A 1981 study in the British Medical Journal examined the effects of husbands’ smoking habits on their wives.
The study, conducted in Japan, was a prospective cohort study; the study groups was tracked forward for 14 years, and the researcher started off knowing exposures and ended with outcomes (exposures included smoking habit and occupation; outcomes were risks for diseases including lung cancer.)
When the researcher examined the relationship between smoking habits among husbands and the risk of their wives dying from lung cancer, the researcher found that wives of heavy-smoking husbands were two-fold more at risk than their counterparts with non-smoking husbands. Interestingly, the researcher found that the risk was dose-dependant (wives’ risks increased when husbands smoked more.)
The researcher also age and occupation adjusted; this adjustment revealed that the relationship was “particularly significant in agricultural families” with husbands between 40 and 59 years of age. This finding was interesting because wives in agricultural families are more in contact with their smoking husbands than their urban counterparts. Are societal roles at play; do more agricultural wives follow tradition than urban wives? That is an interesting question not addressed in this study.
But how did the researcher deal with confounding variables and other possible outcomes? Since the study was a cohort study, there were many advantages that allowed for such questions to be answered. One advantage is that cohort studies allow common exposures to be examined. In addition to examining the relationship between husbands’ smoking and wives’ disease risks, the researcher also compared husbands’ drinking to wives’ disease risks. The researcher concluded that husbands’ drinking habits did not affect wives’ death from lung cancer, so certainly alcohol was not a confounding exposure.
Another advantage is that cohort studies can examine multiple outcomes and an exposure. For example, the researcher compared husbands’ smoking habits to risks for various diseases (lung cancer, asthma, cervical cancer, etc). The researcher found that only husbands’ smoking habits were strongly associated with wives’ death from lung cancer. So we wouldn’t expect something like cervical cancer to be affected by husbands’ smoking habits.
But what is the importance of this study? One reason why this study is important is because of an on-going debate about second-hand smoke and whether it is an environmental health issue. I think it is, and I think that this study supports that because the study clearly shows that the wives’ environments (husbands’ smoking) can affect their risks of dying from lung cancer.
An analogy would be working in an office building in which the floors below had smokers and the floors high up didn’t. (Let’s also assume that smokers don’t know that nonsmokers are above them, and vice-versa.) Since air vents circulate from bottom up, smoke is carried to the nonsmokers above, causing second-hand smoking in the nonsmokers. In this case, the environment of the office building is contaminated by smoke, and air quality and occupational health alarms are triggered. And in short, environmental health is called in to deal with the work-environment crisis.
But is it even a health issue? Association does not mean causation, but it sure can be a strong hint to causation. I think this study does support second-hand smoking being a cause of lung cancer; there is clearly a significant increase in risk of dying from lung cancer in women whose husbands are heavy-smokers.
One has to remember that this study was prompted by “sharply increasing [mortality rates] for men and women in Japan” even though “only a fraction of Japanese women with lung cancer smoke cigarettes.” So the question was, can second-hand smoking account for some of these deaths? Evidently, yes it can.
Eric Wan
Erik- great job in identifying this as a prospective cohort study! When we're trying to determine whether an identified association might have a causal relationship, we often use the Hill's criteria. You mentioned dose response relationship (which is one of Hill's criteria for causality). Other Hill's criteria this article addressed was temporal relationship (women were exposed to second hand smoke first, then developed cancer later), and strength of association (the data showed that women who lived with a smoker were significantly more likely than women who didn't live with a smoker to have lung cancer). It's important for researchers to keep Hill's Criteria in mind when they're trying to confirm a causal relationship. Let me know if you have any questions!
ReplyDelete