While cardiovascular disease (CVD) risk stratification tools exist for use in the general population, they may not accurately estimate risk in persons living with HIV (PLWH). Examining the performance of CVD risk scores in PLWH requires large studies with comprehensive clinical data and well-validated outcomes.
We developed a state-of-the-art screening algorithm and central adjudication protocol for the validation of incident myocardial infarction (MI) in the CFAR Network of Integrated Clinical Systems (CNICS), which harmonizes comprehensive clinical data on PLWH in routine care at multiple US sites. Among PLWH enrolled between 1996-2014, we compared the performance of 3 CVD risk scores developed in the general population (Framingham, ATP-3, and 2013 ACC/AHA ASCVD) and one developed for use in PLWH (D:A:D) using area under the curve (AUC). The Universal Definition of MI classifies MI by type. Type 1 MI (T1MI) result spontaneously from atherosclerotic plaque instability, whereas type 2 MI (T2MI) occur secondary to oxygen demand/supply mismatch of any cause such as sepsis. We compared the AUC for risk scores for T1MI, T2MI, and all MIs combined. Beginning in 2007, CNICS patients completed clinical assessments every 4-6 months that included tobacco use. We repeated analyses among this subset to ensure smoking status was updated for those who quit or started smoking.
There were 243 incident MIs among 11,338 PLWH during a mean follow-up of 4.3 years. ASCVD had a significantly better AUC than other scores for all MI and for T2MI (Table) including the DAD AUC (p<0.001), and was not inferior to the other AUCs for T1MI. Our results were similar in the subset of PLWH with time-updated smoking status.
The large size, comprehensive clinical data and central adjudication of MI by type in CNICS allows for direct comparison of clinical risk scores in PLWH. Some variations across risk scores are to be expected given differences in the outcome (i.e. predicting CVD vs. MI). The addition of HIV-specific variables as in the DAD score did not improve discrimination compared with ASCVD, however inclusion of different HIV-specific measures may lead to improved discrimination and is planned in future analyses. ASCVD performed as well or better than other risk scores across all MI events and the superior performance in detecting T2MI is worthy of additional investigation.