Cardiovascular death is the most important cause of mortality in alcoholics, yet alcohol may protect against ischaemic heart disease. This could be explained if deaths were a consequence of alcohol-related arrhythmias rather than of coronary atheroma. In many conditions, abnormalities of the QT interval are markers of arrhythmia and for risk of sudden death. We examined the relation between QT intervals and mortality in patients with alcoholic liver disease. Simultaneous 12-lead electrocardiographic recordings were obtained from 69 patients with histologically proven alcoholic liver disease (without evidence of structural heart disease), and from 40 healthy non-drinking controls matched for age and sex. Patients were abstinent for at least 7 days before investigation to exclude acute effects of alcohol. QT intervals were corrected for rate with Bazett's and cube root formulae to define QTc and QTcub, respectively. Unlike QTc, QTcub was independent of rate. Patients were followed for up to four years. For those who died, the cause was determined from case records and postmortem reports. Maximum QT intervals were longer in alcoholics than in controls (QTcub 450 vs 439, p = 0.016). This difference was not explained by variations in electrolytes. QT intervals were prolonged in the 14 patients who died compared with survivors (QTcub 471 vs 446, p = 0.007). This difference was mainly due to the long QT intervals in the 6 patients with sudden cardiac deaths (QTcub 493). The only other factor independently associated with death was sex. QT interval prolongation occurs in some patients with alcoholic liver disease and is associated with an adverse prognosis, especially sudden cardiac death. QT measurement should be included in the initial assessment of alcoholic patients, particularly in those considered for liver transplantation.