Comment on: The impact of smoking on prevalence of psoriasis and psoriatic arthritis

Dear Editor, Gazel and colleagues performed an ambitious meta-analysis showing (1) increased incidence and prevalence of psoriasis among smokers compared with non-smokers [pooled odds ratio (OR) 1.84], (2) reduced prevalence of PsA among smokers with psoriasis (pooled OR 0.70), (3) no significant effect of smoking on PsA prevalence among the general population, and (4) increased incidence of PsA among smokers (unanimously reported but not meta-analysed) [1]. This paper serves to highlight a very important issue that readers should be mindful of when interpreting meta-analyses: combining biased results can only give biased conclusions.

It is first worth giving some examples as to why this is important. Repeated observational studies showed that maternal smoking was protective against mortality among underweight infants, despite maternal smoking being detrimental for infants overall. A paradox. It was later shown that this paradox can be entirely explained away by collider bias [2]. Most would agree that, in the face of this methodological explanation and biological implausibility, it would be inappropriate to continue positing biological mechanisms.

Take another example: HRT was shown to be protective against coronary heart disease in repeated observational studies [3]. Randomized controlled trials later showed that HRT in fact increased coronary heart disease risk, particularly during early years of treatment. Briefly, this discrepancy was due to bias from depletion of susceptibles [4]. Again, most would agree that pooling biased observational results would be inappropriate and, in the case of HRT, dangerously misleading.

Coming back to the authors’ meta-analysis of PsA risk among smokers. What protective biological mechanism of smoking can increase incidence of psoriasis, increase incidence of PsA, yet reduce PsA prevalence? Smoking offers no protection here. The authors included the paper by Nguyen et al. in their review [5]. This study convincingly demonstrated that the paradox (smoking being protective of PsA) could be explained by methodological bias: risk of PsA was increased among smokers in the general population [hazard ratio (HR) 1.27], but the effect was reversed when restricting the analysis to people with psoriasis (HR 0.91). This would have been prime opportunity to highlight the importance of methodological literacy for interpreting observational studies. Instead, the authors proceeded to posit several possible biological explanations for the protective effect of smoking.

The reader need not be confused by the contradictory results of this meta-analysis. It should instead whet their appetite for greater causal and methodological literacy, which we hope will increasingly infiltrate the rheumatology literature. As a postscript, the conclusion and abstract reported increased prevalence of smoking in psoriasis patients compared with the general population. While this may be true by deduction, it is not what the results actually reported nor what the title implied, which is increased prevalence of psoriasis among smokers than among non-smokers in the general population. The same applies to the PsA results and figure/table legends.

Funding: No specific funding was received from any funding bodies in the public, commercial or not-for-profit sectors to carry out the work described in this manuscript.

Disclosure statement: The authors have declared no conflicts of interest.

References