Changes in Atherosclerotic Cardiovascular Disease Risk Scores in a Predominantly Black Cohort with HIV and Associated Comorbidities: A Preliminary Study

According to the World Health Organization, HIV/AIDS and ischemic heart disease are each among the leading causes of disease burden [1]. In people with HIV (PWH), the causes of mortality have shifted from predominantly AIDS-defining illnesses to non-AIDS conditions, most notably atherosclerotic cardiovascular diseases (ASCVDs) for which a twofold higher rate of acute myocardial infarction (MI) is observed compared to uninfected people [2‒5].

While the increased risk of cardiovascular disease (CVD) in PWH is believed to be multifactorial, spanning traditional CVD and HIV-related risk factors, as well as effects of antiretroviral therapy (ART) [6‒8], these data have lacked sufficient numbers of vulnerable populations, especially African Americans, women, those who use illegal substances, and associated comorbidities such as hepatitis C [3, 8‒11]. In addition, several previous studies did not adjust for the comprehensive set of established CVD risk factors, including cigarette smoking and family history [10‒12]. Therefore, there remains a need for further studies to assess a more complete set of traditional and nontraditional CVD risk factors in PWH in various demographic groups.

Importantly, there is a paucity of data on change in CVD risk over time, especially among PWH. One prospective follow-up from the Jackson Heart Study conducted in exclusively black people without HIV (PWoH) compared how changes in modifiable risk factors contributed to change in ASCVD risk compared to aging [13]. Another study evaluated the temporal trends of cardiovascular risk factors in PWoH undergoing percutaneous coronary intervention (PCI) and assessed the Framingham Risk Score (FRS) over time after PCI [14]. Among PWH, one study specifically assessed the change in ASCVD risk scores 1 year after switching from tenofovir disoproxil fumarate to tenofovir alafenamide [15]. Recently, a longitudinal cohort from South Africa assessed FRS over 36 months in PWH and PWoH and found that FRS increased in all participants, but that PWH had lower CVD risk than PWoH over the 36-month follow-up [16]. Taken together, there remains a need for continued assessment of the risk score to elucidate how it works in biologically and socially diverse populations and to investigate how nontraditional CVD risk factors not included in current risk scores are driving CVD risk in PWH.

The goal of our study was to evaluate CVD risk using the ASCVD risk score derived from the Pooled Cohort Equation [17], over a 4-year period among high-risk, vulnerable, and underrepresented PWH. The purpose of this study is twofold. We aimed to (1) investigate the change in CVD risk by evaluating the change in ASCVD risk scores over the study period and (2) identify clinical and sociodemographic factors associated with the change in ASCVD risk scores or high-risk scores to add important data to the current literature about risk stratification of high-risk, vulnerable PWH and optimize CVD prevention efforts.

We conducted a retrospective analysis of PWH, between 40 and 75 years old, seen at the Evelyn Jordan Center with at least two routine HIV visits. To accomplish this, we reviewed the medical records of 604 PWH treated at the Evelyn Jordan Center (EJC), an HIV clinic associated with the University of Maryland Medical Center (UMMC), from June 1, 2008, to May 2012. Eligible participants ranged between 40 and 75 years of age, began receiving care at the EJC on June 1, 2008, and had at least two routine visits by May 2012. The Institutional Review Board of the University of Maryland, Baltimore, approved this study (HM-HP-00055406) with a waiver of informed consent because there were no human interactions with the study team.

Data were collected from outpatient routine HIV follow-up visits at least 3 months apart, with a maximum of four visits per year. Abstractors reviewed all notes for every office visit for 604 PWH and determined which ones qualified as routine HIV visits, defined as follow-up for HIV, and included physical exams and assessment of health maintenance. Each abstractor collected data from all eligible visits from medical charts housed in four different electronic record systems (Medicos, Horizon Patient Folder, PowerChart, and EPIC). Diagnoses were ascertained following a review of all clinic and hospital documentation. Major depressive disorder was diagnosed either from mental health provider notes or treatment with appropriate psychiatric medication. Coronary heart disease (CHD) was defined as history of an acute coronary syndrome (e.g., unstable angina), MI, clinically or arteriographically proven coronary artery disease, and ischemic cardiomyopathy. Cerebrovascular disease was defined as a stroke or transient ischemic attack. Cardiovascular disease was defined as either CHD or cerebrovascular disease. Lab values and ranges were obtained from medical charts and based on assays used by the UMMC reference laboratory, LabCorp. All data were entered into a standardized chart abstraction template.

