QuantHealth preprint: the model estimated the event risk in VESALIUS-CV in advance, and the trial result was close

QuantHealth preprint: the model estimated the event risk in VESALIUS-CV in advance, and the trial result was close

On August 31, QuantHealth published a preprint on a computer simulation of VESALIUS-CV. A calculation has been available in the public OSF repository since August 11, 2025. The hazard ratio, which compares the rates at which first events accumulate in two groups, was 0.78. On June 25, 2026, ClinicalTrials.gov published the trial result: 0.75.

VESALIUS-CV was a randomized trial of evolocumab, a drug that lowers LDL, commonly called “bad” cholesterol. Over a median follow-up of 55.2 months, the first primary event occurred in 336 of 6 129 participants who received evolocumab and in 443 of 6 128 participants who received placebo.

In the prediction archived on OSF, an event was defined as cardiovascular death, myocardial infarction, or ischemic stroke. The registry lists the trial’s primary outcome as the first occurrence of one of three events: death from coronary heart disease, myocardial infarction, or ischemic stroke. The values 0.78 and 0.75 therefore refer to two overlapping composite endpoints.

In its preprint, QuantHealth described a model trained on deidentified insurance claims and electronic health records. A knowledge graph represented the proteins, diseases, and biological pathways associated with the drug. Using the protocol criteria, the model selected a virtual patient cohort from existing records and calculated two scenarios for each patient, one with evolocumab and one with placebo, to estimate the time to the first event.

Before applying the method to VESALIUS-CV, the authors evaluated it against 22 completed cardiovascular trials. Across 24 comparisons between two groups, the F1 score, a measure that combines prediction precision with the proportion of successful outcomes correctly identified, was 0.83. For VESALIUS-CV, the estimate was produced before the trial result was published. All authors were QuantHealth employees while conducting the study, which QuantHealth funded.