Effect of time, dose, and fractionation on temporal lobe necrosis following radiotherapy for nasopharyngeal carcinoma

Abstract

Purpose: To study the relative effects of different radiation factors on temporal lobe necrosis (TLN) and predictive accuracy of different biological equivalent models. Methods and Materials: Consecutive patients (1008) treated radically with four different fractionation schedules during 1976-1985 for T1 nasopharyngeal carcinoma were retrospectively analyzed. All were irradiated by megavoltage photons using the same technique. Their age ranged from 18-84 years, and 92% of patients had complete follow-up. The fractional dose to inferomedial parts of both temporal lobes ranged from 2.5-4.2 Gy, total dose 45.6-60 Gy, and overall time 38-75 days. Results: Despite a lower total dose of 50.4 Gy, the 621 patients irradiated with 4.2 Gy per fraction had a significantly higher incidence of temporal lobe necrosis than the 320 patients treated to 60 Gy with 2.5 Gy per fraction: the 10-year actuarial incidence being 18.6% vs. 4.6%, p < 0.001. Multivariate survival analysis showed that fractional effect (product of total dose and fractional dose) was the most significant factor: p = 0.0022, hazard ratio (HR) = 1.044 per Gy2. Overall time and age were both insignificant. The α/β ratio calculated from our data was 2.9 Gy (95% CI: -1.8, 7.6 Gy). Biological effective dose (BED(Gy3)), neuret, and brain tolerance unit all showed strongly significant correlation with the necrotic rate (p < 0.001), and gave similar predictions. The hazard of TLN increased by 14% per Gy3, and it was estimated that 64 Gy (at conventional fractionation of 2 Gy daily) would lead to a 5 % necrotic rate at 10 years. Not only did the nominal standard dose (NSD) show the lowest value in terms of log likelihood and standardized HR, but its predictions on TLN deviated markedly from clinically observed rates. Conclusion: Fractional effect is the most significant factor affecting cerebral necrosis, and overall time has little protective effect. The BED formula, assuming an α/β ratio of 3 Gy, is an appropriate model for predicting late effects on the temporal lobe, and NSD could give seriously misleading predictions.