12.07.2015 Views

Review of the Air Quality Criteria Document for Particulate Matter

Review of the Air Quality Criteria Document for Particulate Matter

Review of the Air Quality Criteria Document for Particulate Matter

SHOW MORE
SHOW LESS
  • No tags were found...

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

dominant, one criticism is that “best lags” were reported, yet this use <strong>of</strong> best lags isjustified later (6-238). Almost all studies explore “data driven” lag structures.6-131 It is bizarre that <strong>the</strong> mortality data are brought up in this section dealing withhospitalizations to “shore up” <strong>the</strong> argument <strong>for</strong> PM & cardiovascular effects.6-134 Why does this summary only include <strong>the</strong> US studies, which incidentally are all “positive”studies, when important international studies, many <strong>of</strong> which are “negative” studies, arenot included?To summarize, ra<strong>the</strong>r than attempting to shore up favored hypo<strong>the</strong>ses, and through doing so,revealing a bias, it would be preferable to do less editorializing during presentation <strong>of</strong> studies,and stand back <strong>for</strong> a more objective look at <strong>the</strong> studies as a whole. This is what we expect froma CD.3. ConfoundingConfounding remains an issue <strong>of</strong> concern. In <strong>the</strong> time-series studies, concerns regardingmeteorology, in <strong>the</strong> absence <strong>of</strong> more innovative approaches to specifying <strong>the</strong> <strong>for</strong>m <strong>of</strong>meteorology in <strong>the</strong> time series regression models, can probably be put to rest given <strong>the</strong> manyattempts to incorporate alternative specifications without significant impacts on <strong>the</strong> PMestimates <strong>of</strong> effect. The CD is probably correct in this regard.Confounding by co-pollutants, a perennial concern, has also been addressed in <strong>the</strong> CD.Several points should be noted. First, it is correctly noted that effects based on attempts tocontrol <strong>for</strong> confounding in two-pollutant or multi-pollutant models are <strong>of</strong>ten difficult to interpretbecause <strong>of</strong> <strong>the</strong> typically strong between-pollutant correlations that are present in <strong>the</strong> time-seriesstudies. However, this does not imply that effects from single-pollutant models <strong>of</strong> PM areunconfounded estimates. The findings regarding PM effects, as well as estimates <strong>of</strong> PM effect in<strong>the</strong> CD, are largely reported only from single-pollutant models (as one example, p.6-142, line17). Second, results from various alternatives to <strong>the</strong> use <strong>of</strong> multi-pollutant models inestimating PM effects unconfounded by co-pollutants are presented. These approaches aremotivated by frustration at interpreting PM effects from multi-pollutant models. In NMMAPS IIgaseous pollutant effects were controlled in a second stage (multiple city) analysis after <strong>the</strong>individual-city single-pollutant PM effects were estimated. This is probably justified in thissetting given <strong>the</strong> relatively large number <strong>of</strong> cities included, although it seems difficult to imaginethat adequate control <strong>for</strong> co-pollutants could be adequately accomplished without attending to<strong>the</strong> seasonal variation in co-pollutant concentrations, variation that itself differs from region toregion across <strong>the</strong> country. A different approach to addressing potential confounding by gaseouspollutants is exemplified by <strong>the</strong> multi-city hospitalization studies, including NMMAPS II(Schwartz 2000, Zanobetti 2000). Firstly, <strong>the</strong> description <strong>of</strong> <strong>the</strong>se methods is difficult to followin <strong>the</strong> CD narrative (6.223-225). Descriptions <strong>of</strong> <strong>the</strong>se alternative approaches to accounting <strong>for</strong>co-pollutant effects are difficult to follow. I still cannot figure out <strong>the</strong> rationale behind some <strong>of</strong><strong>the</strong>se approaches from reading this section, which may mean that o<strong>the</strong>rs cannot ei<strong>the</strong>r. Clearerrationale <strong>for</strong> <strong>the</strong> specific approaches taken is needed. Paren<strong>the</strong>tically, I wonder whe<strong>the</strong>r <strong>the</strong>correct correlation (r) between PM and <strong>the</strong> co-pollutants should be <strong>the</strong> correlation after adjusting<strong>for</strong> long-term trends and meteorology (that is, correlations between <strong>the</strong> effect estimates ra<strong>the</strong>rthan raw correlations). Secondly, we have much less confidence in <strong>the</strong> success <strong>of</strong> this approachgiven <strong>the</strong> much smaller number <strong>of</strong> cities ( and <strong>of</strong>ten smaller size <strong>of</strong> cities [e.g., Boulder,Youngstown) used <strong>for</strong> <strong>the</strong>se analyses. The CD seems to uncritically accept this approach tocontrolling confounding by <strong>the</strong> gaseous pollutants (6-126, line 4, <strong>for</strong> example).There has been discussion <strong>of</strong> <strong>the</strong> potential <strong>for</strong> <strong>the</strong> gaseous pollutants to confound <strong>the</strong>association between PM and health effects from <strong>the</strong> perspective <strong>of</strong> <strong>the</strong> definition <strong>of</strong> confounding.It is argued that some <strong>of</strong> <strong>the</strong> co-pollutants cannot be viewed as confounders since, based onbiomedical knowledge, <strong>the</strong>y should not affect <strong>the</strong> outcomes <strong>of</strong> interest. Nei<strong>the</strong>r SO 2 , sulfate norCO can reasonably be argued to cause many <strong>of</strong> <strong>the</strong> effects with which <strong>the</strong>y are <strong>of</strong>ten associated.It would be true that <strong>the</strong>se pollutants could not confound if in fact <strong>the</strong> ambient co-pollutantconcentrations were truly measuring exposure to <strong>the</strong>se specific pollutants. Realistically,however, <strong>the</strong>y do not. The co-pollutants are likely measuring various aspects <strong>of</strong> <strong>the</strong> pollutionmeteorologymix and acting as surrogate measures <strong>of</strong> important exposures that we do not nowA - 12

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!