Re: "Elevated Lung Cancer in Younger Adults and Low Concentrations of Arsenic in Water"

We read with interest the study by Steinmaus et al. (1), in which they found an association between lung cancer and arsenic concentrations less than 100 μg/L in drinking water. The authors used a matched case-control design to study drinking water in 2 regions in Chile. However, we are concerned about the methodology and the conclusions drawn. Our first concern is with selection bias; cases were ascertained from all pathologists, hospitals, and radiologists in the area, but it was unclear whether the cases constituted all lung cancer cases. Our second concern is with the study base and the possibility that cases and controls were chosen from dissimilar sampling frames. Cases had been diagnosed with lung cancer, but controls were free from lung, bladder, and kidney cancers. Moreover, proxy interviews were conducted with 54% of cases and only 7% of controls, which increased the likelihood of information bias due to differences in data quality (2, 3). Our third concern is with the statistical analysis itself. Because the only statistically significant association between arsenic and lung cancer was for the 40-year lag time, a better description of the analysis is required to understand that all persons who contributed to this analysis were at least 40 years of age. It would also be helpful to present a clear statement of the study design (the study was not a matched case-control study) that would explain the choice to use unconditional logistic regression modeling (4). Lastly, we are concerned that the title of the article has the potential to raise the alarm of a new health risk from local drinking water. Inasmuch as the methodological issues prevent drawing conclusive results about the association under study (5), it is prudent to report findings cautiously to minimize the arousal of public fear. ACKNOWLEDGMENTS