XXIV  SEQ CHAPTER \h \r 1 . Analysis of Impacts

A.  Introduction

     FDA has examined the economic implications of this final rule as
required by Executive Order 12866.  Executive Order 12866 directs
agencies to assess all costs and benefits of available regulatory
alternatives and, when regulation is necessary, to select regulatory
approaches that maximize net benefits (including potential economic,
environmental, public health and safety, and other advantages;
distributive impacts; and equity).  Executive Order 12866 classifies a
rule as significant if it meets any one of a number of specified
conditions, including: Hhaving an annual effect on the economy of $100
million, adversely affecting a sector of the economy in a material way,
adversely affecting competition, or adversely affecting jobs.  A
regulation is also considered a significant regulatory action if it
raises novel legal or policy issues.  FDA has determined that this final
rule will be an economically significant regulation under Executive
Order 12866 because it will have an annual effect on the economy of more
than $100 million. 

	The Small Business Regulatory Enforcement Fairness Act of 1996 (Public
Law 104-121) defines a major rule for the purpose of congressional
review as being likely to cause one or more of the following: An annual
effect on the economy of $100 million; a major increase in costs or
prices; significant adverse effects on competition, employment,
productivity, or innovation; or significant adverse effects on the
ability of United States-based enterprises to compete with foreign-based
enterprises in domestic or export markets.  In accordance with the Small
Business Regulatory Enforcement Fairness Act, OMB has determined that
this final rule will be a major rule for the purpose of congressional
review. 

     FDA has examined the economic implications of this final rule as
required by the Regulatory Flexibility Act (5 U.S.C. 601-612).  If a
rule has a significant economic impact on a substantial number of small
entities, the Regulatory Flexibility Act requires agencies to analyze
regulatory options that would lessen the economic effect of the rule on
small entities.  FDA finds that this final rule will have a significant
economic impact on a substantial number of small entities.

	The Unfunded Mandates Reform Act of 1995 (Public Law 104-4) requires
cost-benefit and other analyses for rules that would cost more than $100
million in a single year. The current (2004) inflation-adjusted
statutory threshold is $1185 million. This final rule qualifies as a
significant rule under the statute.  

1. Summary of the Economic Analysis

     We carry out the cost-benefit analyses required for significant
rules in the Final Regulatory Impact Analysis, in section XXIV.B of this
document.  We perform the Final Regulatory Flexibility Analysis of the
effects on the final rule on small businesses in section XXIV.C of this
document.  We estimate that, once it is fully implemented 36 months
after the date of publication, the quantifiable annual benefits from the
final rule will be about $448 million.  The quantified benefits able to
be quantified are generated by more consistently produced dietary
supplements which will increase product safety, which reduces the number
of acute illnesses and product recalls.  In addition, the final rule
maywill generate benefits that we lack sufficient data to quantify. 
These benefits we cannot quantify arise from dietary supplements
manufactured under a system to ensure quality, which leads to a
reduction in the number of chronic illnesses and conditions.  

The final rule will lead to quantifiable costs of $169 million in the
first year it takes effect, $12036 million in the second year, and
$19067 million in the third year.  After 3 years the annual costs will
be about $16450 million.  If we annualize the benefits and costs over 20
years at a 3 percent rate of discount, the annualized quantifiable
benefits are $404 million and annualized quantifiable costs are $15340
million.  These annualized benefits include only those that we are able
to quantify.  The total annualized benefits maywill be larger than our
estimate of $404 million in quantifiable benefits because of the
benefits that we are not able to quantify. 

	We have determined, based on information contained in this regulatory
impact analysis as well as information contained elsewhere in the
preamble, that the benefits of this final rule justify the costs.  We
believe that dietary supplements are purchased based on product labels
or labeling.  If consumers believe that most dietary supplements are
expected to contain the ingredients as labeled, as with any other
product they purchase, then we should help to ensure that they are
manufactured under a system to ensure quality, including the proper
identity, amount of ingredients and limits on contaminants.  We have
determined that, , the costs of ensuring the quality of dietary
supplements are justified by the benefits including those that are
related to increased safety.

The final rule will have a significant economic effect on small
businesses.  We estimate that the annual costs will be about $461,000
for an establishment with fewer than 20 employees and $18463,000 for an
establishment with 20 to 499 employees.

2.	Summary of Comments on the Economic Analysis

     We received numerous substantive comments on the economic analysis
of the 2003 CGMP Proposal.  In general, comments from the dietary
supplement industry state that we underestimated the cost of the 2003
CGMP Proposal.  Specific comments from the industry target the 2003 CGMP
Proposal’s testing requirements, which the comments characterize as
“burdensome.”  Many comments address our estimate of the number of
batches of dietary supplements firms produce in a year.  Many comments
express the fear that, as a result of this 2003 CGMP Proposal, the
prices consumers pay for dietary supplements would increase
dramatically.  Nearly all economic comments mention potential adverse
effects of the 2003 CGMP Proposal on small businesses, stating that many
firms would have to stop manufacturing.  A few comments state that, if
made final, the 2003 CGMP Proposal would make dietary supplements more
expensive than pharmaceuticals.  Other comments address the following
topics:

•	FDA’s other assumptions, including the number of tests required
for each batch and the number of tests already being performed.

•	Development of analytical methods.

•    Equipment and capital investment costs.

•    Recordkeeping costs.

•	Number of batches produced per year.

•	FDA’s estimation of benefits.

	We will summarize comments on individual substantive issues under the
appropriate subject headings and respond. 

B.  Final Regulatory Impact Analysis

1.  The Need for the Final Current Good Manufacturing Practice Rule

     The final rule is needed because establishments that manufacture,
package, label, or hold dietary supplements may not have sufficient
market incentives to use controls to ensure that the quality of the
supplements is what consumers would choose  to buy if they had adequate
information on product quality.  As a result there may be adulteration
of dietary supplements relating to, among other things, their identity,
purity, strength, and composition.  Inadequate information on product
quality makes it difficult for consumers, manufacturers, or the agency
to determine if products meet quality standards.  Manufacturing,
packaging, labeling, and holding practices that ensure quality can be
costly and Because problems can be undetectable by consumers, so
establishments may not adopt the necessary practices to ensure product
quality them unless required to do so by regulation.  Dietary
supplements are a type of food product for which specific CGMPs are
needed.  Manufacturing process controls are needed to ensure that a
dietary supplement contains what the manufacturer intends.  (See Section
III.A for a discussion of why specific CGMPs are necessary for dietary
supplements.)

     Actions by manufacturers, primarily voluntary quality controls,
domay not provide sufficiently protective industry-wide minimum
requirements for manufacturing, packaging, labeling, or holding of
dietary supplements.  Some firms subscribe to fee-for-service quality
control programs, but these programs do not regularly inspect the
facilities to ensure that they are meeting the program’s standards. 
Without the final rule, products in the dietary supplement market are
likely to be sorted into two types:

•     Higher-priced products with respected brand names or industry
certification that follow several of the CGMP  requirements found in
this final rule;

•     Lower-priced products that contain no private certification or
respected brand name and that follow few of the CGMP requirements found
in this final rule. 

In today’s market for dietary supplements, to find  information on the
properties of products consumers must sort through information from the
Internet, magazines, brochures, popular books, television, and a host of
other sources.  The information on dietary supplements available to
consumers from these sources deals mostly often with the claims for the
products themselves, not with the steps taken by establishments to limit
contamination or to ensure quality.  The inadequate information on
manufacturing practices makes this final rule necessary to ensure the
quality of dietary supplements.   

	(Comment 337)  We received several comments on the need for the 2003
CGMP Proposal.  Four comments specifically support the proposal, stating
in part that they are pleased we are addressing the issue of dietary
supplement manufacturing.  In addition, one comment states that the 2003
CGMP Proposal was a good step toward providing assurance that dietary
supplements are as safe as prescription and OTC drugs.    

     Other comments express concern about the 2003 CGMP Proposal.  One
comment generally supports it but expresses concern that the statements
we make regarding market incentives to prevent adulteration and
misbranding are inaccurate and misleading.  The comment points out that
the incentive exists for firms to prevent adulterated products from
entering the marketplace because of their desire to avoid damage to
their reputations.  In addition, adulterated products are already
illegal to market.  Two other comments support the 2003 CGMP Proposal
only with modifications, and another comment supports current good
manufacturing practice regulations, provided they reflect the current
“best practices of leading manufacturers.”  Two comments assert that
a “more rigorous” enforcement program would be more effective than
dietary supplement CGMP requirements in preventing adulteration.  Two
comments state that a regulation would serve no useful purpose because
of the “low level of harm identified in the industry.” 

     One comment states that the 2003 CGMP Proposal spells out design
standards rather than performance standards.  According to the comment,
the 2003 CGMP Proposal spells out procedures a firm must follow rather
than defining a specific outcome, such as a specified level of
contamination.  This comment maintains that we should set a performance
standard and then allow manufacturers flexibility in how that standard
is reached.  Another comment states that, although certain dietary
supplement ingredients may cause concern, this concern did not justify
imposing “overbearing” and “broad” CGMP regulations for an
entire industry.  Another comment asserts that the CGMPs as presented in
the 2003 CGMP Proposal would serve as an anti-competitive tool by
allowing dominant manufacturers to increase their dominance and make it
more difficult for new firms to enter the industry.

	(Response)  Those comments that disagreed with our analysis provided no
data or evidence to support the comment.  Without such data or evidence,
we have no basis upon which to revise our analysis and continue to use
the analysis.  Thus, we have not made any changes based on these
comments.These comments failed to provide evidence that the final rule
is not needed.  We agree with the comments that point out that existing
statutes and regulations, concern for brand names, and voluntary
industry standards can provide some product safety and quality.
Nonetheless, continuing problems in the industry provide evidence
ondemonstrate the need for this final rule.  For example, Ffrom 2000
througho 20053, there were a total of 7541 recall actions in the dietary
supplement industry, includingcovering class 1, 2, and 3 recalls, of
vitamins and minerals and herbal and botanical supplements.  We will
discuss these recalls, which accounted for about 4 percent of the 1,937
FDA food recall actions in 2000-2005, later in this document.  Most of
tThese recalls were all related occurred because establishments failedto
a failure of the company to adhere to product manufacturing or labeling
specifications.  

	For a class 1 recall, there is a reasonable probability of serious
adverse health consequences or death; for a class 2 recall, exposure to
the product may cause temporary or medically reversible adverse health
consequences; for a class 3 recall, exposure to the product is not
likely to cause adverse health consequences.  Full compliance with the
provisions of this final rule could have prevented mostall of thethose
recalls.  We note also recall classifications only track acute hazards,
not long-term quality problems.  The continued problems with the safety
and quality of dietary supplements show the need for the additional
protection provided by this final rule.  The Rresults from
ConsumerLab.com, and other independent laboratory results, and continued
problems with contaminated supplements provide further evidence of a
need for this final rule (Refs. E26E3-E29E6).  Statistical sampling
methods were not used to collect the data reported in these analyses. 
Therefore, although this information provides anecdotal evidence of
problems, the data may not be representative of overall industry
practices.  The information serves as additional evidence of the
existence of problems.

	Although the final rule will increase the monetary cost of entering the
dietary supplement industry, the industry will remain highly competitive
with more than a thousand competing producers and thousands more
potential entrants.

Although this evidence is helpful, there may also be data on adverse
events known only to the supplement manufacturers.

2.  Regulatory Options

     We considered several regulatory options for dealing with current
manufacturing, packaging, labeling, and holding practices that may not
ensure the quality of the dietary supplement.  The options considered
include: (1) No new regulatory action, (2) fewer requirements for
vitamins and minerals, (3) more restrictive regulations than the final
rule, (4) HACCP without the other elements of the final rule, (5) final
product testing only, (6) a final rule for high-risk products or hazards
only, and (7) the 2003 CGMP Proposal.  As a result of comments on the
2003 CGMP Proposal and our reconsideration of our position on several
provisions, this final rule differs from the 2003 CGMP Proposal.

	(Comment  338)  We received few comments on the option of fewer
requirements for vitamins and minerals, and the comments submitted did
not support this option.  One comment supports one set of CGMPs that
would apply to the entire industry rather than fewer requirements for
vitamins and minerals than for botanicals.  Another comment states that
having fewer requirements for vitamins and minerals would not be wise
because of the large number of people who take multivitamin or mineral
supplements.

     One comment supports more restrictive CGMP requirements, including
further testing and quality assurance requirements.  

     We received two comments that support HACCP without other elements
of the final rule.  One comment echoes an earlier comment made about
stressing outcomes and points to the HACCP systems in the juice and
seafood industries as a way of ensuring effective quality control
design.  The comment asserts that the detailed manufacturing controls
and testing requirements spelled out in the 2003 CGMP Proposal may
actually stifle innovation.  Another comment echoes these thoughts,
adding that a HACCP approach could work in tandem with a more
traditional specification and test approach.

    We received one comment that specifically discusses requiring only
final product testing, but received numerous comments on final product
testing in general.  The specific comment did not support reliance on
final product testing only, stating it is not the best or most
appropriate control.  In addition, the comment claims it is not
technically feasible in many cases and is economically burdensome, a
point repeated in other general comments about final product testing. 
In addition, numerous comments point out that a firm cannot “test in
quality,” meaning that ensuring the quality of the dietary supplement
will not be achieved through rigorous end-product testing, which
emphasizes the wrong stage of production, but by ensuring quality
through an effective process control system.

	Few comments discuss regulation of only high-risk products.  Those that
did note that some ingredients would be of public health concern and it
would be preferable to test these ingredients only rather than all
ingredients.

	(Response)  The comments on the regulatory options did not provide
evidence to directly support or oppose those options but instead
addressed particular issues such as testing or coverage. 

	We took the comments on specific issues into account in the analysis of
this final rule.  We discuss them below in the relevant parts of the
analysis.  

One comment supporting HACCP stated that the detailed manufacturing and
testing requirements of the 2003 CGMP Proposal would, compared with
HACCP, stifle innovation.  Although regulations that impose costs will
can divert resources away from innovation, the diversion caused by costs
of thise final rule represent less than one percent of industry
revenues.  (See table 35).  Because research and development
expenditures account for a small fraction of total expenditures, any
reduced expenditures on research and development associated with this
final rule will be a small fraction of one percent of revenues.will not
be large enough to stifle innovation.  As we explained in the economic
analysis of the 2003 CGMP Proposal, the HACCP option would not specify
detailed manufacturing requirements but would also fail to ensure
product quality (68 FR 12157 at 12222).  In section X.I, we discuss why
HACCP is not appropriate for dietary supplements.  The comment
supporting HACCP failed to provide any data or any evidence to support
its conclusion.  Without such data or evidence, we have no basis upon
which to revise our analysis and continue to use the analysis.  The
comment supporting HACCP did not provide any evidence that would lead us
to change that conclusion.    

3.  Coverage of the Final Rule

     The final rule applies to establishments that manufacture, package,
label, or hold dietary supplements.  Tables 20 and 21 list the estimated
number of covered manufacturers, packagers, labelers, holders, and other
establishments subject to the final rule.  Table 20 shows the number of
establishments categorized as manufacturers, re-packagers or
re-labelers, holders whose primary business is dietary supplements, and
other (although not including other holders and distributors).  Table 21
shows our estimate of the number of general warehouses, wholesalers, and
others that hold dietary supplements, but are not otherwise involved in
the industry.

