used.

The Taskinen study appears to treat formaldehyde exposure as a long term
exposure.  In consequence, I am uncertain that the upper bound estimates
are defensible (see the related point #6 in asthma above).  Nonetheless,
the upper and lower bound WTP estimates differ only by about $57,000
(see Table 12).  This outcome could change of course if the analysis is
redone along lines suggested above.  Also, as with asthma, I would
suggest the possibility to treat “new” women entering the model as
those affected by the contemplated reduction in formaldehyde exposure.  

 

The ages used in this analysis reflect the ages reported in the Taskinen
(1999) study, ages 20-40. However, women older than age 40 obtain
fertility treatment. The value of reduced infertility effects from
reduced formaldehyde exposure is not reflected in this analysis in order
to remain consistent with Taskinen (1999).

11. Please comment on whether the concentration-response function was
applied to appropriate age groups given the age groups in the studies. 

The Taskinen study does not appear to provide a direct basis for adding
women aged 40 and older to the analysis.  However, relatively few of
these women may seek to become pregnant in a given year.  If the
analysis focused on women who seek to become pregnant, exclusion of
women over 40 years of age may not be a grave oversight.  I would defer
to a panel member more knowledgeable about fertility issues to make a
judgment on this issue.

 

III.	SPECIFIC OBSERVATIONS

The document appears to have been put together hastily as there are
numerous typos and other glitches.  Below is a list of typos that I
flagged—I am certain there are more.

Page 31, first line below equation (1).   …risk reduction at age a.

Page 35, next to last line of third full paragraph.  Data should be days

Page 36.  Not sure why the Table is called “Table J-10”.  Other
tables are not numbered this way.  Also, in this table Dickie et al.
needs a date

Page 44, first line under monetization.  Analysis should be analyses.

Page 45, Table IV.  2-17 not numbered consistently with other tables,
also in two places decimal points are mistakenly used for commas

Page 53, last full line.  Should be $4.9 million

Page 56, eighth line up from bottom.  Undergoes should be undergo

Page 57, fourth full paragraph, first line.  Ryan (1996) conducted a
contingent …

Page 58, first full paragraph, line 4.  Respondents were asked to
consider taking…

Page 59, line 8 under equation (19) (and elsewhere).  The word data is
plural, not singular

Page 59, line 12 under equation (19).  Difficulty is misspelled

Page 59, second line up from bottom.  Reduction is misspelled

Page 60, part of Table 9 appears to be missing.  Also, the heading is
missing a number for ppb.

Page 60, Table 10.  The probability values are missing

V.	PEER REVIEWER COMMENT TABLE ON NON-CANCER HEALTH 

EFFECTS – DERIVATION OF CONCENTRATION-RESPONSE FUNCTIONS

I.  GENERAL IMPRESSIONS

NAME	COMMENT	RESPONSE

John R. Balmes, M.D. 	The document is clearly and concisely written. 
The background subsections of Section 2: Non-cancer Health Effects –
Derivation of Concentration-Response Functions for sensory irritation
(2.1.1), asthma (2.2.1), and female reproductive toxicity (2.3.1) are
extremely brief, especially the latter two.  While it is understandable
that the EPA does not wish to be duplicative of the material in the
draft IRIS document on formaldehyde or the NAS review of the IRIS
document, some readers of the “Approach” document will not have read
the IRIS and NAS documents and thus will benefit from a bit more
background context on formaldehyde and asthma or formaldehyde and female
reproductive toxicity.

	David Kriebel, Sc.D.	The overall impression is of a thoughtful and
well-written document. The authors have provided a generally clear and
logical explanation of their reasoning throughout. There are a few
exceptions, noted below, where additional explanation or documentation
could improve the presentation.

	Frederick J. Miller, Ph.D., Fellow ATS	The comments provided are
related to the health assessment and concentration-response evaluation
portion of the above cited document as this reviewer was not asked to
comment on the economic benefits assessment approach. For the most part,
the sections on sensory irritation, asthma, and female reproductive
toxicity are clearly presented and the material is developed in a
logical manner. Some exceptions are noted in this reviewer’s responses
to the specific charge questions. The decision on the
concentration-response functions to take forward to the economic
benefits analyses is best documented for sensory irritation. This
reviewer provides a suggestion for how far below 100 ppb of formaldehyde
the modeling results developed by the Agency based on the Hanrahan et
al. (1984) study can be used. The material included in support of using
the McGwin et al. (2010) meta analysis for asthma in children does not
adequately describe the short comings of the meta analysis. The
uncertainties and conflicting results of the studies make this reviewer
reticent to endorse the concentration-response function for the odds
ratio based on the McGwin et al. (2010) meta analysis. The case to use
background adjusted exposure data for the female reproductive toxicity
concentration-response curve is adequately defended. However, the Agency
does not indicate if the linear risk or the exponential risk curve will
be used.

Overall, the information presented is accurate. To their credit, in a
number of places, the EPA authors have incorporated criticisms included
in the NAS report of the Agency’s draft IRIS assessment document. By
addressing the responses to the charge questions, the Agency will be in
a good position to move forward with the cost benefit analysis required
by Executive Order 12866 related to the Formaldehyde Standards for
Composite Wood Products Act. 

	Rebecca T. Parkin, Ph.D., MPH	The draft is clearly organized and, in
large part, well-written.  Minor inaccuracies are noted below.  Most
sections were clear, although some key issues were not mentioned or are
not clear; these concerns are noted below.  Except in a few instances as
noted below, the conclusions are effectively supported by the text
provided.

Uncertainties are important to discuss.  EPA has captured many of the
important ones, but may need to add more to these sections based on this
and other reviewers’ comments.  Being thorough in all uncertainty
sections is crucial to the foundations of the benefit assessments.

	II.  RESPONSE TO CHARGE QUESTIONS

CHARGE QUESTION 1: The literature base of residential epidemiology
studies of the irritation effects with quantitative
concentration-response data was limited, but data were available from
which a concentration-response function could be estimated for eye
irritation.

Please comment on whether the concentration-response function is
appropriately derived from the studies used in the draft Approach
Document.

NAME	COMMENT	RESPONSE

John R. Balmes, M.D. 	The concentration-response function (CRF) for
sensory irritation is derived from one small field study, Hanrahan et
al. (1984).  Because most individuals exposed to formaldehyde from the
products of interest will be exposed at levels below 100 ppb, the EPA is
appropriately concerned about the form of the CRF in this range. 
Unfortunately, the data for exposures below 100 ppb are not given in the
original Hanrahan paper and the regression of prevalence of eye
irritation and formaldehyde concentration in the paper covers the range
of 100-800 ppb.  Another issue regarding CRF derivation from the
Hanrahan paper is the relatively small sample size (n=61).  The Liu et
al. (1991) study, which involved a much larger population, is supportive
of the concept that there is a CRF for formaldehyde and eye irritation
at relatively low levels of exposure, but was not used in the actual
derivation of such a CRF because of the way the authors aggregated their
data into low, medium, and high levels of exposure.

	David Kriebel, Sc.D.	I see no problems with the derivation of the
concentration-response function itself. In Section 2.1.3, I think it
would strengthen the argument to at least acknowledge the problem of
acclimatization. There is evidence from a number of studies – notably
the studies of studies conducting autopsies – that there is a change
over weeks in the short-term irritant effects. One way that this will
impact the calculations is that cumulative exposure is probably not the
appropriate exposure metric. The authors may have little choice to use
the simplified approach they have taken which assumes a constant
exposure-response relationship over time, but they could acknowledge
that there is a source of uncertainty from the changes in potency of
formaldehyde over time for various endpoints. I suspect that the
irritant effects diminish with time exposed, but the other two chosen
endpoints probably do not.

	Frederick J. Miller, Ph.D., Fellow ATS	The draft Approach Document
adequately discussed the type of data available in the literature from
various studies and presented their strengths and weaknesses. The Agency
wisely stayed away from using the Liu et al. (1991) study as the primary
driver for the concentration-response function for eye irritation. The
categories from that study were formed on using a ppm-hr per week
metric. The eye irritation studies discussed in Section 2.1.2 for a 10 %
additional risk for 560 ppb for 3 hours (i.e., 1,680 ppm-hrs) compared
to 240 ppb for 5 hours (i.e., 1,200 ppm-hrs) clearly show that Haber’s
Law does not apply for the effects of formaldehyde exposure on eye
irritation, which agrees with Miller et al. (Haber’s rule: a special
case in a family of curves relating concentration and duration of
exposure to a fixed level of response for a given endpoint. Toxicology
149:21–34, 2000). Thus, the Agency would have been hard pressed to
defend a translation of the ppm-hrs metric of Liu et al. (1991) as the
basis for developing a concentration-response function for eye
irritation.

Rather, the Hanrahan et al. (1984) study that reported eye irritation
was associated with in-home formaldehyde exposures (p < 0.05) (both as
“burning eyes” and “eye irritation”) was used as the data for
developing a concentration-response relationship for eye irritation.
Moreover, the logistic regression model developed by Hanrahan et al.
(1984) gave insight as to the needed nature of the curve if one were to
extrapolate below the range of exposures reported by those authors
(i.e., from 100 to 800 ppb). In the opinion of this reviewer, the
concentration-response function was appropriately derived by the Agency.



	Rebecca T. Parkin, Ph.D., MPH	First, I will comment on each study and
then on the derivation process.

Liu et al (1991): This paper adequately describes how homes were
randomly selected, but omits some key information related to monitoring
results for the homes included.  The study was conducted throughout
California (p. 93); a very large state with a wide range of climatic
conditions.  “Summer” and “winter” sampling periods were used,
but these seasons are quite different in different parts of the state. 
It is well-known that temperature and relative humidity affect
formaldehyde (HCHO) levels, but there is no mention of how these factors
varied in the study homes or how they may have affected measured HCHO
levels.  Further, the Quality Assurance/Quality Control methods are not
described, and the number of laboratories used to analyze the passive
area monitoring samples is not stated.  These omissions leave this
reviewer without key information to assess the accuracy or validity of
the HCHO levels reported.  

