Dose-response analysis for radiotherapy delivered to patients with intermediate-grade and large-cell immunoblastic lymphomas that have completely responded to CHOP-based induction chemotherapy

Abstract

Purpose: To test the hypothesis that prechemotherapy tumor size affects the dose of radiation that should be delivered to intermediate-grade and large-cell immunoblastic lymphomas that have completely responded to cyclophosphamide, doxorubicin, vincristine and prednisone (CHOP)-based induction chemotherapy. Methods and Materials: From September 1988 through December 1996, 294 patients with newly diagnosed, Stage I–IV, intermediate-grade or large-cell immunoblastic lymphomas were enrolled on 2 prospective protocols at the M. D. Anderson Cancer Center. Treatment consisted of CHOP-based chemotherapy with or without involved field radiotherapy. One hundred seventy-two patients, with 178 nodal sites and 87 nonbony, extranodal sites of disease achieved a complete response to 2–6 cycles of chemotherapy and underwent involved field radiotherapy. Total radiation doses ranged from 30.0 to 50.4 Gy (mean ± standard deviation: 39.7 ± 2.5 Gy) over 22–49 days using a daily fraction size of 1.3–2.3 Gy. Because various fraction sizes were delivered, the linear-quadratic model was used to convert total radiation doses to biologically equivalent doses given at 1.8 Gy per fraction (D1.8). An α/β ratio of 10 Gy was used for the lymphomas, resulting in D1.8 ranging from 29.1 to 50.8 Gy. Regression tree analysis was performed on nodal sites of disease to determine which of the following factors were predictive of local control: age, tumor size, D1.8, total radiation dose, and duration of radiotherapy. Based on the results of the regression tree analysis, Kaplan-Meier analysis was used to determine the probability of local control per site as a function of tumor size and D1.8. Regression tree analysis was also performed on patients with nonbony disease who received D1.8 = 29.1–39.1 Gy to determine if small lymphomas could be locally controlled with relatively low doses of radiation. The log-rank test was used to compare local control curves. Results: The median length of follow-up among survivors was 63 months. Regression tree analysis of nodal sites identified 3 distinct groups: (a) lymphomas ≤ 10 cm and D1.8 = 29.1–39.1 Gy; (b) lymphomas ≤ 10 cm and D1.8 = 39.2–50.8 Gy; and (c) lymphomas > 10 cm. For nonbony lymphomas that measured < 3.5 cm, low doses of radiation resulted in excellent local control (5-year rates: 96% vs. 97% for D1.8 = 29.1–39.1 Gy vs. D1.8 = 39.2–50.8 Gy; p = 0.610). For 3.5–10.0 cm lymphomas, higher doses of radiation resulted in better local control (5-year rates: 40% versus 98% for D1.8 = 29.1–39.1 Gy versus D1.8 = 39.2–50.8 Gy, p < 0.0001). A narrow dose range (D1.8 = 39.2–40.7 Gy) was delivered to the 8 lymphomas measuring > 10 cm that completely responded to 6 cycles of chemotherapy, resulting in a 5-year local control rate of only 70%. There was no difference in local control for nodal versus nonbony, extranodal sites of disease. Conclusion: D1.8 ranging from 29.1 to 39.1 Gy yielded excellent local control for nonbony lymphomas measuring < 3.5 cm that had completely responded to a median of 3 cycles of CHOP-based chemotherapy. D1.8 ranging from 39.2 to 50.8 Gy yielded excellent local control for nonbony lymphomas measuring 3.5–10.0 cm that completely responded to either 3 or 6 cycles of chemotherapy. For nonbony lymphomas measuring > 10 cm that completely responded to 6 cycles of chemotherapy, D1.8 ranging from 39.2 to 40.7 Gy yielded suboptimal local control, suggesting that higher doses of radiation are indicated.