American Journal of Epidemiology Vol. 153, No. 3 : 307-308
Copyright © 2001 by The Johns Hopkins University School of Hygiene and Public Health
LETTERS TO THE EDITOR |
RE: "CIGARETTE SMOKING AND SUICIDE: A PROSPECTIVE STUDY OF 300,000 MALE ACTIVE-DUTY ARMY SOLDIERS"
Canynge Hall Whiteladies Road Bristol BS8 2PR, United Kingdom
Department of Primary Care and Population Science Royal Free and University College Medical School Rowland Hill Street London NW3 2PF, United Kingdom
 |
INTRODUCTION |
|---|
 TOP INTRODUCTION REFERENCESINTRODUCTION REFERENCES |
|---|
Miller et al. (1
) provide yet another replication of the association between cigarette smoking and suicide in the United States. The paper brings to mind Kuller's recent strictures against "circular epidemiology" (2). What is the point of this paper? Many studies have shown the same thing in US populations, and these include studies with better follow-up, better data on potential confounders, and greater statistical power. The authors state, "... several [studies] included too few suicides to effectively control for characteristics common to both smokers and suicides" (1, p. 1060), and cite a paper of ours (intended to be a methodological commentary, rather than of substantive interest) in this regard (3). Their study contains 113 suicides; ours, 601 suicides. We demonstrated a strong association between smoking and suicide, very similar to that seen by Miller et al., which was robust to adjustment for several confounders (socioeconomic position, ethnicity, age, previous illness) and which was therefore a strong candidate for the kind of "independent" risk factor identified in observational epidemiology. It is clear that people who are depressed, poor, alcoholic, and chronically ill, who have stressful modes of life, and who are exposed to a wide range of other risks due to area of residence, occupation, and health-related behaviors persist in smoking more than other people do. The higher suicide rate of cigarette smokers is therefore expected, but uninteresting, and the apparent "independence" of the effect probably reflects the limited ability to capture the wide range of common causal and confounding factors with the simple data available (4).
Our paper was entitled, "Smoking as 'independent' risk factor for suicide: illustration of an artifact from observational epidemiology?" (3). However, it has been widely taken to illustrate that smoking does indeed cause suicide. Biologic mechanisms through which this could act (e.g., smoking suppressing brain serotonin levels) have been advanced, and a meta-analysis has been performed showing that most studies demonstrate the same thing. In our study (3), however, smoking was strongly related to the risk of being murdered, with an "independent" relative risk of over 2 for smokers of two packs or more per day compared with nonsmokers. Unless health promoters have moved on to a direct action phase, during which they shoot smokers, this association is very unlikely to be causal.
We live in a world in which associations are more common than lack of associations, and the former are only worth drawing attention to if they increase our understanding of why the world is the way it is. The "independent" association between smoking and suicide is about as interesting as the equally strong "independent" association in which never wearing a seatbelt apparently increases the risk of dying of all-cause or respiratory disease mortality (5). These findings merely show for the thousandth time that smokers are different from nonsmokers and that the people who wear seatbelts differ from people who don't wear seatbelts. Indeed, lack of associations tell us at least as much as do associations. For example, the lack of association between socio-economic position and several causes of ill health demonstrates that strongly socially patterned exposures, which represent a large majority of the exposures that epidemiologists study, are unlikely to be importantly involved in the etiology of these particular conditions (6).
Similar studies carried out in different social, cultural, geographic, and historical contexts tell us much more than mere replications within the same study population. For example, meta-analyses have demonstrated that smoking appears to be strongly related to cervical cancer risk, yet the few studies that fail to show this effect are the ones in which smoking among women is associated in a different way with the risk of acquisition of sexually transmitted viral infections than in the majority of studies carried out in North America or Western Europe (7). In these studies, smoking will not merely serve as a proxy for an unmeasured underlying causal (probably viral) factor (8).
To prevent further ventures down unrewarding alleyways, perhaps an additional section—which could be titled, "Why did we examine this association and consider it worth reporting?"—should be added alongside the Introduction, Methods, Results, and Discussion in reports of observational epidemiologic studies.
|
REFERENCES |
|---|
|
TOPINTRODUCTIONREFERENCESINTRODUCTION REFERENCES |
|---|
-
Miller M, Hemenway D, Bell NS, et al. Cigarette smoking and suicide: a prospective study of 300,000 male active-duty Army soldiers. Am J Epidemiol 2000;151:1060–3.
[Abstract/Free Full Text] -
Kuller LH. Circular epidemiology. Am J Epidemiol 1999;150:897–903.
[Abstract/Free Full Text] - Davey Smith G, Phillips AN, Neaton JD. Smoking as "independent" risk factor for suicide: illustration of an artifact from observational epidemiology? Lancet 1992;340:709–12.[Web of Science][Medline]
- Phillips A, Davey Smith G. How independent are independent effects? Relative risk estimation when correlated exposures are measured imprecisely. J Clin Epidemiol 1991;44:1223–31.[Web of Science][Medline]
- Rogers RG, Hummer RA, Nam CB. Living and dying in the USA. New York, NY: Academic Press, 1999.