Residents received instructions on how to uncap, place and mail back the
samplers.  The use of the kitchen and master bedroom as monitoring sites
may be most relevant for the homeowner and less relevant for other
occupants.  Although averaging the monitored levels for each home may
have lost some useful variation, it is not likely that in fairly small
homes (such as mobile homes) that this loss – except for peak
exposures – would be particularly important.  

The study appropriately controls the impact of smoking indoors, and does
not discuss gas cooking stoves or appliances, opening of windows and
doors, or other factors that have been reported by other authors as
having non-significant impacts on HCHO levels.  

More importantly, if I am reading the article correctly (p. 92), some
indoor samples in this cross-sectional study were taken after symptoms
were recorded.  This conflicts with determining the appropriate temporal
sequence of cause preceding effect.  I believe my reading is different
from what EPA states on p. 10 of their draft, which indicates that
exposures were recorded in a period preceding the documentation of
symptoms.

Hanrahan et al (1984):  This cross-sectional study, conducted in
Wisconsin during July 1979, sampled a smaller number of mobile homes
(65) than did Liu et al.; it also had a lower percentage of participants
among those contacted (31% vs. 44%).  In this study, personal samplers
were used and 33 outdoor samples were taken (at least 1 in each mobile
home park).  The authors noted that prior to and during sampling periods
home windows were closed, gas appliances were turned off and smoking was
prohibited.  

The Quality Control and specific laboratory methods were noted (unlike
in Liu et al).  The indoor samples for each home were averaged, as in
Liu et al, with similar but less specific sampling locations noted.  It
is not clear whether the residents or the researchers collected the air
samples in this study.  

Smoking was appropriately considered, and indoor and outdoor temperature
and humidity, and appliance and construction characteristics were noted.
 Some but not all of these factors’ influences on HCHO levels were
presented in the article.

Derivation:  The seasonal and cumulative averages and ranges of indoor
levels, the concentration-response data are correctly extracted from the
Liu et al. article.  The method to estimate the average number of hours
indoors seems appropriate; however, this reviewer wonders whether EPA
attempted to determine whether the authors still have the raw data that
could be used for a more accurate assessment of hours spent indoors.

EPA had to make many more adjustments when using the Hanrahan et al.
article.  First, EPA estimated the lower bound for the indoor levels
based on the standard deviation for the outdoor levels; this is a
reasonable approach given the missing data for the indoor levels. 
Second, the Agency assumed that the distribution of indoor levels is
log-normal in order to estimate the number of samples with HCHO levels
below 100 and 50 ppb.  However, if this assumption or the estimated
standard deviation is not correct, the numbers of samples could be quite
different.  

It is not clear in Figure 2 whether the data shown are for EPA’s or
the authors’ use of “eye irritation;” e.g., does “eye
irritation” here mean the “burning eyes” AND “eye irritation”
results in the article?  If combined data were used, did EPA assume that
all of these symptoms were distinct and not co-occurring events?  Are we
looking at data for the number of symptoms reported or the number of
people with eye symptoms (meaning “burning eyes” and “eye
irritation”) at each level of exposure?  This question should also be
asked of the data shown in the draft Table 1.  Without knowing whether
EPA combined data in Figure 2 and Table 1, it is difficult to comment
further on this part of the draft.

The conversion of data in the article to prevalence odds seems
appropriate.  The display of estimates below 150 ppb in Figure 4b are
derived from Equation 2-2 and are likely correct.   This reviewer did
not verify the mathematics.

The EPA authors have noted many key uncertainties in the two studies and
how these affect their estimates of the relational function.  However,
the validity of the HCHO monitored levels remains a question, due to
missing information in the articles; exposure misclassification is
possible in both studies but is not discussed in the draft.  

	CHARGE QUESTION 2: Please comment on the validity of extrapolating the
concentration-response function and using the function to estimate
effects at concentrations below 100 ppb.

NAME	COMMENT	RESPONSE

John R. Balmes, M.D. 	Given the limitations of the data available from
the Hanrahan paper, the method of derivation of the CRF including the
portion of the curve below 100 ppb is appropriate, but the extrapolation
is pushing the envelope of what reasonably can be extracted from the
data.  The assumption that the reported standard deviation of the
outdoor formaldehyde measurements represents the lower bound on the
standard deviation of the indoor concentrations appears reasonable, but
nonetheless the distribution of exposure data below 100 ppb in the
Hanrahan study was assumed rather than based on the actual data.  The
EPA used the graphical display of the regression data over the range of
100-800 ppb in the Hanrahan paper to derive a CRF and then to infer the
shape of the CRF below 100 ppb.  Given that the Hanrahan study only
included 61 subjects and that the formaldehyde measurements in the
mobile homes were only for 30-60 minutes, the validity of the
extrapolated CRF can be questioned even if an appropriate method was
used.

	David Kriebel, Sc.D.	I am comfortable with this extrapolation. However,
I think that it would be appropriate to acknowledge that errors in
exposure estimation in the studies on which the function is based have
not been included in the uncertainty calculations. This is a general
concern – for all 3 endpoints, there were certainly errors in the
exposure estimation, and the impact of these on the uncertainty in the
exposure-response function has not been considered.

It might be noted that the extrapolation downwards assumes a logistic
shape, which is reasonable but not subject to validation. It might be
noted that the excellent R2 for Figure 3b is a bit misleading since it
is fitting a curve to categorical data, while the actual individual
exposure data were of course continuous – the fit of this curve to the
original data would not have been nearly as good. And again, see my
comment just above about failure to point out that the exposure data had
error, which was not taken into consideration in the model error
calculations.

	Frederick J. Miller, Ph.D., Fellow ATS	The Agency appropriately focused
on the data from the Hanrahan et al. (1984) study to develop the
concentration-response curve for eye irritation. Importantly, the Agency
used a power law curve to fit the prevalence odds data. This ensures
that the fitted curve will asymptotically approach zero for a zero
formaldehyde exposure level and makes it reasonable to use Equation 2.2
below 100 ppb. Given (1) the limit of detection (LOD) was 10 ppb at the
time the Hanrahan et al. (1984) study was conducted, (2) the Agency’s
assumption about the outdoor standard deviation of measurements being 30
ppb around a log-normal distribution of measurements, and (3) the known
overall shape of the concentration-response curve below 100 ppb, this
reviewer believes it would be reasonable for the Agency to use the
fitted equation down to a level of about 40 ppb, which would correspond
to the LOD plus one standard deviation.

	Rebecca T. Parkin, Ph.D., MPH	The validity of the extrapolation in the
Hanrahan et al. paper depends in part on what data were used for the
outcome measure.  If “burning eye” AND “eye irritation” data
were combined, more information is needed in the draft to describe what
assumptions were made in the combining process.  (See comments above.)

The validity of the estimates below 100 ppb based on the Hanrahan et al.
data is strengthened by their alignment with the results in the Liu et
al. study.

However, clearer discussion of the Agency’s choice of and
uncertainties in the points of departure would improve this portion of
the draft. 

	CHARGE QUESTION 3: The studies considered for the
concentration-response function are residential studies because
residential exposures are the focus of the rule and the residential
studies avoid some of the limitations of other types of studies (e.g.,
controlled exposure chamber studies used high, acute exposures, and
occupational studies reflect high exposure levels and potential
dose-rate effects from peak exposures).  Are there other studies (e.g.,
chamber or non-residential studies) that should be used as supporting
evidence for the currently derived concentration-response function for
eye irritation?  If so, please provide information about the studies and
discuss how they should be used to support the concentration-response
function.

NAME	COMMENT	RESPONSE

John R. Balmes, M.D. 	I do not know of other studies that would be
better to use for derivation of a CRF for formaldehyde exposure and
sensory irritation given the EPA’s interest in chronic, low-level,
non-peak exposures.

	David Kriebel, Sc.D.	I think you should use the other studies –
particularly the chamber studies and the autopsy studies – to show the
consistency of the findings across many settings. I am comfortable with
the argument that the function should be derived from the residential
studies. But I think that the logic of extrapolating downwards is
supported by some of the other studies. Kriebel et al. Arch Env Hlth
2001; 56:11- 18 contains fairly detailed exposure data linked to eye
irritation scores. While much of the data are above the range you are
concerned with, there is evidence here that the relationship observed
was valid across the full range of exposures.

	Frederick J. Miller, Ph.D., Fellow ATS	Beyond the studies considered by
EPA, this reviewer knows of no chamber or epidemiology studies that
would be useful in developing the concentration-response function for
eye irritation. Given the definition of an inhalation RfC and the focus
on extended periods of exposure, the residential studies were conducted
in a manner that best fits the criteria for developing a RfC and using
it in the benefits analyses that the Agency is being required to
conduct.

	Rebecca T. Parkin, Ph.D., MPH	I am not aware of any other studies that
should be considered for this function.

	CHARGE QUESTION 4:  Are there other sensory irritation endpoints
besides eye irritation for which a concentration-response function can
and should be developed for assessing effects at concentrations below
100 ppb?  If so, please provide references of appropriate studies and a
description of how the studies could be used to develop a
concentration-response function.

NAME	COMMENT	RESPONSE

John R. Balmes, M.D. 	I do not know whether a CRF can be developed for
other sensory irritation endpoints. 

	David Kriebel, Sc.D.	Both nose and throat irritation can be linked to
increased risk of lower respiratory tract infection including
bronchitis. So you could argue that these would have greater economic
impacts. 

