Premium
Level of agreement between frequently used cardiovascular risk calculators in people living with HIV
Author(s) -
Dhillon S,
Sabin CA,
Alagaratnam J,
Bagkeris E,
Post FA,
Boffito M,
Anderson J,
Vera J,
Williams I,
Johnson M,
Sachikonye M,
Babalis D,
Mallon PW,
Winston A
Publication year - 2019
Publication title -
hiv medicine
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.53
H-Index - 79
eISSN - 1468-1293
pISSN - 1464-2662
DOI - 10.1111/hiv.12731
Subject(s) - medicine , interquartile range , framingham risk score , kappa , cohen's kappa , risk assessment , demography , disease , statistics , linguistics , philosophy , mathematics , computer security , sociology , computer science
Objectives The aim of the study was to describe agreement between the QRISK2, Framingham and Data Collection on Adverse Events of Anti‐HIV Drugs (D:A:D) cardiovascular disease (CVD) risk calculators in a large UK study of people living with HIV (PLWH). Methods PLWH enrolled in the Pharmacokinetic and Clinical Observations in People over Fifty (POPPY) study without a prior CVD event were included in this study. QRISK2, Framingham CVD and the full and reduced D:A:D CVD scores were calculated; participants were stratified into ‘low’ (< 10%), ‘intermediate’ (10–20%) and ‘high’ (> 20%) categories for each. Agreement between scores was assessed using weighted kappas and Bland–Altman plots. Results The 730 included participants were predominantly male (636; 87.1%) and of white ethnicity (645; 88.5%), with a median age of 53 [interquartile range (IQR) 49–59] years. The median calculated 10‐year CVD risk was 11.9% (IQR 6.8–18.4%), 8.9% (IQR 4.6–15.0%), 8.5% (IQR 4.8–14.6%) and 6.9% (IQR 4.1–11.1%) when using the Framingham, QRISK2, and full and reduced D:A:D scores, respectively. Agreement between the different scores was generally moderate, with the highest level of agreement being between the Framingham and QRISK2 scores (weighted kappa = 0.65) but with most other kappa coefficients in the 0.50–0.60 range. Conclusions Estimates of predicted 10‐year CVD risk obtained with commonly used CVD risk prediction tools demonstrate, in general, only moderate agreement among PLWH in the UK. While further validation with clinical endpoints is required, our findings suggest that care should be taken when interpreting any score alone.