Table 20. Covered Establishments by Type of Operation from DS-EED 

Establishment Type	Number of Establishments	Percent of Establishments

Manufacturer	1,228	84.1

Re-packager; re-labeler	26	1.8

Holder	114	7.8

Establishments not already classified	92	6.3

Total	1460	100.0



Table 21. Covered Establishments That Hold Dietary Supplements

Type of Holders	NAICS Code	Number of Establishments

General Grocery Wholesalers or Drug Wholesalers	424410	4,036

General Warehouse	493110	4,415

Drug Wholesalers 	42420	7,418

Total

15,869



     We consulted several sources to estimate the number of
establishments reported in this document.  The number, 1,460, is the
estimated number of establishments in the Dietary Supplement Enhanced
Establishment Database (DS-EED) that manufacture, package, label, or
hold dietary supplement products in the United States.  In the analysis
of the 2003 CGMP Proposal, we included an additional 106 U.S.
establishments that supplied dietary ingredients.  Because those
establishments are not covered in this final rule, we exclude them from
the total.  RTI developed the DS-EED using FDA’s Official
Establishment Inventory (OEI) and supplemented that source with
information from trade organizations, trade shows, and electronic
databases (Refs. E2 E1 and E3E2).

     To estimate the total number of establishments that could hold
dietary supplements but do not consider dietary supplements as their
primary business, we first looked for a count of establishments that had
North American Industrial Classification System (NAICS) codes for
wholesalers of groceries or drugs.  Next we looked for a count of firms
that met the description of warehouses for groceries or drugs.  We did
not find a category devoted exclusively to food and drug warehousing, so
we concluded that general warehousing most closely corresponded to the
set of establishments that would hold dietary supplements.  The results
are shown in table 21.  This total differs from the total reported in
the analysis of the 2003 CGMP Proposal because the new classification
system allows us to identify more establishments that would not hold
dietary supplements and therefore exclude them from the total.

Foreign firms that export dietary supplements to the United States must
satisfy the requirements of this final rule.  We do not have data on the
number of foreign firms that export dietary supplements to the United
States.  The small number of foreign products in the FDA dietary
supplement sales database (Ref. E18E7) suggests that relatively few
foreign firms export dietary supplements to the U.S.  The foreign firms
that will be most affected by the final rule are suppliers of dietary
ingredients.  Although suppliers of dietary ingredients are not directly
covered by the final rule, the need of manufacturers to meet the
ingredient specifications required by the final rule will indirectly
affect foreign suppliers (as well as domestic suppliers). 

	(Comment 339)  No comments were received on the economic analysis of
the coverage of the 2003 CGMP Proposal.

4.  Baseline Practices

     a.  Consumption.   Baseline risks depend on baseline consumption of
dietary supplements.  Total sales in 20043 were about $20 billion (Ref.
E4E8).  Vitamins and minerals accounted for about 42 percent of sales. 
Sales of herbal supplements, which have not grown in recent years, were
about half as large as sales of vitamin and minerals, accounting for
about 21 percent of the total.  Amino acids, proteins, animal extracts,
tea-like supplements, and other supplements not otherwise classified
accounted for the remainder of sales in 2003.  

	(Comment 340)  There were no comments on the consumption baseline.

     b. Manufacturing.  We contracted with RTI to conduct a survey of
the dietary supplement industry to learn about both baseline (existing)
manufacturing practices and the existing standards used for
manufacturing dietary ingredients and dietary supplements (Ref. E3E2). 
A sample of 966 dietary supplement establishments from the DS-EED
database was selected from an estimated eligible population of 1,566
firms in the industry (the total number of dietary supplement
establishments included 106 ingredient manufacturers, who are now
excluded from the requirements of the final rule).  The eligibility
criteria and the response rate for the survey are fully explained in the
final report on the survey.  (Ref. E32).   We further classified the
target firms by product and by size.  The product categories were; (1)
Vitamins and minerals; (2) amino acids and proteins; (3) herbals and
botanicals, including extracts; and (4) supplements not already
classified.

	The Small Business Administration classifies companies as “small”
based on the size of the entire company, including both parent and
subsidiaries.  If firms that manufacture dietary supplements have fewer
than 500 employees, they are classified as small.  In addition, for
purposes of this analysis, we classify firms with fewer than 20
employees as very small.  

We received 238 completed surveys.  Table 22 shows the number of
completed surveys by product and by size of establishment.

 

Table 22. Number of Completed Surveys by Sampling Strata

	Size

	Very Small

(fewer than 20 employees	Small

(20 to 499 employees	Large

(500 or more employees	Unknown	Total

Vitamins and minerals	19	39	13	1	72

Amino acids, proteins	8	7	0	5	20

Herbals and botanicals, including extracts 	58	25	0	30	113

Supplements not already classified	14	13	2	4	33

Total	99	84	15	40	238



	(Comment 33941)  We received two comments on manufacturers’ baseline
practices.  One comment expresses concern that, as the information is
over three years old, it may no longer represent current industry
practices.  The second comment questions the way we calculated the
number of dietary supplement establishments that do not follow any CGMP
models.  In the 2003 CGMP Proposal, we state that survey data reflect
that 36 percent of surveyed establishments do not follow any CGMP
models.  The comment points out that 26.5 percent of firms responded
“no” to the question, “Does this plant follow a published GMP
model for the dietary supplement products produced at this plant?” 
Furthermore, of the 63 that answered “no”, “at least” 29 of the
firms provided responses indicating the reason they do not follow a
published GMP is that they did not manufacture dietary supplement
products.

	(Response)  Although the survey responses are now close to over 65
years old, they represent the best information we have on the industry
and its practices.  We have, however, adjusted our estimated costs to
reflect the correction of the results from the original survey. 

5.  Baseline Risk 

     The current number of illnesses caused by poor manufacturing
practices requires data linking illnesses to poor practices.  Because
these data do not exist, we looked for other information to provide
indirect evidence on the problem.Industry is not required to report
these data.   

     We looked at many sources for information, including  medical and
other literature on adverse events, information from poison control
centers, reports to the agency, newspaper and magazine articles, and
surveys of users.  The literature review was conducted using Medline,
Healthstar, Aidsline, Cancerlit, and OldMedline (Ref. E95).  We found
evidence of many adverse events associated with dietary supplements. 
For example, one survey found that 12 percent of consumers (about 11.9
million) who have used an herbal remedy claim to have suffered from side
effects or other adverse reactions (Ref. E6).  In another survey, 46
percent of respondents answered that people get sick from dietary
supplements “often” or “sometimes” (Ref. E1).  Iin 2003, the
American Association of Poison Control Centers received 24,412 reports
on events associated with herbal dietary supplements and 57,801 reports
on events associated with vitamin and mineral supplements, with 8,653 of
the herbal and 5,669 or the vitamin and mineral reports treated in
health care facilities (Ref. E6107).  In addition, we have received many
voluntary reports of illnesses caused by dietary supplements (Ref.
E7118).  

The vast majority of these events and illnesses  those described in
otherthe sources we consulted, however, are reported as associated with
the ingredients used in the products themselves, not with contamination
or other results of poor manufacturing processes.  Most of the reports
from poison control centers on vitamins and minerals, for example,
involved inappropriate ingestion by children (Ref. E6107).  We have no
direct evidence on how many illnesses can be attributed to manufacturing
processes.  The anecdotal evidence described elsewhere in the preamble
suggests that many illnesses could have been caused by poor
manufacturing processes, but there are only a few examples of evidence
that explicitly links illnesses to these manufacturing processes. 
Examples of illness that were linked directly to poor manufacturing
practices include vitamin D toxicity from excessive vitamin D in
multivitamins and cardiac glycoside poisoning from botanical dietary
supplements contaminated with Digitalis Lanata.  (Ref. E812).

With no direct evidence on the number of illnesses caused by poor
manufacturing practices, we had to use an indirect approach.  We based
the approach on our recall records.  Class 1 and class 2 recalls all
involve defective products that could have caused illness if ingested. 
Although the recall data cannot be linked directly to illness data, we
have found anecdotes, surveys, and some medical literature on illnesses
that could be caused by avoidable manufacturing mistakes.  We have
recall data that show that manufacturing mistakes exist, so we can
construct a plausible link between manufacturing mistakes and potential
illnesses or injuries.  The number of illnesses associated with a
manufacturing problem leading to a recall is both variable and
uncertain, and could be anything from zero to quite large. Based on data
from FDA food and dietary supplement recalls, we concluded that one
reported illness per recall is a plausible average, so we assumed that a
recalled product could be a proxy for a single reported illness
associated with a defective product.  We did not receive any comments
disputing the appropriateness of this assumption. In particular, we did
not receive any comments from industry, which may have information not
available to us, as to adverse events associated with their products.

     Because there are no active surveillance well established systems
for identifyingthe notification of adverse health events related to
dietary supplements, we assume that the total number of illnesses caused
by poor manufacturing practices is substantially greater than the number
reported.  Based on data for drug and vaccine reporting rates in other
studies, we one study concluded that for dietary supplements, reported
illnesses represent at best approximately 1 percent of total illnesses
(Ref. E9E13).  We use the associated multiplier, 100, in our baseline
estimate and We assume that reporting adverse health events due to
poorly manufactured dietary supplements occurs at the same rate as
reporting adverse health events caused for other reasons by dietary
supplements.  Other reporting rates and associated multipliers are,
however, plausible.  For some hazards that lead to severe events only,
we have used a multiplier of 10; the Centers for Disease Control and
Prevention have used a multiplier of 38 for Salmonella infections and
similar food-related illnesses.  We show the sensitivity of benefits to
the choice of multiplier below, in the uncertainty and sensitivity
analysis of our results.

     We assume that the reported class 1 and class 2 recalls that
occurred from 1990 through 1999 represent the number and type of recalls
that will occur in the future but for compliance with the final rule. 
From 1990 through 1999, we received reports on an annual average of
11.83 class 1 and class 2 recalls of dietary supplements related to
manufacturing problems.  If we assume that each recall is a proxy for a
reported illness, then the total number of illnesses per year is
approximately 1,180300.  We recognize that our procedure generated
uncertain estimates of the number of illnesses.  With a multiplier of
10, the estimated number of illnesses per year is 118; with a multiplier
of 40, the total number of illnesses per year is 472.  

	We estimate that the monetary value of the health costs losses for the
hazards listed in Table 23 associated with an illness as a weighted
average of the values attached to the different health outcomes
associated with each hazard. We estimate the health losses or fatal
cases as the monetary value of a statistical life, defined as the
willingness to pay for a small change in the probability of death. We
estimate the health losses for non-fatal illnesses as the sum ofcome
from: (1) The imputed value of losts of productivity, (2) the imputed
value of pain and suffering, and (3) actual expenditures on medical
treatment.  We measured lost productivity (defined to include household
and market productivity) indirectly with measures of functional state,
which includes measures of physical function.  We estimated the losses
caused by pain and suffering with a symptom-problem index.  We combine
the functional losses with the pain and suffering into a single index of
lost quality-adjusted life years (measured by the Quality of Well-Being
Index).  We then convert the quality-adjusted life years to dollars by
multiplying the index numbers by the dollar value of a quality-adjusted
life year.  We used direct measures of medical costs, such as payments
to physicians and hospitals.  We obtained data on the cost of a hospital
day and other medical costs from the Health Care Cost and Utilization
Project’s Nationwide Inpatient Sample, administered by the HHS Agency
for Healthcare Research and Quality (Ref. E10E14). 

     Table 23 contains summaries of our measures of the health costs
effects potentially caused by known instances of hazardsdefective
products associated with poor manufacturing processes for the decade
1990-1999.  We estimated the health loss per day for the different
levels of illness severity by summing the lost productivity (as measured
by functional state) and the loss from pain and suffering (as measured
by the symptom-problem index).  These losses per day can be interpreted
as the difference between a day of normal health and a day of suffering
from the health conditions caused by these defective products.  The
numerical scale is a relative baseline that rests on the notion of a
quality-adjusted life day (QALD).  The QALD for a day of normal health
equals 1; the QALD for death equals 0.  The loss of quality adjusted
life days per illness equals the daily loss multiplied by the number of
days the illness lasts.  We converted quality adjusted life days to
dollars by multiplying the index numbers by the dollar value of a
quality adjusted life day.  We computed the monetary value of a quality
adjusted life day using three values derived from three different values
for a quality adjusted life year: $100,000, For our base estimates of
non-fatal cases, we used $300,000 as the value of a quality-adjusted
life year.  For fatal cases, we used $5 million as the value of a
statistical life., and $500,000.  These yield values per day of $274,
$822, and $1,370.  Our base measures use $822; we show the effects of
using other values in the sensitivity analysis.  In the sensitivity
analysis below, we show the effects of different values on the estimated
monetary benefits of this final rule. 

Table 23. Summary of Health Effects Based on Potential Illness
Associated With Recalls between 1990 and 1999

 	Recall class	Number of recalls	Expected value of illness	EV of
illness times number of recalls

Chemical	 	 	 	 

Copper salts	2	1	 $       489	 $           489 

Digitalis	1	33	 $  37,442 	 $ 1,235,599 

Ephedra	1	1	 $ 177,237 	 $    177,237 

Hypervitaminosis A	1	2	 $    1,264 	 $       2,528 

Hypervitaminosis D	2	1	 $    $      1,366 	 $       1,366 

Lead poisoning (class 1)	1	1	 $   15,591 	 $     15,591 

Lead poisoning (class 2)	2	40	 $   10,436 	 $   417,451 

Niacin	2	2	 $     5,802 	 $       11,603 

Pyridoxine (Vitamin B6)	2	1	 $   12,085 	 $     12,085 

Selenium poisoning (class 1)	1	1	 $ 755,338 	 $   755,338 

Selenium poisoning (class 2)	2	6	 $     1,288 	 $       7,731

Stannous fluoride	1	1	 $     1,266 	 $       1,266 

Superpotent Zinc	2	1	 $        389 	 $          389 

Biological	 	 	 	 

Botulism (class 1)	1	1	 $  494,683 	 $    494,683 

Botulism (class 2)	2	1	 $      2,044 	 $        2,044 

Klebsiella Pneumonia	1	1	 $  774,178 	 $    774,178 

Salmonella (class 1	1	4	 $    15,298 	 $      61,191 

Salmonella (class 2)	2	4	 $         778 	 $        3,110 

Allergenic	 	 	 	 

Lactose intolerance	2	1	 $         396 	 $           396 

Undeclared sulfites	1	1	 $         723 	 $           723 

Yellow #5 sensitivity	2	5	 $         723 	 $        3,616 

Yellow #6, red #40, blue #2	2	1	 $      1,595 	 $        1,595 

Physical	 	 	 	 

Glass fragments	2	1	 $      4,241 	 $        4,241 

Other	 	 	 	 

L-tryptophan (EMS)	1	7	 $   1,135 	 $       7,946 

Total	 	118	  

	 $3,992,397



Problem	Class of Recall	No. of Recalls	Outcomes	Frequency of Illness
(percent)	Quality Adjusted Life Day	Duration of Illness (days)	Medical
Cost ($) Per Event	Health Cost ($) Per Event

Hypervitaminosis A	1	2

100	0.472	3	$100	$1,264

Salmonella	1	4	Mild	93.8	0.473	2	0	778



	Moderate	5	0.473	5	952	2,896



	Severe	1.2	0.563	17	10,830	18,697



	Reactive arthritis (short term)	2	0.42	25	119	4,640



	Reactive arthritis (long term)	1	0.42	5,223	632	601,635

	2	4	Death	0.04

	10,830	5,010,830

Klebsiella pneumonia	1	1	Severe	85

	7,420	13,806



	Death	15

	7,420	5,007,420

Selenium poisoning	1	1	Low doses	50	0.482	3	100	1,288



	Severe	35	0.482	3	3,068	5,623



	Death	15

	3,068	5,003,068

Stannous fluoride	1	1	Acute	100	0.473	3	100	1,266

	2	1

	0.473	3	100	1,266

Botulism	1	1	Mild	31

	100	2,044



	Moderate	31

	10,388	19,824



	Severe	30

	89,040	167,747



	Death	8

	89,040	5,089,040

Lead Poisoning	1	1	Chronic



0	0



	Mild	90

	100	10,436



	Severe	10

	17,808	34,789











Eosinophilia-

myalgia syndrome	1	7	Mild	50	0.482	60   	100	23,870

	1	7	Severe	47	0.482	5223	1,400	2,070,567

	 	  	Death	3

	1,400 	5,001,400

Glass fragments	2	1	Dental injury, simple	50	0.231	1	   	190



	Dental injury, complicated	12

	4,452	7,304



	Oral emergency	12

	4,452	 8,123



	Tracheo-

esophageal obstruction	25



396



	Esophageal perforation	1

	17,808	29,255

Hypervitaminosis D	2	1

100	0.473	3	200	1,366

Pyridoxine (vitamin B6)	2	2

100	0.482	30	200	12,085

Super-potent zinc	2	1	Mild	50



389



	Moderate	40



815



	Severe	10

	1,484	4,353

Niacin	2	1

100

	100	5,802

Yellow #5 (undeclared)	2	5	Mild allergic reaction	90	0.44	2	0	723



	Severe allergic reaction	10

	2,968	4,337



	Contact dermatitis	50

	100	1,631

Yellow #6, red #40, blue #2 (undeclared)	2	1	Abdominal cramps	10	0.473	3
100	1,266