	Frederick J. Miller, Ph.D., Fellow ATS	Sensory irritation may be
indirectly reflected in pulmonary function responses measured after
exposure to formaldehyde. However, these studies were not conducted at
levels that would be relevant to what the Agency is being charged with
investigating.

	Rebecca T. Parkin, Ph.D., MPH	I agree with EPA and NAS that eye
irritation is the most sensitive sensory irritation endpoint and
therefore is the most appropriate endpoint to use. 

	CHARGE QUESTION 5:  Several studies of the association between
formaldehyde and asthma have been published and these were
quantitatively summarized in a meta-analysis by McGwin et al. (2010).

Please comment on whether the McGwin et al. (2010) meta-analysis is an
appropriate summary of the association between formaldehyde inhalation
exposures and childhood asthma.

NAME	COMMENT	RESPONSE

John R. Balmes, M.D. 	The McGwin et al. (2010) meta-analysis of the
association between formaldehyde exposure and risk of asthma in
childhood appears to have been well conducted, using state-of-the-art
methods, including both fixed and random effects models and various
sensitivity analyses.  Because the meta-analysis was based on a
systematic review of the literature that appears to have been
comprehensive, the results can be considered an appropriate summary of
what is known about the association from epidemiological studies.

	David Kriebel, Sc.D.	I agree that it is appropriate.

	Frederick J. Miller, Ph.D., Fellow ATS	There are many problems with the
studies used in the meta analysis conducted by McGwin et al. (2010).
Relative to the description of the metal analysis, the Agency did a
reasonable job of pointing out many of the strengths and weakness that
McGwin and colleagues noted in their paper, but this description could
be improved. 

There are various inconsistencies in the studies selected for the meta
analysis that would make this reviewer reticent to put much trust in the
findings for applicability to the U.S. population. First, only one study
was conducted in the U. S. and that study did not find a statistically
significant odds ratio (OR) with exposures being in the range of 30 to
80 µg/m3 of formaldehyde. Moreover, the sample size of the U.S. study
was greater than that of 5 of the other studies contained in Table 1 of
McGwin et al. (2010).

A Swedish study with exposures no greater than 10 µg/m3 had an OR more
than twice that of any other study even though several of those studies
involved formaldehyde exposures in the 5 to 70 µg/m3 range, including
another Swedish study. The Zhao et al. (2008) study results are
inconceivable to this reviewer for inclusion in a reliable meta
analysis. With school exposures (1 to 5 µg/m3) and outdoor exposures (5
to 7 µg/m3), mean OR estimates were found to represent decreased risk,
yet the upper bound of the 95 % confidence interval was 17 for the
school OR and more than 2.3 million for the outdoor OR for the same set
of 1,993 children having a mean age of 12.8 years. 

McGowan et al. (2010) state that to evaluate whether the results they
found were unduly influenced by any individual study and to determine if
there was any publication bias, an influence plot and a funnel plot,
respectively were used. However, they did not present these plots in
their paper. The computed values of the Q and I2 statistics were stated
to support the presence of moderate between-study heterogeneity. The EPA
document cites values for these statistics but does not explain how they
are calculated, but should do so. Q (better known as Cochran’s Q) is a
classical measure of heterogeneity and is calculated as the weighted sum
of squared differences between individual study effects and the pooled
effect across studies, with the weights being those used in the pooling
method. Q is distributed as a chi-square statistic with k (number of
studies) minus 1 degrees of freedom (d.f.). The I2 statistic describes
the percentage of variation across studies that is due to heterogeneity
rather than chance and is computed as 100* (Q – d.f.)/Q.

On page 18 of the EPA document, the overall OR for a fixed effects model
is given as 1.026 but on page 20 the same result is stated as an OR of
1.03. The Agency needs to be consistent. If the data are only sufficient
to give an OR with 2 digits after the decimal point, say so and change
the value on page 18 and vice versa. For the reader to better judge the
reasonableness of the meta analysis results, the EPA document should
include Table 1 of the McGwin et al. (2010) paper in addition to Figure
1 from that paper.

	Rebecca T. Parkin, Ph.D., MPH	The authors clearly built on
well-accepted guidelines for systematic meta-analysis (Stroup et al.,
2000).  The study selection/exclusion criteria and process used by
McGwin et al. is well-documented in their article.  The variables
extracted and utilized are stated.  The determination to use a
consistent metric (odds ratio per 10 µg/m3 with the related 95%
confidence interval) across the 7 studies for which actual HCHO
measurements were available is appropriate.  The methods used to pool
the odds ratios and test for heterogeneity are standard analytic
approaches.  Their evaluation of the potentially undue influence of any
one study, using influence and funnel plots, adds strength to their
analysis.  The results are shown with and without Rumchev et al.,
allowing the reader to judge whether this paper should have been
included or not.  The data in Table 2 indicate that the magnitude of the
HCHO effect on asthma prevalence is significant whether this study is
included or not.  The presentation of the Q and I2 statistics guides the
reader to the key results for interpreting the data in this table.  In
summary, this reviewer finds that the McGwin et al. meta-analysis is an
appropriate summary and synthesis of the eligible studies reporting a
quantitative relationship between HCHO and prevalence of childhood
asthma.

	CHARGE QUESTION 6: Please comment on whether EPA chose the appropriate
function from the McGwin et al. paper to use in the benefits assessment.

NAME	COMMENT	RESPONSE

John R. Balmes, M.D. 	Exponentiating the linear slope of the fixed
effects model from the McGwin et al. paper is appropriate, especially
given that the random effects model gave essentially the same pooled
odds ratio.

	David Kriebel, Sc.D.	I am comfortable with the choice to exclude
Rumchev, and agree with the derivation of the function as described.

	Frederick J. Miller, Ph.D., Fellow ATS	The EPA document does not
specifically state which model result (i.e. fixed effects or random
effects) was selected. Rather they state that the mean OR was 1.24 per
10 ug formaldehyde per m3 for both models. One only determines that the
Agency chose to use the random effects model if one calculates the
transformed 95 % confidence interval for an OR for an increase of 1 ppb
and sees that what is stated on page 18 corresponds to the confidence
interval for the random effects model. 

The lack of significant between-study heterogeneity in the meta analysis
of McGwin et al. (2010) supports using a fixed effects model; however,
the extent of the uncertainties among the studies and the fact that the
random effects confidence interval will always be larger than the fixed
effects confidence interval supports the conservative selection by the
Agency to use the random effects model results.

In view of the additional problems this reviewer discussed earlier
concerning the choice of studies for inclusion in the meta analysis,
this reviewer would be reticent to have the benefits assessment use only
one function (i.e., the OR of 1.03 per one ppb increase in formaldehyde
exposure with a 95 % confidence interval from 1.008 to 1.047). The
single U.S. study should also be evaluated wherein the OR is 1.0083 per
1 ppb increase in formaldehyde exposure with a 95 % confidence interval
of 0.97 to 1.045.  

	Rebecca T. Parkin, Ph.D., MPH	The Agency has chosen the appropriate
function from the McGwin et al. paper.  With the exception of the
discussion of the rationale for excluding Rumchev et al., however, more
insightful comments about the strengths of the meta-analysis would lend
greater support to the Agency’s use of and reliance on the
fixed-effects model’s OR=1.24.  Deepening the discussion would improve
the support for the Agency’s conclusions and recommendation.  For
example, McGwin et al. stated that an I2 value of <50% would point to
relying on the fixed-effect results (p. 314).  Although the Agency uses
the fixed-effect modeling results, correctly stating that the Q and I2
data indicate low heterogeneity, the draft does not include the key
cut-point (I2<50%) for the use of fixed- vs. random-effect modeling
results.   In this reviewer’s opinion, pointing the reader explicitly
to this cut-point and the relevant data in McGwin et al.’s Table 2
would sharpen the Agency’s rationale for choosing the fixed-effect
results

The discussion of uncertainties in this section includes the key issues
(such as the changes in definition and determination of “asthma”
over time) and is appropriately written.

	CHARGE QUESTION 7: The literature base describing reproductive effects
of formaldehyde exposure in humans is limited; however, there are
supportive findings from the animal toxicological literature.  The
epidemiologic literature suggests an association between formaldehyde
and increased risk of spontaneous abortion and a related delay in time
to pregnancy (i.e., reduced fertility).  The only study with
quantitative concentration-response data was Taskinen et al. (1999).

The a priori hypothesis in Taskinen et al. (1999) was that formaldehyde
would be associated with decreased fertility.  The primary analyses
showed that there was a statistically significant reduction of fertility
among the women in the highest exposure group that appeared to be
concentration dependent.  Post-hoc subgroup analyses showed that
stratification of the 39 women in the highest exposure group by reported
glove use yielded effect estimates whose confidence intervals
overlapped.  The FDR for the 17 women who were classified as not wearing
gloves was 0.51 [95% CI: 0.28−0.92].  The FDR for women who were
classified as always or sometimes wearing gloves was 0.79 [95% CI:
0.47−1.23].  

In light of the results regarding glove use, in what way(s) can the
Taskinen et al. results be used to support calculating a concentration
response for formaldehyde inhalation risk to time to pregnancy?

NAME	COMMENT	RESPONSE

John R. Balmes, M.D. 	The results of the stratified analysis regarding
glove use are suggestive of the possibility that the dermal route may
contribute to total exposure, but are not adequate to consider
modification of a CRF for decreased fertility based on inhalational
exposures, especially given the small size of the two strata and the
wide confidence intervals of the FDR point estimates.

