To directly estimate how much smoking contributes to cause-specific mortality in Bangladesh.

A case–control study was conducted with surveillance data from Matlab, a rural subdistrict. Cases ( n = 2213) and controls ( n = 261) were men aged 25 to 69 years who had died between 2003 and 2010 from smoking-related and non-smoking-related causes, respectively. Cause-specific odds ratios (ORs) were calculated for “ever-smokers” versus “never-smokers”, with adjustment for education, tobacco chewing status and age. Smoking-attributable deaths among cases, national attributable fractions and cumulative probability of surviving from 25 to 69 years of age among ever-smokers and never-smokers were also calculated.

The fraction of ever-smokers was about 84% among cases and 73% among controls (OR: 1.7; 99% confidence interval, CI: 1.1–2.5). ORs were highest for cancers and lower for respiratory, vascular and other diseases. A dose–response relationship was noted between age at smoking initiation and daily number of cigarettes or bidis smoked and the risk of death. Among 25-year-old Bangladeshi men, 32% of ever-smokers will die before reaching 70 years of age, compared with 19% of never-smokers. In 2010, about 25% of all deaths observed in Bangladeshi men aged 25 to 69 years (i.e. 42 000 deaths) were attributable to smoking.

Smoking causes about 25% of all deaths in Bangladeshi men aged 25 to 69 years and an average loss of seven years of life per smoker. Without a substantial increase in smoking cessation rates, which are low among Bangladeshi men, smoking-attributable deaths in Bangladesh are likely to increase.