- Davey Smith G, Gunnell D, Ben-Shlomo Y. Life-course approaches to socioeconomic differentials in cause-specific adult mortality. In: Leon D, Walt G, eds. Poverty, inequality, and health. Oxford, UK: Oxford University Press, 2000.
-
Davey Smith G, Phillips AN. Declaring independence: why we should be cautious. J Epidemiol Community Health 1990;44:257–8.
[Free Full Text] -
Phillips AN, Davey Smith G. Cigarette smoking as a potential cause of cervical cancer: has confounding been controlled? Int J Epidemiol 1994;23:42–9.
[Abstract/Free Full Text]
THE AUTHORS REPLY
Department of Health Policy and Management Harvard University School of Public Health Boston, MA 02115
Department of Social and Behavioral Sciences Boston University School of Public Health Boston, MA 02118
US Army Research Institute of Environmental Medicine Natick, MA 01760-5007
|
INTRODUCTION |
|---|
|
TOPINTRODUCTIONREFERENCESINTRODUCTION REFERENCES |
|---|
Drs. Davey Smith and Phillips (1
) assert that they expected the dose-response relation between smoking and suicide that we found (2) but claim this association is uninteresting. They then ask the reader to accept their reasoning by narrow analogy: Since they previously demonstrated a relation between smoking and homicide, we should dismiss the possibility of a causal relation between smoking and suicide (what about between smoking and lung cancer?). They state this despite being aware of some of the posited mechanisms by which smoking may plausibly lead to depression, such as suppression of serotonin levels. Cigarettes contain myriad chemicals that may affect the brain. Why are Davey Smith and Phillips so sure that none may be a causative factor for depression and for suicide?
A recent study of several thousand teenagers suggests that the conclusion of Davey Smith and Phillips may have been premature. In this study, teenagers were followed longitudinally to assess the nature and direction of the relation between cigarette smoking and depression (3). Among nondepressed teens, cigarette smoking was the strongest predictor of developing depressive symptoms, but among teens who did not smoke at baseline, baseline depressive symptoms were not predictive of heavy smoking.
The arguments of Davey Smith and Phillips reiterate those they and their colleague, J. D. Neaton, made in their 1992 paper, where they expressed general disdain for observational epidemiology (4). They assert that associations are of value only if they increase our understanding of why the world is the way it is and imply that the purpose of science is only to determine causality. We disagree. The fact that the cock's crowing doesn't cause the sun to rise did not make the cock any less useful as an alarm clock throughout the ages. Similarly, smoking, regardless of whether or not there is a causal link with suicide, may prove an easily observable and useful marker of persons at risk.
We are not apologists for a causal relation between smoking and suicide nor do we claim that our paper supports this particular explanation for the relation we describe. It may well be that third variable explanations account for some of the relation we observe, that some soldiers who were depressed self-medicated with tobacco, and that others were depressed as a consequence of smoking. These explanations are not mutually exclusive. What troubles us is not that uncertainty remains about the likely explanation(s) for the relation we describe. Rather, we are troubled by the implication of Davey Smith and Phillips that they are properly the arbiters of what avenues investigators ought to pursue.
|
REFERENCES |
|---|
|
TOPINTRODUCTIONREFERENCESINTRODUCTION REFERENCES |
|---|
-
Davey Smith G, Phillips AN. Re: "Cigarette smoking and suicide: a prospective study of 300,000 male active-duty Army soldiers." (Letter). Am J Epidemiol 2001;153:307.
[Free Full Text] - Miller M, Hemenway D, Bell NS, et al. Cigarette smoking and suicide: a prospective study of 300,000 male active-duty Army soldiers. Am J Epidemiol 2000;151:1060–3.
-
Goodman E, Capitman J. Depressive symptoms and cigarette smoking among teens. Pediatrics 2000;106:748–55.
[Abstract/Free Full Text] - Davey Smith G, Phillips AN, Neaton JD. Smoking as "independent" risk factor for suicide: illustration of an artifact from observational epidemiology? Lancet 1992;340:709–12.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  J. Macleod Cannabis use and psychosis: the origins and implications of an association Adv. Psychiatr. Treat., November 1, 2007; 13(6): 400 - 411. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 N. Breslau, L. R. Schultz, E. O. Johnson, E. L. Peterson, and G. C. Davis Smoking and the Risk of Suicidal Behavior: A Prospective Study of a Community Sample Arch Gen Psychiatry, March 1, 2005; 62(3): 328 - 334. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. Hemmingsson and D. Kriebel Smoking at age 18-20 and suicide during 26 years of follow-up--how can the association be explained? Int. J. Epidemiol., December 1, 2003; 32(6): 1000 - 1004. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 G. Davey Smith, A. N. Phillips, M. Miller, D. Hemenway, N. Bell, M. Yore, and P. Amoroso RE: "CIGARETTE SMOKING AND SUICIDE: A PROSPECTIVE STUDY OF 300,000 MALE ACTIVE-DUTY ARMY SOLDIERS" Am. J. Epidemiol., February 1, 2001; 153(3): 307 - 308. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||