Role of principal component analysis in predicting toxicity in prostate cancer patients treated with hypofractionated intensity-modulated radiation therapy.

PURPOSE: To determine if principal component analysis (PCA) and standard parameters of rectal and bladder wall dose-volume histograms (DVHs) of prostate cancer patients treated with hypofractionated image-guided intensity-modulated radiotherapy (hypo-IMRT) can predict acute and late gastrointestinal (GI) toxicity.

METHODS AND MATERIALS: One hundred twenty-one patients underwent hypo-IMRT at 3 Gy/fraction, 5 days/week to either 60 Gy or 66 Gy, with daily online image guidance. Acute and late GI and genitourinary (GU) toxicity were recorded weekly during treatment and at each follow-up. All Radiation Therapy Oncology Group (RTOG) criteria toxicity scores were dichotomized as <2 and ≥2. Standard dosimetric parameters and the first five to six principal components (PCs) of bladder and rectal wall DVHs were tested for association with the dichotomized toxicity outcomes, using logistic regression.

RESULTS: Median follow-up of all patients was 47 months (60 Gy cohort = 52 months; 66 Gy cohort = 31 months). The incidence rates of ≥2 acute GI and GU toxicity were 14% and 29%, respectively, with no Grade ≥3 acute GU toxicity. Late GI and GU toxicity scores ≥2 were 16% and 15%, respectively. There was a significant difference in late GI toxicity ≥2 when comparing the 66 Gy to the 60 Gy cohort (38% vs. 8%, respectively, p = 0.0003). The first PC of the rectal DVH was associated with late GI toxicity (odds ratio [OR], 6.91; p < 0.001), though it was not significantly stronger than standard DVH parameters such as Dmax (OR, 6.9; p < 0.001) or percentage of the organ receiving a 50% dose (V50) (OR, 5.95; p = 0 .001).

CONCLUSIONS: Hypofractionated treatment with 60 Gy in 3 Gy fractions is well tolerated. There is a steep dose response curve between 60 Gy and 66 Gy for RTOG Grade ≥2 GI effects with the dose constraints employed. Although PCA can predict late GI toxicity for patients treated with hypo-IMRT for prostate cancer, it provides no additional information over using more standard DVH parameters.