Multivariate analysis of dissemination relapse of medulloblastoma and estimation of its time parameter for craniospinal irradiation.

PURPOSE

To examine in retrospect prognostic factors influencing meningeal dissemination relapse of medulloblastoma (MB) and to estimate time parameter gamma/alpha of the biologically effective dose for fractionated craniospinal irradiation (CSI).

MATERIALS AND METHODS

Fifty-eight patients with MB who had been treated at our six hospitals from 1980 to 1990, were analyzed by the proportional hazards model consisting of radiation factors of both CSI and local irradiation (LI), sequential CSI time-lag, and eight non-radiation factors (gender, age, performance status, T-stage, dissemination score, extent of resection, and use of chemotherapy and immunotherapy). The gamma/alpha for CSI was estimated by the profile likelihood method using the maximum value of conditionally calculated time-incorporated biologically effective dose, tBEDmax, of the field treated with the least dose.

RESULTS

Dissemination relapse was seen in 23 (40%) patients. Nineteen disseminations occurred within four years, and the cumulative dissemination-free rate was 64% at five years. The site of initial dissemination relapse was both cranial and spinal in 13 patients (57%). Dissemination relapse was accompanied with local failure in 43% (10/23) of patients, and four of them were seen later than four years. In the multivariate analysis, significant prognostic factors were dissemination score (p=0.0008) and total dose of CSI (p=0.018). The estimate of gamma/alpha for CSI was about 0.2 Gy/day in BED units. In another multivariate analysis including the best-fitted tBEDmax, significant prognostic factors were dissemination score and the tBEDmax of both CSI (p=0.021) and LI (p=0.024).

CONCLUSION

This analysis indicated that the dissemination score, total dose of CSI, and tBEDmax of both CSI and LI were significantly prognostic for dissemination relapse of MB. The estimate of gamma/alpha for CSI was smaller than that derived from our previous analysis for LI. However, in order to estimate the time factor for CSI more precisely, a larger group of patients treated with concurrent CSI is needed.