Summary Background Alcohol is an important determinant of the high and fluctuating adult mortality rates in Russia, but cause-specific detail is lacking. Our case–control study investigated the effects of alcohol consumption on male and female cause-specific mortality. Methods In three Russian industrial cities with typical 1990s mortality patterns (Tomsk, Barnaul, Biysk), the addresses of 60 416 residents who had died at ages 15–74 years in 1990–2001 were visited in 2001–05. Family members were present for 50 066 decedents; for 48 557 (97%), the family gave proxy information on the decedents' past alcohol use and on potentially confounding factors. Cases (n=43 082) were those certified as dying from causes we judged beforehand might be substantially affected by alcohol or tobacco; controls were the other 5475 decedents. Case versus control relative risks (RRs; calculated as odds ratios by confounder-adjusted logistic regression) were calculated in ever-drinkers, defining the reference category by two criteria: usual weekly consumption always less than 0·5 half-litre bottles of vodka (or equivalent in total alcohol content) and maximum consumption of spirits in 1 day always less than 0·5 half-litre bottles. Other ever-drinkers were classified by usual weekly consumption into three categories: less than one, one to less than three, and three or more (mean 5·4 SD 1·4) bottles of vodka or equivalent. Findings In men, the three causes accounting for the most alcohol-associated deaths were accidents and violence (RR 5·94, 95% CI 5·35–6·59, in the highest consumption category), alcohol poisoning (21·68, 17·94–26·20), and acute ischaemic heart disease other than myocardial infarction (3·04, 2·73–3·39), which includes some misclassified alcohol poisoning. There were significant excesses of upper aerodigestive tract cancer (3·48, 2·84–4·27) and liver cancer (2·11, 1·64–2·70). Another five disease groups had RRs of more than 3·00 in the highest alcohol category: tuberculosis (4·14, 3·44–4·98), pneumonia (3·29, 2·83–3·83), liver disease (6·21, 5·16–7·47), pancreatic disease (6·69, 4·98–9·00), and ill-specified conditions (7·74, 6·48–9·25). Although drinking was less common in women, the RRs associated with it were generally more extreme. After correction for reporting errors, alcohol-associated excesses accounted for 52% of all study deaths at ages 15–54 years (men 8182 59% of 13968, women 1565 33% of 4751) and 18% of those at 55–74 years (men 3944 22% of 17 536, women 1493 12% of 12 302). Allowance for under-representation of extreme drinkers would further increase alcohol-associated proportions. Large fluctuations in mortality from these ten strongly alcohol-associated causes were the main determinants of recent fluctuations in overall mortality in the study region and in Russia as a whole. Interpretation Alcohol-attributable mortality varies by year; in several recent years, alcohol was a cause of more than half of all Russian deaths at ages 15–54 years. Alcohol accounts for most of the large fluctuations in Russian mortality, and alcohol and tobacco account for the large difference in adult mortality between Russia and western Europe. Funding UK Medical Research Council, Cancer Research UK, British Heart Foundation, International Agency for Research on Cancer, and European Commission Directorate-General for Research.