Division of Clinical Epidemiology
Hôpitaux Universitaires de Genève
1211 Genève 14
To the Editor:
The publication by Egan et al.1 of the findings of the Nurses’ Health Study on the relation of breast cancer to active and passive smoking is an important addition to the epidemiologic literature on this controversial topic. These are the first results from a large cohort study with incident cases, questionnaire assessment of both active and passive exposure to tobacco smoke, and adjustment for major potential confounders. 1 These results speak against a role of passive smoking in breast cancer.
Egan and colleagues 2 further speculate that the inconsistency between the Nurse’s Health Study and previous case-control studies can be explained by recall or selection bias affecting case-control studies. Even though this is true, in theory, I argue here that neither speculation nor data support the allegation that these biases are, as a rule, major issues in case-control studies of the association of smoking and breast cancer.
Consider recall bias first. It is reasonable to expect recall of past exposure to passive smoking to be influenced by occurrence of disease, but not necessarily in the direction of an overestimation of past exposure. When the results of the Geneva case-control study showing an association between passive smoking and breast cancer 3 became publicized, some women with breast cancer contacted us because they were concerned that their report would point to their husband or relatives being responsible for their disease. Thus, women in the study might have underreported their true exposure if they had suspected that we were investigating a link between passive smoking and breast cancer at the time of the study; however, at that time, participants did not know that we were studying passive smoking.
One way to quantify a possible propensity to search one’s memory for past exposures to passive smoke is to determine if cases are more preoccupied with passive smoke in their everyday life than controls. The Geneva study 3 included in the interview the following question, among others related to health behaviors. “What is your reaction to other people’s smoke?: (a) it leaves you indifferent; (b) it worries you”. Forty percent of the controls and 43% of the cases reported worrying about passive smoking. The percentage of women who worried about passive smoke was highest in those who were neither active nor passive smokers and the lowest in current smokers, but these differences in proportions across smoking categories were similar in cases and controls, arguing against any recall bias.
The other criticism of Egan et al. 2 refers to selection bias, that is, differential participation of cases and controls according to smoking status. This source of bias can be quantified by assessing the smoking status of women who refused to participate. In the Geneva study, 3 the proportions of ever-active-smokers in cases and controls, respectively, were, 48% and 40% in participants and 49% and 33% in refusals. Whatever the cause for the small selection differential, it would have resulted in a weaker association of smoking and breast cancer in participants than in non-participants.
There is one strength of several case-control studies on smoking and breast cancer (eg, Morabia et al. 3 and Johnson et al. 4) which Egan et al. 2 do not mention. These studies did elicit detailed lifetime histories of exposure to passive smoke. This is in contrast with the Nurses’ Health Study, 2 in which only three questions were asked on passive smoke. In the case-control studies, respondents were invited to remember and quantify their exposures at different epochs. Such data can generate more specific definitions of a passive smoker (eg, having been exposed for one hour daily or more for at least one year) and better minimize the role of casual exposures than would be feasible with a less detailed questionnaire.
It will be difficult to match the standards of Egan et al.2 in future cohort studies. Nevertheless, fairly weighing the relative strengths and weaknesses of the different study designs is crucial when it comes to reviewing all of the evidence. 5
1. Morabia A. Smoking (active and passive) and breast cancer. Epidemiologic evidence up to June 2001. Environ Mol Mutagens 2002; 38: 89–95.
2. Egan KM, Stampfer MJ, Hunter D, et al
. Active and passive smoking in breast cancer: prospective results from the Nurses’ Health Study. Epidemiology 2002; 13: 138–145.
3. Morabia A, Bernstein M, Heritier S, Khatchatrian N. Relation of breast cancer with passive and active exposure to tobacco smoke. Am J Epidemiol 1996; 143: 918–928.
4. Johnson KC, Hu J, Mao Y. Passive and active smoking and breast cancer risk in Canada, 1994–97. The Canadian Cancer Registries Epidemiology Research Group. Cancer Causes Control 2000; 11: 211–221.
5. Morabia A, Ambrosone CB, Baron J, Phillips DH, Russo IH. What do we currently know about the epidemiological and biological plausibility of the association of smoking and breast cancer? J Women’s Cancer 2001; 3: 5–8.