Repair halftimes estimated from observations of treatment-related morbidity after CHART or conventional radiotherapy in head and neck cancer

Background and purpose: The CHART (Continuous Hyperfractionated Accelerated Radiotherapy) head and neck cancer fractionation schedule delivered 54 Gy in 36 fractions on 12 consecutive days and this was compared in a randomise d controlled trial with conventional fractionation delivering 66 Gy in 33 f ractions over 6-7 weeks. Patients receiving CHART experienced statistically significantly less treatment-related morbidity after 6 months than patient s receiving conventional Fractionation. However, this improved tolerance wa s much less than anticipated from existing knowledge of dose-fractionation effects on late-responding normal tissues. Here, the experience from the CH ART study is analysed and repair halftimes for three types of late treatmen t-related morbidity of human tissues an estimated. Patients and methods: The CHART trial was open for patient accrual from Mar ch 1990 to April 1995 and a total of 918 patients in 11 participating centr es were randomised. All patients were followed at regular intervals for a m inimum of 5 years or until the time of death. At each follow-up, a number o f treatment-related morbidity items were evaluated and scored prospectively . Data for three late endpoints are analysed here: laryngeal oedema, skin t elangiectasia and subcutaneous fibrosis. Differences in the incidence of th ese endpoints in the two trial arms were quantified by means of the ratio o f hazard rates in a Cox proportional hazards model. Monte Carlo sampling wa s performed from distributions of fractionation sensitivity (quantified by the alpha/beta-ratio) and steepness of the dose-response curve (quantified by the normalised dose-response gradient, gamma(50)) with means and standar d deviations derived from the literature. Each pair of values were used to convert a Monte Carlo sampled estimate of the difference in biological effe ct into an estimate of the repair halftime. From the distribution of 1000 M onte Carlo samples, the mean repair halftime and its 95% confidence interva l were estimated. Results: The estimated repair halftimes, with 95% confidence intervals in p arentheses, were 4.9 h (3.2, 6.4) for laryngeal oedema, 3.8 h (2.5, 4.6) fo r skin telangiectasia and 4.4 h (3.8, 4.9) for subcutaneous fibrosis. Calcu lations show that these repair halftimes are consistent with the observatio ns from two published randomised controlled trials of altered fractionation in head and neck cancer, the EORTC 22791 and 22851 trials. Conclusions: These long repair halftimes for late effects in human normal t issues have to be considered in order to gain the full benefit from fractio nation schedules employing multiple fractions per day. (C) 1999 Elsevier Sc ience Ireland Ltd. All rights reserved.