Session Item

Our recent individual patient-based meta-analysis of phase III studies on prophylactic cranial irradiation (PCI) in stage III non-small cell lung cancer (NSCLC) showed a reduced incidence of brain metastases (BM) (hazard ratio [HR] 0.38), without a significant effect on overall survival (OS). In these RCTs, different fractionation schedules were used. We here performed subset analyses to explore the impact of different PCI fractionation schedules on brain metastases-free survival (BMFS) and neurocognitive toxicity.

We compared two different PCI fractionation schedules (30 Gy in 10 fractions and 30 Gy in 15 fractions) to observation. All analyses were performed based on the intention-to-treat principle. Patients intended to be treated with other schedules (e.g. 30 Gy in 12 fractions or 36 Gy in 18 fractions) were excluded. The log-rank observed minus expected number of events and its variance as well as the HRs with 95% confidence intervals (95% CIs) were calculated. Interaction was tested using a chi-square test.

The highest reported toxicity grade among all reported adverse events per patient in the PCI arm during follow-up was analysed and categorized into no toxicity (grade 0), mild toxicity (grade 1 or 2) and severe toxicity (≥ grade 3). A mixed-effects logistic regression model with a random intercept for study was used to compare any toxicity (≥ grade 1) to no toxicity (grade 0) between the two fractionation schedules. Furthermore, severe toxicity (≥ grade 3) was compared to mild/no toxicity (< grade 3) using the same approach.

In total, 580 patients were included from three RCTs: RTOG 0214 using 30 Gy in 15 fractions; Guangzhou 2005 and NVALT-11 using 30 Gy in 10 fractions. For both PCI fractionation schedules, BMFS was statistically significantly higher than for observation (30 Gy in 10 fractions: HR 0.26, 95% CI [0.15 to 0.44], 30 Gy in 15 fractions: HR 0.52, 95% CI [0.31 to 0.86]). No statistically significant interaction was found between the two fractionation schedules (p = 0.07)(Figure 1).

Among the 285 patients studied, those included in the 30 Gy  in 15 fractions arm had a statistically significantly lower risk of neurocognitive toxicity compared to patients included in the 30 Gy in 10 fractions arm (Odds ratio [OR] 0.38, 95% CI [0.23 to 0.62]). A similar trend was observed when comparing high grade neurocognitive toxicity to mild/no toxicity, although not statistically significant (OR 0.21, 95% CI [0.02 to 2.45]).

The 30 Gy in 10 fractions schedule seemed more effective in reducing BM, although no statistically significant interaction was observed, but the 15 fractions schedule was statistically significantly less toxic. This trade-off should be taken into account when considering PCI for NSCLC.