Defining the optimal radiation dose with three-dimensional conformal radiation therapy for patients with nonmetastatic prostate carcinoma by using recursive partitioning techniques

BACKGROUND. The objective of this study was to determine the effect of dose and its interaction with known prognostic variables, including pretreatmen t prostate specific antigen (PSA), Gleason score (GS), and T classification , on patients with nonmetastatic prostate carcinoma treated with three-dime nsional conformal radiation therapy (3DCRT) alone using recursive partition ing analysis. METHODS. Between November 1987 and November 1997, 939 patients with nonmeta static prostate carcinoma were treated with 3DCRT alone at Fox Chase Cancer Center. Biochemical no evidence of disease (bNED) control was defined usin g the American Society of Therapeutic Radiology and Oncology Consensus defi nition. Recursive partitioning analysis was used to identify subgroups with similar risks of bNED failure. Prognostic factors used in the model includ ed pretreatment PSA, GS, T classification, and radiation dose. The median f ollow-up was 47 months (range, 2-133 months). RESULTS. Twelve terminal nodes of the decision tree were merged to form fou r prognostic groups with similar bNED control rates. The 5-year actuarial r ates of bNED control rates for Groups I, II, III, and IV were 84%, 41%, 16% , and 67%, respectively (P < 0.0001). Increasing the dose to greater than 7 235 centigray (cGy) improved bNED control rates for patients with PSA level s of 10-19.9 ng/mL and T1/2a classification disease. Increasing the dose to greater than 7629 cGy improved bNED control rates for patients with T2b/3 classification disease with PSA levels less than 20 ng/mL. Patients with PS A levels greater than or equal to 20 ng/mL need high-dose 3DCRT. For those patients with GS 2-6 and T1/2a classification disease, treatment with great er than 7400 cGy resulted in 67% bNED control rate versus 16% at 5 years fo r treatment with less than 7400 cGy. High radiation dose (> 7700 cGy) impro ved bNED control rate from 16% to 41% for patients with high-risk disease ( PSA 20 ng/mL and GS 7-10) at 5 years. CONCLUSIONS, The authors showed that with recursive partitioning techniques radiation dose continues to be an important predictor of bNED control rate and that a radiation dose response for patients with clinically localized prostate carcinoma exists. Patients with one or more prognostic feature (PS A > 10 ng/mL, classification T2b/T3, GS 7-10, or the presence of perineural invasion) achieve similar rates of bNED control compared with those patien ts with lower volume disease when radiation dose is increased. (C) 2001 Ame rican Cancer Society.