OBJECTIVES: KEYNOTE-048 compared pembrolizumab +/- chemotherapy (P and P+CT, respectively) with cetuximab + platinum-based chemotherapy (‘EXTREME’) for patients with R/M SCCHN. Differences in progression-free survival (PFS) and overall survival (OS) were observed for various combined positive score (CPS) subgroups. To improve external validity of within-trial CPS 1-19 outcomes, we investigated differences in PFS and OS between P/P+CT and ‘EXTREME’ by utilizing external trial data.

METHODS: Three representative published trials (EXTREME [NCT00122460], TPEXTREME [NCT02268695], and GORTEC [NCT01289522]) reporting on ‘EXTREME’, and no additional trials reporting on P/P+CT, were identified. Parametric models were fitted to re-created OS and PFS data to extrapolate over a lifetime horizon and combined with published utility values to estimate quality-adjusted life years (QALYs). We compared incremental QALYs for KEYNOTE-048 (results for P/P+CT) with estimates for each alternative trial (results for ‘EXTREME’).

RESULTS: Applying models with the best statistical goodness-of-fit using only re-created KEYNOTE-048 data suggested +0.05 and +0.18 QALYs gained for P and P+CT, respectively, versus ‘EXTREME’. However, when switching the comparison data to estimates derived via the three aforementioned trials, QALY gains ranged from -0.12 to +0.27 (P) and -0.01 to +0.38 (P+CT). In some exploratory analyses, QALY gains were positive (i.e., favoring P/P+CT), while life-year gains were negative (i.e., favoring ‘EXTREME’).

CONCLUSIONS: Through our analyses, we conclude that because outcomes for ‘EXTREME’ differ markedly between trials, incremental QALY estimates for subgroup analyses are uncertain and vary depending on the data used to inform them. Hence, outcomes of KEYNOTE-048 should be interpreted with caution. While CPS is expected to have little bearing on patients’ outcomes with ‘EXTREME’, a limitation of our analysis is that historic trials did not report outcomes by CPS to explore similar subgroups. Further research is required to conclusively determine the relative effectiveness of P/P+CT versus ‘EXTREME’, for patients with CPS 1-19.