The Fraction Size Sensitivity of Late Genitourinary Toxicity: Analysis of Alpha/Beta (α/β) Ratios in the CHHiP Trial
Douglas Brand, Sarah C. Brüningk, Anna Wilkins, Olivia Naismith, Annie Gao, Isabel Syndikus, David P. Dearnaley, Nicholas van As, Emma Hall, S. Gulliford, Alison Tree
Abstract
PURPOSE: Moderately hypofractionated external beam intensity modulated radiation therapy (RT) for prostate cancer is now standard-of-care. Normal tissue toxicity responses to fraction size alteration are nonlinear: the linear-quadratic model is a widely used framework accounting for this, through the α/β ratio. Few α/β ratio estimates exist for human late genitourinary endpoints; here we provide estimates derived from a hypofractionation trial. METHODS AND MATERIALS: The CHHiP trial randomized 3216 men with localized prostate cancer 1:1:1 between conventionally fractionated intensity modulated RT (74 Gy/37 fractions (Fr)) and 2 moderately hypofractionated regimens (60 Gy/20 Fr and 57 Gy/19 Fr). RT plan and suitable follow-up assessment was available for 2206 men. Three prospectively assessed clinician-reported toxicity scales were amalgamated for common genitourinary endpoints: dysuria, hematuria, incontinence, reduced flow/stricture, and urine frequency. Per endpoint, only patients with baseline zero toxicity were included. Three models for endpoint grade ≥1 (G1+) and G2+ toxicity were fitted: Lyman Kutcher-Burman (LKB) without equivalent dose in 2 Gy/Fr (EQD2) correction [LKB-NoEQD2]; LKB with EQD2-correction [LKB-EQD2]; LKB-EQD2 with dose-modifying-factor (DMF) inclusion [LKB-EQD2-DMF]. DMFs were age, diabetes, hypertension, pelvic surgery, prior transurethral resection of prostate (TURP), overall treatment time and acute genitourinary toxicity (G2+). Bootstrapping generated 95% confidence intervals and unbiased performance estimates. Models were compared by likelihood ratio test. RESULTS: The LKB-EQD2 model significantly improved performance over LKB-NoEQD2 for just 3 endpoints: dysuria G1+ (α/β = 2.0 Gy; 95% confidence interval [CI], 1.2-3.2 Gy), hematuria G1+ (α/β = 0.9 Gy; 95% CI, 0.1-2.2 Gy) and hematuria G2+ (α/β = 0.6 Gy; 95% CI, 0.1-1.7 Gy). For these 3 endpoints, further incorporation of 2 DMFs improved on LKB-EQD2: acute genitourinary toxicity and prior TURP (hematuria G1+ only), but α/β ratio estimates remained stable. CONCLUSIONS: Inclusion of EQD2-correction significantly improved model fitting for dysuria and hematuria endpoints, where fitted α/β ratio estimates were low: 0.6 to 2 Gy. This suggests therapeutic gain for clinician-reported GU toxicity, through hypofractionation, might be lower than expected by typical late α/β ratio assumptions of 3 to 5 Gy.