Re: Change in mental health after smoking cessation: systematic review and meta-analysis
We read with interest the paper by Taylor et al. (BMJ, 12 Feb 2014) in which they showed a clear association between smoking cessation and improved mental health, with quitters reporting more favourable mental health outcomes than continuing smokers. We agree that physicians and others may perhaps find this helpful in reassuring patients that quitting smoking need not necessarily have a lasting adverse effect on their mental health, and that if anything the opposite may be true. Nevertheless, while the authors were careful in their use of causal language in the BMJ article, readers of the subsequent press release and media coverage could be forgiven for believing that the study provided evidence of a causal effect of smoking cessation leading to improved mental health, which it did not.
Taylor et al. state that their hypothesis was that smokers who gave up would experience a resulting improvement in mental health because they would no longer experience multiple episodes of negative affect induced by withdrawal. However, their study design clearly could not support or refute the second half of this hypothesis. The authors reasonably present three possible explanations of their findings: 1) that smoking cessation causes the improvement in mental health; 2) that improving mental health causes cessation; or, 3) that a common factor explains both improved mental health and cessation. We see little in their findings to support the authors’ apparent endorsement of the first over the second and third of these explanations in subsequent media coverage.
Causal inference from observational data is problematic, due to well known problems of confounding and reverse causality. The use of within individual changes in mental health measures may be less subject to confounding than between group comparisons but, given the clustering of health behaviours (1), it is not implausible that there will be changes in other lifestyle factors in addition to smoking cessation which may also influence mental health. In addition, while these data may not suggest that mental health affects likelihood of making quit attempts, reverse causality could still be an explanation if change in mental health influences ability to successfully sustain abstinence following a quit attempt.
For similar reasons, we feel that comparisons of the magnitude of the association between smoking cessation and mental health measures obtained from this systematic review, which is largely based on observational data, with effect sizes obtained from randomised controlled trials of anti-depressants, should be treated with caution. Observational studies are generally more prone to bias, and hence effect size estimates may well be inflated (2, 3). Thus, even if the observed association is causal, any true effect of smoking cessation will not necessarily be of the same magnitude. While we recognize that the term ‘effect size’ is a statistical one, we suggest that care is required when comparing effect sizes obtained from observational data to those from experimental data.
We agree with the authors’ suggestion that additional methodologies will shed light on whether smoking cessation does causally affect mental health. Such methodologies may include the use of instrumental variable analyses including Mendelian randomisation (4).
1. Heroux M, Janssen I, Lee DC, Sui X, Hebert JR, Blair SN. Clustering of unhealthy behaviors in the aerobics center longitudinal study. Prev Sci 2012;13(2):183-95.
2. Deeks JJ, Dinnes J, D'Amico R, Sowden AJ, Sakarovitch C, Song F, et al. Evaluating non-randomised intervention studies. Health Technol Assess 2003;7(27):iii-x, 1-173.
3. MacMahon S, Collins R. Reliable assessment of the effects of treatment on mortality and major morbidity, II: observational studies. Lancet 2001;357(9254):455-62.
4. Smith GD, Ebrahim S. 'Mendelian randomization': can genetic epidemiology contribute to understanding environmental determinants of disease? Int J Epidemiol 2003;32(1):1-22.
Competing interests: No competing interests