	David Kriebel, Sc.D.	Yes, but you could strengthen your argument.
First, there is a trend in both FDR and spontaneous abortion across
levels of formaldehyde, so you should not only focus on the effect
modification by glove use in the high exposure group. While the high
exposure is the only group for which the authors provided the results
stratified by glove use, the fact that there is a trend across air
concentrations suggests that glove use – not constant across exposure
levels – could not be a very important source of protection or you
would not see the trend. Second, you have not correctly dealt with the
question of whether or not there is effect modification by glove use.
The overlap of confidence intervals is not the correct way to do this.
Construct the null hypothesis that glove users and non-users have the
same FDR and then calculate a p-value testing this null. You can do it
with the data provided – back calculate the standard errors from the
confidence intervals, or ask the authors for the standard errors. I am
pretty sure that this p-value will be large, and then you will be in a
strong position to say that the appropriate conclusion is to act as if
glove use has not modified the FDR. 

	Frederick J. Miller, Ph.D., Fellow ATS	The results after stratifying
for glove use are indirectly supportive for an inhalation effect of
formaldehyde on the fecundity density ratio (FDR). Those women who did
not wear gloves had a statistically significantly decreased FDR of 0.51
(i.e., 95 % confidence interval did not include 1). Women classified as
always or sometimes wearing gloves had a FDR closer to 1 and their
confidence interval included 1. Thus, the inclusion of women wearing
gloves would tend to move the overall FDR closer to 1, where FDR values
lower than 1 can be interpreted as an adverse effect in that the time to
achieve conception is delayed. This was indeed the case since a FDR of
0.64 was found for all women in the highest exposure group. This
provides evidence in support of the hypothesis that effects on FDR were
most likely due to inhalation exposure rather than to dermal exposure to
formaldehyde.

	Rebecca T. Parkin, Ph.D., MPH	The article (p. 208) is not clear as to
whether the post-hoc analysis was adjusted for the variables noted in
the footnote of Table VI; if not, the post-hoc results would not be
directly comparable to the data presented there.  That said, the
reported post-hoc high exposure group glove/not odds ratios and
confidence intervals expand and are consistent with the trend shown in
Table VI.  However, all of the confidence intervals overlap each other,
yielding a set of non-significant odds ratios.  If the adjusted post-hoc
odds ratios are available, it would be interesting to see the results of
a test for trends comparing high exposure-gloves not used, high
exposure-gloves used, medium exposure, and low exposure.  The trend
probably is not significant, but testing its significance may offer more
information for interpreting the concentration-response relationship.

	CHARGE QUESTION 8: Please comment on whether the concentration-response
function is appropriately derived from the Taskinen et al. (1999) study.

NAME	COMMENT	RESPONSE

John R. Balmes, M.D. 	The CRF for formaldehyde exposure and decreased
fertility appears to be appropriately derived from the Taskinen et al.
(1999) study.  The adjustment for probable exposure to formaldehyde
outside of the workplace is appropriate given the interest of the EPA in
potentially controlling indoor exposures due to off-gassing from certain
products.  The use of reasonably contemporaneous background exposure
data from Finland (Jurvelin et al., 2001) for this adjustment is also
appropriate.

	David Kriebel, Sc.D.	Can’t you ask Taskinen to confirm your (and my)
assumption that age was the conditioning time variable in their FDR
calculations?

The process of adjusting for background exposure is OK, but I don’t
think it strengthens the presentation. First, I doubt very much that the
relevant background formaldehyde exposure for the women in this plant
could have been so much higher than the levels inside the plant. My
guess is that the mean (one mean for an entire country over a long
period of time? How useful is that?) had a wide range around it. Second,
any of my graduate students could have told you that adding a constant
to all the x values in a regression model will not change the slope.

	Frederick J. Miller, Ph.D., Fellow ATS	A concentration-response
function can be derived from the Taskinen et al. (1999) study, and the
selected function must be one for which there was an adjustment for
background exposure levels; however, the EPA document does not actually
state whether they will use the linear function of risk or the
exponential function of risk equations that are discussed in the
document (i.e. will the Agency use Eq. 2-5 or 2-6?).

The two curves presented in Figures 6 and 7 of the Agency document are
not explained either in the figure legend or in the text. While they
obviously relate to the regression of risk on formaldehyde
concentration, their difference is not explained. The Agency clearly
picked up on most of the criticisms provided by the NAS review of the
draft IRIS document related to the Taskinen et al. (1999) study.
However, the Agency’s section on the discussion of uncertainties does
not comment at all on the clear possibility of confounding due to xylene
exposures in the Finnish worker study. Also, there is no discussion of
the possible magnitude or direction of any exposure misclassification
bias in the Taskinen et al. (1999) study. If the Agency is going to use
the Taskinen et al. (1999) study in their economic analysis, more
attention to a description and discussion of the study’s strengths and
weaknesses is needed. 

	Rebecca T. Parkin, Ph.D., MPH	Exposure measurements were available only
for the middle of the study period, but the authors were able to verify
“most” women’s self-reported exposure levels based on data from
their workplaces.  The article does not state whether this verification
was completed to the same extent across the entire study period.  This
reviewer also wonders whether for some women (outside of the undefined
“most”) the authors assumed that self-reported HCHO exposure levels
were accurate and did not change during the study period.  The Agency
draft notes general exposure measurement concerns and agrees with
Taskinen et al. regarding their modest response rate to the mailed
questionnaire.  

 

Although Taskinen et al. note that reporting biases may be present in
their study, they did not report any evaluation of their data to address
that concern.  For example, the response rates for women with
pregnancies in the 1985-90 vs. 1991-95 periods were quite different. 
Although the authors note that “only a few women could not report the
length of time to pregnancy,” they do not indicate whether more women
in the earlier vs. later period were non-reporters.  Also, there is no
mention to what extent the authors believe the recall of the number of
months-to-pregnancy was accurately recalled by either group of women. 
Neither Taskinen et al. nor EPA adequately discuss whether self-reported
time-to-pregnancy (or spontaneous abortion) data are reliable for use as
an outcome measure.  If outcome misclassification is an issue as well as
exposure misclassification, the odds ratios could be further compromised
toward the null.  Discussion of the reliability of self-reported outcome
data should be considered for the next draft.

The Agency’s adjustment for non-occupational HCHO exposures, based on
contemporaneous data, shifts the relationship to the right (from Figure
6 to Figure 7) but does not consider the variation in exposure levels
found by Jurvelin et al.  The draft assumes that all of the women in the
study had the same non-occupational HCHO exposures.  Considering the
range, not just the mean, of Jurvelin et al.’s measurements would be
more informative and would allow for the identification of upper and
lower bounds on the concentration-response relationship.  

	III.  SPECIFIC OBSERVATIONS ON THE DOCUMENT

NAME	COMMENT	RESPONSE

John R. Balmes, M.D. 	Page 10, 1st full paragraph; page 14, figure 4a
legend; page 15, 2nd full paragraph.  The use of “inhalation”
exposure in the context of eye irritation seems inappropriate.  I would
eliminate the use of “inhalation” in all three sentences.

Page 16, paragraph. 2.2.1.  Asthma is defined by the Global Initiative
for Asthma (GINA) as a disease characterized by inflammation of the
airways, not lungs.  I would eliminate “and lungs” in the first
sentence of this paragraph.  To avoid confusion, I would not use
“sensitization” when discussing non-allergic, neurogenic mechanisms
of irritant chemical-induced asthma in the third sentence.

Page 20, paragraph 2.3.1.  “Increased relative risk estimates (RRs)
were in the range of 1.7 to more than 3.0.” To avoid confusion, I
would cite the specific papers that reported these increased relative
risks.

Page 22, last sentence of text.  The text here refers to “estimated
concentration-response functions” but does not explain why there is
more than one CRF in Figure 6.  It is not until page 24 that one learns
that “Regression equations for linear and exponential fits are also
provided.”  If the two lines in Figures 6 and 7 are for both types of
fits then this should be clearly indicated.

Page 26, last full sentence.  The use of a contraction here
(“didn’t”) is unprofessional.

	David Kriebel, Sc.D.	Page 20, section 2.3.1, 6th and 7th lines. The
statement “Increased relative risk estimates (RRs) were in the range
of 1.7 to more than 3.0” is not helpful – without explaining whose
risk is being compared to whose, the absolute value of an RR is not
useful. 

Section 2.0 Background on the Identification of Non-Cancer Health
Effects, last sentence. It would be helpful to summarize at least
briefly the logic behind focusing on these 3 endpoints. If I have
understood the NAS comments correctly, that Committee also raised
questions about these choices, and in particular on the decision not to
use pulmonary function evidence. Given this concern from the NAS, I
think a bit more explanation of the perceived inadequacies in the data
for the other endpoints would strengthen this document. 

	Frederick J. Miller, Ph.D., Fellow ATS	None.

	Rebecca T. Parkin, Ph.D., MPH	Eye Irritation 

Page 10, last paragraph, line 4 and Figure 2, line 2.  The EPA draft
states that these samples were taken from 30-60 minutes, but the authors
only mention pumps running “for approximately one hour” (p. 1026). 
Does EPA have information in addition to what is in the article or is
the 30-minutes an error?  

Pages 11-12.  Figure 2 in the draft has a more detailed caption, but
otherwise seems to be exactly the same as the one in the Hanrahan et al.
paper (which is data for “burning eyes”); is it the same as in the
paper?  If so, shouldn’t it clearly indicate that it is “burning
eye” data?  

Page 12.  Table 1 may need a clearer caption indicating whether the data
shown are for “burning eyes” AND “eye irritation” or “eye
irritation” alone.  Wherever combined data are used, EPA should
clarify this usage and possibly create a new name for combined eye
symptom data.   Further, the title of Table 1 would be more accurate if
it included “…extracted and calculated from…”  

Asthma

Page 16, last paragraph, last 2 lines and top of page 1.  The
descriptive language in the Agency draft about using the McGwin et
al.’s I2 statistic would be stronger if made more specific; e.g.,
noting that McGwin et al. used a cut-point of <50% for choosing fixed-
vs. random-effects data (page 314, top of middle column).  It appears
that EPA agreed with this cut-point.  

