In our study, nivolumab showed poor survival outcomes regarding poor PS among gastric cancer patients. In the ATTRACTION-2 trial, poor PS (PS 1), low serum sodium concentration, high neutrophil-to-lymphocyte ratio, and no prior ramucirumab suggested poor prognosis among AGC patients who received nivolumab . The results of subgroup analysis in phase 2 and 3 trials of pembrolizumab showed that better PS was associated with a higher RR and longer OS [6, 7]. However, no patients had PS 2–4 in these trials. Spigel DR et al. reported on the safety and efficacy of nivolumab among non-small cell lung cancer (NSCLC) patients, including those aged ≥70 years or with poor PS (CheckMate 153). In the CheckMate 153 trial, the levels of safety in the overall population and patients with PS 2 were almost the same (serious treatment adverse events: 5–6%). However, the median OS of the overall population was 9.1 months and that of patients with PS 2 was only 4.0 months . Although, there were few studies of nivolumab in patients with poor PS (PS 2,3, and 4).
Mishima S et al. reported significant improvements in objective RR (ORR) and PFS among patients with PS 0 compared with those with PS 1 or 2 (ORR: 30% vs. 3%, p < 0.01; median PFS 3.0 vs.1.1 months; hazard ratio [HR] 0.30, 95% CI 0.18–0.52, p < 0.01) . For other cancers such as NSCLC and malignant melanoma, several studies reported that poor PS was associated with poor survival outcome. Fujimoto D et al. reported that smoking status, EGFR mutation/ALK rearrangement and poor PS were independent poor prognostic factors among NSCLC patients in a multicentre retrospective cohort study (PS 0–1 vs 2–4; HR 0.41, p < 0.001) . Katsura H et al. studied the efficacy and safety of nivolumab among NSCLC patients with poor PS. The OS durations of patients with PS 0–1 and 2–4 were 412 and 32 days, respectively (p < 0.001) . Our study is the first to focus on nivolumab for AGC patients with poor PS. In our study, the OS among patients with poor PS was significantly shorter than that among those with good PS (83 vs. 177 days, p = 0.0255). The same trend was observed in our study.
In a previous study of NSCLC (CheckMate 153 trial), irAEs were similar for the overall population (6%) and patients with an ECOG PS of 2 (9%) . Katsura H et al. reported that the incidence of pneumonitis in the group with poor PS was significantly higher than that in the group with good PS (35% vs. 9%, p = 0.028) . Fujimoto D et al. reported that the incidence rates of severe irAEs were similar between those with good PS scores (0–1) and poor PS scores (2–4) within 2 months after commencing nivolumab therapy (6.1% vs. 6.3%, respectively; p = 0.918). However, 3 out of 4 patients who developed toxicities of grade 5 had poor PS . In our study, there were similar frequencies of treatment-related adverse events between the Good and Poor groups, but 1 patient with poor PS developed grade 5 toxicity. These results suggested that severe toxicity needed to be noted in the Poor PS group.
In our study, patients with poor PS received limited benefit from nivolumab; only 1 patient with poor PS achieved PR due to complications such as brain infarction. Only 2 patients achieved a PFS of 6 months in the Poor group; the poor PS among both patients was due to brain infarction and osteoarthritis of the hip. In the Poor group, 20 patients recorded poor PS due to gastric cancer; the RR was 0% and the DCR was 5%. These results suggested that a high tumour burden was associated with poor outcome among gastric cancer patients who received nivolumab.
In AGC patients, no biomarker has been established to predict the efficacy of nivolumab therapy. We investigated whether JCOG index, RMH index and mGPS could be predictive factor for poor PS patients, but none could be a predictive factor. The reason may be that poor PS was already included as a factor in JCOG index and RMH index. In addition, only 4 cases had mGPS of 0 in poor PS patients. We consider the need for studies to validate new predictive factors of nivolumab efficacy in poor PS AGC patients.