Leukocyte Count and Coronary Artery Disease Events in People With Human Immunodeficiency Virus: A Longitudinal Study

Abstract Background People with human immunodeficiency virus (HIV; PWH) have increased cardiovascular risk. Higher leukocyte count has been associated with coronary artery disease (CAD) events in the general population. It is unknown whether the leukocyte-CAD association also applies to PWH. Methods In a case-control study nested within the Swiss HIV Cohort Study, we obtained uni- and multivariable odds ratios (OR) for CAD events, based on traditional and HIV-related CAD risk factors, leukocyte count, and confounders previously associated with leukocyte count. Results We included 536 cases with a first CAD event (2000–2021; median age, 56 years; 87% male; 84% with suppressed HIV RNA) and 1464 event-free controls. Cases had higher latest leukocyte count before CAD event than controls (median [interquartile range], 6495 [5300–7995] vs 5900 [4910–7200]; P < .01), but leukocytosis (>11 000/µL) was uncommon (4.3% vs 2.1%; P = .01). In the highest versus lowest leukocyte quintile at latest time point before CAD event, participants had univariable CAD-OR = 2.27 (95% confidence interval, 1.63–3.15) and multivariable adjusted CAD-OR = 1.59 (1.09–2.30). For comparison, univariable CAD-OR for dyslipidemia, diabetes, and recent abacavir exposure were 1.58 (1.29–1.93), 2.19 (1.59–3.03), and 1.73 (1.37–2.17), respectively. Smoking and, to a lesser degree, alcohol and ethnicity attenuated the leukocyte-CAD association. Leukocytes measured up to 8 years before the event were significantly associated with CAD events. Conclusions PWH in Switzerland with higher leukocyte counts have an independently increased risk of CAD events, to a degree similar to traditional and HIV-related risk factors.

People with human immunodeficiency virus (HIV; PWH), have an increased risk for coronary artery disease (CAD) events compared with the general population [1,2]. CAD risk in PWH is related to traditional CAD risk factors, HIV-related factors including chronic inflammation [3,4], immunosuppression [5,6], potential deleterious effects of certain antiretroviral therapy (ART) agents [7,8], and individual genetic background [9]. An increased CAD risk may persist in PWH with suppressed HIV viremia [1,2]. This suggests a role for low-level inflammation and immune activation in the pathogenesis of CAD in PWH and has generated considerable interest in inflammatory biomarkers for CAD event prediction in PWH [4,10,11].
Leukocytes are implicated in the pathogenesis of atherosclerosis, and ever since the 1980s, studies in the general population have shown leukocyte count in the peripheral blood to be an independent risk factor for CAD events [12][13][14][15]. Whether blood leukocytes are associated with CAD events in PWH has not been verified. Therefore, the aim of this report is to assess an independent association of leukocyte count with CAD events in participants of the Swiss HIV Cohort Study (SHCS), analyzed in the context of traditional and HIV-related CAD risk factors. We also considered multiple factors that may influence leukocytes, including ethnicity [16], smoking [17], infections, and alcohol intake [18].

Study Population
We included PWH enrolled in the SHCS (http://www.shcs.ch [19]), an observational study that has prospectively enrolled PWH since 1988, and has captured rich cardiovascular, metabolic, genetic, and other data since 1999. Participants provided written informed consent. The study was approved by the local ethics committees. Cases had a first CAD event and controls were CAD event-free during the study period (1 January 2000-31 October 2021).

CAD Events
CAD events were defined per the Data Collection on Adverse events of Anti-HIV Drugs study and the World Health Organization's Monitoring Trends and Determinants in Cardiovascular Disease Project [20], as we have previously published [9,21]. CAD events included myocardial infarction, coronary angioplasty/stenting, coronary artery bypass grafting, and fatal cases (confirmed at autopsy or ascertained by the treating HIV physician as sudden death with no other likely cause plus evidence of CAD before death).

Case-Control Matching
As in our previous CAD case-control studies [9,21], we used incidence density sampling [22], aiming to select 1 to 3 eventfree controls for each case. We used risk-set sampling [23 (ie, we matched controls at the CAD event date of the corresponding cases [matching date] on similar observation duration, and their observation period was during similar calendar periods to account for differences in ART [with different CAD risk associations] [8,24]) in use during different periods and other differences. Matching criteria were sex, age ± 4 years, and date of SHCS registration ± 4 years. Observation time started at SHCS registration; observation ended for cases at the matching date (CAD event date) and for controls ended at the first regular SHCS follow-up visit after the matching date, respectively.

