Tobacco Documents Online

Letters to the Editor Regarding "Environmental Tobacco Smoke and Lung Cancer Risk in Non-smoking Women," B.G. Stockwell, A.L. Goldman, G.H. Lyman, C.I. Noss, A.W. Armstrong, P.A. Pinkham, E.C. Candelora, and M.R. Brusa, Journal of the National Cancer Institute 84(18): 1417-1422, 1992. The Journal of the National Cancer Institute published four letters concerning the Stockwell, et al., paper, which reported on a Florida case-control study of household and spousal smoking and lung cancer. Stockwell, et al., reported statistically significant risk estimates for 40 or more "smoke-years" of household exposure during adulthood and for 22 or more "smoke-years" of exposure during childhood and adolescents. Letters by Peter N. Lee, Maxwell W. Layard, Paul Switzer, and Heather G. Stockwell and several of her co-authors appear in the Journal of the National Cancer Institute 85(9): 748-751, 1993. In his letter, Lee states that the Stockwell, et al., paper "adds little to the data on environmental tobacco smoke and lung cancer." He cites several potential sources of bias that could have affected the reported results. Lee comments on the method of control selection, the high proportion of surrogate respondents among cases, the interviewing process, and the potential for misclassification of smoking habits. Lee also notes that the possibility of dietary confounding was not considered, which he calls "remarkable," as Stockwell and colleagues have elsewhere reported a "protective effect" of vegetable and carotene consumption on lung cancer risk. Lee also criticizes the presentation of data in the Stockwell, et al., paper, noting that risk estimates were not given for exposure indices for which no association was claimed. He also presents a table of meta-analysis results of the spousal smoking, and lung cancer studies. Layard's letter includes a discussion of the Candelora, Stockwell, et al., paper on dietary factors that was referenced by Lee. Layard notes that "strong inverse associations" were reported for lung cancer and total vegetable consumption and total carotene intake. Layard notes that the diet analyses did not take into account ETS exposure. He suggests that Stockwell, et al., should explore the possibility of associations between diet and ETS exposure that could lead to confounding. After mentioning another potential confounder, a history of nonmalignant lung disease, Layard notes that the "weakness of the overall epidemiologic data" on spousal smoking and lung cancer makes adjustment for potential confounders important. Switzer references an editorial by David Burns which supported the Stockweil, et al., study, and then states: "[T]he evident inconsistencies pointed out in the Stockwell report should give one pause." In particular, Switzer notes the contrast between the adenocarcinoma data reported in the Stockwell, et al., study (no

association was reported) and the 1992 Fontham, et al., study (statistically significant risk estimates were reported). Switzer writes: "[H]unting expeditions through the data . . . can easily produce inconsistent artifacts." Switzer also comments on the large number of risk estimates presented by Stockwell, et al. He proposes that all the risk estimates reported to statistically significant may be related to only one statistically significant estimate, because the risk estimates are "overlapping." Switzer calls for investigators to describe their choices in reporting data, and to publish study protocols and reporting procedures in advance of data collection. Switzer also notes that the Stockwell, et al., paper did not include data on numbers of cases for the individual exposure categories, nor actual risk estimates for workplace and social exposures, calling the latter "a fine example of a publication bias." In their response, Stockwell, et al., indicate that they are analyzing data on dietary factors in persons reportedly exposed to ETS. They say that the question of a "protective effect" of diet should be considered separately from the question of ETS exposure. With regard to Layard's comment on prior lung disease as a confounder, Stockwell, et al., propose that "a shared common exposure to ETS" is "a more likely explanation" for prior lung disease in persons with lung cancer. Commenting on Lee's concern about surrogate respondents, Stockwell, et al., suggest that their risk estimates based on surrogates were lower than those based on "self reports"; they suggest that "an even stronger association" would have been reported had fewer surrogates been used. In conclusion, Stockwell, et al., stress that their study had "positive findings," and that "dismissal of all such findings" on ETS exposure is becoming "increasingly difficult." )