	Contact dermatitis	90

	100	1,631

Copper salts	2	1

100	0.473	1	100	489

Problem	Class of Recall	No. of Recalls	Outcomes	Frequency of Illness
(percent)	Quality Adjusted Life Day	Duration of Illness (Days)	Medical
Cost ($) Per Event	Health Cost ($) Per Event

Digitalis	1	33	Mild	94.9	0.473	3	100	1,266



	Severe (heart block)	5

	     1,484	622,547



	Death	0.1



5,000,000

Ephedra (undeclared)	1	1	Cardiovascular	14

	1,684	4,573



	CVS w/chronic	2

	3,084	624,147



	Nervous system	14	0.47	2	1,584	2,362



	NS w/chronic	2

	2,984	621,936



	Liver impairment	4

	200	5,902



	Exfoliative dermatitis	7

	100	1,387



	Other	54	0.29	1	0	238















	Death	3

	$2,984	$5,002,984

Lactose (undeclared) intolerance	2	1	Mild	100	0.48	1	0	396.16

Iron poisoning	2	1	Mild	100	0.48	1	100	496.16

Sulfites (undeclared)	1	1	Mild allergic reaction	100	0.44	2	0	723



     The hazards that occurred between 1990 and 1999 are not necessarily
the same hazards that would occur today.  For example, botulism is rare
and may no longer be a hazard associated with dietary supplements, but
recalls involving botulism represent generic examples of adulteration
that could occur with other substances in the absence of good
manufacturing practices.  Also, we base our cost estimates on
information from 1999, so it is appropriate to estimate benefits from
the same time.

	(Comment 3402)  We received a comment that took issue with the way the
recalls are counted.  The comment asserts it is more appropriate to
count each recall action as a separate recall, regardless of the number
of different products affected.

     The same comment criticizes the inclusion of the outbreak of
eosinophilia-myalgia syndrome (EMS) in the table of what is
characterized as “ordinary” recalls, since this case is analyzed
separately as an example of a “rare catastrophic event”.  The
comment states that the outbreak of digitalis should also have not been
included in the recall list because it also was a rare event.  The
comment asserts that FDA announcements and media attention should have
led to full reporting of any adverse events.

	Other comments generally refer to risk associated with dietary
supplements.  One comment states that botanical supplements pose minimal
risk if dispensed directly to a patient rather than used in an
unsupervised setting, and that toxicology and adverse event reports
indicate that end-of-process adulteration in herbal clinics is rare.  By
contrast, another comment states that adverse events related to dietary
supplement use led to hospital admissions at one location and that
reports of misbranded and adulterated dietary supplements are common.

	(Response)  We are not changing the way we count recalls.  Each
different product recalled under a recall action will continue to be
counted as a separate product recall.  How product recalls are counted,
however, does not affect the analysis.  The method used in this analysis
corresponds to an average of about one reported illness per recall
action.  A particular event can lead to many recall actions.  If we
changed the way we counted recalls so as to reduce the number of
baseline recalls to correspond to events, the average reported illnesses
per recall would rise in proportion. The estimated benefits would not
change.

     We are no longer including the outbreak of EMS in our analysis of
benefits.  The product recalls associated with eosinophilia-myalgia
syndrome occurred several years after the outbreak that we are now
excluding.  The continued benefit associated with preventing EMS is
associated with incorporating quality controls aimed at such hazards. 

6.  Benefits

	The benefits of this final rule come from ensuring the quality of
dietary supplements.  Dietary supplements should contain the listed
ingredients in the listed amounts in product forms that disintegrate and
dissolve.  Dietary supplements should not contain any contaminants that
would adulterate the product under section 402(a)(1), (a)(2), (a)(3), or
(a)(4) of the act. 	

Estimating the benefits of preventing adulteration and contamination is
straightforward, at least in theory.  These benefits are the value of
reducing the risk of the acute illnesses and longer-term complications
associated with physical, chemical, and microbiological contamination
(see table 23).  The direct value of preventing recalls is another
source of benefits from preventing adulteration and contamination.  We
estimate the benefits of preventing adulteration and contamination by
first estimating (based on recall data) the number and kinds of
illnesses prevented, and then placing a value on preventing those
illnesses.  We include the recall costs avoided by industry as
additional benefits of preventing adulteration and contamination.

	Estimating the value of ensuring the quality of the dietary supplements
and that they are manufactured according to their specifications is
difficult in practice because we lack the necessary data on what is
missing and how what is missing affects public health.  These benefits
are the value of reducing the risk of chronic illnesses and chronic
conditions associated with nutritional deficiencies (see table 25). 
Some dietary supplements have authorized health claim labeling that
allows them to state their products may reduce the risk of chronic
illnesses or conditions.  Ensuring that those supplements are
manufactured consistently according to the appropriate specifications
will increase their effectiveness in reducing the risk of chronic
illnesses.  In this analysis, we describe those benefits but are not
able to do not quantify them.

     The benefits from the final rule, then, will be:

•     Reduced health costs associated with a reduced number of acute
illnesses (quantified);

•     Fewer product recalls (quantified); and

•     Reduced health costs associated with a reduced number of chronic
illnesses and conditions(not quantified).

	(Comment 3413)  We received many comments on the estimated benefits. 
Although we did receive comments that stated the rule would benefit
consumers by enhancing public confidence in dietary supplements, many
comments state that the estimated benefits in the 2003 CGMP Proposal
were overstated.  In addition, one comment states that our estimates of
benefits are double counted, because the outbreak of EMS was included in
the measure of benefits from preventing a large catastrophic event as
well as total benefits of reduction of illnesses measured by recalls. 
Furthermore, comments critical of the benefits state the search cost
model used in the analysis is not applicable or the benefits of reduced
search costs do not exist, we lack evidence with which to base the
estimate of reduced health care costs from elimination of rare
catastrophic events, and recalls will not fall to zero as a result of
implementing CGMPs.  

	(Response)  We agree with the comment that benefits were overstated
because of the inclusion of the outbreak of EMS.  We no longer include
the value of preventing that or similar outbreaks in our estimate of
benefits.  Although we do not agree with the comments on the
applicability of the search model as a measure of benefits, the
empirical difficulties associated with quantifying those benefits have
led us to replace the search model with a qualitative description. 

	We now explain each of the 3 sources of benefits: Reduced acute
illnesses, fewer recalls, and reduced chronic illnesses and conditions. 

     a.  Reduced health costs associated with a reduced number of acute
illnesses.  The final rule will help ensure the quality of dietary
supplements, which will lead to improved safety of dietary supplements,
reducing the probability of acute illness or deaths caused by
manufacturing problems.  We estimated the reduction of acute illnesses
by using our recall records as evidence of possible illnesses; class 1
and class 2 recalls of dietary supplements all involved adulterated
products that could have caused illness if ingested.  In the 2003 CGMP
Proposal we estimated the reduction of illnesses from preventing
catastrophic events by using the public health effects of the outbreak
of EMS that resulted from consumption of contaminated L-tTryptophan.  We
agree with comments questioning the applicability of this outbreak to
CGMP, so we are no longer including the value of preventing this
outbreak as a benefit of this rule.

	We estimated the annual expected health benefits for acute illnesses
prevented by taking the values of preventing particular illnesses and
weighing them by their likely incidence as indicated by recall data. 
The acute illnesses prevented that we use to estimate benefits are not
actual illnesses, but statistical illnesses (defined as the probability
of illness multiplied by the population at risk) prevented by the
reduction in risk associated with this final rule.  These recalls
indicate recurring failures to ensure the quality of dietary
supplements.  Although each class 1 and 2 recall is estimated to have
resulted in some illnesses (which may have triggered the recall), there
may also be other manufacturing problems that did not lead to recalls
but that did lead to illness. Both situations are part of the baseline
number of illnesses and deaths estimated.

	We computed the expected health benefits from preventing a single
illness (of any type) associated with a recall as a weighted average of
all potential illnesses.  We then calculated the average health benefits
of preventing a single illness associated with a non-fatal class 1 or a
class 2 recall as:

Health costs prevented = (QALYD x days ill x value per QALYD) + medical
costs

We define QALYD as the average quality-adjusted life day year per
illness; as explained earlier, we computed the average by weighting the
quality adjusted life yearsdays lost for the probability of each health
outcomelevel of severity (mild, moderate, severe, or fatal) by the
expected frequency of that outcomelevel of severity. 

     To estimate the number of acute illnesses prevented, we started
with the average number of recalls per year for the decade 1990 to 1999.
 The yearly averages for the decade were six class 1 recalls and seven
class 2 recalls.  As discussed above, we then assumed that these recalls
represented about 1 percent of all acute illnesses caused by the
manufacturing problems leading to the recalls.  With that assumption, we
estimated that the recalls represented about 530 acute illnesses from
class 1 recalls and 650 acute illnesses from class 2 recalls.  The
illnesses used to estimate the benefits of the final rule represent a
sample of acute illnesses that could occur without this final rule.  We
assume that the benefits computed for the average year from the decade
1990-1999 represent the annual average benefits we should expect in the
future.  We do not assume that the acute illnesses prevented in the
future will be identical to those that occurred during 1990-1999.  

We calculated the weighted average benefit for all class 1 illnesses and
multiplied it by the estimated number of class 1 illnesses prevented. 
We then repeated the procedure for class 2 illnesses prevented.  The
health benefits are therefore estimated as:

Health benefits = (average benefit per class 1 illness prevented) x
(estimated annual number of class 1 illnesses prevented) + (average
benefit per class 2 illness prevented) x (estimated annual number of
class 2 illnesses prevented). 

Table 24. Health Benefits Estimated Using Recall Data from 1990-1999

Estimated total annual number of acute illnesses prevented

(530 class 1 and 650 class 2 recalls)	1,180300

Estimated number of acute illnesses associated with class 1 recalls
prevented	600

Estimated number of acute illnesses associated with class 2 recalls
prevented	700

Dollar estimate of average health benefit for preventing an acute
illness associated with a class 1 or class 2 recall	$33,800$65,000

Dollar estimate of health benefit for preventing an acute illness
associated with a class 2 recall	$6,500

Estimated dollar estimate of annual health benefits, recall base	$404
million



     

     To estimate the number of acute illnesses prevented, we started
with the average number of products recalled per year for the decade
1990 to 1999.  The yearly averages for the decade were six class 1
recalls and seven class 2 recalls.  As discussed above, we then assumed
that these recalls represented about 1 percent of all acute illnesses
caused by the manufacturing problems leading to the recalls.  With that
assumption, we estimated that the recalls represented about 600 acute
illnesses from class 1 recalls and 700 acute illnesses from class 2
recalls.  The illnesses used to estimate the benefits of the final rule
represent a sample of acute illnesses that could occur without this
final rule.  We assume that the benefits computed for the average year
from the decade 1990-1999 represent the annual average benefits we
should expect in the future.  We do not assume that the acute illnesses
prevented in the future will be identical to those that occurred during
1990-1999.  In the sensitivity analysis, we show the effects of
substituting the annual average for 2000-2003.

	The estimated benefits are indeed sensitive to the choice of years. 
For 2000-2005, there were 75 recalls: 29 class 1, 25 class 2, and 21
class 3.  The annual averages for 2000-2005 are therefore 4.8 class 1,
4.2 class 2, and 3.5 class 3 recalls.  We estimate that about 80 percent
of the class 1 and class 2 recalls were related to manufacturing
problems (for 1990-1999 over 95 percent of class 1 and class 2 recalls
stemmed from manufacturing problems).  With an average of 9 class 1 and
class 2 recalls per year, our baseline estimate of total associated
illnesses using 2000-2005 data is 900 (9 x 100).  If this final rule
prevents 80 percent of these events, then 720 illnesses will be
prevented.  We do not use this estimate to calculate  baseline benefits
for this final rule because we do not have a comparably recent estimate
of costs.  If the reduced number of recalls reflects increased controls
in the industry, then the benefits and costs of this final rule will be
lower than what we have estimated.

	(Comment 3424)  We received comments critical of the estimates of
reduced illness due to recalls.  One comment points out that drugs,
despite having stringent CGMP requirements, have a higher rate of
recalls than dietary supplements, thus providing evidence that such
requirements do not necessarily reduce recalls.  Expanding on this
thought, other comments state that we seem to assume that new CGMP
requirements will reduce human error to zero and no more recalls will
occur, which is said to be unrealistic.  

     As discussed earlier in this document, the agency received a
comment disagreeing with how the agency counts recalls.  Again, this
comment expresses belief that each recall action should be counted as a
single recall, regardless of the number of products affected. 

     Other comments express concern about the 100-fold multiplier used
to estimate the costs related to recall-associated illnesses.  The
comment states that we, besides referencing Walker (2000)(Ref. E9E13;
Ref. E16 in the 2003 CGMP Proposal), providedhave no other information
to substantiate the use of the 100-fold multiplier and therefore are
being arbitrary.  Any other number could be as accurate.  In addition,
other comments state that it is difficult to believe that the multiplier
would be applicable to recalls associated with Klebsiella pneumonia and
selenium poisoning, and L-tryptophan, because the severity of the
illnesses would certainly have been associated with the highly
publicized recalls; that is, they would not have gone unreported.

     Some comments present recalculated benefits.  One comment estimates
benefits from fewer illnesses as a result of the 2003 CGMP Proposal to
be $10.9 million, rather than our estimate in the analysis of the 2003
CGMP Proposal of $39 million.  This new estimate was arrived at by
taking into account what was characterized as double-counted benefits
which, as mentioned earlier, were characterized as the inclusion of EMS
in the measure of benefits from preventing a large catastrophic event as
well as total benefits of reduction of illnesses measured by recalls. 
Another comment re-estimates the benefits as $16 million.  This estimate
was calculated assuming 100 percent of potential illnesses related to
Klebsiella pneumonia were classified as severe (with none classified as
deaths), and 50 percent of illnesses associated with the selenium recall
were classified as serious and none were classified as deaths.  This
comment also disagrees with the assumption that 3 percent of the 100
potentially ill from the recall associated with undeclared ephedra would
have died.  Furthermore, this comment adjusts the benefits to take into
account recalls that this commenter felt were erroneously included in
the calculation of benefits from reduced illnesses.

	(Response)  We have not seen any new data or other information that
would lead us to change the 100-fold multiplier for our basic estimate. 
We recognize that the multiplier is uncertain; different multipliers
lead to different estimated numbers of illnesses and different estimated
benefits. With a multiplier of 10, estimated benefits are 10 percent of
our baseline; with a multiplier of 40, estimated benefits are 40 percent
of our baseline.  The estimated benefits of this final rule thus move in
proportion to the assumed multiplier.  We recognize this uncertainty and
show how it affects the estimated benefits in the sensitivity analysis. 
The multiplier implicitly assumes that the more severe illnesses are
more likely to be reported; the average reporting rate for all adverse
events is assumed to be about 1 percent.  The average incorporates
higher reporting rates for more severe illnesses, and lower reporting
rates for less severe illnesses.  