We calculated the ASCVD risk score using the Pooled Cohort Equation [17], which calculates the 10-year risk for developing CVD. The components needed to calculate the risk score are age, sex, race, smoking status, total cholesterol, high-density lipoprotein cholesterol, diabetes, SBP, and treatment for blood pressure. PWH were eligible for inclusion in analyses if they were 40–75 years old; had a minimum of two visits at least 1 year apart that had all laboratory, demographic, and behavioral data required for the ASCVD risk score calculation; and had no prior history of ASCVD. Using the first eligible visit, each consecutive visit with a risk score was included if it was at least 1 year after the preceding visit to EJC. Of the 604 people with 3,685 visits, 187 people with 431 visits during the study period were included in the final sample (online suppl. Fig. 1; for all online suppl. material, see https://doi.org/10.1159/000540526).

Univariate analyses examined the distribution of clinical, demographic, and behavioral characteristics of people at baseline. These included but were not limited to variables that comprised the risk score, use of aspirin and statins, illicit drug use, alcohol use, antiretroviral use, CD4 cell count, and HIV viral load. We categorized risk scores into the following categories: <7.5%, 7.5% to <20%, and ≥20% to represent low, intermediate, and high risk based on American College of Cardiology risk classifications for CVD risk [18]. Exploratory analyses examined whether people changed risk categories throughout the study period. Unadjusted analyses used the Wilcoxon rank sum, ANOVA, and repeated measures ANOVA tests to examine the relationship between risk scores over time and potential confounders. Variables with a p value ≤0.2 from the unadjusted analyses or determined to be clinically relevant (BMI, aspirin, statin use) were eligible for inclusion in the final models [19]. Online supplementary Table 1 lists all variables assessed for potential confounding. Given the skewed distribution of the risk scores, we examined plots of raw and Pearson scaled residuals and predicted values to assess the assumptions of linearity and normality. Final adjusted analyses evaluated the relationship between time (as a continuous variable) and ASCVD risk score calculated for each person, using linear mixed models with the compound symmetry covariance structure. Given the variation in baseline risk scores, we evaluated the utility of a random intercept model; however, this did not improve the final model fit. We assessed model fit by comparing the Akaike Information Criterion (AIC) statistic of the nested models. Variables were added one at a time to the model, and differences in the AIC were calculated. A smaller AIC indicated a better model fit for the data. Variables in the final model were determined based on a p value ≤0.05 in the multivariable model, clinical relevance, and impact on model fit. Though statin therapy and aspirin use were clinically important variables, they were not included in the final model due to small numbers of participants receiving either medication. Finally, we conducted sensitivity analyses to determine if participants 40–75 years old who were excluded from the analyses differed from those with eligible visit data. All analyses were conducted using SAS (v.9.4, SAS Institute Inc., Cary, NC, USA). Testing was two sided and done at the 0.05 level of significance.

Within this PWH cohort were 108 males and 79 females (43%); the vast majority (89%, n = 166) were African American (Table 1). A relatively high percentage were active cigarette smokers (55%), current or past users of illicit drugs (74%), and were unemployed (67%). Most (59%) were on antiretrovirals, many of whom were on regimens containing a protease inhibitor (n = 76, 40%). The median CD4 cell count was 241, and nine people had viral suppression, though 44% of the study population was missing viral load data. Over half had a diagnosis of viral hepatitis (B or C) with 67 (35.8%) being dually infected with hepatitis B and C. Median follow-up was 1.78 years, with approximately 50% (n = 94) having a maximum of 2 years between their first and last visit.

The median ASCVD risk score at baseline was 6% (IQR 3, 11), and most people (n = 105, 56.2%) had a risk score <7.5%. Statin (n = 18) and aspirin (n = 17) use was low in our study population during the study period, with approximately 7% being prescribed either medication.

