Streamlining event ascertainment in cancer randomised controlled trials: accuracy of the SCORE-CVD-MI algorithm using health system data in a large prostate cancer trial
R Gonzalez-Manzanares, S Love, S Lessels, J Farrell, R E Langley, D Gilbert, N James, L C Brown, M Murray, M SydesAbstract
Background
Cardiovascular safety is critical in oncology trials. Health system data (HSD) can facilitate event detection while minimising site burden.
Purpose
We aimed to evaluate the feasibility of ascertaining myocardial infarction (MI) in a prostate cancer clinical trial using HSD and a recently developed phenotyping algorithm.
Methods
Participants from a large prostate cancer trial were linked to Hospital Episode Statistics – Admitted Patient Care (HES-APC), which contains all NHS inpatient admissions in England. MI events were identified from HES-APC using two approaches: conventional code lists (cHES) and the SCORE-CVD-MI phenotyping algorithm (SCVD). Agreement with trial-reported MI was assessed.
Results
A total of 6,901 trial participants with 44,405 HES-APC episodes were included. Over a median follow-up of 21.2 months (IQR 5.7-65.9), 91 MIs were reported in 90 (1.3%) trial participants. SCVD identified 84 (92.3%) of these events, and cHES 83 (91.2%). Agreement between SCVD and the trial was moderate-to-strong (kappa: 0.79, 95% CI: 0.77 to 0.82), and higher than between cHES and the trial (kappa: 0.72, 95% CI: 0.70 to 0.75) (p=0.001). Sensitivity and specificity of SCVD were 0.99 (95% Credible Interval 0.99-0.99) and 0.95 (95% Credible Interval 0.88-0.99). Time-to-first-event analysis were comparable between SCVD and trial [HRpatient-strata (95% CI): 1.23 (0.93-1.64), p = 0.148].
Conclusions
Identification of MI admissions using HES-APC was feasible and accurate in a prostate cancer trial. The SCVD phenotyping algorithm outperformed the conventional approach, offering a promising and reproducible method for cardiovascular safety monitoring in oncology trials.Figure 1 Figure 2