     The comments on the severity weights for Klebsiella pneumonia and
ephedra did not persuade us to change these estimates.  We based the
estimates on the outcomes for severe events associated with these
hazards. The Klebsiella weights come from the medical literature (Ref.
E5E9); the ephedra weights are based on adverse events involving
ephedrine alkaloids. 

	The comparison of drug recalls to dietary supplement recalls does not
provide data that would cause us to change our analysis is not
convincing.  The drug industry is far larger than the dietary supplement
industry and any such comparison would have to account for that
difference as well as other differences.  Expenditures on prescription
drugs exceeded $200 billion in 20043.  More to the point, most drug
recalls are not the result of a failure of CGMPs.  

	(Comment 3435)  We received many comments regarding the use of the
outbreak of EMS in 1989 as a basis for estimating health benefits from
preventing a catastrophic event.  The majority of the comments assert
that CGMPs would not have prevented the outbreak.  One comment expands
this assertion by stating our claim that testing requirements would
reduce the probability that contaminated ingredients would be released
to the public is incorrect, because it was not known what, if any,
contaminants caused the outbreak.  Secondly, the comment states that our
claim that complaint files would allow for fast identification of an
adverse health event is also incorrect because the victims of EMS did
not know the L-tryptophan was the cause of their illnesses.

     Two other comments question the periodicity for a cycle of
potential catastrophic events due to dietary supplements.  One comment
suggests a period of 70 years rather than our 30 years.  The other
comment does not suggest a period but rather states that, since we have
no data to support the cycle of 30 years, and we admit it is difficult
to know how likely rare events are, it is possible that the total
projected benefit could be zero. 

     Lastly, other comments state that the benefits from preventing a
rare catastrophic event are double counted.  These comments state these
benefits are double counted because they are also included in the
estimation of benefits from reduced recalls.

	(Response)  As stated above, we are no longer including estimated
benefits from preventing a rare catastrophic event in the analysis of
benefits.  We continue to include the benefits of preventing statistical
cases of EMS in the annual health benefits, because several recalls of
L-tryptophan, which could be associated with EMSthis hazard took place
during the 1990-99 period.  

     b.  Fewer products recalled.  Implementation of the final rule will
reduce the number of adulterated products distributed to the public,
which will reduce the number of products recalled.  Process controls and
better record keeping will increase the ability of establishments to
produce dietary supplements according to specifications and to identify
problems before distribution.  If adulterated products are caught before
they are distributed or earlier in the production process, they will not
need to be recalled.

     To estimate the direct benefits from fewer recalled adulterated
dietary supplements, we estimate the number of annual recalls of dietary
supplements that would be prevented by adherence to CGMP requirements in
the final rule.  From 1990 to 1999, FDA received reports on 195 recalls
related to manufacturing problems, an average of 19.520 recalls per year
(Ref. E5E9).  The average figure reported here includes class 3 recalls.
 The number of units of dietary supplements for each recalled product
varied, so we used a distribution per recalled product of 1,000 units to
34,000 units (Ref. E5E9).  Product price (updated to 2004) also varyies,
with most prices falling between $65 per unit and $119 per unit; we used
a most likely price of $87.570 per unit.  We also include an adjustment
for the goodwill lost by the establishment as a result of the recall. 
We multiply the direct cost of the recall by two in order to include the
lost goodwill.  We also adjust for recalls that would likely not be
prevented by the final rule.  The result is an estimated savings of
$1.85 million in direct costs and $1.85 million in goodwill, for a total
savings of about $3.6 million per year.

	(Comment 3446)  We received several comments on our estimates of the
reduction in recalls.  As noted previously, a comment generally states
that drugs, despite having stringent CGMP requirements, have a higher
rate of recalls than dietary supplements, thus providing evidence that
CGMPs do not necessarily reduce recalls.  Again, other comments state
that we seem to hold the unrealistic assumption that the final rule will
reduce human error to zero and no more recalls will occur.  Another
comment points out that the assumption that the final rule would cause
the discovery of all adulteration is inconsistent with the requirement
that firms keep complaint files.  If the rule eliminates adulteration,
the comment states, then there should be no complaints to report.

     Again, we received a comment disagreeing with how we count recalls,
which expresses the belief that each recall action should be counted as
a single recall, regardless of the number of products affected. 

	(Response)  As stated earlier in this section, we did not receive
sufficient information to warrant changing the way we counts recalls. 
Each separate product affected by a recall action will be counted as a
single recall. We do not believe that recalls will fall to zero.  We
assume that the recalls identified as being preventable by this final
rule will fall to zero, but that mistakes and other hazards will
continue to generate recalls.  In the sensitivity analysis, however, we
show the effects of a lowerdifferent levels of effectiveness in
preventing recalls associated with manufacturing problems.

As we explained, recalls of drugs are not comparable to recalls of
dietary supplements and do not typically involve failure of the CGMP
requirements.

      c. Reduced health costs associated with a reduced number of
chronic illnesses and conditions.  We cannot quantify the value of
ensuring that dietary supplements contain everything in the established
specifications (and nothing that is not in the specifications) because
we lack the necessary data on what is missing and how what is missing
affects public health.  The public health benefits are derived from the
reduced number of chronic illnesses and conditions.  These benefits may
arise from known nutritional effects or from uncertain nutritional
effects. 

	d. Benefits From Known Nutritional Effects.  Many of the nutritional
benefits of vitamins and minerals are known and well-documented.  For
example, the Dietary Guidelines for Americans, 2005 states that dietary
supplements can be used to help meet the recommended intakes of vitamin
B12, folic acid, and vitamin D (Ref. E11E15).  The Institute of
Medicine’s Dietary Reference Intakes include statements that
supplements can be sources of several vitamins and minerals (Ref.
E12E16).  We have recognized the use of supplements in authorized health
claims for calcium and osteoporosis (§ 101.72) and folic acid and
neural tube defects (§ 101.79). 

	In the following table, we list some of the health benefits associated
with the consumption of various dietary supplements.

Table 25. Selected Health Benefits from Certain Dietary Supplements

Dietary supplement

	Who should take itUser	What it doesBenefit

Folic acid	Women of child-bearing age	Reduces the risk of neural tube
defects

Calcium	Children and adults	Reduces the risk of osteoporosis

Iron	Adolescent females and women of child-bearing age	Reduces the risk
of anemia

Vitamin D	Children and adults; persons with dark skin, or with too
little exposure to sunlight	Reduces the risk of osteoporosis

Vitamin B12	Persons over the age of 50

	Reduces the risk of anemia



e. Benefits from uncertain nutritional effects.  We do not know the full
range of effects (or lack of effects) of most dietary supplements.
Vitamins and minerals with known nutritional effects in supplement form
may have other effects that we have yet to discover.  Our uncertainty is
particularly large with respect to the nutritional effects of herbal and
botanical supplements.  The evidence is still too mixed and incomplete
to determine the effects of most of these substances.  If, however,
herbal dietary supplements do indeed have significant beneficial effects
on the risk of chronic illnesses and conditions, then if the final rule
ensures that the supplements consistently meet their specifications, we
should add those benefits to those from supplements having known
nutritional effects. 

	The benefits of this final rule that we can identify are those
associated with the known effects.  The product deficiency might be, for
example, that packages contain some percentage less or more of the
necessary ingredient (such as calcium) than what is listed on the label.
 The relationship between the shortagereduced or excess amount of the
ingredient and the probability of chronic illness would also have to be
taken into account in order to determine the risk associated with the
product deficiencies.  The increase in the probability of chronic
illnesses may be negligible, less than, the same, or more than the
shortage or excessreduction in the amount of the ingredient.  The
increase in the probability of chronic illness would also depend on how
long the supplement contained a shortage or excess amount of was
deficient in the ingredient.  Suppose, for example, that a calcium
supplement contains 10 percent less calcium than it should for one year.
 If the average consumer takes calcium supplements for 20 years, would
the one-year deficiency of 10 percent increase the probability of
osteoporosis by more or less than 0.5 percent (10 percent x
(1/20))?With sufficient data and other information, we would estimate
risk reduction for some dietary supplements based on the chronic illness
or condition that would help prevent.  The model for risk reduction
would start with the situation before the final rule, where the number
of chronic illness prevented by dietary supplements is:

Number of persons at risk for the chronic illness x percent of persons
at risk who take dietary supplements (before the final rule) x reduction
in the risk of chronic illnesses if person at risk consumes dietary
supplements (before the final rule)  

After the final rule takes effect, the number of chronic illnesses
prevented would be:

Number of persons at risk for the chronic illness x percent of persons
at risk who take dietary supplements (after the final rule) x reduction
in the risk of chronic illnesses if person at risk consumes dietary
supplements (after the final rule)  

The number of chronic illnesses prevented by the final rule would be: 

Number of persons at risk for the chronic illness x change in percent of
persons at risk who take dietary supplements x change in the reduction
in the risk of chronic illnesses if person at risk consumes dietary
supplements 

For some chronic illnesses or conditions, the final rule can be expected
to change the numbers of persons consuming dietary supplements and the
reduction in the probability of chronic illness brought about by
consuming dietary supplements.  

	If we could determine the change in the number of chronic illnesses
prevented by dietary supplements as a result of this final rule, we
could estimate benefits by multiplying the additional number of chronic
illnesses prevented by the value of preventing those illnesses.  The
values consumers place on preventing illness differ across illnesses and
across consumers, and are related to the reasons they use dietary
supplements.  We will illustrate the method with two examples: Calcium
and osteoporosis and folic acid and neural tube defects.

Calcium and osteoporosis.  Many consumers take calcium supplements to
reduce the probability of osteoporosis, which afflicts as many as 10
million people over age 50 (about 8 million women and 2 million men). 
An additional 34 million men and women may be at risk for developing
osteoporosis (Ref. E14E17). If ensuring that calcium supplements contain
what they should reduces the risk of osteoporosis, the total
osteoporosis health benefits associated with the final rule will be the
number of cases prevented multiplied by the health costs per case.  We
estimated the health costs per case as the sum of the direct medical
costs, the value of functional disability, and the value of the pain and
suffering associated with the illness.  Cases range in severity from
mild to severe.  A mild case, for example, might lead to a loss of
utility (measured as quality adjusted life years – a year of life
adjusted for the individual’s health status) of 0.14 per year for 9
years.  If we apply a discount rate of 7 percent to the years the
condition lasts, the loss of quality adjusted life years is about 0.9
(6.5 discounted years x 0.14 lost utility per year).  In other
rulemakings we have used a range of values for a quality adjusted life
year; the range has been from $100,000 to $500,000, with a medium
monetary value of $300,000 (68 FR  41434, July 11, 2003).  With a value
per year of $300,000, the value of preventing a mild case is about
$270,000 (0.9 x $300,000). 

A severe case, by contrast, can lead to fractures and permanent
disability.  Also, osteoporosis in women can occur at early ages and
last decades.  If someone suffers from osteoporosis for 30 years, the
discounted quality adjusted life years lost would be 6.9 (12.4
discounted years x 0.56 lost utility per year).  We estimate that
medical costs for a severe case can be over $17,000.  The value of
preventing a severe, long-lasting case is therefore about $2.1 million
((6.9 x $300,000) + $17,000). 

Folic acid and neural tube defects.  Many women of child-bearing age
take dietary supplements to help ensure their own health and the health
of their children should they become pregnant.  For example, 40 percent
of women aged 18 to 45 take supplements containing folic acid, which may
reduce the probability that children will be borne with neural tube
defects (Ref. E15E18).  Neural tube defects affect the spine (spina
bifida) and the brain (anencephaly).  About 3,000 pregnancies are
affected each year (Ref. E15E18). 

	The benefit of ensuring that folic acid supplements contain what they
should equals the population at risk multiplied by the reduction in the
probability of neural tube defects, multiplied by the value of
preventing a neural tube defect.  Neural tube defects involve large
medical expenses, and either early death or permanent disability.  The
lifetime medical costs alone are between $400,000 and $500,000 for spina
bifida (Ref. E17E19, with values updated).  In recent rulemakings, we
have used $5 million to $6.5 million as the value of a statistical life,
defined as the willingness to pay for reductions in small risks of
premature death.  Preventing a statistical death from anencephaly would
therefore generate benefits of $5 million to $6.5 million.  For spina
bifida, one estimate is that an average case leads to a loss of more
than 15 quality adjusted life years, for a monetized loss of close to $5
million for a non-fatal case if valued at $300,000 per quality adjusted
life year (Ref. E16E20).  The value of preventing a case of spina
bifida, then, is the sum of medical costs and the value of a saving the
quality adjusted life years, or about $5 million ($450 million value of
quality adjusted life years + $500,000 direct medical costs). 

	Estimating the total benefits of this final rule requires estimates of
the numbers of chronic illnesses and conditions whose incidence can be
further reduced by ensuring that dietary supplements contain what they
should.  Because we have no information on the baseline number of
chronic illnesses caused by deficient or excessive ingredients, or on
the change in the likelihood of chronic illness that will occur as a
result of the provisions of this final rule, we cannot estimate the full
benefits of ensuring that dietary supplements contain what they should. 
Our quantified benefits for this final rule must therefore consist
entirely of the benefits from reducing the risks of acute illnesses and
reducing the number of product recalls.  The total benefits will be
larger by an amount we are not able to quantify.

     (Comment 3457)  We received many comments about the estimated
benefits as measured by the value of hypothetical search time. 

	(Response)  We are no longer using the search model. 

	f. Total benefits.  The total benefits from the final rule are the sum
of the value of health benefits from fewer acute illnesses, the value of
fewer product recalls, and the value of the health benefits from fewer
chronic illnesses.  Table 26 shows the total benefits.

	(Comment 3468)  One comment states that our total estimated benefits
could be as little as $21 million.

	(Response)  Our current estimate of total quantified benefits is $448
million per year, once the final rule takes full effect.  In addition,
as discussed above, there are benefits to this rule that have not been
quantified.  The unqualified benefits estimate is the mean of a range of
estimates based on assumptions about reporting rates and the
effectiveness of the final rule. 

In the analysis of benefits for this rule there are two large
uncertainties: quantified underreporting of acute illnesses and injuries
and non-quantified benefits associated with chronic illnesses.  Despite
the best efforts by public health authorities, there will always be
underreporting of illness and injuries.  Where fatalities are concerned,
unless there are litigation problems or the potential for the spread of
infectious disease, there is no incentive to do extensive forensic work
to determine whether a fatality is related to the ingestion of a dietary
supplement.  This leads to reporting most fatalities under the most
general International Classification of Diseases codes.  We acknowledge
the large uncertainties in our estimate because of these factors.

The degree of prevention of chronic illnesses due to preventing
super-potent or sub-potent dietary supplements depends on two factors,
both of which are highly uncertain.  The first factor concerns product
benefit: how many dietary supplements have any beneficial effect on
chronic illnesses and how strong are those effects?  Recent work in this
area so far has examined only a few dietary supplements, with mixed
results.  Of course, ensuring the potency of an ingredient that has
adverse effects or has adverse interactions with drugs would subtract
from the benefits.  The second factor is the incidence and effects of
sub-potency and super-potency across products and over time: how much of
a difference in the product need there be to generate a substantial
adverse health effect?  Because of these uncertainties, it is virtually
impossible to make any sort of quantitative statement about likely
effects of a regulation ensuring against super-potency and sub-potency.

	Because of the uncertainties in estimating the benefits associated with
both chronic and acute illnesses associated with manufacturing practices
for dietary supplements, the decision to implement regulatory
requirements becomes an exercise in weighing precautionary measures for
public health against expenditure of scarce resources. 

