Published online Apr 28, 2020. doi: 10.13105/wjma.v8.i2.119
Peer-review started: February 5, 2020
First decision: March 21, 2020
Revised: April 8, 2020
Accepted: April 21, 2020
Article in press: April 21, 2020
Published online: April 28, 2020
A systematic review of the relationship between smoking and incident type 2 diabetes, based on 88 epidemiological prospective studies, was published in 2015. Much new evidence on this relationship has become available since then.
To obtain up-to-date evidence relating smoking to type 2 diabetes.
To systematically review available evidence from prospective studies on the relationship of type 2 diabetes onset to ever, current or former smoking of cigarettes or of any tobacco product, including dose-response data.
Attention was restricted to prospective studies of populations free of type 2 diabetes at baseline which related subsequent incidence of the disease to one or more defined major or dose-related smoking indices. The major indices compared ever, current or former smokers to never smokers and current smokers to non-current smokers. The dose-related indices concerned amount currently smoked and years quit. Literature searches identified relevant papers from previous reviews, from Medline searches and from references lists of relevant papers identified. Data were extracted on study details and on the relative risks required, estimated if required using standard methods. Care was taken to avoid overlap of data from the same study from multiple publications. Fixed-effect and random-effects meta-analyses were conducted, including tests of heterogeneity and publication bias. Where a study provided multiple estimates, a preference scheme was used involving factors such as level of adjustment for confounding factors, length of follow-up and age range considered. Sex-specific results were used, if available. Effect estimates were derived based on all the selected RRs, and also for those subdivided by various categorical variables – sex, continent, year of publication, basis of diagnosis of diabetes, initial diabetes status of the population, age, length of follow-up, definition of smoking, and whether a range of different variables were adjusted for.
The literature searches identified 157 relevant publications providing results from 145 studies. Overall random-effect RR estimates were 1.33 [95% confidence interval (CI): 1.28-1.38] for current vs never smoking, 128 (95%CI: 1.24-1.32) for current vs non-smoking, 1.13 (95%CI: 1.11-1.16) for former vs never smoking and 1.25 (95%CI: 1.21-1.28) for ever vs never smoking, each combined estimate being based on at least 99 individual estimates. Estimates were generally elevated in each subdivision of the data by the categorical variables considered, though in some cases RR estimates varied significantly (P < 0.05) by level. The dose-response analysis showed that risk increased with increasing amount smoked, and reduced with increasing time quit.
Our analyses confirmed and extended reports of a modest dose-related association of current smoking and a weaker dose-related association of former smoking with risk of type 2 diabetes. The evidence suggests smoking may contribute to the risk of type 2 diabetes, though our estimates may be affected by publication bias and some uncontrolled confounding. Although reduction of smoking is clearly important to limit risk of diseases such as lung cancer, chronic obstructive pulmonary disease and cardiovascular disease, the worldwide rise in incidence of type 2 diabetes, coupled with a decline in smoking, suggests that control of other factors, such as diet, may be much more beneficial in reducing type 2 diabetes risk.
Our analyses suggest strongly that there is a modest increased risk of type 2 diabetes associated with current smoking which is greater in heavier smokers and reduced following quitting. Further large prospective studies could characterize this more precisely by more detailed assessment of smoking history and by more fully accounting for the range of other factors known to be related to type 2 diabetes. Care should be taken to determine the accuracy of all the data used, and to assess the effect that any possible inaccuracy might have on the estimated association.