Exploratory analyses examined the change in ASCVD risk score category over time. Longitudinally, the mean number of eligible visits per person over time was 2.3. Of the 187 PWH, 34 (18.2%) had a change in risk score category by their last visit; 19 (10.2%) had an increase. In the final adjusted model, there was no significant change in ASCVD mean risk score over time (Table 2). We found that with each year, the change in risk score was 0.18% (95% CI: −0.29, 0.64; p = 0.46). However, there was a significant relationship between the ASCVD risk score and diagnosis of viral hepatitis and family history of CVD. Specifically, those with viral hepatitis C had significantly higher ASCVD risk scores (7.34%; 95% CI: 2.59, 12.09, p < 0.003) compared to non-affected. Moreover, people with combined hepatitis B and C experienced higher ASCVD risk score of borderline statistical significance (p = 0.07). We examined the interaction between each mono-infected and dually infected hepatitis group and time and found that ASCVD risk score remained relatively unchanged in all hepatitis groups over time (online suppl. Fig. 2). People with a family history of CVD had a mean ASCVD risk score that was 3.77% (p = 0.05) higher than those without family history of CVD. Clinical and demographic factors across people with hepatitis B, C, and B/C dual infection were comparable across the study population, except for higher rates of alcohol and illicit drug use in people with hepatitis C and B/C dual infection (online suppl. Table 2). Sensitivity analysis among those with risk score data, comparing clinical or demographic differences between those included and excluded in the final analyses, showed that those included in the analyses were comparable across all factors to those excluded, including no significant differences in diagnoses of viral hepatitis or family history of CVD (online suppl. Table 3). Additional sensitivity analysis comparing those who did not have risk score data to those included in the final analysis found that groups were comparable across most factors, including family history of CVD. However, 4.5% of those excluded from the analysis had hepatitis C compared to 16% of our final sample (online suppl. Table 4).

Our study found that there was no significant change in the mean ASCVD risk score over 4 years in PWH with at least two routine visits spaced at least 1 year apart. However, we found a significant relationship between ASCVD risk scores and diagnosis of viral hepatitis. People with hepatitis C had a significantly higher ASCVD risk score than those without hepatitis B or C.

To our knowledge, there have been only a few studies that evaluated changes in CVD risk over time, and even fewer in PWH. Bress et al. [13] investigated the change in ASCVD risk after 3 years in PWoH and compared how aging versus modifiable risk factors contributed to change in risk. In this exclusively black cohort, they found that in addition to age, increases in SBP and initiation of antihypertensive medication contributed significantly to the development of high ASCVD risk over time [13]. Of note, this study did not evaluate illicit drug use as a risk factor for ASCVD. In contrast, in our study of predominantly black PWH with a mean age of 49 years, we found no change in risk and no significant association between high SBP and an increased ASCVD risk. Our smaller sample size, longer study duration, and exclusively PWH cohort may partially account for these differences. Schafer et al. [15] also found an increase in ASCVD risk score over time, though their study specifically evaluated change in PWH 1 year after switching from tenofovir disoproxil fumarate to tenofovir alafenamide. It included fewer females than our study and did not evaluate factors including smoking history, family history of CVD, and illicit drug use [15]. Volpe et al. [20] investigated potential risk factors associated with progression of internal carotid artery intima-media thickness and coronary artery calcium over a 6-year period in PWH and found that both carotid artery intima-media thickness and coronary artery calcium progressed during the study period. The association between FRS and changes in surrogate markers was evaluated in this study; however, the change in FRS over the study period was not [20].

Unlike the previous studies, Lee et al. [14] evaluated the change in CVD risk factors over time using FRS in PWoH undergoing PCI and found that the FRS risk score declined over time in these patients. Important limitations to this study were the lack of reporting on clinical outcomes, such as cardiac events or mortality, and the use of FRS to measure change in risk, which has been replaced by the ASCVD risk score [14]. In the previous three studies, a change in CVD risk was seen over time in patients with CVD risk factors but in differing directions, with Schafer et al. [15] and Volpe et al. [20] observing an increase in risk and Lee et al. [14] observing a decrease in measured risk. Our data found no change in risk over the 4-year period. These differential results may have been due to (1) differences in study population (e.g., sociodemographic groups and traditional CVD risk factors), (2) differences in study design (e.g., follow-up time intervals and related study objectives), and (3) different assessment measures of CVD risk. Regardless of these differences, our results warrant reconsideration of the dogma that ASCVD risk increases over time due to age. Our findings from this study suggest there may be other factors affecting CVD risk that are underappreciated in PWH, thereby underscoring the potential limitation of using ASCVD risk score as an accurate measure of CVD risk in this population [21].

