Rajiv Bahia, Peggy Lopipero and Allan H. Smith

Debra T. Silverman

In 1993, EPA labeled secondhand smoke a "known human carcinogen." This was based on its analysis of about 30 epidemiologic studies of secondhand smoke and lung cancer. But 80 percent of the studies did not support EPA's decision.

So how did EPA justify its conclusion?

EPA performed a "meta-analysis" of the studies. That is, the relative risks from the 30 studies were weighted, pooled and an "average" relative risk of 1.19 was calculated. And EPA concluded that secondhand smoke increased lung cancer risk 19 percent.

But this result was criticized because for a number of reasons, including:

• Epidemiologic studies are not generally capable of reliably identifying small relative risks (i.e., less than 2.0).
• None of the 30 studies used quantitative exposure data. All the studies used "guesstimated" exposure data.
• The relative risk of 1.19 was not statistically significant at the conventional 95 percent level.
• EPA underadjusted for the effect of smoking misclassification (i.e., the tendency for smokers to claim they are nonsmokers).

EPA (and the rest of the junk science world) chose to ignore these criticisms.

Now, consider a new study just published in the journal Epidemiology on diesel exhaust and lung cancer.

Researchers from the University of California (San Francisco) and the University of California (Berkeley) did a meta-analysis of 23 epidemiologic studies of diesel exhaust and lung cancer (Note: 7 other diesel exhaust/lung cancer studies were excluded from the meta-analysis, 6 of which did not support the researcher's ultimate conclusion).

The researchers reported a relative risk of 1.33 (95 percent confidence interval 1.24. - 1.44).

But in an accompanying editorial, the National Cancer Institute's Debra T. Silverman wrote:

[The authors] conclude that the data support a causal association between diesel exhaust and lung cancer in humans. Has science proven causality beyond any reasonable doubt? Probably not. The repeated finding of small effects, coupled with the absence of quantitative data on historical exposure, precludes a causal interpretation.

How would Mr. Rogers would put it? "Can you say 'double-standard'?"