Why Is There a Link Between Smoking and Suicide? In Reply

In Reply: We welcome the critique of our study by Dr. Davey Smith and Dr. Munafo and agree that our data do not demonstrate causation. We disagree with their conclusion that the smoking-suicide link is entirely accounted for by depression or other causes. The parallel association between homicide and smoking does not explain away the relationship between suicide and smoking. Homicide decedents and suicide decedents have similar neurobiological profiles (1), which justifies our using homicide decedents as a control group. Dr. Davey Smith and Dr. Munafo correctly observe that there was an age difference between the homicide and accident decedents—with the former group being older. To address this concern, we ran regression models limited to only suicide and accident decedents, whose ages were similar. The association of smoking with suicide persisted among males in both raw and fully adjusted analyses. [A table presenting these results is available in an online data supplement to this letter.] Of interest, significant associations were also found among females in unadjusted models—something that did not show up in the results of our analysis that included homicide decedents.

The omission of nonsmokers from our analysis was intentional. This group was not included because we wished to examine a dose-response relationship (in durations of smoking and of abstinence). Including nonsmokers in this case would bias the dose-response relationship (2) because of a different case-control ratio among lifetime nonsmokers [see Table 2 in the online supplement]. An interesting question is whether among people who die violently, lifetime smoking is associated with death from suicide (rather than from homicide or accident). We performed this analysis and entered lifetime smoking and depressive symptoms as independent variables. The analysis indicated that both variables significantly predicted suicide among males [see online Table 3]. Therefore, the possibility remains that smoking makes a residual contribution to suicide, as other recent studies have indicated (3,4). The absence of a causal association between smoking and depression in studies using Mendelian randomization (5) does not contradict our result. What is implied by our study is precisely that smoking has an association with suicide independent of depression. Collectively, our results do not prove causation, but ruling it out on the basis of parsimony seems, in our opinion, to be premature.