Development and Validation of Imaging-Free Myocardial Fibrosis Prediction Models, Association with Outcomes, and Sample Size Estimation for Phase 3 Trials.

Abstract

Background: Phase 3 trials testing whether pharmacologic interventions targeting myocardial fibrosis improve outcomes require myocardial fibrosis measurement that does not rely on tomographic imaging with intravenous contrast.

Methods: We developed and externally validated extracellular volume (ECV) prediction models incorporating readily available data (comorbidity and natriuretic peptide variables), excluding tomographic imaging variables. Associations between predicted ECV and incident outcomes (death or hospitalization for heart failure) were tested in survival analysis. We created various sample size estimates for a hypothetical therapeutic clinical trial testing an antifibrotic therapy using (1) predicted ECV, (2) measured ECV, or (3) no ECV.

Results: Multivariable models predicting ECV had reasonable discrimination (optimism corrected C-statistic for predicted ECV ≥27%, 0.78 [95% CI, 0.75-0.80] in the derivation cohort [n=1663] and 0.74 [95% CI, 0.71-0.76] in the validation cohort [n=1578]) and reasonable calibration. Predicted ECV associated with adverse outcomes in Cox regression models: ECV ≥27% (binary variable) hazard ratio 2.21 (95% CI, 1.84-2.66). For a hypothetical clinical trial with an inclusion criterion of ECV ≥27%, use of predicted ECV (with probability threshold of 0.69 and 80% specificity) compared with measured ECV would obviate the need to perform 3940 cardiac magnetic resonance scans, at the cost of an additional 3052 participants screened and 705 participants enrolled.

Conclusions: Predicted ECV (derived without tomographic imaging) associates with outcomes and efficiently identifies vulnerable patients who might benefit from treatment. Predicted ECV may foster the design of phase 3 trials targeting myocardial fibrosis with higher numbers of screened and enrolled participants, but with simplified eligibility criteria, avoiding the complexity of tomographic imaging.