Obesity, underweight, and smoking are associated with an increased mortality. We investigated the joint effects of body mass index and smoking on all-cause and cause-specific mortality. Data of the Third National Health and Nutrition Examination Survey (1988-1994) including mortality follow-up until 2011 were used (n = 17,483). Cox proportional hazards models were used to estimate hazard ratios (HRs) and 95% confidence intervals (CIs) for all-cause, cardiovascular disease (CVD), and cancer mortality with BMI, smoking, and their combinations as exposure, stratified by sex. Normal weight never smokers were considered as reference group. Compared to normal weight never smokers, obese and underweight current smokers were the two combinations with the highest mortality from all-causes, CVD, and cancer. Among underweight current smokers, the HR of death from all-causes was 3.49 (95% CI 2.42-5.02) and for obese current smokers 2.76 (2.12-3.58). All-cause mortality was particularly high in women who were underweight and current smoker (3.88 [2.47-6.09]). CVD mortality risk was the highest among obese current smokers (3.33 [2.98-5.33]). Cancer mortality risk was the highest among underweight current smokers (5.28 [2.68-10.38]). Obese current smokers in the middle age group (between 40 and 59 years old) had the highest risk of all-cause mortality (4.48 [2.94-7.97]). No statistically significant interaction between BMI and smoking on all-cause and cause-specific mortality was found. The current study indicates that obesity and underweight in combination with smoking may emerge as a serious public health problem. Hence, public health messages should stress the increased mortality risk for smokers who are underweight or obese. Also, health messages regarding healthy lifestyle are aimed at maintaining a healthy body weight rather than just "losing weight" and at not starting smoking at all.
Keywords: Mortality; NHANES III; Obesity; Smoking; Underweight.