Page 18, first complete paragraph, line 6.  The Agency states that the
I2 statistic is 11.2 when the Rumchev et al. paper is excluded, but
Table 2 in McGwin et al. shows 11.3.

Female Reproductive Toxicity

Page 22.  The title on Table 2 should state which “response” (TTP)
is presented.

	VI.	PEER REVIEWER COMMENT TABLE ON 

THE BENEFITS ASSESSMENT APPROACH

I.  GENERAL IMPRESSIONS

NAME	COMMENT	RESPONSE

Glenn C. Blomquist, Ph.D.	Overall, the draft Approach Document is
well-written.  Refreshingly, Standard English language is used
extensively and technical terms introduced when necessary.  Examples to
explain terms such as odds ratios should make the document
understandable to educated lay readers with varied backgrounds.  The key
concentration-response functions are clearly presented in the form of
equations and graphs or tables.  This draft is readable.  Information
presented appears to be accurate and representative of studies that are
the source of the information.  Brief descriptions of key studies are
given in addition to the findings relevant to the topic of this draft
Approach Document.  The extensive list of references allows the reader
to learn more about the studies from which crucial estimates are taken. 
Conclusions regarding concentration-response functions and valuation
(monetization) as reflected in their use in the examples follow from the
information and models presented in a logical and reasonable way.  My
assessment of presentation applies to the entire document.  My
assessment of accuracy and soundness applies mostly to Section 3
Benefits Assessment Approach because of my background in environmental
and health economics.

	A. Myrick Freeman III, Ph.D.	The general model for aggregating over
time and across individuals that is outlined in Section 3.1 is correct. 
For other “impressions,” see my responses to specific charge
questions below.

	Shelby Gerking, Ph.D.	The report is, for the most part, complete and
clearly written.  I had no problem following the narrative, however, an
executive summary, a conclusion and a map of the climate zones would
have been helpful.   Also, in Section 3, there are at least seven issues
that warrant additional thought.  First, the choice of the 3% discount
rate, while not unreasonable, appears from out of nowhere.  Some
attention needs to be given to this choice as well as to the assumption
of exponential discounting.  Other types of discounting considered by
behavioral economists may also be appropriate.   Second, the contingent
valuation estimates of WTP for eye irritation are rather dated—those
obtained by Tolley et al. are 25 years old.  EPA still uses these
estimates, but there have been a number of advances over the years in
the application of stated preference methods aimed at reducing
hypothetical bias which causes stated preference estimates to be too
high.  Third, the Dickie et al. estimates in Table J-10 may not be
directly comparable to those of Tolley et. al. and Loehman et. al. in
that they are based on the averting behavior method (which looks at
actual purchases of goods), rather than on contingent valuation.   In
any case, the narrative might usefully reflect the fact that sound
willingness to pay estimates are elusive and that different methods of
estimating WTP often give (very) different results.  Fourth, both asthma
and eye irritation estimates rely on transferring adult WTP estimates to
children.  This procedure, which may be the only available option, may
well lead to substantial inaccuracies.  Parents, for example, may be WTP
more to reduce health symptoms for their children than they are WTP to
reduce those same symptoms for themselves.  In any case, the issue of
WTP for adults vs. WTP for kids (with citations to recent literature)
should be a discussion point in the report.  Fifth, the WTP estimates
presented implicitly assume that switching to different building
materials, or treating wood products to cut down formaldehyde
off-gassing will not occur for the 30 year horizon analyzed.  Is this
assumption reasonable?  In this same vein, could health effects from
formaldehyde off-gassing be lessened by providing particularly residents
of new homes with information that would lead them to install better
ventilation systems?  If so, the marginal cost of better ventilation (if
such a strategy is appropriate) might then form the basis of a simple
WTP estimate along the lines of the headache example given on p. 31. 
Sixth, the population by housing type and climate zone apparently are
assumed not to change over the next 30 years, even though there have
been major changes in population by housing type and climate zone over
the past 30 years.  This assumption seems hard to defend, although as
noted on p. 42 it would be a difficult task to predict these movements. 
 Seventh, it would be helpful to develop the main features of the cohort
x housing type x climate zone model in introductory Section 3.1.  It
would be helpful in this section to explain about the assumptions
underlying this approach (some of these are identified above).  This
reorganization will clarify the narrative at a number of points and
avoid misunderstandings created when readers simply encounter model
assumptions in the explanation of results.

	II.  RESPONSE TO CHARGE QUESTIONS

CHARGE QUESTION 1: The concentration-response function for the
prevalence of eye irritation is based on Hanrahan et al. (1984). While
Hanrahan does not identify the relevant time period for this
relationship, Liu et al. (1991) produced qualitatively similar results
for a two-week time period. We assumed that the concentration-response
relationship for eye irritation is for a two-week period and multiplied
the concentration-response function times 26 to obtain the number of
cases of eye irritation experienced in one year. 

The concentration-response curve used is non-linear, which means that
the annual approach may not produce an exact measure of the benefits,
but it is believed that the bias is probably relatively small given the
small change in exposure.

Please comment on whether the concentration-response function is used
appropriately in the benefits assessment approach.

NAME	COMMENT	RESPONSE

Glenn C. Blomquist, Ph.D.	The change in exposure expected from the
policy appears to be small relative to typical exposure.  It follows
that the linear approximation implied by multiplication by 26 to obtain
annual changes in number of cases of eye irritation should produce only
a small bias.  It seems as reasonable as other approaches.

	A. Myrick Freeman III, Ph.D.	It was used appropriately.



	Shelby Gerking, Ph.D.	The non-linear concentration response function is
appropriately used in the benefits assessment approach.  However, the
approach used here appears to take no account of other sources of
formaldehyde exposure, such as workplace exposures.  In any case, the
policy-related reductions in formaldehyde exposure should be treated as
reductions in total exposure, not simply reductions in home exposure. 
This refinement may alter the benefits calculation given the use of a
non-linear concentration response function.

	CHARGE QUESTION 2: The concentration-response curve is for an acute
effect occurring at a time less than the one-year time step of the
exposure model. The willingness to pay to avoid cases of eye irritation
is calculated for each year of analysis by multiplying the number of
cases in a year by the marginal willingness to pay value to avoid one
case. This may produce biased results, but the direction of that bias is
unknown.

On one hand, some studies have suggested that the willingness to pay per
symptom-day declines as the number of symptom-days avoided increases. In
other words, the value to reduce two symptom-days of some effect is less
than two times the value for one symptom-day. Therefore, this approach
may overstate the total value. 

On the other hand, this analysis only applies values to one symptom, eye
irritation, largely because this is the only symptom for which an
effective concentration-response curve was derived. However, there is
evidence that there are other forms of minor eye, nose, and throat
irritation associated with formaldehyde exposure. Since eye irritation
is the largest of the median values from the Tolley et al. (1986) study,
the values used here likely represent a lower bound on the true effect
of multiple symptoms. A reasonable upper bound of the total effect for
one symptom-day would be sum of the willingness to pay for one
symptom-day of each effect individually.  

Please comment on whether the annualized approach as opposed to the
lifetime approach is appropriate in this context.

NAME	COMMENT	RESPONSE

Glenn C. Blomquist, Ph.D.	I admit to being confused by the statement
that eye irritation has the largest of the median values in the Tolley
et al. (1986) study.  (For full disclosure, I am one of et al.)  Much of
that study was published in Tolley, Kenkel, and Fabian Valuing Health
for Policy (1994) particularly in Chapter 4 by Kenkel, Berger, and me. 
In Table 4.3A on page 88, the median values for willingness to pay to
avoid an additional symptom day are $11 for coughing, $12.50 for eye
irritation, $13 for throat irritation, and $14 for sinus congestion. 
All three are similar, but eye irritation is not the largest.  It will
not make much difference, however, if either of the other two is used
instead of $12.50 for eye irritation.

Comparing results reported in Table 4.3A and Table 4.3B suggests that
willingness to pay per symptom-day declines as the number of symptom
days avoided increases.  For example, the median value for avoiding 30
symptom-days of eye irritation is $100 which is less than 30 times
$12.50 (or $375).  This suggests that multiplying the $12.50 times the
number of days of eye irritation avoided will bias the annual estimate
for avoiding eye irritation upwards if there are multiple days for
individuals.  Declining marginal willingness to pay for avoided
illnesses is also found in the recent study,   SEQ CHAPTER \h \r 1
Bosworth, Ryan, Trudy Ann Cameron, and J.R. DeShazo. “Demand for
Environmental Policies to Improve Health: Evaluating Community-Level
Policy Scenarios” Journal of Environmental Economics and Management 57
(2009): 293-308.

Table 4 also shows willingness to pay values for avoiding combinations
of symptoms in a day.  The willingness to pay values for avoiding
combinations of symptoms in a day are less than the sums of the
willingness to pay values for avoiding the symptoms separately.  The
declining marginal value of adding another light symptom suggest that a
reasonable upper bound of the annual value of the total effect for one
symptom-day would be the sum of the willingness to pay for one
symptom-day of each effect individually.

On page 34, the paragraph that begins with Tolley et al. (1986) is
mostly correct.  (I am a coauthor on that Tolley report and the Berger
et al. (1987) article.)  It is correct, as stated, that the mean
willingness to pay to relieve one symptom-day of a light symptom was $25
- $50.  It is mostly correct, as stated, that Berger et al. report
higher willingness to pay values for these same symptoms.  The
difference is due to different samples.  The Tolley et al. values are
based on a sample of 176 respondents, some of whom had experienced the
symptoms and some of whom had not.  The Berger et al. values are based
on a subsample of 119 who actually had experienced a symptom.  The mean
values for the subsample that had experienced the symptoms are greater
than for the full sample for five of the seven light symptoms, nearly
the same for one, and less for the remaining one.  The mean value for
eye irritation was $48.48 for the 16 who reported eye irritation whereas
the mean value was $27.73 for all 176 who responded.  The values
reported are in 1984 dollars in both the Tolley et al. report and the
Berger et al. article.  The median value for eye irritation of $12.50
that Weitzel (1990) found is also reported in Tolley, Kenkel, and Fabian
Valuing Health for Policy (1994) in Table 4.3 on page 88.  It is correct
that it too is in 1984 dollars. 

