Inclusion of clinical risk factors into NTCP modelling of late rectal toxicity after high dose radiotherapy for prostate cancer

Background and purpose: To fit an NTCP model including clinical risk factors to late rectal toxicities after radiotherapy for prostate cancer. Methods and materials: Data of 669 patients were considered. The probability of late toxicity within 36 months (bleeding and incontinence) was fitted with the original and a modified Logit-EUD model, including clinical factors by fitting a subset specific TD 50s: the ratio of TD 50s with and without including the clinical variable was the dose-modifying factor (D mod). Results: Abdominal surgery (surg) was a risk factor for G2-G3 bleeding, reflecting in a TD 50 = 82.7 Gy and 88.4 Gy for patients with and without surg (D mod = 0.94; 0.90 for G3 bleeding); acute toxicity was also an important risk factor for G2-G3 bleeding (D mod = 0.93). Concerning incontinence, surg and previous diseases of the colon were the clinical co-factors. D mod(surg) and D mod(colon) were 0.50 and 0.42, respectively for chronic incontinence and 0.73 and 0.64, respectively for mean incontinence score \ue2\u89\ua51. Best-fit n values were 0.03-0.05 and 1 for bleeding and incontinence, respectively. The inclusion of clinical factors always improved the predictive value of the models. Conclusions: The inclusion of predisposing clinical factors improves NTCP estimation; the assessment of other clinical and genetic factors will be useful to reduce parameter uncertainties