The aim of the study was to investigate the UK prevalence of late, severe side-effects associated with radical radiotherapy for cancer of the cervix and try to identify associated factors. All patients treated for cancer of the cervix with radical radiotherapy in 1993 were identified and retrospective case notes studied to determine mortality and severe complications occurring following treatment. Of the 55 radiotherapy departments in the UK that were treating gynaecological malignancy in 1993, 53 participated in the study. There were 1558 patients with carcinoma of the cervix receiving radical radiotherapy as part of their treatment regimen in 1993, whose patterns of treatment were assessed. The main outcome measures were the development of late severe complications as defined by the Franco-Italian Glossary and mortality. Of the patients receiving surgery and radiotherapy, 58.5% underwent Wertheim's procedure. The crude rate of late severe complications in all patients with cervical cancer treated with radical radiotherapy in 1993 was 6.1% (actuarial rate 8%) at 5 years, and only four of the 91 patients who developed complications died as a result of their morbidity. There was no significant correlation of stage, centre size, surgery or radiotherapeutic approach with late morbidity in univariate analysis. The overall survival at 5 years was 47% and was lower than that of the European data from FIGO's 1990-92 cohort, for all stages. Increasing FIGO stage was the only factor significantly associated with mortality. The absence of variables that were significantly associated with late complications may well be related to the relatively low event rate compared to the sample size. Differences in surgical treatment prior to radiotherapy and radiation technique may be confounding the comparison of outcomes. The relatively poor survival for locally advanced disease and the difficulty with which these data were collated indicates that national prospective data collection is urgently required to monitor performance and hence derive best practice.