BMJ  2003;327:501 (30 August), doi:10.1136/bmj.327.7413.501-a

Letter

Passive smoking

Wider evidence needs to be interpreted

Editor—Enstrom and Kabat's analysis has several omissions.¹ First they accept that most epidemiological studies have found positive but not statistically significant relationships between environmental tobacco smoke, coronary heart disease, and lung cancer, but then argue against meta-analysis to establish a causal relation. This is precisely where systematic reviews, and sometimes meta-analysis, show considerable benefit by increasing power. Enstrom and Kabat say that publication bias may explain positive results in reviews; however, larger cohort studies, unlike small trials and reports, are more likely to be published, regardless of results.² They do not explain heterogeneity between their findings and others, simply arguing that their cohort is large, and has more strengths. In fact, large prospective cohort studies like this may have greater losses to follow up, or more misclassification, over time.³

Misclassification, mentioned by the authors, may explain the apparent lack of association. Furthermore, the relative risks reported for active smoking and coronary heart disease (relative risk 1.5, table 10 in the paper) are lower than other cohort studies, which may be sufficient to obscure a modest but important increase in risk.^{4 5} They further assume an (unlikely) linear relation between cigarette smoking and mortality to validate their main results (extrapolating a very low estimate of a relative risk of 1.03 for coronary heart disease, by implying that environmental tobacco smoke is equivalent to smoking one cigarette per day). This analysis is unclear and unconvincing.

One study is insufficient to overturn established relations between environmental tobacco smoke and mortality, and I think that the authors overemphasise their negative findings.

International Health Research Group, Liverpool School of Tropical Medicine, Liverpool L3 5QA juliac{at}liverpool.ac.uk

---

Competing interests: None declared.

References

1. Enstrom JE, Kabat, GC. Environmental tobacco smoke and tobacco related mortality in a prospective study of Californians, 1960-98. BMJ 2003;326: 1057. (17 May.)[Abstract/Free Full Text]
2. Sutton AJ, Duval SJ, Tweedie RL, Abrams KR, Jones DR. Empirical assessment of effect of publication bias on meta-analysis. BMJ 2000;320: 1574-7.[Abstract/Free Full Text]
3. Critchley JA, Unal B. The health effects associated with smokeless tobacco use: a systematic review. Thorax 2003;58: 435-43.[Abstract/Free Full Text]
4. National Center for Chronic Disease Prevention and Health Promotion. Tobacco Information and Prevention Source. Report of the Surgeon General. Women and smoking, 2001. www.cdc.gov/tobacco/sgrpage.htm#2000s (accessed 14 Aug 2003).
5. Jousilahti P, Vartiainen E, Korhonen HJ, Puska P, Tuomilehto J. Is the effect of smoking on the risk for coronary heart disease even stronger than was previously thought? J Cardiovasc Risk 1999;6: 293-8.[ISI][Medline]

Related Articles

Environmental tobacco smoke and tobacco related mortality in a prospective study of Californians, 1960-98

James E Enstrom and Geoffrey C Kabat
BMJ 2003 326: 1057. [Abstract] [Full Text] [PDF]

Empirical assessment of effect of publication bias on meta-analyses

A J Sutton, S J Duval, R L Tweedie, K R Abrams, and D R Jones
BMJ 2000 320: 1574-1577. [Abstract] [Full Text] [PDF]

This article has been cited by other articles:

• Wen, W., Shu, X. O., Gao, Y.-T., Yang, G., Li, Q., Li, H., Zheng, W. (2006). Environmental tobacco smoke and mortality in Chinese women who have never smoked: prospective cohort study. BMJ 333: 376- [Abstract] [Full text]