Efficacy and a novel clinicopathologic-genomic nomogram of atezolizumab in advanced non-small cell lung cancer (POPLAR and OAK): A combined analysis of two multicenter, randomized, phase II/III trials.

Abstract

2573 Background: Atezolizumab, a programmed death ligand 1 (PD-L1) inhibitor, prolonged overall survival (OS) compared with docetaxel among patients with previously treated advanced non–small-cell lung cancer (NSCLC) in two independent multicentre, randomized trials (POPLAR and OAK). We conducted a combined analysis of the two trials to evaluate its efficacy and genomic biomarkers, and to further developed a novel predictive clinicopathologic-genomic nomogram of immunotherapy in NSCLC. Methods: Patients (N = 1,137) with stage IIIB/IV NSCLC and disease progression after previously platinum-based chemotherapy were randomly assigned (1:1) to receive atezolizumab (1200 mg/kg every 3 weeks) or docetaxel (75 mg/m2 every 3 weeks). The primary endpoint was OS. We applied a two-stage meta-analysis of pooled individual patient data in the intention-to-treat population. In OAK trial, patients treated with atezolizumab were randomly assigned (1:1) to the training group or the validation group to develop a predictive clinicopathologic-genomic nomogram of immunotherapy. POPLAR and OAK were registered with ClinicalTrials.gov, numbers NCT02008227 and NCT01903993. Results: In the pooled analysis, the median overall survival was 13.49 months (95% confidence interval [CI], 11.95 to 15.22) with atezolizumab versus 9.66 months (95% CI, 8.73 to 10.70) with docetaxel. The risk of death was 28% lower with atezolizumab than with docetaxel (hazard ratio [HR], 0.72; 95% CI, 0.62 to 0.83; P < 0.001). The race, sex, tumor histology, eastern cooperative oncology group performance status, PD-L1 expression, and especially pretreatment mutation (TP53, DNMT3A and KEAP1) were significantly associated with OS, and were used for the development of the predictive nomogram. The clinical use of the nomogram showed a closer association with 3-year OS than the blood-based tumor mutational burden (bTMB) or PD-L1 expression alone (nomogram, AUC = 0.818; bTMB, AUC = 0.701; PD-L1, AUC = 0.526) among NSCLC patients who had received atezolizumab. The superior predictability of the nomogram was further confirmed in the validation and entire OAK cohorts. Conclusions: Among patients with advanced, previously treated NSCLC, OS was significantly better with atezolizumab than with docetaxel. Furthermore, we constructed a novel and powerful clinicopathologic-genomic nomogram for personalized immunotherapy options.