Secondhand smoke and statistical analysis
CMAJ 1999;160:180
In their letter about secondhand smoke and cancer,1 Dildar Ahmad and W. Keith Morgan correctly note that there is a lack of proof that secondhand smoke causes lung cancer. Indeed, there is growing awareness that many of the "facts" about environmental tobacco smoke have been exaggerated for what appear to be political purposes.
Publication bias is troublesome in meta-analyses based solely on the published scientific literature. The "publication threshold" for peer-reviewed journals appears to have fallen in recent years, especially for topics concerning public health and "risky" personal behaviour, because studies deemed to be "of great reader interest" are more likely to be reported in the mass media.
In addition, there is a selection bias favouring publication of positive results. Studies with no statistically significant association or a negative correlation are not published. Foreign-language publications, another wealth of material, are also frequently ignored. Responsible researchers should be urged to take the time and trouble to include these studies or to use "trim and fill" algorithms2 to account for their absence.
A larger problem is the troubling trend toward reporting "positive correlation" for relative risks of less than 2.0 — particularly when the lower bound is less than 1.0. In a press release3 accompanying publication of a study on breast cancer and abortion,4 the US National Cancer Institute noted that "In epidemiological research, relative risks of less than 2 are considered small and are usually difficult to interpret. Such increases may be due to chance, statistical bias or effects of confounding factors that are sometimes not evident." Thus, the relative risk of 1.16 (confidence interval 0.931.44) reported by the World Health Organization regarding environmental tobacco smoke and lung cancer is meaningless.
Even the best and most rigorous calculations of risk are but shaky estimates, providing only an upper bound for the effect of a variable. Although it is possible to account for some confounders, multiple factors are often simply not recognized. Unrecognized confounders are important in the issue of environmental tobacco smoke as well as smoking itself, given that smoking or being in the presence of environmental tobacco smoke is often just one in a cluster of risky behaviours.
Anne Fennell
Littlewood & Fennell, P.A.
Independent Public and Health Policy Research
Austin, Tex.
annef@io.com
Competing interests: None declared.
 |
Send a letter to the editor Envoyez une lettre à la rédaction |
References
- Ahmad D, Morgan WK. Secondhand smoke and cancer: Where's the proof? CMAJ 1998;159(5):441-2.
- Taylor S, Tweedie R. Trim and fill: a simple funnel plot based method of adjusting for publication bias in meta-analysis. Denver: Department of Statistics, Colorado State University; 1998.
- Abortion and possible risk for breast cancer: analysis and inconsistencies [press release]. Washington (DC): US National Cancer Institute; 1994 Oct 26.
- Daling JR, Malone KE, Voigt LF, White E, Weiss NS. Risk of breast cancer among young women: relationship to induced abortion. J Natl Cancer Inst 1994;88(21):1584-92.
