Radiation-induced liver disease after three-dimensional conformal radiotherapy for patients with hepatocellular carcinoma: dosimetric analysis and implication.

PURPOSE: To analyze the correlation of radiation-induced liver disease (RILD) with patient-related and treatment-related dose-volume factors and to describe the probability of RILD by a normal tissue complication probability (NTCP) model for patients with hepatocellular carcinoma (HCC) treated with three-dimensional conformal radiotherapy (3D-CRT). METHODS AND MATERIALS: Between November 1993 and December 1999, 93 patients with intrahepatic malignancies were treated with 3D-CRT at our institution. Sixty-eight patients who were diagnosed with HCC and had complete 3D dose-volume data were included in this study. Of the 68 patients, 50 had chronic viral hepatitis before treatment, either type B or type C. According to the Child-Pugh classification for liver cirrhosis, 53 patients were in class A and 15 in class B. Fifty-two patients underwent transcatheter arterial chemoembolization with an interval of at least 1 month between transcatheter arterial chemoembolization and 3D-CRT to allow adequate recovery of hepatic function. The mean dose of radiation to the isocenter was 50.2 +/- 5.9 Gy, in daily fractions of 1.8-2Gy. No patient received whole liver irradiation. RILD was defined as Grade 3 or 4 hepatic toxicity according to the Common Toxicity Criteria of the National Cancer Institute. All patients were evaluated for RILD within 4 months of RT completion. Three-dimensional treatment planning with dose-volume histogram analysis of the normal liver was used to compare the dosimetric difference between patients with and without RILD. Maximal likelihood analysis was conducted to obtain the best estimates of parameters of the Lyman NTCP model. Confidence intervals of the fitted parameters were estimated by the profile likelihood method.
RESULTS: Twelve of the 68 patients developed RILD after 3D-CRT. None of the patient-related variables were significantly associated with RILD. No difference was found in tumor volume (780 cm(3) vs. 737 cm(3), p = 0.86), normal liver volume (1210 cm(3) vs. 1153 cm(3), p = 0.64), percentage of normal liver volume with radiation dose >30 Gy (V(30 Gy); 42% vs. 33%, p = 0.05), and percentage of normal liver volume with >50% of the isocenter dose (V(50%); 45% vs. 36%, p = 0.06) between patients with and without RILD. The mean hepatic dose was significantly higher in patients with RILD (2504 cGy vs. 1965 cGy, p = 0.02). The probability of RILD in patients could be expressed as follows: probability = 1/[1 + exp(-(0.12 x mean dose - 4.29))], with coefficients significantly different from 0. The best estimates of the parameters in the Lyman NTCP model were the volume effect parameter of 0.40, curve steepness parameter of 0.26, and 50% tolerance dose for uniform irradiation of whole liver [TD(50)(1)] of 43 Gy. Patients with RILD had a significantly higher NTCP than did those with no RILD (26.2% vs. 15.8%; p = 0.006), using the best-estimated parameters.
CONCLUSION: Dose-volume histogram analysis can be effectively used to quantify the tolerance of the liver to RT. Patients with RILD had received a significantly higher mean dose to the liver and a significantly higher NTCP. The fitted volume effect parameter of the Lyman NTCP model was close to that from the literature, but much lower in our patients with HCC and prevalent chronic viral hepatitis than that reported in other series with patients with normal liver function. Additional efforts should be made to test other models to describe the radiation tolerance of the liver for Asian patients with HCC and preexisting compromised hepatic reserve.