Authors and BMJ editors are urged to publish dataset
Professor Hackshaw and colleagues claim, based on a systematic review of observational studies, that “smoking only about one cigarette per day carries a risk of developing coronary heart disease and stroke much greater than expected: around half that for people who smoke 20 per day.” Unfortunately, the study suffers from two major limitations. First, none of the included studies measured cardiovascular risk for one or two cigarettes per day (CPD), but rather from broader and higher exposure categories (e.g. 1-9, or up to 1-19 CPD). This crucial limitation was not fully explained in the paper and can only be detected by exploring the underlying observational studies. Second, confounding may bias the observed cardiovascular (CVD) risk estimates. In a classic randomised trial, the authors detected a major difference in CVD risk between placebo compliers and non-compliers (NEJM 1980; 303: 1038-41). Presumably, the difference could not be explained by the placebo pills, and confounding was the likely cause of the difference. After controlling for 40 (!) confounders, there was still a mortality difference of 15.0% versus 24.6% (p=0,00011). In Hackshaw and colleagues review, 15 studies did not adjust for any confounding beyond age and sex. Additionally, several commentators have issues about the statistical modelling that lies behind the risk estimate for one CPD. Surprisingly, Hackshaw has, despite two requests, still not published their dataset to allow other researchers to re-analyse the data. Surprisingly, BMJ’s editor claim that such publication is not in line with their editorial policies. I urge the authors and the BMJ to publish the dataset. Hopefully this may re-establish trust in the research group and the editorial process.
Competing interests:
No competing interests
21 April 2018
Ivar S Kristiansen
Professor
University of Oslo
Dept of Health Management and Health Economics, P.O. Box 1089, Blindern, NO-0317 Oslo, Norway
Rapid Response:
Authors and BMJ editors are urged to publish dataset
Professor Hackshaw and colleagues claim, based on a systematic review of observational studies, that “smoking only about one cigarette per day carries a risk of developing coronary heart disease and stroke much greater than expected: around half that for people who smoke 20 per day.” Unfortunately, the study suffers from two major limitations. First, none of the included studies measured cardiovascular risk for one or two cigarettes per day (CPD), but rather from broader and higher exposure categories (e.g. 1-9, or up to 1-19 CPD). This crucial limitation was not fully explained in the paper and can only be detected by exploring the underlying observational studies. Second, confounding may bias the observed cardiovascular (CVD) risk estimates. In a classic randomised trial, the authors detected a major difference in CVD risk between placebo compliers and non-compliers (NEJM 1980; 303: 1038-41). Presumably, the difference could not be explained by the placebo pills, and confounding was the likely cause of the difference. After controlling for 40 (!) confounders, there was still a mortality difference of 15.0% versus 24.6% (p=0,00011). In Hackshaw and colleagues review, 15 studies did not adjust for any confounding beyond age and sex. Additionally, several commentators have issues about the statistical modelling that lies behind the risk estimate for one CPD. Surprisingly, Hackshaw has, despite two requests, still not published their dataset to allow other researchers to re-analyse the data. Surprisingly, BMJ’s editor claim that such publication is not in line with their editorial policies. I urge the authors and the BMJ to publish the dataset. Hopefully this may re-establish trust in the research group and the editorial process.
Competing interests: No competing interests