On page 36, if the BLS Inflation Calculator is used to inflate the
median value of $12.50 for eye irritation from 1984 dollars to 1999
dollars, the value is $20.03 as shown in Table J-10.  If the same is
done for 1984 to 2006, the value is $24.25, which rounds to $24, as
reported below Table J-10.

Two recent studies find that different lifetime risk profiles are valued
differently;  see   SEQ CHAPTER \h \r 1 Nielsen, Jytte Seested, Susan
Chilton, Michael Jones-Lee, and Hugh Metcalf. “How Would You Like Your
Gain in Life Expectancy to Be Provided? An Experimental Approach”
Journal of Risk and Uncertainty 41 (December 2010): 195-218 and Cameron,
T.A., J.R. DeShazo, and P. Stiffler. “Demand for Health Risk
Reductions: A Cross-national Comparison between the U.S. and Canada”
Journal of Risk and Uncertainty 41 (December 2010): 245-273.  Given the
limited information on the symptoms of interest it is not clear how
incorporate these finding into the Approach Document.

	A. Myrick Freeman III, Ph.D.	I don’t understand the question.  What
do you mean by the “lifetime approach”?  If you mean the results
shown in Table 3 on p. 42, then either approach is appropriate.  They
are just different ways of summarizing the data.

	Shelby Gerking, Ph.D.	A search of the review document turned up no
references to “lifetime approach”, so the issue to be addressed here
is not entirely clear.  Nonetheless, it is likely that WTP per symptom
day of eye irritation falls as the number of symptom days increase. 
This outcome would be a general implication of utility maximization
under a fixed budget.  As additional resources are spent to relieve eye
irritation, the opportunity cost of foregone consumption will increase. 
If eye irritation is coupled with other symptomatic discomforts, then we
may have a different problem.  On the one hand, suppose that there
exists a good that will reduce eye irritation together with all of the
other symptoms.  Then, the marginal cost (inclusive of both time and
money) of using the good to reduce these symptoms will be the WTP for
all combined.  On the other hand, if more than one good needs to be
purchased to reduce all of the symptoms, then the WTP would be the
marginal cost of the combination of goods that will do the job.  In any
case, it is problematic to simply add up CVM estimates of WTP for
individual symptoms as this is likely to yield a total WTP value that
really is upward biased.  Note that the CVM estimates themselves are
probably upward biased.  

	CHARGE QUESTION 3: The concentration-response function for the
prevalence of eye irritation is for a single two-week period.  The
annualized approach used an adjusted concentration-response relationship
that consisted in multiplying the concentration-response function
obtained from Hanrahan et al. (1984) by 26 two-week periods.

Please comment on whether the adjustment of the concentration-response
function is appropriate in the annualized approach.

NAME	COMMENT	RESPONSE

Glenn C. Blomquist, Ph.D.	It seems as reasonable as other possible
approaches.



	A. Myrick Freeman III, Ph.D.	Yes, it is appropriate.



	Shelby Gerking, Ph.D.	An issue here is that there appear to be
different ways to interpret the Total WTP expression on p. 39.  Should
we assume that the same people experience eye irritation repeatedly for
each 26 two-week periods in the year?  If this (Groundhog Day)
assumption is correct, then the WTP value applied may well be on the
high side.  On the other hand, if we assume that a different set of
people experience these symptoms in each of the two week periods, then
the calculations presented may be more defensible.  

	CHARGE QUESTION 4:  The concentration-response relationship for asthma
incidence is given as an odds ratio for children exposed to
formaldehyde. An odds ratio is a way of comparing the difference in the
probability of an event, in this case asthma diagnosis, occurring in
groups exposed at different levels. The odds ratio is used to determine
the willingness to pay for any change in emissions exposure over a 16
year period. We assume that the probability of being diagnosed with
asthma is uniformly distributed across these 16 years. 

The use of an odds ratio implies that the concentration-response
relationship is non-linear. However, given that the change in exposure
is relatively small, the results from using this implied non-linear
function are anticipated to be similar to the results if we were to use
a linear function.

Please comment on whether the concentration-response function is used
appropriately in the benefits assessment approach.

NAME	COMMENT	RESPONSE

Glenn C. Blomquist, Ph.D.	It seems as reasonable as other possible
approaches.



	A. Myrick Freeman III, Ph.D.	It was used appropriately.

	Shelby Gerking, Ph.D.	I agree that the bias from using the implied
non-linear function is likely to be small, but please see #1 in the eye
irritation section for a possible qualification.

	CHARGE QUESTION 5:  The valuation of asthma effects from formaldehyde
exposure is based on valuing the incidence of asthma rather than the
more standard valuation based on the exacerbation of asthma symptoms.
The reason for this is that a statistical relationship between
formaldehyde exposure and asthma exacerbation has not been sufficiently
well established, but there is a statistical relationship between
formaldehyde exposure and asthma occurrence or incidence. However, the
only time that EPA has valued asthma exacerbation is for the first
prospective study of the benefits and costs of the Clean Air Act (EPA
1999). This is probably due to a lack of an appropriate
concentration-response relationship. The EPA’s Office of Air does
include recommended values for the monetization of chronic asthma in its
BenMAP documentation (EPA 2011b), which is based on an average of the
results from Blumenschein and Johannesson (1998) and O’Conor and
Blomquist (1997).

For this analysis, EPA treats prevalence as a measure of cumulative
incidence by assuming that the disease duration is the rest of the
child’s life. However, the EPA (2000a) has previously assumed that a
portion of asthmatic children will become asymptomatic as they move into
adulthood.

Please comment on whether the annualized approach as opposed to the
lifetime approach is appropriate in this context.



NAME	COMMENT	RESPONSE

Glenn C. Blomquist, Ph.D.	The approach used seems reasonable except for
ignoring that some of the asthmatic children are likely to become
asymptomatic as adults.  If the incidence and prevalence data can
support outgrowing asthma, then it should be incorporated in the
estimates of valuing the incidence of asthma.  Otherwise, the estimates
will be biased upward.

I agree with the choice of the best estimate from O’Conor and
Blomquist (1997) for the implied annual value of asthma control (page
47).  It is $1,474 and is based on the risk-risk tradeoff and the
assumed value of statistical life of roughly $6 million.  The $1,474 is
close to the nonparametric estimates of $1,500 for the 5 unit change and
$1,440 for the 10 unit change.  It is also close to the parametric
estimates assuming a value of statistical life of $6 million of $1,468
for Model 3 and $1,504 for Model 4.  They are the more complete
specifications of the drug choice logistic regression.  Since surveys
were done in 1995, the values are in 1995 dollars.

The implied annual value of asthma control of $1,474 in 1995 dollars is
also close to the best estimates from our more recent work; see Glenn C.
Blomquist, Mark Dickie, and Richard M. O’Conor. “Willingness to Pay
for Reducing Fatality Risks and Asthma Symptoms:  Values for Children
and Adults of All Ages” Resource and Energy Economics 33 (May 2011):
410-425. In this recent work, we are able to estimate the willingness to
pay without assuming a value of statistical life.  As Table 5 shows, for
the ordered logistic regression for willingness to pay for the more
effective drug we have a larger sample for the preferred specification
(263 in 2011 > 75 in 1997) and the ratio of the coefficient for efficacy
in controlling asthma to its standard error is much larger (3.14 in 2011
> 0.70 in 1997).  This facilitates estimation directly from the
willingness to pay logit rather than indirectly from the risk-risk logit
and an assumed value of statistical life.  We were also able to test for
selection bias related to the choice of drug at the first stage of the
hybrid contingent valuation.  (We found we could not reject the null of
no correlation between errors and so based estimates of willingness to
pay on the logit regression without selection.)  Lastly, we made
estimates based on using only “definitely yes” as a “yes”
instead of both “definitely yes” and “probably yes” as a
“yes” as in O’Conor and Blomquist (1997).  In Table 6 we report
values that turn out to be comparable to the best estimate from our
earlier work.  For the average age adult (45 years) our estimate for the
annual value of asthma control is $1,960 in 2007 dollars (or $1,441 in
1995 dollars).  From Table 6 which reports annual values of asthma
control by age we cautiously summarize saying that the pattern of
results hints that values are higher for children and young adults.  For
the average age child (11 years) our estimate for annual value of asthma
control is $2,842 in 2007 dollars (or $2,089 in 1995 dollars.)  This
pattern is relevant for estimating the asthma related benefits because
the ages considered are 0 – 16 and point estimates for values for
children are greater than the value for the average age adult.  The
pattern suggests that estimates based on the average value for adults
are probably underestimates, at least with respect to this factor.

	A. Myrick Freeman III, Ph.D.	Again, I don’t understand the question. 
What do you mean by the “lifetime approach”?  The term is not used
in the document.

