Smoking, smoking cessation, and risk for type 2 diabetes mellitus: a cohort study

Abstract

Background: Cigarette smoking is an established predictor of incident type 2 diabetes mellitus, but the effects of smoking cessation on diabetes risk are unknown.

Objective: To test the hypothesis that smoking cessation increases diabetes risk in the short term, possibly owing to cessation-related weight gain.

Design: Prospective cohort study.

Setting: The ARIC (Atherosclerosis Risk in Communities) Study.

Patients: 10,892 middle-aged adults who initially did not have diabetes in 1987 to 1989.

Measurements: Smoking was assessed by interview at baseline and at subsequent follow-up. Incident diabetes was ascertained by fasting glucose assays through 1998 and self-report of physician diagnosis or use of diabetes medications through 2004.

Results: During 9 years of follow-up, 1254 adults developed type 2 diabetes. Compared with adults who never smoked, the adjusted hazard ratio of incident diabetes in the highest tertile of pack-years was 1.42 (95% CI, 1.20 to 1.67). In the first 3 years of follow-up, 380 adults quit smoking. After adjustment for age, race, sex, education, adiposity, physical activity, lipid levels, blood pressure, and ARIC Study center, compared with adults who never smoked, the hazard ratios of diabetes among former smokers, new quitters, and continuing smokers were 1.22 (CI, 0.99 to 1.50), 1.73 (CI, 1.19 to 2.53), and 1.31 (CI, 1.04 to 1.65), respectively. Further adjustment for weight change and leukocyte count attenuated these risks substantially. In an analysis of long-term risk after quitting, the highest risk occurred in the first 3 years (hazard ratio, 1.91 [CI, 1.19 to 3.05]), then gradually decreased to 0 at 12 years.

Limitation: Residual confounding is possible even with meticulous adjustment for established diabetes risk factors.

Conclusion: Cigarette smoking predicts incident type 2 diabetes, but smoking cessation leads to higher short-term risk. For smokers at risk for diabetes, smoking cessation should be coupled with strategies for diabetes prevention and early detection.

Figure 1

Nine-year adjusted hazard ratio (1990–1998) for incident diabetes in 10,892 middle-aged adults by tertile of pack-years of smoking at baseline (1987–1989). Estimates are simultaneously adjusted for race, sex, ARIC study center, level of education, baseline age, body mass index, waist circumference, physical activity, triglycerides, HDL cholesterol, and systolic blood pressure. Bars indicate 95% confidence intervals. Never-smokers comprise the reference group. Summary relative hazards [95%CI] were estimated in a model in which pack-years were handled as a continuous variable.

Figure 1

Study flow diagram Individuals may have met more than 1 exclusion criteria.

Figure 2

Nine-year adjusted hazard ratio (1990–1998) for incident diabetes in 10,892 middle-aged adults by years of quitting before baseline (1987–1989). Estimates are simultaneously adjusted for race, sex, ARIC Study center, level of education, baseline age, body mass index, waist circumference, physical activity, triglycerides, HDL cholesterol, and systolic blood pressure. Bars indicate 95% confidence intervals. Never-smokers comprise the reference group.

Figure 3. Adjusted hazard ratios for incident self-reported diabetes over 12 years by smoking status in 10,406 middle-aged adults without diabetes at baseline and 3-year follow-up

The diamonds connected by the solid line represents hazard ratios of adults who quit smoking between baseline (Visits 1, 1987–89) and Visit 2 (1990–1992), and continued cessation up to the interval prior to the clinic follow-up visits or telephone interviews where hazard ratios were estimated. The circles connected by the dotted line represents hazard ratios of smokers who continued smoking up to the prior interval. Non-smokers comprise the reference group (not depicted on the graph). Estimates are simultaneously adjusted for race, sex, ARIC study center, level of education, baseline age, body mass index, waist circumference, physical activity, triglycerides, HDL cholesterol, and systolic blood pressure. The bars indicate upper and lower confidence intervals; (for the sake of clarity, only one interval is shown for each estimate).