In addition, numerous studies have investigated the role of hepatitis C virus (HCV) infection on the risk of developing CVD. In a previous cross-sectional analysis, we found that the association between HCV infection and ASCVD risk was significant [22], and this relationship remains in this longitudinal analysis as well. Chew et al. [23] also found that HCV infection was associated with a higher risk of an ASCVD event; however, while their study assessed cardiovascular risk in HCV mono-infected men, our cohort included both males and females who were infected with HIV and/or HCV, offering more insight into the specific risk of coinfection. Similarly, the Veterans Aging Cohort Study found that among 8,579 male PWH, those who were coinfected with HCV had significantly higher risk of CHD compared to mono-infected male PWH [24], consistent with the findings of our study. However, data have been conflicting across studies examining the role of HCV coinfection among PWH, with Williams-Nguyen et al. [25] finding a twofold greater risk of type II MI, but not type I MI in people with HCV/HIV coinfection, while Lang et al. [26] found no significant association between HCV/HIV coinfection and increased risk of type I MI (they did not assess type II MI). A meta-analysis of four cohort studies consisting of 33,723 participants demonstrated that individuals with HIV/HCV coinfection had an increased risk compared to those with HIV mono-infection [27]. There are several possible reasons for the potential increased ASCVD risk due to HCV infection alone and with HIV coinfection such as increased burden of systemic inflammation [28], but the underlying mechanisms remain poorly understood.

Analogous to the Data collection on Adverse events of Anti-HIV Drugs (D:A:D) cohort study which found no association between hepatitis B virus (HBV) or HCV coinfection and the development of MI [22], our previous cross-sectional study found no significant association between HBV coinfection and elevated CVD risk [29]. Additionally, while Riveiro-Barciela et al. [30] found chronic hepatitis B to be an independent risk factor for carotid plaques and subclinical atherosclerosis, other studies found no significant association between HBV infection and increased ASCVD risk in HBV mono-infected individuals [31‒33]. Our current longitudinal study found no statistically significant association between ASCVD risk and HBV/HCV dual infection or HBV infection among PWH but a trend toward significance for HBV/HCV dual infection, raising the notion that perhaps HCV, not HBV, may contribute greater to CVD risk among PWH.

Our study had several limitations. Our modest sample size and follow-up time along with a relatively young study population may have limited our ability to find a significant change in ASCVD risk over time and may have prevented us from detecting other significant associations. However, given that Schafer et al. [15] demonstrated ASCVD risk score change in a smaller sample size with a shorter follow-up period, our study has likely identified important risk factors associated with a high ACSVD score that are under-recognized. In addition, we were unable to evaluate the relationship between changes in ART medication regimens and changes in ASCVD risk over time because very few people changed their ART regimens during the study period. There have also been significant changes to ART regimens and ASCVD treatments in the time elapsed since our study completion, and thus those represented in our study may differ from those currently in widespread use. Furthermore, current ART and ASCVD treatment guidelines allow PWH to receive these medications earlier than in our study period, and thus our population may not be reflective of current PWH in many treatment settings. However, recent research reports mixed findings on the association between ARTs and increased ASCVD risk. Reassuringly, Verstraeten et al. [16] found that there was no evidence that ART was associated with a change in risk score and Rein et al. [34] from the HIV-CAUSAL Collaboration and the Antiretroviral Therapy Cohort Collaboration estimated that INSTI use did not result in a clinically meaningful increase of CVD in either ART-naïve or ART-experienced PWH. These findings highlight a need to focus on other nontraditional CVD risk factors regardless of the types of medications that PWH are receiving. Importantly, we found that our risk score did not change over time and that hepatitis C infection and a family history of ASCVD are associated with an increased risk score and clinically deserve attention regardless of the types of medications PWH may be receiving. Furthermore, the time interval between person visits was irregular, and though we accounted for this in our model, our results may have been driven toward the null since many people had shorter total follow-up times. Additionally, there may be unmeasured confounding due to factors related to length of follow-up time and missing variables for those included in the study. A significant proportion (44%) of our participants did not have viral load measurements, which is consistent with data from the National Centers for Disease Control and Prevention which showed that in most states and among all races and ethnicities, less than 60% of PWH were retained in care and had regular viral load measurements obtained [35]. Other studies have examined the socioeconomic factors that may contribute to retention in medical care or lack thereof for PWH [36, 37]. While this scenario reflects real-world challenges of an inner-city clinic (vs. research) population, the missing measurements may have skewed viral suppression rates of the study population. Finally, people who frequently missed clinic visits did not have enough data to calculate ASCVD risk scores and were excluded from analysis; it is possible they may have higher ASCVD risk. Sensitivity analyses revealed that there was no difference in prevalence of family history of CVD in those without risk score data compared to those included in the final analysis, but there was significantly less prevalence of hepatitis C infection in those who did not have risk score data compared to our final sample. Given our findings of HCV infection and ASCVD risk score, it is possible that exclusion of these participants interjected bias into our study results. Additional sensitivity analysis among those with risk score data found no significant clinical or demographic differences between those included and excluded in the final analyses, suggesting that among participants with eligible data, any potential bias was not significant and therefore unlikely to adversely impact study findings.