	Shelby Gerking, Ph.D.	A search of the review document turned up no
references to “lifetime approach” so the issue to be addressed here
is not entirely clear.  Nonetheless, the report makes clear that
valuation of asthma effects is based on valuing the incidence of asthma,
rather than symptom exacerbation.   This approach is crude for two
reasons.  First, there is substantial heterogeneity in cases of asthma. 
For some people, asthma is little more than an annoyance, while for
others this illness is debilitating.  Willingness to pay to avoid asthma
would be expected to vary according to how serious a case is
contemplated.  Second, willingness to pay to reduce additional cases of
asthma will not include the willingness to pay to reduce exacerbated
symptoms of those who already have asthma.   Two other issues: (1) While
estimates of increased incidence of asthma pertain to children (see p.
19), estimates of WTP to reduce asthma incidence are based on two
studies of adults.  Parents may well be willing to pay more to reduce
asthma incidence among their children than they are WTP to protect
themselves from the same illness.  (2) It might be useful to perform a
sensitivity analysis on the estimates provided to determine how much
difference it would make to assume that a percentage of childhood
asthmatics become asymptomatic when they become adults, rather than
assume that childhood asthmatics carry their illness forward to
adulthood.  

	CHARGE QUESTION 6: The concentration-response relationship is based on
an odds ratio for a period of time longer than the one year. The
exposure model produces results for one-year of exposure to formaldehyde
for each age, housing type, and climate cohort. As a consequence, an
adjustment to the simple model needed to be made to model the effect on
individuals who are currently alive when the regulation takes effect.

As an upper bound estimate, we assume that the odds ratio can be applied
to the one year reduction in the same fashion as it was applied for the
16 year time period. The problem with this approach is that it
effectively treats the exposure as an acute effect. If the odds ratio
indicates a particular percentage reduction in asthma cases from a
reduction in formaldehyde exposure over 16 years, the upper bound
approach would suggest the same percentage reduction in asthma case in
one single year if we were to witness the same reduction in formaldehyde
exposure in one year. The lower bound approach assumes that a particular
exposure reduction for one year produces the same effect as if that
exposure reduction were spread out over the entire 16-year time period. 

Please comment on the upper and lower bound methodology used.

NAME	COMMENT	RESPONSE

Glenn C. Blomquist, Ph.D.	It seems as reasonable as other possible
approaches.



	A. Myrick Freeman III, Ph.D.	The approach seems reasonable.  The bounds
are relatively narrow, as shown in Table 7.  How will the results be
presented in the final report, as upper and lower bounds or averaged to
give a “best” estimate?

	Shelby Gerking, Ph.D.	The upper bound estimates are obtained by
assuming that asthma onset is an acute effect of formaldehyde exposure
particularly among older teenagers.  I am uncertain as to whether this
assumption is at variance with scientific information regarding the
development of this disease and would defer to another panel member with
greater expertise on this point.  If asthma onset from formaldehyde
exposure cannot be considered an acute effect, then the upper bound
estimates would be less credible.  Nonetheless, the upper bound
estimates and the lower bound estimates do not differ greatly.  Table 7
reports that for the three housing profiles total discounted upper bound
WTP is $4,939,166 and the corresponding lower bound figure is
$4,340,740.  So, the upper bound estimate is about 14% larger than the
lower bound estimate.  An alternative to the lower bound estimate might
be to assume that currently living children are unaffected by reduced
formaldehyde exposure.  Instead, we might assume that the reduced
exposure applies only as “new” children as they enter the model. 
These “new” children are then assigned cases of asthma with the
incidence rates described on p. 48 of the review document.  

	CHARGE QUESTION 7: The prevalence data suggest that 7.2% of children
age 0-4 have been told that they have asthma. 16.4% of children aged
5-14 have been told that they have asthma. We assume that the
probability of being diagnosed is uniformly distributed across the 0 to
16 age bracket, so that the incidence rate would be 1.44% for the first
five years and 1.49% for the next eleven years. Allocating cases
uniformly across the age brackets provides us with a way to calculate
the annual benefits of any exposure reduction that occurs for the entire
time period, but may not be appropriate if the diagnosis of asthma for
very young children (e.g., age 0-2) differs from older children.

Is the assumption about applying the concentration-response function to
children aged 0 to 16 appropriate, given the asthma onset studies’
focus on school-aged children and adolescents, and given the potential
uncertainty of the Rumchev study involving younger children?

NAME	COMMENT	RESPONSE

Glenn C. Blomquist, Ph.D.	It seems as reasonable as other possible
approaches.  A question about this and the other concentration-response
functions and exposure is whether or not they take into account averting
behavior.  Smell leads to open windows and more time outside of house
and less exposure.

	A. Myrick Freeman III, Ph.D.	I can’t answer this.  It lies outside my
area of expertise.



	Shelby Gerking, Ph.D.	I defer to another panel member here as I have no
expertise in this area.  

	CHARGE QUESTION 8: The linear concentration-response function
accounting for background exposures based on Taskinen et al. (1999) is
used for the annual approach.

Please comment on whether the concentration-response function is used
appropriately in the benefits assessment approach.

NAME	COMMENT	RESPONSE

Glenn C. Blomquist, Ph.D.	It seems as reasonable as other possible
approaches.



	A. Myrick Freeman III, Ph.D.	It was used appropriately.

 ?  If so, the notation should be changed.  If not, the equation needs
additional explanation.  Second, in these equations, population refers
to women who have difficulty conceiving.  The population of women who
are actively attempting to become pregnant would instead be more
appropriate.  Page 59 indicates that for the proposed rule this latter
definition would be incorporated.  I agree that this change would be
appropriate.  

	CHARGE QUESTION 9: Monetization of delayed fertility has been discussed
in the literature but has never been included in primary analyses of EPA
rules because it was believed that the scientific knowledge on
reproductive and developmental health effects was not strong enough to
quantify risk in the primary benefits analysis.

While there appears to be good data on the cost of treatment, there are
few studies on the willingness to pay to reduce delayed conception. Most
economic valuation studies regarding fertility focus on an
individual’s or a couple’s willingness to pay for infertility
treatment, particularly in-vitro fertilization (IVF).  Conceptually, the
willingness to pay approach is preferable and could be deduced from the
valuation studies. However, given that there has been less research on
these willingness to pay values than other values and the fact that the
values have not been used in previous economic analyses, we used the
more straight-forward cost of illness approach for this analysis.

As an upper bound, we assume that any woman who has difficulty
conceiving (i.e., time to pregnancy exceeds 12 months) will obtain some
type of fertility service. This may be a reasonable upper bound, but
there are certainly women who have trouble becoming pregnant and do not
seek fertility treatment. 

The willingness to pay for fertility treatment is treated as an expected
value. The average cost of various cycle-based fertility treatments is
multiplied times the probability of obtaining that treatment at various
ages. There is probably a small opposite effect associated with an
increase in pregnancy from reduced formaldehyde exposure by women who do
not want to become pregnant, but we do not consider this effect

Please comment on whether the annualized approach as opposed to the
lifetime approach is appropriate in this context.

NAME	COMMENT	RESPONSE

Glenn C. Blomquist, Ph.D.	It seems as reasonable as other possible
approaches.

On page 58, the choice is made to use the cost-of-illness estimates from
Katz et al. (2011) for valuing difficulty in conceiving due to
formaldehyde exposure.  A strength of these estimates is that apparently
most fertility treatment costs are paid for out-of-pocket by
consumer/patients rather than by third parties through private or public
insurance.  As such, consumers reveal that the treatments are worth at
least as much as they paid.  Their maximum willingness to pay would be
expected to be greater than these costs because of consumer surplus. 
(This does not take into account any utility or disutility of
treatments.)  If data were available and demand for fertility treatment
could be estimated, then a better estimate of maximum willingness to pay
could be made.  This suggests using cost-of-illness will produce an
underestimate.  However, if the values are applied to all women who are
exposed to formaldehyde, benefits will tend to be overestimated.  The
reason is that not all the women whose fertility is affected will pay
the market prices for the treatments.  In other words, the price is
known for the various packages of treatments, but the number of women
who are willing to pay the price is not.  The maximum willingness to pay
for treatment for the women whose fertility is affected but choose to
forgo treatment and the number of them are unknowns.  Their values are
likely to be positive, but they are less than the costs.  Katz et al.
report treatment costs per successful outcome of more than $70,000. 
They also report that 72% of the women had college education.  A hunch
is that those women have above average household incomes and spend most
of their time in single family detached homes where exposure reduction
is smallest.  Less educated women with lower household incomes are
probably less likely to pay for treatments at the going prices. 
Assuming that women who have below average incomes and live in mobile
homes will pay the market prices for the same distribution of fertility
treatments is likely to bias the estimate of benefits of reducing
exposure to formaldehyde upward.

	A. Myrick Freeman III, Ph.D.	Again, I don’t understand the question. 
What do you mean by the “lifetime approach”?  The term is not used
in the document. Perhaps the charge question should be "Please comment
on the annual approach to calculating a total WTP per year as described
in Section 3.4.2."  My answer to that question is that if a valid
individual WTP per year can be obtained, equation (18) is the
appropriate way to calculate aggregate WTP per year.  But for reasons I
point out in the Specific Observations section below (see my comments to
question #3), I have serious doubts about the ability to obtain a valid
individual WTP per year.

	Shelby Gerking, Ph.D.	Cost-of-illness estimates are conceptually
inappropriate in this context for two reasons.  First, cost-of-illness
is not a conceptually sound approach for estimating willingness to pay. 
This point is noted in the document on p. 58.  In consequence, I would
suggest that cost-of-illness estimates might be used only as a last
resort.  The WTP estimates obtained by Smith and van Houtven may well be
a more defensible choice.  Careful thought should be given to whether
this alteration would be an improvement.  Second, the COI estimates
utilized (see equation (20) and Table 10) do not really pertain to the
delay in conception addressed in the Taskinen study.   Women (or
couples) may be willing to pay to reduced time to conceive, but may at
the same time be reluctant to seek fertility treatment.  At a minimum,
this potential misalignment should be addressed.  

