Responses

Extract

We were bemused by Dr. Cohen'ss letter regarding our meta-analysis of eight case-control studies of indoor radon and lung cancer, since we did not criticize his ecologic regression analyses (1) . In the penultimate paragraph of our discussion, we made the simple observation that his negative regression relationship between age-adjusted lung cancer rates and mean indoor radon levels for counties (2) was not consistent with, and indeed contradicted, results from nearly all analytic case-control and cohort studies. Specifically, the relative risks (RRs) from our meta-analysis of eight indoor radon studies, from projections based on a pooled analysis of 11 cohort studies of radon-exposed underground miners, and from miner data restricted to exposures comparable to residential exposures were consistent with a positive exposure-response trend.

Limitations of ecologic studies have been widely discussed in general (3) and for radon in particular (4) . In a telling example, data from a national population survey were used to demonstrate that a regression slope estimated from area means could be biased upward, downward, or even reverse trend, compared with the regression slope estimated from individuals (3) . Ecologic studies of lung cancer and indoor radon and confounding by cigarette smoking were the topics of a recent exchange in the American Journal of Epidemiology(5–8) , which included a commentary from Cohen (7) . There is no need to revisit these issues. The RR relationship between radon exposure and lung cancer risk is complex, depending on age, time since exposure, and exposure rate, and smoking is an overwhelming risk factor. The 5%-6% of homes in the United States with radon concentrations over 150 Bq ³ are estimated to increase lung cancer risk by 20%-30%. In comparison, the 40%-50% of the population who are current or former smokers incur an estimated 1500% excess lung cancer risk. The complexity of the relationship between lung cancer risk and indoor radon exposure and smoking leads us to concur with the conclusion of Piantadosi (>6) :