Our study had several important strengths. We included many groups that have been underrepresented in prior research and are particularly vulnerable to poor health outcomes including black patients, females, and past or current illicit drug users. In addition, our study focused on change in CVD risk using ASCVD risk score, a risk stratification tool that serves as the current gold standard and is easily calculated by clinicians with readily available patient history and laboratory measurements. Finally, we identified a potential role for both HBV and HCV coinfection among PWH, not previously noted, suggesting that viral hepatitis in addition to HIV and HCV may further compound ASCVD risk in PWH.

In this study, we found that there was no change in ASCVD risk score over a 4-year period in PWH receiving care in an inner-city clinic. Though our results differed from findings in prior studies, they represent a lack of change in risk in a relatively young cohort over a 4-year period that is important to note. We present results from a predominantly black cohort including large numbers of females and commonly recognized underserved, vulnerable populations (such as injecting drug users and men who have sex with men) fueling new infections in the current HIV epidemic. We have identified associations between ASCVD risk score and viral hepatitis that are underappreciated and may be potentially additive or synergistic with HIV infection in contribution to CVD risk. HCV and potentially HBV infections should be considered when ascertaining ASCVD risk as neither (along with HIV) are currently unaccounted for in the ASCVD risk score calculation. Importantly, HCV infection is a substantial public health problem with the numbers affected outnumbering PWH and for which there is cure, underscoring the unmet need to screen for this infection and expand HCV treatment as a potential intervention to address potentially increased CVD risk [38‒41]. Given that there are few studies looking at the change in ASCVD risk score over time in both PWH and PWoH, this longitudinal relationship warrants further investigation in larger studies that include high-risk demographic groups such as ours to allow more dynamic, accurate assessment and intervention of CVD risk prior to onset of clinical events. These investigations should include the potential influence of different ART regimens and hepatitis coinfection on ASCVD risk score over time.

We thank Drs. Pankti Patel, Ahmed Babiker, Alexandra Ward, Rawan Faramand, Ruth Adekunle, and Andrew Hickey for their assistance with abstraction of chart data.

The study protocol was reviewed and approved by the University of Maryland Baltimore Institutional Review Board (HP00074189). All recruitment and enrollment procedures began after IRB approval was obtained for study conduct. The Institutional Review Board of the University of Maryland, Baltimore, approved this study (HM-HP-00055406) with a waiver of written informed consent because there were no human interactions with the study team.

The authors have no conflicts of interest to declare.

The work was supported by the National Heart, Lung, and Blood Institute of the National Institutes of Health, 5K23HL133358 of Dr. Bagchi. The funder had no role in the design, data collection, data analysis, and reporting of this study.

Study design: S.A.B.B., E.Z., A.W., and M.M. Database creation: S.A.B.B., L.F., and M.M. Data management: S.A.B.B. Statistical analysis: S.A.B.B. and Q.B. Drafting of the manuscript: S.A.B.B., Q.B., E.Z., and A.W. Data collection: E.Z. and A.W. Statistical analyses: J.S. Editing of the manuscript: S.A.B.B., Q.B., E.Z., J.S., L.F., M.M., and A.W. Database creation: L.F. and M.M. Participation in all aspects of the study and overall project management and supervision for the project: S.B.

Shana A.B. Burrowes, Erin Zisman, Quoc Bui, and Angela Wu contributed equally to this work.

Data are not publicly available due to containing information that could compromise the privacy of study participants but may be made available upon reasonable request to the corresponding author.