	In the uncertainty and sensitivity analyses in section XXIV.B.9, we
show how uncertainty and different assumptions generate higher or lower
quantifiableed benefits.  Using plausible assumptions about the
uncertain variables, we estimate that total quantified benefits (using
1990-1999 data) most likely fall within a range of $810 million to $6480
million per year.

Table 26. Summary of Annual Benefits

Benefits	Mean

Fewer acute illnesses 	$404 million

Fewer product recalls 	$43 million

Fewer chronic illnesses 	Not quantified

Total quantified benefits	$448 million



7.  Costs

     The same changes in manufacturing practices that produce benefits
also have costs, the opportunity costs of reducing some of what they are
now doing.  For example, firms may spend fewer resources on worker
safety, product development and marketing, or testing the efficacy of
their products.  The final rule will require dietary supplement
establishments to adopt some new practices in order to manufacture,
package, label, or hold their products in compliance with CGMP
requirements.  In some cases, establishments will make capital
improvements to the physical plant, add or replace equipment or
controls, perform additional maintenance, establish written procedures,
keep records, carry out tests, monitor production and process controls,
or execute a variety of additional tasks that they may not have
previously performed.  Not all firms will comply – some will go out of
business or move their plants to other countries and not sell their
product in the United States.  We estimated the additional costs of
production associated with the final rule and the leading regulatory
options using the survey (Ref. E3E2) to estimate baseline manufacturing
practices.

     a. Description of the costs.   To estimate costs for the dietary
supplement industry, we initially divided the industry into four product
categories and three size categories.  Because the survey showed that
there were only a few establishments in some categories, we consolidated
the size and product into three size categories.  The size categories
were:

•     Very small (fewer than 20 employees)

•     Small (20 to 499 employees)

•     Large (500 or more employees)

Although this consolidation glosses over the important differences
across products, the purpose is to estimate the broad average costs of
the rule.

     For each size category, we constructed a cost model that included
every provision of the final rule.  We then attached a cost to each
provision that had an additional activity associated with it.  Most
provisions did not have costs attached to them, because they were either
descriptive or the costs were included elsewhere.  

The costs will be the marginal, or additional, costs of the activities
producers undertake in response to the provisions of the final rule. In
the cost model, we expressed the cost as cost per unit, with the unit
being the establishment, the number of employees, or the annual number
of batches produced or affected.	

	b.	Summary of general comments on costs.  We received many comments on
the costs of the 2003 CGMP Proposal.  Many of the comments were general
in nature and addressed the belief that our economic analysis
underestimated the total costs of the 2003 CGMP Proposal, both first
year costs and annual costs.  Numerous comments point to the rule’s
testing requirements as the main cause of the high costs.  Comments also
state that the analysis underestimates costs of hiring new workers,
capital equipment, and holding and distributing costs.  In addition,
some comments point out that the economic analysis did not include
estimates of costs of holding reserve samples and tracking product
complaints.  

     As a result of the 2003 CGMP Proposal, comments assert, product
choice would decline, prices of existing products would increase, and
many businesses -- particularly small businesses -- would be forced to
shut down.  One comment states there could be a decrease in spending on
research and development.  Some comments state that the burden on
business could be alleviated by allowing the use of certificates of
analysis for incoming raw materials and using a statistical, or more
flexible, testing regime instead of requiring final product testing on
all batches.  

     A comment from a trade association representing ingredient
suppliers and manufacturers in the dietary supplement industry accepts
our assumptions on the following variables:

•     The number of control points

•     The average number of ingredients per product

•     The average cost per test

Other comments, however, state that the average number of ingredients is
higher than estimated and that the average cost per test is higher than
estimated; one comment from a manufacturer states that its average cost
was 2.5 times our estimate.  These comments came from self-described
small firms.

	(Comment 3479)  One comment states that we failed to consider start-up
costs.

	(Response)  We include start-up costs (also referred to as set-up or
one-time costs) throughout this analysis. 

	(Comment 34850)  Many comments on the regulatory impact analysis
targeted our estimates of firms’ batches per year.  Nearly all
comments about batches state that our batch estimates are too low.  For
example, an industry trade groups claims our estimate of 309 batches per
year for large firms is “implausibly low.”  The same comment states
that the distribution of the number of batches per firm of 309, 554, and
223 for large, small, and very small firms, respectively, is
“illogical” because it does not make sense that large firms would
have fewer batches per year than small firms.  

	(Response)  Due to a contractor’s error, we used an inaccurate
estimate of the annual number of batches in the analysis of the 2003
CGMP Proposal.  The analysis of the final rule corrects for this error. 
The corrected mean numbers of batches per firm are 444 for large, 2436
for small, and 1164 for very small firms.  The corrected estimates of
the number of batches continue to show that small firms produce more
batches than large firms.  Comments from self-described small firms
suggest that this distribution of batches is reasonable.  These comments
state that many small firms produce many small batches of product using
machinery with smaller capacity than that used by large firms. 

	(Comment 34951)  One comment states that we used faulty data in the
economic analysis.

	(Response)  In accordance with our information quality guidelines, we
have used the best available data in this analysis.  As explained in the
response to the previous Ccomment 348, the survey results used in the
analysis of the 2003 CGMP Proposal included an inaccurate estimate of
the number of batches of dietary supplements produced.  We use the
corrected estimate in the analysis of this final rule.  

	(Comment 3502)  Some comments dispute the estimated testing costs.  In
particular, comments question our assumptions on:

•     The number of tests required per batch;

•     The number of tests already being performed;

•     The costs to perform specific analytical tests; and

•     The development of analytical methods.

	(Response)  The final rule reduces the number of required tests.  In
the final rule, we account for tests where no analytical methods have
been developed.  We now require fewer tests, although we anticipate that
some testing will take place associated with the creation of
certificates of analysis required for component specifications and as
verification for process controls.  We now assume that the tests will
be: 

•  One identity test for each shipment lot of incoming dietary
ingredients (e.g., Vitamin C),

•  Tests of subsets of shipment lots by supplier firms to create
certificates of analysis for identity of other components (e.g., sugar),

• Tests of subsets of shipment lots for other specifications in the
certificates of analysis,

• Tests of subsets of batches of dietary supplements for microbial,
chemical, or physical contaminants,

• Tests of subsets of batches of dietary supplements for
specifications, and

• Tests for meeting requirements that water used to manufacture
dietary supplements complies with Federal, State, and local requirements
and does not contaminate the dietary supplement.

     We are not changing our estimate of the current prevalence of
testing, which is based on the survey of manufacturers (Ref. E3E2).  We
would only revise this estimate in light of new data of comparable
quality to that provided by the survey. 

	(Comment 3513)  We did receive two comments favorable to recordkeeping,
stating that master and production batch records were good to adopt and
that associated costs will be minimal.  One of the comments states that
the level of detail may be unrealistic for a small firm, but also states
that any final regulation could be made more flexible for small
manufacturers.

     Although there were favorable comments, we received several
comments critical of the recordkeeping requirements.  These comments
make general statements that the economic analysis underestimates the
recordkeeping burden and some added that these requirements go beyond
the CGMPs for food.  In addition, several of the comments include
firms’ own estimates of costs of complying with the recordkeeping
requirement.  Comments estimate costs in the range of $11,000–$64,000.

	(Response)  The recordkeeping requirements in the final rule differ
from the 2003 CGMP Proposal and are addressed in the revised costs below
and in the Paperwork Reduction Act analysis.

	(Comment 3524)  We received a favorable comment regarding the
requirements for physical plant and equipment, saying that, although the
costs would be moderate, the result would be higher quality products. 
Another comment states that, although not unrealistic, the provision
would be very costly. 

     Other comments are more critical.  One comment estimates that
renovation expenses would amount to approximately $600 million over the
entire industry, as opposed to our estimate of $45 million.  This
comment states that the reason our estimates in the 2003 CGMP Proposal
were too low is that we apply a reduction factor which assumes that 18
percent of very small firms, 10 percent of small firms and 1 percent of
large firms will have to make capital improvements.  It is more
appropriate, the comment states, to assume that most facilities will
need to renovate about 10 percent of their plant, regardless of firm
size.  In addition, the requirement that plants have smooth, hard
surfaces on all floors, walls and ceilings is unrealistic and would add
quite a bit of cost.  The comment asserts no company will have such
surfaces throughout the plant and, this is not a requirement in either
the food or drug CGMP requirements.  Other comments echo the belief that
capital expenditures would be greater than our estimates and would be
excessively burdensome.  One comment estimates this cost at
approximately $83,000 per facility.

     The comment also estimates that a large firm that needs to expand
its capacity could expect to incur costs of $240,000, as opposed to the
$2,000 that the comment says we estimated for large firms.  In addition,
it is pointed out that equipment costs could be burdensome to small
firms, which likely do not have well-equipped labs.  This thought is
affirmed by other comments that estimate that new equipment could cost
anywhere from $50,000 to $1 million, with annual costs estimated between
$15,000 and $100,000.  In addition, expansion of laboratory space is
estimated at $200 per square foot, as opposed to the agency’s estimate
of $50 per square foot.  Lastly, one comment suggests we work with the
Internal Revenue Service to allow for more rapid depreciation of
facility costs to help small business make facility upgrades.

	(Response)  In the analysis of the 2003 CGMP Proposal, we estimated the
number of firms needing to make capital expenditures associated with the
rule as a distribution, with the parameters of the distribution
determined by the size of the facility.  We assume that if a firm does
make a capital investment in response to the rule, it would affect about
10 percent of the plant.  With an estimated cost of $50 per square foot,
and the average size of a very small plant of about 25,000 square feet,
the cost per very small establishment making a capital investment would
be about $125,000.  With the average size of a small plant of about
70,000 square feet, the cost per small establishment making a capital
investment would be about $350,000.  With the average size of a large
plant of about 600,000 square feet, the cost per large establishment
making a capital investment would be about $3 million (Ref. E3E2).  We
assume that most facilities will not need to make capital investments to
meet the sanitation requirements of this final rule as, according to the
survey results, most establishments already meet the sanitation
standards of this final rule.  This would not be possible if their
facilities were inadequate.  We note that the final rule does not
require smooth and hard surfaces throughout the plant. 

	We estimated the capital costs as the costs of minor renovations to
help meet sanitation requirements, not as the cost of (say) expanding
the size of a laboratory or some other technically sophisticated change.
 Although some facilities may choose to expand laboratories, the testing
requirements of this final rule should be able to be met by existing
laboratory facilities within or outside of the manufacturing facilities.
  

     Working with the Internal Revenue Service on depreciation is beyond
the scope of our authority.  We will provide advice on financing capital
improvements through our small business representatives in the Office of
Regulatory Affairs.

	(Comment 3535)  Many comments address costs resulting from what
industry describes as the exhaustive testing requirements outlined in
the 2003 CGMP Proposal.  Comments point out that the requirement to test
every ingredient would be very costly for firms large and small, with
many firms stating that they risk going out of business.  In addition,
several comments add that the testing requirements would do little to
enhance product quality.  Many comments assert that allowing the use of
a certificate of analysis would reduce the amount of tests performed on
a shipment of incoming raw materials, reducing redundant testing, and
also reducing the risk that a firm may go out of business.  Other
comments state that allowing statistical testing regimes would also cut
down on testing costs.

	(Response)  As we already have discussed in this section, we have
reduced the amount of required testing in this final rule.  The final
rule requires testing the identity of every incoming dietary ingredient.
 However, the final rule allows for use of certificates of analysis in
place of identity tests of other components and other tests of incoming
dietary ingredients and other components.  The final rule also allows
sound statistical testing regimes for finished products.

	(Comment 3546)  One comment states that our cost estimates are based on
the assumption of only two ingredient tests, an assumption which the
comment calls into question.  For multivitamins, one comment estimates
about 8 separate tests and 16 separate assays, depending on the
nutrients present.

	(Response)  In the analysis of this final rule, we assume one identity
test per incoming shipment lot of dietary ingredients based on the
revisions made to the final rule compared with the 2003 CGMP Proposal.  

	(Comment 3557)  Many comments include individual estimates of testing
costs.  For example, two comments estimate an average cost of about $100
per test, and other comments estimate averages as high as $360, as
opposed to our estimate of about $60 per test.  Several other comments
claim that the costs of finished product testing alone would be “at
least 100 times” greater than our estimates; other comments state that
testing costs would almost equal the costs of manufacturing.  One
comment estimates testing costs for firms of all sizes at $245 million,
as opposed to our estimate of $24 million, although another contains
estimates as high as $13.6 million annually for one firm.  Two comments
concede that some finished product testing may be necessary.  

     In addition, some comments state that our estimate of finished and
raw material testing is off by a multiple of 3 to 6.  One comment states
that, for companies that have products which contain a large number of
ingredients expensive to test, very large costs will be incurred.  This
comment also states that our cost estimates do not include in-process
testing, which they claim the rule would clearly require.  Specifically,
our analysis suggests that an average of 2.5 in-process tests per batch
are likely to be needed at critical control points.  In addition, the
comment maintains that our analysis showed that additional testing may
be required for an average of 2.5 components of herbal products and 7.5
components of herbal products, but our estimates do not include costs of
the tests.  Finally, comments point out that, if the production system
is properly controlled, then a “reduced schedule” of final product
testing is justified and that focusing excessive resources on
end-product testing does not constitute GMP.  Quality controls should be
built into the production and process system from the beginning of the
manufacturing process.

     A comment also states that our estimates of firms that already test
are inaccurate.  The comment asserts that our estimates are overstated
and they also think we have understated that proportion of finished
batches not currently being tested.  In addition, the comment claims
that “even large firms that are testing 90 percent of their products
are unlikely to be performing the exhaustive level of testing required
by the 2003 CGMP Proposal, namely testing every component of every batch
of finished product”.

     The comments point out that our cost estimates do not include
estimates for the cost of developing methods of analysis for
ingredients.  At a minimum, one comment states this estimate should be
$2 million (the cost of 100 methods at a minimum of $20,000 each). 
Several comments point out that often there are no existing
scientifically valid analytical methods to test finished products,
especially botanical products.  Another comment states that costs of
analytical testing are at least three times our estimate, and could be
as high as eight times our estimate.  Because of this, many comments
call for the use of a certificate of analysis in place of analytical
testing.

     Another comment states some unintended consequences could occur in
the industry due to the testing requirements, including stress on the
current contract laboratory facilities and in-house laboratories, and
also increases in holding costs, due to changes in turn-around time at
outside labs.  Other comments point to the loss of product choice that
could occur if the testing requirements force manufacturers to go out of
business or discontinue certain products.

	(Response)  In response to comments, we have revised the testing
requirements in the final rule.  We also have revised our estimates of
the costs of testing.  In what follows, we describe the estimated number
and costs of tests required by this final rule. 

     The final rule requires tests for identity for each incoming
shipment lot of dietary ingredient.  Estimating the number of tests per
batch is complicated, because the tests are required on the shipment
lots and we have data only on the number of batches of dietary
supplements produced.  For example, if a shipment lot of some dietary
ingredient is used in 6 batches of final products, it would need to be
tested for identity only once.  The number of required identity tests
per batch of final product will equal the number of dietary ingredients
per batch, divided by the average number of batches per shipment lot (to
account for the production of multiple batches of dietary supplements
from single lots of components).  In addition to the required identity
tests, a subset of other components will be tested for identity (these
tests are likely to be the responsibility of suppliers and need only be
done once per batch no matter how many recipients of those batches). 

	The quantity and quality of evidence on the variables used to estimate
the number of tests varies greatly.  In this section, we explain the
evidence and assumptions we used to construct the formulas for the
number of tests.

     Number of dietary ingredients.  We based our measure of the number
of dietary ingredients per product on a sample of almost 3,000 dietary
supplement labels (Ref. E18E7).  Although some ingredients may be
missing from the labels and some listed ingredients may be missing from
the products, the ingredient list represents the best evidence we have
on what ingredients are used in dietary supplements.  Although comments
claimed that we underestimated the number of ingredients, they offered
no evidence that would persuade us to change our estimates, which are
based on a sample representing at least 10 percent of the products in
the market. 