Power Calculation
To capture odds ratios of ≥1.6, we would need 255 cases and 2 controls per case [30], assuming an exposure correlation between pairs in the case-control set of 0.2 [30].

Leukocyte Count
The SHCS database routinely includes total leukocytes, total lymphocytes, CD4, and CD8 counts. For the main analysis, we compared latest leukocyte count before the matching date in cases and controls and. In addition, we considered leukocyte count at increasing intervals before the matching date. In exploratory analyses, we obtained neutrophils, eosinophils, and the neutrophil-lymphocyte ratio retrospectively for participants at University Hospital Zurich, where approximately 20% of SHCS participants are followed.

Clinical CAD Risk Factors
Covariables were defined a priori based on their CAD association in the general population, as reported previously [9,21,25], and were ascertained at the latest SHCS visit before the matching date except for CD4 nadir (lowest CD4 value during the study period). Covariables included age (per 10 years older, added to detect any residual effect of suboptimal matching, as we have done previously [21]), family history of CAD, smoking, diabetes mellitus, hypertension, and dyslipidemia (total cholesterol >6.2 mmol/L or high-density lipoprotein < 1 mmol/L [men] and <1.2 mmol/L [women] or use of lipid-lowering drugs [25]). HIV-related covariables included HIV RNA < or ≥50 copies/mL, CD4 nadir, and ART exposures until the matching date, based on their CAD association in the Data Collection on Adverse events of Anti-HIV Drugs study [8,24], including recent (past 6 months) abacavir, didanosine, and integrase inhibitors; and cumulative (>1 year) exposure to lopinavir, indinavir, boosted darunavir, and stavudine [9]; hepatitis C [26]; and cytomegalovirus seropositivity [27].

Potential Confounding Variables Associated With Leukocyte Count
These were defined a priori, based on reported associations in the general population. We considered both current smoking (vs past/never [28]) and daily cigarettes smoked (never, not currently, ≤5/d, 6-20/d, >20/d, unknown [29]); ethnicity (White/Black/Hispanic/Asian) [16]; and alcohol (none/mild vs moderate/heavy; defined in the SHCS until 2012 as </≥40 g [men], </≥20 g [women]), and using the Alcohol Use Disorders Identification Test-C questionnaire beginning in 2013 (</≥4 points [men], </≥2 points [women]; hepatitis.va.gov/alcohol/treatment/audit-c.asp#S1X) [18]. We also tested each variable in the CAD event model for a potential interaction with leukocytes (Supplementary Methods, Supplementary Table 1). We did not analyze corticosteroid use and non-HIV inflammatory conditions because these were recorded before the event in only 8 cases/36 controls and in 3 cases/5 controls, respectively, and because of insufficient available details (eg, specific diagnoses, date, corticosteroid duration/dose).

Infection Episodes
Because infections may influence leukocytes, we assessed nonopportunistic infections (recorded in the SHCS since 2017, defined as leading to hospitalization or antibiotic use for ≥5 days) and opportunistic infections in the year before matching date in cases and controls.

Sensitivity Analyses
To test the robustness of the leukocyte-CAD association; (1) we replaced all risk factors by the 10-year Framingham risk score (FRS) for CAD or (2) by FRS risk category (<10% vs ≥ 10% risk); (3) analysis restricted to participants with suppressed HIV RNA at matching date; and (4) after adding the latest estimated glomerular filtration rate (eGFR) before CAD event to the model (note that kidney function is available in the SHCS after 1 January 2002).

Statistical Analyses
Characteristics of cases and controls were compared using Fisher exact test (categorical variables) and Wilcoxon ranksum test (continuous variables). Univariable, bivariable, and multivariable conditional logistic regression analyses were used to estimate associations of the different risk factors with CAD and their interactions. We decided a priori to stratify leukocyte counts into quintiles for better visualization of potentially nonlinear associations with CAD events. Variables were entered into the multivariable model if their association in the univariable model had a P level < .2. Model fit and interactions were analyzed using Akaike and Bayesian information criteria and likelihood ratio tests. The effect of potential confounders on the leukocyte-CAD association was tested on a 1:1 basis (bivariable models including interaction terms). Trajectories of total leukocytes, leukocyte subtypes, and smoking over the past 15 years were created using local polynomial smoothing with the Epanechnikov kernel. We used Stata/SE 17.0 (StataCorp, College Station, TX, USA).

