Predictive ability of a drug-based score in patients with advanced non-small-cell lung cancer receiving first-line immunotherapy.

ABSTRACT

BACKGROUND: We previously demonstrated the cumulative poor prognostic role of concomitant medications on the clinical outcome of patients with advanced cancer treated with immune checkpoint inhibitors, creating and validating a drug-based prognostic score to be calculated before immunotherapy initiation in patients with advanced solid tumours. This 'drug score' was calculated assigning score 1 for each between proton-pump inhibitor and antibiotic administration until a month before cancer therapy initiation and score 2 in case of corticosteroid intake. The good risk group included patients with score 0, intermediate risk with score 1-2 and poor risk with score 3-4. METHODS: Aiming at validating the prognostic and putative predictive ability depending on the anticancer therapy, we performed the present comparative analysis in two cohorts of advanced non-small-cell lung cancer (NSCLC), respectively, receiving first-line pembrolizumab or chemotherapy through a random case-control matching and through a pooled multivariable analysis including the interaction between the computed score and the therapeutic modality (pembrolizumab vs chemotherapy). RESULTS: Nine hundred fifty and 595 patients were included in the pembrolizumab and chemotherapy cohorts, respectively. After the case-control random matching, 589 patients from the pembrolizumab cohort and 589 from the chemotherapy cohort were paired, with no statistically significant differences between the characteristics of the matched subjects. Among the pembrolizumab-treated group, good, intermediate and poor risk evaluable patients achieved an objective response rate (ORR) of 50.0%, 37.7% and 23.4%, respectively, (p < 0.0001), whereas among the chemotherapy-treated group, patients achieved an ORR of 37.0%, 40.0% and 32.4%, respectively (p = 0.4346). The median progression-free survival (PFS) of good, intermediate and poor risk groups was 13.9 months, 6.3 months and 2.8 months, respectively, within the pembrolizumab cohort (p < 0.0001), and 6.2 months, 6.2 months and 4.3 months, respectively, within the chemotherapy cohort (p = 0.0280). Among the pembrolizumab-treated patients, the median overall survival (OS) for good, intermediate and poor risk patients was 31.4 months, 14.5 months and 5.8 months, respectively, (p < 0.0001), whereas among the chemotherapy-treated patients, it was 18.3 months, 16.8 months and 10.6 months, respectively (p = 0.0003). A similar trend was reported considering the two entire populations. At the pooled analysis, the interaction term between the score and the therapeutic modality was statistically significant with respect to ORR (p = 0.0052), PFS (p = 0.0003) and OS (p < 0.0001), confirming the significantly different effect of the score within the two cohorts. CONCLUSION: Our 'drug score' showed a predictive ability with respect to ORR in the immunotherapy cohort only, suggesting it might be a useful tool for identifying patients unlikely to benefit from first-line single-agent pembrolizumab. In addition, the prognostic stratification in terms of PFS and OS was significantly more pronounced among the pembrolizumab-treated patients.