     According to the sample of listed ingredients, vitamin and mineral
products contain about 13 listed dietary ingredients.  Other dietary
supplements, mainly herbals, contain about 4 listed dietary ingredients
(Ref. E18E7).

     Number of unlisted components.  Dietary supplements are
manufactured using solvents, binders, and lubricants that may not show
up in the final product.  An industry source (Ref. E20E21) says that 4
to 6 unlisted components are typical per product, although fewer are
certainly possible.  The minimum number is zero.  Based on industry
data, we assume that the number of unlisted components would be zero to
6 for vitamins and minerals, and zero to 4 for herbal and other
products. 

     Number of shipments (that is, shipment lots) of ingredients and
unlisted components.  We have no direct information on the number of
shipment lots of dietary ingredients and other components.  We also have
no information on the number of shipments per lot or on the number of
shipments per batch.  It is costly to store components, so some
establishments may buy many small lots of dietary ingredients and other
components rather than a few large lots.  Crude botanical and other
ingredients are inherently unstable and may lose their stability in even
a short time unless costly temperature, humidity, and light controls are
in place.  We also know, however, that some dietary ingredient suppliers
produce and ship ingredients in large lots.  For dietary supplements
produced using part of a large production run of a dietary ingredient,
the number of batches per shipment lot could be large.  Also, some
producers buy a single large shipment lot of a raw material and use it
in many batches.  We assume that as many as 12 batches per shipment lot
of dietary ingredient is a plausible maximum.  Based on consultation
with industry (Ref. E20147), we assumed, in the cost calculation, that 1
was the minimum and 12 the maximum number of batches produced per lot,
with 6.5 the average.  We received no comments on our use of the
assumption in the analysis of the proposed rule and continue to use it
in our analysis of the final rule.  In the sensitivity analysis, we show
how costs change when we change the assumption. 

     Number of batches produced.  We have survey results (Ref. E3E2) on
the number of batches produced per establishment.  Several comments
stated that we underestimated the number of batches produced, which we
found to be the case because of an erroneous calculation in the
contractor’s report.  In the revised contract survey results, very
small establishments produce an average of 444 (revised from 223)
batches per year, small establishments produce an average of 2,436
(revised from 554) batches per year, and large establishments produce an
average of 1,164 (revised from 309) batches per year. 

     Number of final product tests per batch.  We have reduced the
number of tests required for final products.  We assume that
establishments will test a representative sample of batches to ensure
that the final products meet specifications. We do not specify any
particular statistical sampling plan. 

     Costs per test.  We estimate the costs per test partly with
published prices of independent laboratories as posted on the Internet
(Refs. E21 E22 and E22E23), and partly from our conversations with FDA
and industry experts on testing.  Testing costs vary according to
frequency and complexity.  The more frequently technicians perform
tests, the lower are the costs per test.  Many tests require
sophisticated equipment, such as gas chromatography, high pressure
liquid chromatography, distillation, extraction, various
spectrophotometers, and other types of equipment.  Using sophisticated
equipment requires trained personnel.  Even simple physical or
organoleptic testing requires training or experienced personnel.  The
type of ingredient, compound, or product can also affect the cost
because some are easily identified using routine or single step
techniques and others require multiple steps or complex techniques,
especially if there are similar products that can be mistaken for the
products being identified.  The type of defect tested for affects the
cost; some defects can be found visually if they are found on the
surface, but others are latent.  Some tests require multiple samples or
multiple steps.  In addition, tests require taking and preparing
samples, whose cost can vary.

By assuming a single distribution for testing costs, we may overestimate
testing costs for sectors or products with below-average costs and
underestimate testing cost for sectors with above-average costs.  In the
cost model, for example, we distinguish between botanical ingredients
and non-botanical ingredients in the number of tests but not in average
testing costs.  If the average cost of testing botanical products is
higher than the average cost for vitamins and minerals, the distribution
of costs may underestimate total testing costs for botanical products. 
We do not have sufficient information on the range of testing costs for
botanical ingredients to determine if the average cost of testing is
higher or lower than for other ingredients.  

     The average cost per test is about $60, based on a range of costs
we found on the Internet.  This cost represents the full cost of
carrying out a test, including collecting and storing the sample, the
time for training the personnel who carry out the test, and any
associated records.  We assume that $20 per test represents a lower
bound.  Although some Internet prices for tests are as high as $300, we
assumed that with frequent testing $150 would be a more plausible upper
bound average cost.  The majority of listed prices fell into the $20 to
$80 range, so we selected $50 (the midpoint) as most likely.       

	The number and cost of tests: summary.  We estimate the number of tests
required of the representative manufacturer as a weighted average of the
number of tests required for vitamins and minerals and the number of
tests required for all other supplements (which were mainly herbal
products).  The weights, as shown below, differ by size of manufacturer:

•     24 percent of very small manufacturers produce vitamins and
minerals; 76 percent produce other dietary supplements.

•     42 percent of small manufacturers produce vitamins and minerals;
58 percent produce other dietary supplements.

•     69 percent of large manufacturers produce vitamins and minerals;
31 percent produce other dietary supplements.

     Most establishments already conduct some tests, or send samples out
for testing.  We therefore adjusted the estimated testing costs of the
final rule to include only required tests and to account for the testing
costs currently borne voluntarily by manufacturers.  The survey results
showed how many respondents were conducting various types of tests.

Table 27. Values Used to Estimate Testing Costs

Name	Value or Distribution Used	Source

Number of dietary ingredients per product batch	Vitamins and
minerals--13

All other categories—4	Sample from 3,000 dietary supplement labels
(Ref. E18E7)

Number of identity tests per dietary ingredient lot	1 identity test per
ingredient lot	Based on requirements of final rule 

Number of identity tests per other component lot	1 identity test per
subset of component lots	Assumption based on use of certificates of
analysis for ingredients

Number of tests for specifications per ingredient lot	1 to 5 tests per
subset of ingredient lots	Assumption based on use of certificates of
analysis for ingredients

Number of unlisted components	0 to 6 components for vitamins and
minerals, 0 to 4 for herbal and other products	Ref. E20E21

Number of shipments (lots) of ingredients and unlisted components	1 to
12 batches per shipment lot of dietary ingredients	Assumption based on
discussions with industry

Number of batches produced per year	Very small establishments–444

Small establishments–2,436

Large–1,164	Ref. E3E2

Number of final product tests per batch	1 test per subset 	Based on
requirements of the final rule

Costs per test	Beta pert distribution skewed rightward between $20 to
$150; $50 most likely; $60 average	Refs. E21 E22 and E22E23



	(Comment 3568)  We received comments on labor costs that would be
incurred as a result of the 2003 CGMP Proposal.  All comments state that
personnel costs will increase significantly.  One comment states that
the average manufacturing wage we used to estimate labor costs, $15.65,
does not reflect the true cost of additional labor, since higher skilled
employees, such as quality control engineers and, as one comment
asserts, PhD-level employees, will need to be hired to comply with the
rule.  This comment states that, including benefits, the wage actually
ranges between $23.28 and $72.00 per hour, depending on skill.  Other
comments estimate additional annual labor costs ranging between $25,000
and $350,000.  

	(Response)  We used more recent estimates to the average manufacturing
wage cost of $216 per hour to estimate the cost of labor (Ref. E19E24). 
The comment that asserted Ph.D.-level employees are needed to comply
with the rule, provided no basis for this assertion.  We disagree that
Ph.D. level workers are needed for the tasks required by this final rule
because most of the costs estimated as labor costs all involved ordinary
labor tasks such as sanitation, monitoring, and record keeping.  For
more difficult or complicated tasks, we assume that more skilled workers
hours maywill be required but the overall average labor cost represents
the best overall estimate for valuing the average cost of labor in the
industry.  We assume that various tasks required by the final rule would
take some number of hours per year, per batch of product, or per square
foot of physical plant.

 	Estimating costs.  We initially gathered information and made
assumptions about the full cost of a provision.  We then adjusted these
estimates to account for the many activities already being carried out,
as well other activities that would not have to be carried out by all
establishments.  We used the survey to estimate the likelihood that an
establishment will incur a cost.  To get an estimate of the average cost
of a provision (adjusted for baseline activities) for each category, we
multiply the average cost per establishment by the probability that the
establishment will need to undertake the expense (one minus the
probability that the establishment is already doing it).  For each
provision of the final rule, the simulation carried out the following
calculation:

Cost per unit of analysis for each provision = 

number of units of analysis per establishment x 

probability that establishment incurs cost x 

cost per provision per establishment.

	We estimate both a setup cost (a one-time fixed cost) of the provision
and an annual recurring cost.  To get the total costs of the rule, we
multiply the number of establishments in each size category (from the
survey) by the average costs per establishment in that category.  We
then adjust for the establishments that did not respond to the survey
but are believed to be in the industry.  Two hundred thirty-eight
establishments responded to the survey; we estimate that 1,566 firms are
in the industry, including ingredient suppliers. The number of firms
covered by most of the provisions will therefore be about 1,460.  

We estimate total costs with the following calculation:

(Number of very small establishments x costs per very small
establishment)  +  

(number of small establishments x costs per small establishment)  + 

(number of large establishments x costs per large establishment)]  +x 

(number of warehouses x costs per warehouse)adjustment for
establishments not in survey

     The rule is complex and the industry is made up of very different
kinds of firms, so cost estimates are averages with, in some cases,
large variances.  The cost per unit, number of batches and employees,
and probability that the establishment would incur the cost all contain
uncertainty.  The values in the table are used in the cost estimates,
and are generated from multiple sources.

Table 28. Additional Values Used in Cost Calculations

Name	Value or Distribution Used	Source

Average wage per hour	$216	Employment Index, Bureau of Labor Statistics
(Ref. E19E24)

Average size of establishments in square feet	very small = 24,674;

small = 71,354;

large = 596,000	Ref. E3E2

Average number of employees	very small = 7.6;

small = 95;

large = 1,005	Ref. E3E2

Procedures	8 to 16 hours set-up time for small firms; 30 to 40 hours for
large firms; annual cost is 10 percent of setup time per provision	Ref.
E23E25

Personnel sanitation	1 hour per week per worker	Assumption, based on
requirements of final rule

Sanitation time for physical plant	1 hour per week for very small
establishments; costs per small and large plants scaled in proportion to
size of plant	Assumption, based on difference in average physical plant
size



Sanitation supervisor	1 hour per week	Assumption, based on requirements
of final rule

Pest control setup costs	$1,500 to $2,000 for very small establishments;
$1,800 to $2,400 for small establishments; $2,600 to $3,400 for large
establishments.  Average for each size establishment was midpoint
($1,750, $2,100, $3,000)	Ref. E24E26

Pest control annual costs	$400 to $600 per month for very small
establishments; $480 to $720 for small establishments; $700 to $1,000
for large establishments.  Average for each size establishment was the
midpoint ($500, $600, $850)	Ref. E24E26



Renovation cost	$50 per square foot; with 0 to 20 percent of physical
plant to be renovated, with 10 percent most likely	Based on construction
costs and square feet (Ref. E3E2)



Equipment replacement	For very small establishments, 0 to $1,000;costs
per small and large plants scaled in proportion to size of plant
Assumption, based on size of establishments (Ref. E3E2)

Setup costs for automatic equipment	$500 for hardware, 16 hours	Software
costs and assumptions about labor hours



Annual costs for automatic equipment	10 percent of set-up costs
Assumption based on requirements of the final rule

Sanitation of equipment and surfaces	5 hours per week for very small
establishments; costs per small and large plants scaled in proportion to
size of plant	Assumption based on average sizes of establishments (Ref.
E3E2)



Holding products and dietary ingredients: capital requirements	Same as
costs of equipment upgrades	Based on average sizes of establishments
(Ref. E3E2)

Default probabilities that establishments are not currently acting in
accordance with a provision	For very small establishments, 0.2; for
small establishments, 0.05, for large establishments, 0.01	Based on
results of survey for other practices (Ref. E3E2)



     The total set-up costs for this final rule will be $4137 million,
spread out over the 36 months following the publication date of the
final rule.  The annual costs, once the final rule is fully implemented,
will be $16450 million, with the 2 largest costs being $5260 million for
testing and $2417 million for records.  The estimated total cost is the
mean of a range of estimates based on the data and assumptions described
in tables _(8)27 and 28(9).  In the uncertainty and sensitivity analyses
below, we show how uncertainty and different assumptions generate higher
or lower estimated costs.  Using plausible assumptions about the
uncertain variables, we estimate that total quantified costs most likely
will fall within a range of $10490 million to $32200 million per year.

8.  Summary of Benefits and Costs

     We estimate that, once it is fully implemented, the annual
quantified benefits from the final rule will be $810 million to $6480
million, with a mean estimate of $448 million.  However, there are
potentially large benefits of the rule that we were not able to
quantify.  The annual costs will be $10490 million to $32200 million,
with a mean estimate of $16450 million.  The rule will not be fully
effective until 36 months after the publication date.  Table 29 shows
how the phase-in of the final rule will generate the costs and benefits
for the first 4 years.  Table 30 shows the present and annualized values
of costs and benefits over 20 years, calculated at discount rates of 3
percent and 7 percent.  We have determined, based in part on the
analysis presented here, that the benefits of this final rule justify
the costs.

Table 29. Costs and Benefits by Year

	1st year	2nd year	3rd year 	4th year

Costs

(in millions)	$169	$12036	$19067	$16450

Benefits

(in millions	$32	$2938	$448	$448



Table 30. Present and Annualized Values of Costs and Benefits

	Present value at 3 percent(in billions)	Present value at 7 percent(in
billions)	Annualized Value over 20 years at 3 percent (in millions)
Annualized Value over 20 years at 7 percent (in millions)

Costs

	$2.31	$1.65	$15340	$14938

Benefits

	$0.66	$0.46	$404	$3943



     In table 31 we show the annual costs for each subpart of the
regulation.  We identify selected costs for particular activities for
some of the subparts.  We are unable to estimate benefits by subpart,
because we estimate the benefits by type of benefit rather than by
provision of the final rule.  We classify benefits as arising from
preventing contamination and adulteration and from ensuring more
consistently manufactured products.  The provisions of the final rule
contribute to both sources of benefits, but some contribute
disproportionately to one or the other source.  In the table, we
indicate which source of benefits each subpart supports, but we do not
quantify benefits by subpart.

Table 31. Costs and Benefits by Subpart

Subpart	Set-up cost (in millions)	Annual cost (in millions)	Benefit

A. General provisions	not applicable	not applicable	not applicable

B. Personnel	$1.50	$15.74.3	Preventing contamination and adulteration 

C. Physical plant and grounds	$34.02.5	$17.48	Preventing contamination
and adulteration 

D. Equipment and utensils	$0.9	$2.31.4	Preventing contamination and
adulteration 

E. Establish production and process control system

  (1)Subtotal for identity testing

  (2)Subtotal for all other testing	$0.53

negligible

$0.3	$66.170

$45.0

$ 6.8	Preventing contamination and adulteration and ensuring more
consistently manufactured products 

F. Quality control unit	negligible	$21.1	Preventing contamination and
adulteration and ensuring more consistently manufactured products 

G. Components, packaging, labels and dietary supplements received 
negligible	$31.623.9	Ensuring more consistently manufactured products 

H. Master manufacturing record	negligible$0.1	negligible	Ensuring more
consistently manufactured products 

I. Batch production record	negligible	$5.43.7	Ensuring more consistently
manufactured products 

J. Laboratory operations	$0.21	negligible	Ensuring more consistently
manufactured products 

K. Manufacturing operations	negligible	$2.21.5	Preventing contamination
and adulteration and ensuring more consistently manufactured products 

L. Packaging and labeling operations	$0.1	$10.8.0	Ensuring more
consistently manufactured products 

M. Holding and distributing	$2.71.6	$0.53	Preventing contamination and
adulteration and ensuring more consistently manufactured products 

N. Returned dietary supplements	negligible	$0.2	Preventing contamination
and adulteration and ensuring more consistently manufactured products 

O. Product complaints	$0.1	$4.53.4	Preventing contamination and
adulteration and ensuring more consistently manufactured products 

P. Records and recordkeeping

   Paperwork cost for all subparts	not applicable

$3.7	not applicable

$24.2	not applicable



	(Comment 3579)  We received several comments on the summary of the
costs and benefits.  In general, the comments state that we
overestimated the benefits of the 2003 CGMP Proposal and underestimated
the costs.  Other comments assert that total estimated benefits of the
2003 CGMP Proposal would not be $216.6 million, as estimated by us, but
as low as $13.9 million.  Comments also estimate first year costs as
high as $629 million, as opposed to our estimate of $54.5 million, with
annual costs estimated as high as $860 million, as opposed to our
estimate of $300.6 million.  Other comments predict product prices will
increase, and consumers will decrease consumption of dietary
supplements.