	CHARGE QUESTION 10: The time period associated with the
exposure-response function is longer than the time step of the exposure
model. The Taskinen (1999) study examined the fecundity of women age
20-40. This causes a problem if we have an individual who does not
experience the exposure reduction for the full time period. We can form
an upper and lower bound of this willingness to pay as we did for asthma
prevalence. The upper bound is based on the assumption that the linear
concentration-response function holds for an acute exposure of one year.
The lower bound assumes that an exposure reduction for one year
represents 1/20th of an exposure reduction for the entire 20-year time
period. 

Please comment on the upper and lower bound methodology used.

NAME	COMMENT	RESPONSE

Glenn C. Blomquist, Ph.D.	It seems as reasonable as other possible
approaches.



	A. Myrick Freeman III, Ph.D.	The approach seems reasonable.  How will
the results be presented in the final report, as upper and lower bounds
or averaged to give a “best” estimate?

	Shelby Gerking, Ph.D.	The Taskinen study appears to treat formaldehyde
exposure as a long term exposure.  In consequence, I am uncertain that
the upper bound estimates are defensible (see the related point #6 in
asthma above).  Nonetheless, the upper and lower bound WTP estimates
differ only by about $57,000 (see Table 12).  This outcome could change
of course if the analysis is redone along lines suggested above.  Also,
as with asthma, I would suggest the possibility to treat “new” women
entering the model as those affected by the contemplated reduction in
formaldehyde exposure.

	CHARGE QUESTION 11: The ages used in this analysis reflect the ages
reported in the Taskinen (1999) study, ages 20-40. However, women older
than age 40 obtain fertility treatment. The value of reduced infertility
effects from reduced formaldehyde exposure is not reflected in this
analysis in order to remain consistent with Taskinen (1999).

Please comment on whether the concentration-response function was
applied to appropriate age groups given the age groups in the studies. 

NAME	COMMENT	RESPONSE

Glenn C. Blomquist, Ph.D.	It seems as reasonable as other possible
approaches.



	A. Myrick Freeman III, Ph.D.	Given my reservations about using the
cost-of-fertility treatment to monetize this effect (see  my comments to
question #3 in the Specific Observations section below), I have no
problem in limiting the quantification of this effect to the age group
covered in the Taskinen study.

	Shelby Gerking, Ph.D.	The Taskinen study does not appear to provide a
direct basis for adding women aged 40 and older to the analysis. 
However, relatively few of these women may seek to become pregnant in a
given year.  If the analysis focused on women who seek to become
pregnant, exclusion of women over 40 years of age may not be a grave
oversight.  I would defer to a panel member more knowledgeable about
fertility issues to make a judgment on this issue.

	III.  SPECIFIC OBSERVATIONS ON THE DOCUMENT

NAME	COMMENT	RESPONSE

Glenn C. Blomquist, Ph.D.	Page 53, bottom.  $4.9 instead of $4,9.

Page 56, 3rd paragraph.  Monetization is misspelled.

Page 58.  Van Houtven and Smith (1999) instead of 1997 and $12,500
instead of $12,50.

Page 59.  Is the probability of obtaining cycle-based fertility
treatment shown in Table 8, the best projection for future treatment? 
The utilization percentages given on page 916 in Katz et al. seem
different.

Page 68.  O’Conor instead of O’Connor.

	A. Myrick Freeman III, Ph.D.	There are a number of typographical and
editorial errors.  This is not a complete list:

Table of Contents.  Page numbers are wrong.

Page 17.  Figure 2 should be labeled figure 5.

	Section 3.  Why do the table and figure numbers start over instead of
continuing throughout the report?  And some of the table numbers seem to
be wrong (e.g., J-10 on page 36).

Page 6.  The report gives the aim of developing concentration-response
relationships as being “ ... to inform the economic benefits
assessment of potential cost savings ...”  This is not about “cost
savings;” it is about welfare gains.

	In some places, coefficients in the concentration-response
relationships are Greek letters.  In other places they are not.

Section 2.1.1, 3rd sentence in the first paragraph.  Needs editing.

Page 53.  Where it says “($93,095 + $58,662 + $90,852= $242,602),”
it should be = $242,609.

Page 58.  It says “willingness to pay of between $3,750 and $12,50 for
fertility treatment.”

Page 59.  The probability of using IVF should be 2.5%, not .25%.

Page 61, last paragraph.  There are two references to Table 10, which
should be Table 12.

Page 63.  The sentence, “This is probably the largest category of
value associated fertility issue from a reduction in formaldehyde
exposure,” needs editing.

I have added the following three charge questions based on my expertise
in non-market valuation:

(1) Please comment on the validity of the monetization of sensory
irritation. The approach to the monetization of this effect is the same
as that used in the First Prospective Section 812 Report.  In my
judgement, this is a valid approach and makes the best use of the
available empirical evidence.  However, it should be noted that the
effect actually valued in the  First Prospective Section 812 Report as
“Upper Respiratory Symptoms” defined as “two or more of the
following symptoms: runny or stuffy nose; coughing; and eye
irritation.”  Since this document states that formaldehyde can cause a
variety of types of sensory irritation in the upper respiratory system
(p. 7), this is appropriate.  But it should be made clear that more than
simply eye irritation is involved.

(2) Please comment on the validity of the monetization of chronic
asthma.  The approach to the monetization of this effect is the same as
that used in the First Prospective Section 812 Report.  In my judgement,
this is a valid approach and makes the best use of the available
empirical evidence. 

(3) Please comment on the validity of the monetization of reduced
fertility. I have serious reservations about the approach taken here to
monetize this effect.  First, “reduced fertility” is not a
well-defined effect.  There is ambiguity as to both the degree of
reduction and its duration.  This makes determining the proper valuation
difficult.  

I agree that conceptually, the willingness to pay (WTP) approach is
preferable to a cost-of-illness or cost-of-treatment approach.  However,
as I read the available WTP studies, they apply to the more serious
infertility rather than to the “subfertility” (p. 21) associated
with exposure to formaldehyde.  Thus, the available WTP studies are not
suitable for benefits transfer.

The report takes a cost of treatment approach to valuation.  But it is
not at all clear to me that the “cycle-based treatments” described
here are either appropriate treatments or likely to be effective in the
face of continued formaldehyde exposures.

For these reasons, I recommend that this effect be described and
quantified to the extent that evidence supports quantification, but that
it be acknowledged that there is not an adequate basis for monetization
of this effect.



	Shelby Gerking, Ph.D.	The document appears to have been put together
hastily as there are numerous typos and other glitches.  Below is a list
of typos that I flagged—I am certain there are more.

/

0

1

œ

ö

Y

Ã

ð

ñ

ò

ó

ô

õ

ö

÷

ø



 

!

"

6

7

8

R

S

T

V

W

X

X

Y

Z

[

w

x

y

z

 

¡

¢

¼

½

¾

À

Á

Â

Ã

Ä

Å

á

â

ã

ä

j¸

j;

 hÚ4

 hÚ4

  hÚ4

 hÚ4

 hÚ4

␃愃̤摧棁w

ü

Æ

Ç

ü

ý

 hÚ4

 hÚ4

@

@

@

옍)

옍)

 h

$

 h

 hÚ4

 hÚ4

Ð

Ñ

å

æ

|

}

 hÚ4

 hÚ4

 hÚ4

 hÚ4

 hÚ4

 hÚ4

 hÚ4

 hÚ4

옍)

 h

uation (1).   …risk reduction at age a.

Page 35, next to last line of third full paragraph.  Data should be days

Page 36.  Not sure why the Table is called “Table J-10”.  Other
tables are not numbered this way.  Also, in this table Dickie et al.
needs a date

Page 44, first line under monetization.  Analysis should be analyses.

Page 45, Table IV.  2-17 not numbered consistently with other tables,
also in two places decimal points are mistakenly used for commas

Page 53, last full line.  Should be $4.9 million

Page 56, eighth line up from bottom.  Undergoes should be undergo

Page 57, fourth full paragraph, first line.  Ryan (1996) conducted a
contingent …

Page 58, first full paragraph, line 4.  Respondents were asked to
consider taking…

Page 59, line 8 under equation (19) (and elsewhere).  The word data is
plural, not singular

Page 59, line 12 under equation (19).  Difficulty is misspelled

Page 59, second line up from bottom.  Reduction is misspelled

Page 60, part of Table 9 appears to be missing.  Also, the heading is
missing a number for ppb.

Page 60, Table 10.  The probability values are missing

	

 Available at  HYPERLINK
"http://www.arb.ca.gov/toxics/compwood/compwood.htm"
http://www.arb.ca.gov/toxics/compwood/compwood.htm 

 Available at  HYPERLINK
"http://www.gpo.gov/fdsys/pkg/FR-2008-06-27/pdf/E8-14618.pdf"
http://www.gpo.gov/fdsys/pkg/FR-2008-06-27/pdf/E8-14618.pdf  

 Available at  HYPERLINK
"http://www.gpo.gov/fdsys/pkg/FR-2008-12-03/pdf/E8-28585.pdf"
http://www.gpo.gov/fdsys/pkg/FR-2008-12-03/pdf/E8-28585.pdf 

 Available at  HYPERLINK
"http://www.arb.ca.gov/toxics/compwood/compwood.htm"
http://www.arb.ca.gov/toxics/compwood/compwood.htm 

 Available at  HYPERLINK
"http://cfpub.epa.gov/ncea/iris_drafts/recordisplay.cfm?deid=223614"
http://cfpub.epa.gov/ncea/iris_drafts/recordisplay.cfm?deid=223614 

 Available at  HYPERLINK
"http://www.nap.edu/catalog.php?record_id=13142"
http://www.nap.edu/catalog.php?record_id=13142 

 Available at
http://cfpub.epa.gov/ncea/iris_drafts/recordisplay.cfm?deid=223614

 Available at http://www.nap.edu/catalog.php?record_id=13142

 PAGE   

 PAGE   2 

September 20, 2011

