Mitiperstat altered 28 plasma proteins and was projected to reduce all-cause mortality by 28% (HR 0.72, 95% CI 0.53-0.98, p=0.037) using proteomics-based digital twins.
Does a simulated treatment effect based on mitiperstat-induced proteomic changes predict improved long-term clinical outcomes in patients with HFrEF?
A digital twin approach using proteomic signatures from a phase II trial projected a significant reduction in all-cause mortality for mitiperstat in HFrEF patients, demonstrating a novel method for phase II/III translation.
Abstract Background The development of new HF therapeutics critically depends on early-phase clinical trials. Surrogate markers in HF trials including biomarkers, imaging or self-reporting, are often used as interim endpoints to assess efficacy and predict long-term clinical outcomes. However, there is a lack of consistent association between phase II results and morbidity and mortality rates in phase III trials. Modern analytical methods such as proteomics provide more comprehensive information regarding changes induced by a therapeutic intervention. These methods potentially provide better surrogate markers in early HF trials. MYSTERY-HF (Myeloperoxidase inhibition in patients with ischemic or non-ischemic cardiomyopathy and heart failure with reduced ejection fraction (HFrEF), EudraCT: 2022-003797-23) was a phase II multicenter randomized, double-blind placebo-controlled trial: Patients diagnosed for HFrEF with an LVEF ≤40% were randomized 1:1 to receive either the MPO inhibitor mitiperstat or matching placebo for 12 weeks. Proteomic changes measured with the Olink Explore 3072 platform were an exploratory endpoint. MyoVasc is a large prospective cohort study that uses deep phenotyping and long-term follow up to analyze the development and progression of HF. Purpose To evaluate a new approach using a protein signature as surrogate marker for drug treatment in a phase II HF trial. Methods Proteomic changes induced by mitiperstat in MYSTERY-HF were compared with and translated to MyoVasc. The proteome in MyoVasc was measured using the Olink Explore HT platform. The creation of "digital twins" derived from MyoVasc enabled us to estimate long-term treatment effects of mitiperstat on symptomatic HF patients (see A). For this, mitiperstat treatment was mimicked by altering protein levels in MyoVasc in accordance with the average treatment effects on proteins observed in MYSTERY-HF. The projected change in predicted outcome between mitiperstat and placebo was estimated using G-computation with 1000 bootstrap runs to estimate 95% confidence intervals and p-values. Results Mitiperstat induced a significant change in expression of 28 plasma proteins (see B). After adjusting for sex, age and various cardiovascular diseases, translation of this proteome pattern into the comprehensively phenotyped MyoVasc cohort projected a 28% lower all-cause mortality (average treatment effect (ATE(HR) = 0.72, 95% CI: 0.53;0.98, p = 0.037), see C+D). No significant reduction was predicted by the simulated treatment for cardiac death, HF hospitalization and worsening HF. Conclusions We integrated the proteomics datasets from a phase II trial into a rigorously characterized, much larger cohort of similar patients with available long-term follow-up. Using this -omics approach to create digital twins has the potential to supersede classical phase II surrogate markers and thereby reduce the risk for unsuccessful phase II/III translation of clinical trials.
Braumann et al. (2025) studied this question. Mitiperstat altered 28 plasma proteins and was projected to reduce all-cause mortality by 28% (HR 0.72, 95% CI 0.53-0.98, p=0.037) using proteomics-based digital twins.