	(Response)  We agree with the comments stating that we underestimated
the costs and overestimated the quantified benefits of the 2003 CGMP
Proposal.  We have increased our estimate of costs in this final rule
compared with the estimate in the 2003 CGMP Proposal.  We have decreased
our estimate of quantified benefits of the final rule compared with the
estimate in the 2003 CGMP Proposal.  As explained above, we are unable
to quantify all of the benefits of the final rule.  These changes in the
estimated benefits and costs of this final rule reflect both the changes
in the 2003 CGMP Proposal and the changes in our analysis in response to
comments. 

     We agree with the comment that part of the costs of this final rule
will be passed on to consumers as higher prices for dietary supplements.
 The annual costs of this final rule are less thanabout 1 percent of
total spending on dietary supplements.  We expect that the majority of
these costs will be borne by consumers of dietary supplements, who will
likely respond to the increase in prices by reducing consumption.  

    The comments suggesting very high costs and very low benefits did
not persuade us that those extreme values were more likely than our
estimates.  We recognize, however, that the uncertainties in our
analysis make a broad range of benefits and costs possible.  In the
analysis of uncertainty we will show the range of predicted benefits and
costs.  We also will show the sensitivity of costs and benefits to
certain key assumptions used in the analysis, and how changes in those
assumptions can generate the extreme values cited in some comments.

9. Benefits and Costs of Regulatory Options

	We considered several regulatory options, including (1) no new
regulatory action, (2) fewer requirements for vitamins and minerals, (3)
more restrictive regulations than the final rule, (4) HACCP without the
other elements of the final rule, (5) final product testing only, (6) a
final rule for high-risk products or hazards only, and (7) the 2003 CGMP
Proposal.  Although we received no comments on our analysis of the
benefits and costs of options (2) through (6), we received many comments
on the estimated benefits and costs of the 2003 CGMP Proposal.  We have
now revised the estimated quantifiable benefits and costs of the 2003
CGMP Proposal.  The revised estimates are based on the comments received
and the corrections made to the data. 

	Using the same method as used in this final rule to determine benefits,
we estimate that the quantifiable benefits of the 2003 CGMP Proposal
would be approximately the same as the quantifiable benefits of the
final rule, $44 million per year. 

	With the corrected estimated number of batches produced, we estimate
that the set-up costs of the 2003 CGMP Proposal would be $51 million. 
If the 2003 CGMP Proposal had been finalized, the annual costs of
complying with the requirements would be $282 million, or about $118
million more than this final rule.  The 2003 CGMP Proposal relied more
on testing final products and other controls closer to the end-product. 
Under the 2003 CGMP Proposal, for example, annual testing costs would be
about $97 million.  

109.	Uncertainties in the Analysis 

     We used indirect measures of the benefits of this final rule which
required several key assumptions that are critical for our estimates. 
With the exception of the recall benefit, which is based directly on our
recall records, each component of the estimated benefits involves
assumptions that reflect our uncertainty.  The estimated costs also
embody key assumptions that reflect our uncertainty., but the costs are
less uncertain because of the greater availability of data on the
components of cost. 

     One assumption that affects both estimated costs and estimated
benefits is that manufacturing practices in the industry will persist in
the absence of additional regulation. If the trend in the market is
toward the adoption of more manufacturing controls than we are proposing
here, then both the costs and benefits of the rule will be less than we
estimate.  If the market trend is toward fewer voluntary controls, then
both the costs and benefits of the regulation will be greater than we
estimate.

         In addition to the general assumption about the effects of the
rule, we rely on several The key assumptions. in the benefits model are:


•     We assume tThere is an average of one reported illness for each
recall.  We assume that for each class 1 and 2 recalled product there
was on average one illness that was reported to the public health
authority.

•     The frequency of actual illnesses is 100 times the frequency of
reported illnesses.  We recognize that there is considerable uncertainty
about the factor of 100., although it has empirical support, might be
wrong and that there is likely to be considerable uncertainty about this
estimate.  The estimated health benefits will vary proportionally with
the assumed multiplier. Direct health benefits double if we increase the
factor multiplier from 100 to 200; lowering the multiplier to 50 halves
the health benefits. 

•     ForIn the baseline estimates, we assumeused $5 million ias the
value of a statistical life and $300,000 ias the value of a quality
adjusted life year. The estimated health benefits change with changes in
those valuations.

•     Finally, wWe assume that the reported class 1 and 2 recalls that
occurred from 1990 through 1999 represent the number and type of recalls
that would have occurred in the future but for the implementation of
this regulation.  If the number or types of recalls are not
representative, then we over or under estimated the benefit of avoiding
recalls. 

     The cost model also relies on several key assumptions.  of the rule
chiefly depend on our assumptions about the amount and cost of
additional testing.  The amount of testing is highly uncertain; we have
tried to model the number of tests based on number of ingredient
shipments and the numbers and types of tests. 

     The key assumptions of the cost model are:

•     We assume that sSingle shipment lots of ingredients and unlisted
components will be used for many batches of final dietary supplement
products.  We assume that the average number of batches per incoming
ingredient lot will be between 1 and 12, with 6.5 the most likely. 

•     Aall testing other than identity testing of incoming ingredients
will be done on representative samples. The number of batches or lots
tested will be the square root of n + 1, with n equal to the total
number of batches or lots. 

•     We also assume that Tthe costs per test, which include the labor
costs of selecting the samples and arranging for the tests, will be
between $20 and $150, with $50 most likely. 

•     We assume that the survey results (Ref. E2) represent the best
estimate available of current practices in the industry. 

     We first characterized the uncertainties as a probability
distribution.  We ran 5,000 computer simulations to estimate both
benefits and costs.  The simulations used distributions (given in the
tables and the text) in place of point estimates.

Table 32. Distribution of Simulation Results for Annual Benefits and
Costs

	5th Percentile	Mean	95th Percentile

Annual costs

(in millions)	$1090	$16450	$260

Annual quantified benefits

(in millions)	$3624	$448	$5479



     The Monte Carlo computer simulations give the distributions of
estimated benefits and costs.  If the underlying distributions fully
capture the uncertainty of the estimates, then the simulation results
give a full picture of the uncertainty.  With uncertain distributions
used in the simulations, however, the ranges reported in the tables may
not fully capture the uncertainties of the analysis.  An alternative way
to show the uncertainty is to see how sensitive the results are to
plausible changes in certain key assumptions.  We start with benefits.

    For our baseline estimated benefits of this final rule, we use a $5
million value for a statistical life and a $300,000 value for a
quality-adjusted life year.  In the sensitivity analysis, we use values
of $3 million for a statistical life and $100,000 for a quality adjusted
life year to generate a “low” estimate of health benefits and values
of $7 million and $500,000 to generate a “high” estimate.   

	The reporting rate of illnesses associated with dietary supplements is
unknown, which makes our estimate of the total number of illnesses
highly uncertain.  The uncertainty is compounded by the additional step
of assuming an average of one reported illness per recall in our
estimated baseline of reported illnesses.  This assumption, however, is
based on the fact that we do observe an average of about one reported
illness per recall.  We find much larger uncertainty in the likely
number of total illnesses associated with reported illnesses.  We use 1
percent as the average reporting rate, which implies that total illness
are 100 times our estimate of reported illnesses.  Although we
assumebelieve that this reporting rate is the most plausible rate for
illnesses associated with dietary supplements, the evidence supporting
it is not strong.  We show the effects of varying the reporting rates of
2.5 percent and 10 percent.  

	(Comment 358)  Several comments questioned our assumption that the
final rule will eliminate the recalls used to estimate benefits.  

	(Response)  We do not assume that all recalls will be eliminated;, we
only assume that the recalls caused by manufacturing problems identified
above will be eliminated if the rule is fully effective.  If the rule is
not fully effective, then the quantified benefits will be less than we
have estimated.  In the following discussionIn table 33 below we show
the effects of different assumptions about the effectiveness of the
final rule., along with the effects of different assumptions about the
reporting rate.  Our base assumptions are that the reporting rate is one
percent (for a multiplier or 100) and the effectiveness of the final
rule is 100 percent. 

Table 33. Sensitivity of Quantified Benefits to Changes in Assumed
Reporting rate and Effectiveness of the Final Rule (in millions)

Prevention

Rate	Reporting Multiplier

	1	10	50	100	200

20%	$0.7	$1.5	$5	$10	$18

40%	$1.4	$3	$10	$19	$37

60%	$2.2	$5	$15	$29	$55

80%	$2.9	$6	$20	$38	$74

100%	$3.6	$8	$25	$48	$92



     In the economic analyses of recent regulations, we have used values
of $5 million and $6.5 for a statistical life. Some estimates place the
value of a statistical life at $3 million.  We have used values of
$100,000, $300,000, and $500,000 for a quality adjusted life year.  For
our baseline estimated benefits of this final rule, we use a $5 million
value for a statistical life and a $300,000 value for a quality adjusted
life year.  In the sensitivity analysis, we use values of $3 million for
a statistical life and $100,000 for a quality adjusted life year to
generate a “low” estimate of health benefits equal to $25 million,
and values of $6.5 million and $500,000 to generate a “high”
estimate equal to $66 million.  

The quantified benefits depend on the hazards found in recalled products
between 1990 and 1999.  The 69 recalls in 1998 dominate the estimate,
accounting for 58 percent of class 1 and class 2 recalls, and 35 percent
of all recalls for the decade.  In this sensitivity analysis we estimate
the effect of excluding 1998 from the data used to estimate average
annual benefits.  We also consider the effects of using the annual
average number of recalls from 2000-2005 to estimate benefits.  

Table 33. Sensitivity of Benefits

Description	Estimated Annual Benefits (after 3 years)

(in millions)

Final rule	$44

VSL = $3 million and $/QALY = $100,000 (baselines are $5 million and
$300,000)	$24

VSL = $7 million and $/QALY = $500,000(baselines are $5 million and
$300,000)	$64

Each recall represents one illness, with reporting rate of 10 percent
(baseline is 1 percent)	$8

Each recall represents one illness, with reporting rate of 2.5 percent
(baseline is 1 percent)	$20

Final rule reduces manufacturing recalls by 80 percent (baseline is 100
percent)	$35

Exclude 1998 recalls from estimate, so average annual number of
manufacturing recalls is 14 (baseline is 19.5) 	$27

Average annual number of manufacturing recalls = 2000-2005 average, so
average per year is 10 (baseline is 19.5)	$26



	In the sensitivity analysis of annual costs, we change assumptions
about the numbers covered by the rule, the number of batches produced
per establishment, the number of lots per batch, the average costs per
test, and the rate of verification amount of testing.  

	The number of establishments covered is uncertain because we based it
on voluntary survey responses and other evidence from the 1990s.  If the
number of establishments has increased or decreased, or if our original
data overstated or understated the correct number, then the estimated
costs will be either too low or too high.  We show the effects of
different numbers for one arbitrarily lower number covered and  one
arbitrarily higher number covered. 

The number of batches produced is our basic measure of output. Annual
costs therefore vary directly with this measure and its components.  We
show how the costs depend on the number of batches by estimating costs
for 50 percent less and 50 percent more batches than estimated from the
survey.

	The number of shipment lots and the cost per test determine identity
testing costs, We look at the sensitivity of changing assumptions about
testing because it is the single largest contributor to annual costs. 
We show how the costs vary if the average number of batches per lot is 1
or 12.  We vary the average cost per test from $20 to $100.   ; we
estimate that annual testing costs will be about $60 million, about 40
percent of total costs.  The second largest contributor to costs is
paperwork, including written procedures and records.  The annual cost of
written procedures and record keeping for this final rule is about $17
million.  The estimated costs of this final rule are not highly
sensitive to the estimated costs of record keeping.  Recordkeeping
costs, and other costs as well (including testing costs) are driven by
the number of batches produced, our index of output.  To show how
changes in the estimated number of batches changes total costs, we
include estimated costs for the 5th and 95th percentiles of batches
produced, according to the survey of the industry.

Table 34. Sensitivity of Costs

Description	Estimated Annual Cost (after 3 years)

(in millions)

Final rule	$164

Number of covered establishments is 1300 (baseline is (1460)	$148

Number of covered establishments is 1600 (baseline is 1460)	$178

Number of batches are 50 percent of baseline (baseline is 444, 2436, and
1164)	$104

Number of batches are 150 percent of baseline (baseline is 444, 2436,
and 1164)	$224

1 batch of dietary supplements per shipment lot of a dietary ingredient
(baseline is 6.5)	$322

12 batches of dietary supplements per shipment lot of a dietary
ingredient (baseline is 6.5)	$136

Average cost per test is $20 (baseline is $60)	$129

Average cost per test is $100 (baseline is $60)	$197



Description	Estimated Annual Cost (after 3 years)

(in millions)

Final rule	$150

1 batch of dietary supplements per shipment lot of an ingredient
(baseline is 6.5)	$340

12 batches of dietary supplements per shipment lot of an ingredient
(baseline is 6.5)	$116

Test 1 percent of batches or lots (baseline is square root of batches or
lots plus 1)	$145

Test every batch or lot (baseline is square root of batches or lots plus
1)	$424

Average cost per test is $20 (baseline is $60)	$109

Average cost per test is $100 (baseline is $60)	$189

5th percentile number of  batches (baseline is distribution of survey
results)	$100

95th percentile number of  batches (baseline is distribution of survey
results)	$216



	We combine the results of the sensitivity analyses to generate overall
ranges for benefits and costs. We estimate that, once it is fully
implemented, the annual benefits, able to be quantified, from the final
rule will be $8 million to $64 million; the annual costs will be $104
million to $322 million.  

C.  Initial Final Regulatory Flexibility Analysis

1.  Introduction 

     We have examined the economic implications of this final rule as
required by the Regulatory Flexibility Act (5 U.S.C. 601-612).  If a
rule has a significant economic impact on a substantial number of small
entities, the Regulatory Flexibility Act requires agencies to analyze
regulatory options that would lessen the economic effect of the rule on
small entities.  We find that this final rule will have a significant
economic impact on a substantial number of small entities.

2.  Economic Effects on Small Entities

     a. Number of small entities affected.  The final rule will affect
many small entities.  A small business in this industry is any
establishment with fewer than 500 employees.  For purposes of the
cost-benefit analysis, we also looked at the category we call very small
establishments: eEstablishments with fewer than 20 employees.  Based on
the survey (Ref. E32), we estimated that 774 establishments, 53 percent
of the total establishments, could be classified as very small (under 20
employees) and 526 as small (20 to 499 employees), which is 36 percent
of the total establishments.  Based on the results of the survey (Ref.
E32), wWe estimated the total number of warehouses, wholesalers, and
other holders likely to be covered by this regulation to be 15,689, of
which 15,421 are small businesses.

