BACKGROUND : A hazard ratio network meta-analysis (NMA) is a commonly applied method for conducting an indirect treatment comparison. More recently, also PNMAs have been introduced in the NICE DSU 14 guidelines. PNMAs allow treatment differentiation in long-term survival predictions.

OBJECTIVES : To estimate expected PFS for D-VMP versus ESMO/EMN recommended treatments for newly-diagnosed MM patients SCT ineligible through a PNMA.

METHODS : A SLR identified 30 randomized controlled trials (RCT), of which nine investigated EMEA relevant treatments. Numbers of RCTs evaluating D-VMP was one (n = 1), MP-thalidomide (MPT) (n=5), MP (n = 6), VMP (n=2), lenalidomide-dexamethasone (Rd) 18 cycles (n=1), Rd continuous (n=1), MPR-R (n = 1) and CTD (n = 2). In the absence of patient-level data, the PNMA relied mostly on reconstructed survival data and estimate long-term results. The PNMA allows differences in the shape and scale parameters for each treatment options. The considered survival functions were Weibull, Exponential, Lognormal, Gompertz, and Loglogistic. Best fit was determined based on Watanabe-Akaike Information Criterion.

RESULTS : Weibull distribution provided the best fit. D-VMP had best mean and median estimated PFS (39 months; 51 months), followed by MPR-R (25 months; 39 months), Rd continuous (25 months; 35 months), VMP (19 months; 24 months), Rd18 (19 months; 22 months), MPT (19 months; 22 months), CTD (14 months; 17 months) and MP (12 months; 15 months).

CONCLUSIONS : D-VMP showed the most favourable PFS for newly-diagnosed MM ASCT ineligible patients versus the comparators in the network, with a longer median and mean PFS irrespectively of extrapolation function, except with the Gompertz. Over time, Gompertz predicted hazard rates to become zero and is therefore considered not clinically plausible. Despite the order of treatment options being stable over all functions, the estimated mean and median survival times were sensitive to the reference trial and the applied distribution.