OBJECTIVES: The complexity of clinical trial data requires us to be innovative in conducting indirect treatment comparisons. This study undertakes a formal comparison of three distinct indirect treatment comparison (ITC) methods, namely conventional Bayesian NMA (c-BNMA), matching adjusted indirect comparisons (MAIC) and multilevel-network meta-regression (ML-NMR). BNMA fails to accurately quantify the impact of baseline characteristics with limited study data. In such cases, MAIC incorporating individual patient data (IPD) and employing reweighting techniques emerges as a promising alternative. Furthermore, ML-NMR, a novel ITC approach, relies on a network of evidence and at least one study with IPD.

METHODS: ML-NMR integrated simulated IPD (VIRTUAL) and aggregated data from CARD and PROfound to evaluate the efficacy of NewTech, cabazitaxel and olaparib compared to Androgen Receptor Pathway Inhibition (ARPi) in metastatic castration-resistant prostate cancer. Bernoulli likelihood with logit link is used to regress age, ECOG and PSA level on PSA response along with treatment interaction. Markov chain Monte Carlo (MCMC) simulations estimated relative treatment effects as odds ratios (OR). For BNMA, sufficient MCMC iterations along with vague prior were used. The propensity weighting method was used to perform MAIC.

RESULTS: BNMA showed that NewTech is associated with higher PSA response vs ARPi (OR = 4.32; 95% credible interval (CrI): 2.5-7.51). MAIC and 2-stage MAIC results for NewTech vs ARPi are (OR=9.83, 95% CI: 2.69-35.98 and OR=11.64, 95% CI: 3.12-43.48) respectively. ML-NMR also showed benefit in favor of NewTech vs ARPi (OR=5.87, 95% CrI: 3.00-12.80).

CONCLUSIONS: Our findings indicate that the BNMA, MAIC and ML-NMR generated similar results with ML-NMR depicting a more promising assessment of PSA response by incorporating baseline characteristics and providing an opportunity to overcome the conventional ITC limitations.