     The small establishments that will be affected by the final rule
are those establishments that will have to perform the various required
activities, and would not have done so without the rule.  We determined
estimated baseline (pre-CGMP requirements) manufacturing practices with
the survey of the industry (Ref. E3E2).  The survey asked representative
respondents to answer a series of questions, including how many
employees they had and what their existing practices were.  From the
survey, we determined that small establishments do not now follow all of
the provisions of the final rule.  Those that do not follow all the
provisions will incur a cost to do so.

     b. Costs to small entities.  Implementation costs vary across
establishments depending on current practices and the types of products
manufactured, packaged, labeled, or held.  We estimated the range of
current practices using the survey of the industry.  The cost model,
which we describe in detail in section XXIV.B.7, divided establishments
by size, which allowed us to estimate the distribution of costs per
establishment for each size and product class.  Table 35 shows the cost
per establishment for very small and small establishments.  For
comparison, we include the estimated average cost per large
establishment and the median revenues for each size category.  As the
table shows, costs per establishment are proportionally higher for very
small than for large establishments.  The table’s most striking result
is that annual costs are highest for small (20 to 499 employees)
establishments. 

Table 35. Costs per Establishment, by Size





	Set-up Costs per Establishment	Annual Costs per Establishment	 Median
Annual Revenue per Establishment

Very small establishments	$26,000	$4637,000	Under $ 1 million

Small establishments	$2019,000	$1844,000	$5 to $10 million 

Large establishments	$310,000	$690,000	$10 to $50 million

Warehouses, wholesalers, and other holders	$360250	$1,000950	Not
applicable



Table 36. Total Costs by Establishment Size

	Set-up Costs 	Percent of total set-up costs	Annual Costs 	Percent of
total annual costs

Very small establishments	$20 million	49 percent	$38 million	23 percent

Small establishments	$10 million	24 percent	$98 million	60 percent

Large establishments	$5 million	12 percent	$11 million	7 percent

Warehouses, wholesalers, and other holders	$6 million	15 percent	$17
million	10 percent



     Small establishments that do not perform a substantial number of
the actions required by the final rule will bear relatively high costs
for compliance with the provisions of this final rule.   Although the
final rule will raise product prices, the price increase (which would
largely be determined by changes made by large establishments) may be
much smaller than the increase in the average costs of very small
producers.  The average burden to very small establishments will be
about 4 percent of annual revenue.  The average burden to small
establishments will be 1.5 to 3 percent of annual revenue. 
Establishments with above average costs, and even establishments with
average costs, could be hard pressed to continue to operate.  Some of
these may decide it is too costly and either change product lines or go
out of business.

     We use a model developed under contract by Eastern Research Group
to estimate the effects of FDA regulations on small businesses (Ref.
E25E27).  The model is designed to assess the effects of a wide range of
potential regulatory activities, ranging from HACCP to product labeling.
 CGMP regulations are included as a potential regulatory activity.  The
model allows us to predict the probability and frequency of small
business failure as a result of our regulations. 

     We ran the model for the final rule.  The model predicts that, as a
result of the final rule, 140 very small, and 32 small dietary
supplement manufacturers will probably be at risk to go out of business.
The model estimates the number of workers in those firms to be about
2,2500.

	The regulatory costs of this final rule will also discourage new small
businesses from entering the industry.  The dietary supplement has been
characterized by substantial entry of small businesses.  Although we
cannot quantify how much that will change, we expect that the rate of
entry of very small and small businesses will decrease.  

3.  Regulatory Options

     a. Exemptions for small entities.  The burden on small
establishments would be reduced if they were exempt from some provisions
of the final rule.  Most entities – we estimate close to 90 percent
– affected by this final rule, however, are small.  Exempting small
establishments from some or all of its provisions would substantially
reduce benefits.

     b. Longer compliance periods.  Lengthening the compliance period
provides some regulatory relief for businesses with fewer than 500
employees.  The longer compliance period will allow additional time for
setting up recordkeeping, making capital improvements to the physical
plant, purchasing new or replacement equipment, and other one-time
expenditures.  It will also delay the impact of the annual costs of
compliance.  We have given businesses with fewer than 20 employees an
additional 24 months and businesses with fewer than 500 but 20 or more
than 20 employees an additional 12 months for compliance.  The final
rule, then, will be phased-in over 36 months, with firms with 500 or
more employees complying after 12 months, firms with under 500 but 20 or
more than 20 employees after 24 months, and firms with fewer than 20
employees after 36 months.  The cost savings of delay may well be larger
than simply the present value of the delay because the firms with fewer
than 500 employees may also be able to reduce their compliance costs by
taking advantage of increases in industry knowledge and experience in
implementing these regulations. 

c. Reduced requirements in the final rule.  The modification of
requirements in this final rule, compared with the 2003 CGMP Proposal,
significantly reduce the costs borne by small businesses.  We estimate
the average set-up costs under the proposal to be $25,000 for very small
establishments and $40,000 for small establishments.  We estimate that
the annual costs of the 2003 CGMP Proposal would be $90,000 for very
small establishments and $300,000 for small establishments.  The final
rule therefore reduces annual costs for very small establishments to
about half of the estimated costs of the proposed rule and reduces costs
for very small establishments to about 60 percent of the estimated costs
of the 2003 CGMP Proposal.  Under the 2003 CGMP Proposal, 216 very
small, and 50 small businesses would be at risk of going out of
business, over 50 percent more than under the final rule.

4.  Description of Recordkeeping and Reporting

     The Regulatory Flexibility Act requires a description of the
recordkeeping and reporting required for compliance with this final
rule.  This final rule will require the preparation of records.  As
described in the Preliminary Regulatory Impact Analysis, written records
or electronic documents must be kept that demonstrate that specific
actions occurred in the manufacturing process in compliance with the
final rule.  Records that will be required in this final rule will
demonstrate that corrective actions were taken; that equipment,
instruments, and controls used in laboratory operations and quality
control were installed and calibrated properly; that maintenance
programs were followed; and that the results of any testing show that
components or dietary supplements meet the established specifications.

     The compliance cost of recordkeeping is the sum of both the initial
design and printing of the recordkeeping documents and the recurring
costs of maintaining the records.  The cost of training personnel to use
the new documents is a recurring cost depending on how frequently
documents are modified, how often personnel turn over, and how
complicated the tasks are that are being recorded.  The recurring costs
are measured by the workers’ wage rate multiplied by the expected
labor hours necessary to for a written or electronic record and the time
necessary for management to review the records to see that actions are
documented accurately.  In addition, electronic records necessitate
recurring time spent ensuring that the equipment is serviced and
maintained properly.

5.  Summary

     The final rule will have a significant economic impact on a
substantial number of small entities.

	(Comment 35960)  We received many comments on the Initial Regulatory
Flexibility Analysis.  Nearly all of the comments addressing small
business state that the requirements of the 2003 CGMP Proposal, the
testing requirements in particular, would be an enormous burden on small
business.  Other comments assert that, because of this burden, the rule
is in violation of the Regulatory Flexibility Act.  In addition,
comments assert small business will be particularly burdened by the rule
and that consumers will see little improvement in product safety as a
result.

     Some small firms estimate annual testing costs for the 2003 CGMP
Proposal as high as $600,000, as opposed to the $60,000 per year
estimated by us.  Another firm estimates set-up costs in the range of 4
to 7 times our estimate and annual costs 8 to 30 times our estimate. 
Comments also express concern that we have underestimated the number of
businesses forced to close if this rule is made final as proposed - one
comment states that the rule would cause 50 percent of small businesses
to shut down.  Some comments assert that this rule is anti-competitive:
that is, the comments claim that this rule will make dietary supplement
manufacturing so expensive that only large companies will survive.  In
addition, a few comments note the loss of product choice, innovation,
and domestic employment that accompany firm closures, in addition to the
increase in prices of products made by remaining firms.  In addition,
another comment suggests that foreign manufacturers will be at an
advantage because they will not have to comply with the rule’s
requirements.

     Some comments reiterate the points made earlier that the use of
statistical sampling and supplier certificates of analysis could help
reduce the burden on small business.

     One comment states that it would be extremely costly for small
firms to come into compliance with the 2003 CGMP Proposal, especially
because, as several firms pointed out, small firms often produce batches
that are small in size.  A few comments, however, say that small firms
should be made to comply with the new rule at the same time as large
firms.

     We received many comments on the compliance period of this rule. 
Some of these comments favor the extended compliance periods granted to
small and very small firms.  Other comments do not support the
compliance periods, stating that they are not long enough for firms to
set up recordkeeping systems, make capital improvements, and so on.  

     Other comments do not favor granting small firms more time to
comply.  Three comments state that granting small firms a longer
compliance period defeats the purpose of the rule, by making it
difficult for consumers to determine which dietary supplements comply
with the CGMPs and which do not yet comply.  Another comment suggests
that products made by firms not in compliance one year after the
rule’s effective date be labeled to say, “This product may not
conform to government standards for purity and potency”.  Other
comments propose a single compliance period for all firms.  

	(Response)  We disagree with comments that the burden of this final
rule violates the Regulatory Flexibility Act.  The act requires agencies
to consider the burden of their regulatory proposals on small entities,
analyze and consider effective alternatives that reduce the burden on
small entities, and make their analyses available for public comment. 
We have considered the burden of this final rule on all covered firms,
including small businesses, and as a result have reduced modified
certain requirements to reduce the costs of the final rule compared with
the 2003 CGMP Proposal.certain testing requirements.  In addition, small
businesses are allowed more time to comply with the rule.  The burden on
small businesses remains large, but the Regulatory Flexibility Act does
not require agencies to adopt regulations that impose the least burden
on small entities.  In addition, the Data Quality Act has been fulfilled
by using the most objective data available.  In this analysis, we used
data from surveys and from other Federal agencies.  Although more data
are desirable, we consider the quality of the data used in this analysis
and in the references to be the best available and sufficient to fulfill
the requirements of the Regulatory Flexibility and Data Quality Acts.

     We have reduced the amount of testing required in this final rule
in response to comments on the burden of testing costs on the 2003 CGMP
Proposal.  As explained in Section XXIV.B.7, we underestimated costs in
the proposed rule because of an error in a contractor’s report.  We
have corrected the cost calculations, including estimated testing costs,
for this final regulatory flexibility analysis. 

We note that foreign firms that sell dietary supplements in the United
States are required to be in compliance with the final rule. 

     In response to comments on the number of firms likely to go out of
business, we have used our small business model to estimate that 172
small and very small firms will probably go be at risk of going out of
business.  Many other small firms – some of them already experiencing
financial difficulties – will may see their financial condition worsen
as a result of this final rule. 

     We disagree with the comments that oppose longer compliance periods
for small businesses.  The additional time will only slightly delay the
full implementation (and full benefits) of this final rule, and may
provide the margin of survival for some small businesses. 

D.  Unfunded Mandates 

     The Unfunded Mandates Reform Act of 1995 (Public Law 104-4)
requires cost-benefit and other analyses for rules that would cost more
than $100 million in a single year.  The current (2004)
inflation-adjusted statutory threshold is $1185 million.  This final
rule qualifies as a significant rule under the statute.  Most of the
requirements of the Unfunded Mandates are fulfilled in the Executive
Order 12866 analysis.  The requirements under the Unfunded Mandates
Reform Act of 1995 include assessing the rule’s effects on future
costs; productivity; particular regions, communities, or industrial
sectors; economic growth; full employment; job creation; and exports.

     The future costs from the rule are the recurring costs, which reach
their long-term value in the third year after this rulee 2003 CGMP
Proposal would become final.  These costs would be incurred, directly or
indirectly, by the establishments that manufacture, process, pack,
label, transport, distribute, receive, hold, or import dietary
supplements or ingredients or dietary products.  Recurring costs from
the regulatory requirements will be incurred in each future year.  Table
29 summarizes the annual future recurring costs. 

     The costs, direct and indirect, of the rule will be shared among
manufacturers, processors, packagers, transporters, receivers, holders,
and importers of dietary supplements or ingredientsproducts, as well as
domestic consumers.  Much of the higher costs incurred by domestic
suppliers of dietary supplement products as a result of these
regulations will be passed on to consumers as higher prices.  The higher
prices will be offset by the benefits from these regulations.  

     Although this final regulation is significant, we do not expect it
to substantially affect national productivity, growth, jobs, or full
employment.  The total costs will be small relative to the economy, and
will be offset by benefits.  The improved safety and quality of dietary
supplements leads to less illness and fewer sick days taken by
employees, and lower adjustment costs by firms that would otherwise need
to hire replacement employees.  Furthermore, Tthe dietary supplement
industry is too small a part of the domestic economy to influence
overall economic activity.  According to our model of small business
effects, about 2,500 jobs may be lost in the short run as a result of
displaced workers in firms leaving the industry. 

     This final rule will require additional controls throughout the
production and distribution chain for the manufacture of dietary
supplements.  The additional costs will increase the total costs of
production and distribution for all of the regulated products, including
products sold within the United  States and across national borders. 
These increased costs will be partly passed on to consumers in the form
of higher prices, which will tend to reduce the quantity demanded of the
regulated products.  The increased prices of U.S. exports could reduce
the quantity of U.S. exports demanded, particularly in comparison with
exports from countries that do not implement similar regulations.  We
expect this effect to be insignificant, because under the final rule the
increases in the price of U.S. exports (and resulting decreases in
quantity demanded) will be quite small.

 Options 1-6 were discussed in detail in the 2003 CGMP Proposal (68 FR
12221-12223; March 13, 2003) and analyses of costs were provided when
possible.  The principles of the options discussion have not changed and
are still relevant for purposes of the requirements of the final rule. 
The 2003 CGMP proposal also included an Analysis of Impacts which
contained some errors from a contractor’s report.  We have corrected
the analysis and have recalculated the costs of the 2003 CGMP Proposal. 
These corrections and recalculations are discussed in Section XXIV.B.9. 
  

	Functional Status Code is a measure of lost mobility (MOB), physical
activity (PAC) and social activity (SOC).  Lost MOB might mean an
inability to drive a car.  Lost PAC might mean walking with physical
limitations.  Lost SOC might mean self-care is not possible. 
Symptom-problem health utility index is a weighted measure of the cost
of each symptom.  For example, a sick or upset stomach leads to an
estimated loss of 0.290 per day.

 In the uncertainty analysis below, we used a probability distribution
to represent the uncertainty associated with the number of illnesses. 
We modeled the number of illnesses prevented for each class as the
average number of recalled products plus a negative binomial
distribution representing unknown cases.  The negative binomial
distribution estimates the number of failures (unknown cases) that will
occur before some number of successes (known cases) for a given
probability of success.  In the negative binomial distribution, we
assumed that the numbers of recalls represented reported cases and that
the probability of reporting equaled 1 percent (Ref. E913).  The mean
estimated number of illnesses is 100 times the reported number of
recalls.

	 We used a probability distribution to represent the uncertainty
associated with the number of illnesses.  We modeled the number of
illnesses prevented for each class as the average number of recalled
products plus a negative binomial distribution representing unknown
cases.  The negative binomial distribution estimates the number of
failures (unknown cases) that will occur before some number of successes
(known cases) for a given probability of success.  In the negative
binomial distribution, we assumed that the numbers of recalled products
were reported cases and that the probability of reporting equaled 1
percent (Ref. E9).  The result is that the mean estimated number of
illnesses is 100 times the reported number of recalls.

 We recognize, however, that the presence of L-trypotophan only
indicates a small probability of EMS. The estimates in Table 23 assume
that L-trypotophan represents a 0.1 percent probability of EMS.

FDA Draft Internal Confidential Deliberative

2-4-05

page   PAGE  30 

HHS/FDA-Internal-Confidential-Deliberative

Final RIA 9-28-05, 6-2628-06

page   PAGE  972 

FDA Draft Internal Confidential Deliberative

2-4-05

page   PAGE  52 

FDA Draft Internal Confidential Deliberative

2-4-05

page   PAGE  122 

