Consolidating risk estimates for radiation-induced complications in individual patient: late rectal toxicity

To test the feasibility of a new approach to synthesize published normal tissue complication data using late rectal toxicity in prostate cancer as an example.

A data survey was performed to identify the published reports on the dose-response relationships for late rectal toxicity. The risk estimates for Grade 1 or greater, Grade 2 or greater, and Grade 3 or greater toxicity were obtained for a test cohort of patients treated at our institution. The influence of the potential factors that might have affected the reported toxicity levels was investigated. The studies that did not conform to the general data trends were excluded, and single, combined risk estimates were derived for each patient and toxicity level.

A total of 21 studies of nonoverlapping patient populations were identified. Three studies provided dose-response models for more than one level of toxicity. Of these 21 studies, 6, 14, and 5 were used to derive the initial risk estimates for Grade 1, 2, and 3 or greater toxicity, respectively. A comparison of risk estimates between the studies reporting rectal bleeding and rectal toxicity (bleeding plus other symptoms) or between studies with follow-up <36 months and ≥36 months did not reveal significant differences (p ≥ .29 for all comparisons). After excluding three reports that did not conform to the general data trends, the combined risk estimates were derived from 5 reports (647 patients), 11 reports (3,369 patients), and 5 reports (1,330 patients) for Grade 1, 2, and 3 or greater toxicity, respectively.

The proposed approach is feasible and allows for more systematic use of published dose-response data to estimate the complication risks for the individual patient.