A Network Meta-Analysis of Hereditary Transthyretin Amyloidosis With Polyneuropathy Supports a Class Effect, Not a Ranking of Individual Drugs

Abstract

Duarte and colleagues [1] have synthesized the randomized evidence in ATTRv-PN with care, and their central finding is sound: gene silencers outperform stabilizers. The effect is large, consistent across both efficacy outcomes, and biologically coherent.

Our concern is the finer claim: that the individual gene silencers can be ranked against one another, and in particular that the RNAi agents outperform the antisense oligonucleotides. It is a different statement, resting on different evidence, and that evidence does not support it. The authors grade the key comparisons as low certainty, then present a definite ranking. The abstract and rankings read as established fact; the evidence behind them does not.

The ranking rests on a single transformed outcome. The network compares each trial against the others, and it contains only two direct head-to-head comparisons: the two RNAi agents against each other, and the two antisense oligonucleotides against each other. In the authors' own league table, neither shows a difference. Every comparison between an RNAi agent and an antisense agent is indirect, and the authors grade these low, downgraded for risk of bias and imprecision.

The ranking also appears on only one outcome. On Norfolk-QoL-DN, reported on a single scale, vutrisiran, patisiran, and eplontersen do not separate. The ordering appears only on mNIS+7, where two versions of the composite scale, differing in range and component weighting, had to be pooled using standardized mean differences.

The eplontersen comparison carries a further, structural problem. In NEURO-TTRansform, inotersen patients crossed to eplontersen at Week 35 [2], after which no inotersen comparator remains, so the eplontersen-versus-inotersen contrast derives only from that interim. Patisiran and vutrisiran enter the network at 18 months [3, 4]. Silencer effects grow with continued treatment, so a comparison anchored early understates the drug measured late. The network has no closed loop, so this bias cannot be detected or corrected.

For the patisiran-versus-eplontersen comparison, the authors downgrade for imprecision, noting credible intervals that “include widely disparate conclusions.” The abstract nevertheless states that vutrisiran and patisiran “demonstrated improvements compared with diflunisal, eplontersen, and inotersen.” A comparison the authors flag as spanning opposite conclusions cannot at the same time establish a ranking. A standardized mean difference rescales range but does not reconcile different constructs; Samjoo et al. [5] found that even the two mNIS+7 versions weight their components differently, yet the analysis pools them.

Three entries in the baseline-characteristics table of Duarte et al. do not change the main finding but affect whether the trials are similar enough to be compared fairly. NEURO-TTR is listed as 57%/43% stage 1/2, whereas the trial reports 67%/33% [6]. The NEURO-TTRansform disease-stage entry reflects the polyneuropathy disability score, not Coutinho stage, mixing two scales across rows [2]. And the disease-duration mean of 46.8 months is not readily reconciled with the published trial, which reports medians from two different anchors [2].

The signal that gene silencers outperform stabilizers is large, consistent, and biologically grounded; these are robust findings, and we expect them to hold. The ranking among them is another matter: indirect, graded low by the authors themselves and downgraded for imprecision, absent on one primary outcome, and drawn from a network that cannot test its own consistency. A ranking the authors grade as low-certainty should not be presented as an established hierarchy. The choice among these agents awaits head-to-head trials.

Marco Luigetti: conceptualization, investigation, formal analysis, writing – review and editing. Nicholas Streicher: conceptualization, data curation, investigation, validation, formal analysis, project administration, writing – original draft, writing – review and editing. Michelle Mezei: conceptualization, data curation, validation, writing – review and editing, supervision.

The authors have nothing to report.

N.S. reports research funding from Alnylam Pharmaceuticals and has served as a paid speaker for AstraZeneca. M.M. has received consultant, speaker, and advisory board fees from Alnylam, Akcea-Sobi, AstraZeneca, and Pfizer, and research investigator support from Alnylam, AstraZeneca, and Ionis. M.L. has received honoraria and speaking fees from Akcea, Alnylam, AstraZeneca, and Pfizer, and travel grants from Akcea, Alnylam, Sobi, Pfizer, Kedrion, CSL Behring, and Grifols. AstraZeneca had no role in the content of this letter or in the decision to submit it.

This article is linked to Duarte, papers. To view this article, visit https://doi.org/10.1111/ene.70627.

Data sharing not applicable to this article as no datasets were generated or analysed during the current study.