All D:A:D and individual cohort procedures are developed in accordance with the revised 1975 Helsinki Declaration. We assessed the following individual endpoints in these analyses: Antidiabetic Compound Library MI (including fatal and nonfatal cases); coronary heart disease (CHD; MI plus invasive coronary artery procedure, including coronary artery bypass or angioplasty, or death from other CHD); CVD (CHD plus carotid artery endarterectomy, or stroke); and all-cause mortality. All endpoints are protocol defined, audited for completeness and centrally validated. In the D:A:D study, smoking status is reported as current smoker (yes/no) and ever smoker

(yes/no) at each visit. Dates of stopping or starting smoking are not recorded. Patients were therefore categorized as never smokers, previous smokers, current smokers or smokers who had stopped smoking during D:A:D follow-up. Without dates of stopping or starting cigarette smoking, reasonably accurate times since stopping smoking could only be calculated for those subjects who stopped smoking during the

D:A:D study follow-up period. We calculated time since stopping smoking from the mid-point between the last visit where a subject reported being a current smoker and the HSP cancer first visit where a subject reported being a current nonsmoker. Similarly, subjects who reported

that they started smoking again were taken to do so at the mid-point of the respective visits. Where smoking status was reported to be missing, previous smoking status was carried forward. A sensitivity analysis was also performed omitting all periods of follow-up where smoking status was missing. These analyses were limited to D:A:D patients who ever reported smoking status at enrolment (cohort entry) or during D:A:D follow-up, and had not reported a previous CVD event. Follow-up started at the later of D:A:D cohort entry (enrolment in D:A:D commenced in December 1999) and first reported smoking status, and finished at the earlier of date of death, 6 months after the patient’s last clinic visit or 1 February 2008, whichever occurred first. else Event rates for each endpoint were calculated for never, previous and current smokers, and smokers who stopped during D:A:D follow-up. Event rates for smokers who stopped during D:A:D follow-up were calculated in annual increments (<1, 1–2, 2–3 and >3 years). Smoking status for individual patients could change during D:A:D follow-up. For example, never smokers may become current smokers and then stop smoking, while previous smokers may restart smoking during follow-up. Crude unadjusted event rates for each smoking status group were calculated.