A Bayesian Network Meta-Analysis (NMA) of First-Line Systemic Chemotherapies for Metastatic Pancreatic Cancer (mPC)

Abstract

Background: The NAPOLI 3 trial added NALIRIFOX as a new first-line treatment option for mPC to the existing options of FOLFIRINOX, Nab-paclitaxel plus gemcitabine (NABPplusGEM) and Gemcitabine (GEM). In contrast to conventional NMA using hazard ratios (HR), we applied an innovative method using reconstructed pseudo individual patient-level data (IPD) from published survival curves without the constraint of the proportional hazards assumption to conduct a Bayesian indirect treatment comparison of these regimens.Methods: Eligible phase III randomized clinical trials directly comparing either NALIRIFOX, FOLFIRINOX, NABPplusGEM or GEM were identified from PubMed and Embase. The pseudo-IPD were reconstructed using the Liu et al method. A fixed-effects model Bayesian NMA was performed using HRs of overall survival (OS) and progression free-survival (PFS) curves with GEM as the comparator. Additionally, a Bayesian NMA was also conducted using OS and PFS curves to estimate the HRs over a time frame of 1 month to 36 months, with the assumption that the true survival functions follow parametric distributions such as Weibull, Gompertz, Exponential, Log-logistic, and Log-normal. The best-fitting model was selected based on the lowest deviance information criterion (DIC). Fixed effects log-logistic and log-normal models were retained for OS and PFS curves, respectively, with GEM as the comparator. Results: The estimated HRs for the OS NMA of HRs were 0.53 (95% CrI: 0.42, 0.67) for FOLFIRINOX, 1.2, 0.72 (95% CrI: 0.62, 0.83) for NABPplusGEM and 0.60 (95% CrI: 0.48, 0.75) for NALIRIFOX, all compared to GEM. For PFS, the HR for FOLFIRINOX was 0.47 (95% CrI: 0.37, 0.59), 0.69 (95% CrI: 0.58, 0.82) for NABPplusGEM, 0.48 (95% CrI: 0.37, 0.61) for NALIRIFOX, all relative to GEM. For OS NMA of survival curves, the HRs of FOLFIRINOX at months 1 and 36 for the survival curves NMA were 0.71 (95%CrI=0.58-0.86) and 0.84 (95%CrI=0.65-1.04), respectively. The corresponding estimates for NABPplusGEM were 0.78 (95%CrI=0.69-0.90) and 0.86 (95%CrI=0.74-1.00); and 0.67 (95%CrI=0.56-0.80) and 0.76 (95%CrI=0.60-0.94) for NALIRIFOX. For PFS, the HRs of FOLFIRINOX at months 1 and 36 were 0.56 (95%CrI=0.47-0.66) and 0.57 (95%CrI=0.43-0.76). The corresponding estimates for NABPplusGEM were 0.68 (95%CrI=0.58-0.80) and 0.68 (95%CrI=0.53-0.85); and 0.50 (95%CrI=0.41-0.61) and 0.44 (95%CrI=0.32-0.60) for NALIRIFOX. Conclusion: This Bayesian NMA of survival curves and HRs for first-line treatment regimens for mPC indicates that FOLFIRINOX and NALIRIFOX have comparable survival benefits. Both regimens are associated with peripheral neuropathies and hematological toxicity, with NALIRIFOX being comparably more tolerable. NABPplusGEM is a suitable option for patients with lower performance status. All three regimens prevailed in efficacy over GEM.