_____________________

Ex Parte Meeting Summary:  Ozone NAAQS

Date:  December 20, 2007, 10:30 – 11:30 am 

Attendees: 

Visitors: 

Dr. Suresh Moolgavkar

Exponent Corp

Seattle, WA

(representing Utility Air Regulatory Group, UARG)

Dr. Peter Valberg

Gradient Corp.

Cambridge, MA

(representing UARG)

EPA attendees: 

George Gray, ORD

Pamela Williams, ORD

Bob Fegley, ORD

John Vandenberg, ORD

Ila Cote, ORD

Lydia Wegman, OAR

Karen Martin, OAR

Harvey Richmond, OAR

Dave McKee, OAR

Mary Ross, ORD

Debra Walsh, ORD

Jim Brown, ORD

Tom Bateson, ORD

Summary of discussion:  

Dr. Moolgavkar stated that the EPA doesn’t seem to appreciate the
difficulty of teasing out small signals in data, due to noise.  He
provided two examples where in the past conclusions had been developed
based on limited data which, with time and new studies, did not hold up
(i.e., hormone replacement therapy and antioxidants).  With respect to
air pollution, he stated the studies are very complex and the
epidemiological data was a weak basis for decision making due to
uncertainty. 

Dr. Gray asked Dr. Moolgavkar, specific to ozone, about the relevance of
human chamber studies to interpretation of epidemiological evidence. 
Dr. Mookgavkar stated that we can’t expect chamber studies to observe
mortality and illness, which were major endpoints in epidemiological
studies. Dr. Valberg then noted that chamber studies had shown effects
at 80 ppb, but not at lower levels.  Dr. Valberg stated the agency needs
to consider, for subsequent cost/benefit analyses, the difficulty in
estimating mortality from ozone exposures due to societal covariates
that affect epidemiological observations e.g., day of week. 

Dr. Moolgavkar indicated that a paper published by James et al (2007) on
PM was relevant to ozone in that it provided a new way of looking at air
quality data.  He stated that in this paper no association with
mortality was found in long-term studies (little detail was provided). 

Dr. Valberg provided comments on the ozone FR proposal.  He stated that
presenting the data using alternative scaling factors, as was done for
PM, for 1, 8 and 24 hours would illustrate the day-to-day and seasonal
variability in ozone monitoring data. He concluded that associations of
health outcomes with central monitoring data are uncertain and the
potential exists for misclassification and confounding of data
interpretation.

Dr. Gray asked Dr. Moolgavkar to discuss model choice issues.  Dr.
Moolgavkar provided some background regarding model choice, noting the
issue arose from investigation of GAM issues by researchers at Johns
Hopkins University.  He stated that model sensitivity analyses were
important, that temporal trends were a source of uncertainty, and that
how variables enter the analysis could alter results.   His perspective
is that we should ask “how can we explain time series of mortality and
hospital admissions?”, for example, rather than “does ozone produce
effects on mortality and hospital admissions?”. Various methods to
evaluate goodness of fit were available, and he noted that Dr. Clyde had
found that pollutants need not be included in some models, and that
choice of model fit test can make a large difference in results
interpretation. 

Dr. Gray asked Drs. Moolgavkar and Valberg about consistency in modeling
approaches among air quality researchers.  Dr. Moolgavkar noted that
various lag models are often evaluated, including various model
adjustments for weather, but only the most significant are reported. 
AIC and BIC approaches also are often used.   Dr. Moolgavkar noted that
the Johns Hopkins researchers were consistent in their approach to model
evaluation, but most other researchers were variable. 

Dr. Valberg stated that the variation in patterns of death by day of
week and month were a source of uncertainty in interpretation of
results.  Dr. Mary Ross (EPA) asked about the linearity of dose-response
relationship in that plotting response versus ozone concentration
appeared linear. Dr. Valberg agreed that with increased concentration
the gradation of responsiveness among individuals in the population
would appear linear as more individuals would be affected as ozone
concentration increased.  

Dr. Gray asked about the interpretation of data given air conditioner
use. Dr. Moolgavkar stated some studies have considered variation in air
conditioner usage, though he couldn’t recall at the moment which ones.
 Dr. Moolgavkar noted is it likely some information has been missed due
to publication bias, in that relevant information may be buried in
publications and not evident as to its relevance to the ozone review. 
He also pointed to model choice bias and that some changes had been made
around 2003 in the air pollution literature.  He concluded that EPA
needed to recognize the limitations in approaches used to evaluate ozone
effects.  

Dr. Gray thanked the visitors for their time, noted that the
Administrator had not decided about the NAAQS, and that their comments
had been heard and the comments played and important role in the
decision process.  Dr. Moolgavkar noted he was [pleasantly] surprised by
the evident attention paid by EPA staff and management to the docket
comments. 

