A method of analysis taking into account competing events: application to the study of digestive complications following irradiation for cervical cancer.

Over the past 20 years, mortality rates from uterine cervix carcinoma have been decreasing owing to earlier diagnosis and improved treatment. The incidence of secondary effects from radiation therapy has spurred some interest and injuries resulting from 'over'-treatment have been of some concern. The present study concerns the evaluation of bowel complications following treatment for the 272 stage IIb and III patients treated in our centre between 1967 and 1973. Competing risk methodology is applied and the influence of radiotherapy parameters is evaluated by a log-linear model of the event-specific failure rates. A single parameter (NSD) which summarized total dose, total number of sessions and total treatment time is found to be related to the occurrence of complications.