A cardiovascular risk-scoring system known as ASCVD appears to be a better predictor of myocardial infarction (MI) among people with HIV than other risk scores, including Framingham and D:A:D, according to new research. The findings, presented at CROI 2016 on Feb. 23, are likely far from the last word on the subject of risk score value. But, they may help refine HIV care providers' understanding and use of these risk-scoring tools to reduce their patients' heart risks.
"Cardiovascular disease risk scores are an important, even key driver of cardiovascular disease prevention in the general population," said Heidi Crane, M.D., M.P.H., of the University of Washington, the lead author and presenter of the study findings at CROI 2016. However, she added, "How this should be applied in HIV, where we have an increasingly aging cohort, is just not clear."
Crane's presentation detailed the most recent attempt to provide clarity: a study comparing the expected outcomes from four cardiovascular risk-scoring systems to observational findings from CNICS (short for CFAR Network of Integrated Clinical Systems) a massive, diverse cohort of HIV-positive people in the U.S. Five CNICS sites were utilized for this study, providing data on 10,832 individuals who experienced a total of 229 myocardial infarction (MI) events over four-plus years of follow-up, Crane said.
The four risk-scoring systems assessed in this study were:
Of these, only the D:A:D system was specifically developed for use within an HIV-positive population. Yet, when comparing expected versus observed rates of MI across tools, Crane and her colleagues found that the ASCVD score outperformed its peers in its level of discrimination (ability to codify individuals as belonging to lower- or higher-risk groups) and calibration (ability to accurately predict actual MI events).
Using Harrell's C-statistic to assess discrimination, Crane et al found that ASCVD's risk score had statistically significant superiority over the other three studied scores for all types of MI (0.74 for ASCVD versus 0.68 for FRS-CHD, 0.69 for ATP3 and 0.68 for D:A:D), including Type 1 (0.77 for ASCVD versus 0.73 for FRS-CHD, 0.74 for ATP3 and 0.73 for D:A:D), the type most commonly associated with coronary plaque.
Crane et al then used Hosmer-Lemeshow "goodness of fit" testing to depict a comparison of predicted and observed MI events. For all types of MI, plotted data revealed a tendency for D:A:D and ATP3 to heavily underpredict MIs among individuals who were deemed to have low to moderate risk, while FRS-CHD slightly underpredicted MIs among those it deemed a low risk and significantly overpredicted MIs among those it deemed moderate risk. ASCVD generally slightly underestimated actual risk, but overall offered a "much better-calibrated region," Crane said, particularly at the low end of the risk continuum where the need for statin treatment may be uncertain and the need for accurate risk prediction may thus be greater.
"Even the best of these has room for improvement," Crane admitted. However, "While the ASCVD was far from perfect and definitely had limitations, it definitely performed better than the other risk scores for use in patients with HIV, although questions clearly still remain about whether we can do better," she said.
Although the risk-assessment tools have considerable overlap in their assessment methods, they differ in the cardiovascular events they attempt to predict and in the variables they include in their calculations, making comparison additionally challenging. D:A:D, for instance, is the only tool solely intended to predict MI events (as opposed to predicting also, e.g., stroke incidence or death from coronary heart disease) -- and, because it is the only HIV-exclusive tool, it is also the only one that attempts to take antiretroviral use into account (albeit a generally older generation of antiretrovirals). The tools also differ in the extent to which they incorporate racial variables and individuals' use of antihypertensives. They sometimes significantly differ in the population of individuals whose experiences feed the data set: D:A:D is an exclusively European cohort, for instance, while ASCVD has the highest minimum age at entry (40 years old) of the U.S. cohorts.
Crane also noted that none of the tools take into account key HIV-specific laboratory values such as CD4+ cell count and HIV viral load, "which arguably may be more important predictors of MI risk," she said. "So questions remain about whether you could develop a better HIV risk score that would more accurately capture risk in these patients."
Despite the limitations of both the study and the MI risk assessment tools it assessed, Crane said she was surprised and pleased by the results. "The ASCVD performed better than my a priori expectation," she said during her presentation. "It performed really well, and sort of wiped the floor with several of the others in a way that was shocking and nice, given that I think most of us are probably using it."
Myles Helfand is the editorial director of TheBody.com and TheBodyPRO.com.
Follow Myles on Twitter: @MylesatTheBody.