Re: Low cigarette consumption and risk of coronary heart disease and stroke: meta-analysis of 141 cohort studies in 55 study reports
Despite Professor Hackshaw’s clear restatement of his regression model (15 March), Professor Nord still feels that the regression line should extend down to 0 cpd, corresponding to the non-smokers, for whom the relative risk is 1.
But this is wrong – the regression line stops at 1 cpd because it is restricted to current smokers. The risks for each smoking category are expressed relative to non-smokers in the same study, and the meta-regression line is fitted through the resulting relative risks for current smokers. To be entirely clear, non-smokers are excluded from the regression.
Nord, like previous studies, assumes that non-smokers can be plotted on the dose-response curve for smokers. The novelty of Hackshaw’s analysis is that it explicitly tests this assumption by separating the smokers and non-smokers, and shows the assumption to be false.
Competing interests:
Already stated.
11 April 2018
Tim J Cole
BMJ statistics editor, professor of medical statistics
Rapid Response:
Re: Low cigarette consumption and risk of coronary heart disease and stroke: meta-analysis of 141 cohort studies in 55 study reports
Despite Professor Hackshaw’s clear restatement of his regression model (15 March), Professor Nord still feels that the regression line should extend down to 0 cpd, corresponding to the non-smokers, for whom the relative risk is 1.
But this is wrong – the regression line stops at 1 cpd because it is restricted to current smokers. The risks for each smoking category are expressed relative to non-smokers in the same study, and the meta-regression line is fitted through the resulting relative risks for current smokers. To be entirely clear, non-smokers are excluded from the regression.
Nord, like previous studies, assumes that non-smokers can be plotted on the dose-response curve for smokers. The novelty of Hackshaw’s analysis is that it explicitly tests this assumption by separating the smokers and non-smokers, and shows the assumption to be false.
Competing interests: Already stated.