Diagnosis of subclinical atherosclerosis in HIV-infected patients: higher accuracy of the D:A:D risk equation over Framingham and SCORE algorithms

Abstract

While the detection of subclinical atherosclerosis may provide an opportunity for the prevention of cardiovascular disease (CVD), which currently is a leading cause of death in HIV-infected subjects, its diagnosis is a clinical challenge. We aimed to compare the agreement and diagnostic performance of Framingham, SCORE and D:A:D equations for the recognition of subclinical atherosclerosis in HIV patients and to adjust the D:A:D equation using HIV and CVD variables. Atherosclerosis was evaluated in 203 HIV-infected individuals by measuring the carotid intima-media thickness (IMT). The CVD risk was calculated using the Framingham, SCORE and D:A:D risk equations. Framingham, SCORE and D:A:D equations showed a low agreement with the IMT (Kappa: 0.219, 0.298, 0.244, respectively; p = 0.743) and a moderate predictive performance, (area under the curve [AUC] = 0.686, 0.665 and 0.716, respectively; p = 0.048), with the D:A:D equation being the most accurate. Atherosclerosis was demonstrated in a significant proportion of subjects with low predicted CVD risk by all three algorithms (16.3%, 17.2%, 17.2%, respectively; p = 0.743). In patients with an estimated low CVD risk atherosclerosis was associated with older age (p = 0.012) and low CD4 counts (p = 0.021). A model was developed to adjust the D:A:D equation; a significant increase in accuracy was obtained when CD4 counts and low-grade albuminuria were included (AUC = 0.772; p < 0.001). The D:A:D equation overperforms Framingham and SCORE in HIV patients. However, all three equations underestimate the presence of subclinical atherosclerosis in this population. The accuracy of the D:A:D equation improves when CD4 counts and low-grade albuminuria are incorporated into the equation.