Secondary Logo

Journal Logo

Re: Testing Differential Associations Between Smoking and Chronic Disease Across Socioeconomic Groups

Diderichsen, Finn; Andersen, Ingelise; Lange, Theis; Rod, Naja Hulvej

doi: 10.1097/EDE.0000000000001100
Letters
Free

Department of Public Health, University of Copenhagen, Copenhagen, Denmark, fidi@sund.ku.dk

Back to Top | Article Outline

To the Editor:

De Mestral et al.1 report that they in a pooled dataset of British cohorts find no differential susceptibility across socioeconomic status (SES) groups between smoking and chronic disease outcomes.1 They suggest that the studies, which have recently come to a different conclusion, may be flawed due to biased information of self-reported smoking.

De Mestral et al.1 estimate hazard ratios, that is, relative measures of association. It is for many chronic disease outcomes not surprising that the relative effect of smoking is equal across levels of SES. Let us take cardiovascular outcome as an example. The SCORE chart used in clinical prevention2 indicate 10 year risk of fatal cardiovascular disease and shows that the relative effect of smoking is constant across levels of the other risk factors—hypertension, cholesterol, and diabetes mellitus. This means, however, that absolute measures of the effect of smoking differ across levels of the other interacting causes. It is biologically expected that causes acting on different stages in a common pathway may interact in observational studies. Abnormalities in plasma lipoproteins generate atherosclerosis and smoking then influences the propensity to form thrombi. When cardiovascular risk factors increasingly tend to cluster among people with low SES, it will show up as a departure from additivity between smoking and SES. That is what have been found in studies using additive models for CVD and other tobacco-related diseases.

Self-reported measures of smoking are indeed an important source of bias. In particular, in the case when misclassification of one exposure is differential across levels of the other interacting exposure, it might lead to biased results.3 If smoking is for example more underreported in the low-SES group it will, in the terminology of VanderWeele, 4 imply that the reference interaction will be overestimated, but the mediated interaction will be underestimated. The sum of the two will determine in what direction the estimate of differential susceptibility is biased.5 Because an open-ended category of heavy smokers may include more people who smoke, for example, 50 cigarettes per day in the low SES group the effect estimate of smoking categories will be higher in low-SES groups. And that will lead to an overestimation of both reference interaction and mediated interaction.

If differential susceptibility can be verified it is important, because it has both clinical and policy implications.5 It would have been interesting if de Mestral et al.1 had included an analysis using additive hazard models thereby increasing the policy relevance and comparability to previous study.

Finn Diderichsen

IngeliseAndersen

TheisLange

Naja Hulvej Rod

Department of Public Health, University of Copenhagen, Copenhagen, Denmark, fidi@sund.ku.dk

Back to Top | Article Outline

REFERENCES

1. de Mestral C, Bell S, Stamatakis E, Batty GD. Testing differential associations between smoking and chronic disease across socioeconomic groups: pooled data from 15 prospective studies. Epidemiology. 2019;30:48–51.
2. Piepoli MF, Hoes AW, Agerwall S, et al. 2016 European guidelines on cardiovascular disease prevention in clinical practice. Eur Heart J. 2016;37:2315–2381.
3. Lundberg M, Hallqvist J, Diderichsen F. Exposure-dependent misclassification of exposure in interaction analyses. Epidemiology. 1999;10:545–549.
4. VanderWeele TJ. A Unification of Mediation and Interaction. A 4-way Decomposition. Epidemiology. 2014;25:749–761.
5. Diderichsen F, Hallqvist J, Whitehead M. Differential vulnerability and susceptibility: how to make use of recent development in our understanding of mediation and interaction to tackle health inequalities. Int J Epidemiol. 2019;48:268–274.
    Copyright © 2019 Wolters Kluwer Health, Inc. All rights reserved.