Latest Leukocyte Count: Observed Data
Median time from the latest leukocyte measurement to CAD event (matching date) was 56 (IQR, 30-94) days in cases and 60 (IQR, 29-91) days in controls. Latest median leukocyte count before the matching date was higher in cases than controls (P < .01; Table 1). Leukocytosis (>11 000/µL) was uncommon but more frequent in cases than controls (4.3% vs 2.1%; P = .01). Figure 2 shows the range of leukocytes in each leukocyte quintile and how the number of cases increases and the number of controls decreases in the higher leukocyte quintiles. Supplementary Table 2 shows leukocytes for cases and controls in each quintile.

Leukocyte Count and CAD Events: Univariable Model
In the latest sample before a CAD event, leukocyte count was associated with CAD events (per 1000 leukocytes higher,   Figure 4 and Supplementary Table 3.

Longitudinal Leukocyte Counts and CAD Events: Univariable Model
Leukocyte count (fifth vs first quintile) remained significantly associated with CAD events when measured at year −1 (CAD-OR =

Leukocyte Count and CAD Events: Potential Confounders
Median leukocyte count was higher in cases than controls in most confounder categories (Figure 4). In individual 1:1 bivariable analyses ( Longitudinal observed neutrophil count showed divergent trajectories in cases and controls up to 12 years before a CAD event ( Figure 3G). Zurich participants in the fifth versus first leukocyte quintile had CAD-OR = 4.78 (2.31-9.87), and in   Table 8). Because of the high correlation between leukocytes and neutrophils (Spearman rho = 0.85, P < .01), results from simultaneous modeling (leukocytes and neutrophils in the same model) cannot be interpreted. We found no evidence of an association of eosinophil count or neutrophil:lymphocyte ratio with CAD events (data not shown).

DISCUSSION
Multiple studies have recorded associations of CAD with biomarkers of inflammation and coagulation in PWH [4,10,11] and multiple studies document associations of CAD with  leukocyte count in the general population [12][13][14][15]. To our knowledge, this is the first report of an independent association of leukocyte count with CAD events in PWH. Our study has 3 main findings: first, participants with the highest leukocytes (top quintile, >7810/µL) had a 1.59-fold increased CAD event risk in the final multivariable model. This effect size of high leukocytes was similar to the effect of established CAD risk factors, including hypertension, diabetes, dyslipidemia, or recent abacavir exposure. Second, as in the general population, leukocyte count within normal range values was a predictor of CAD events and overt leukocytosis was infrequent. Third, the leukocyte-CAD association was in part explained by smoking, a well-recorded CAD risk factor known to increase leukocytes [17]. Although the association of black ethnicity or alcohol with lower leukocytes is well established [16,18], these factors only minimally modified the leukocyte-CAD association in our study. The contribution of leukocyte count to CAD events in PWH may demonstrate the potential clinical value of monitoring leukocytes, a cheap, routinely available biomarker with short turnaround time.
Although this was beyond the scope of our study (this will require prospective trials), our findings suggest how knowledge of chronically elevated leukocytes increasing CAD event risk by >50% in the 20% PWH in the top leukocyte quintile may motivate clinicians to place even more emphasis on the optimization of cardiovascular risk factors, and, perhaps, primary prevention of CAD with statins in such persons. Our result of an independent association of leukocyte count with CAD events in PWH appears robust because it persisted after consideration of traditional and HIV-associated CAD risk factors, and in sensitivity analyses adjusting for FRS. Additional strengths of our study are the inclusion of only leukocyte values taken until the day before the CAD event to address the issue of reverse causation (ie, leukocytes being elevated because of a CAD event). In addition, we included all CAD events that occurred in the wellestablished SHCS over a >21-year period, and all CAD events were validated using internationally standardized procedures [20,24].
Additional support for a true leukocyte-CAD event association in PWH is provided by the increase in leukocyte count in CAD cases versus controls that can already be shown 8 years before the CAD event. This suggests the association of high leukocytes with CAD event risk is not attributable to short-term inflammatory/infectious illness immediately before the CAD event that might cause bursts of inflammation and thereby contribute to plaque rupture and CAD events. Our results stand in contrast mechanistically to the association in the general population of acute pneumonia or influenza with increased shortterm CAD event risk [31]. Indirect support for the relevance of high leukocytes to CAD risk is afforded by data showing that adding leukocyte count to the Veterans Aging Cohort Study Index improved prediction of mortality [32].
In our Zurich subpopulation, high leukocytes had a larger CAD-odds ratio than high neutrophils. Leukocytes may provide a pathogenetic link between atherosclerosis and activation of procoagulatory mechanisms, and some general population literature [33,34] points to a stronger neutrophil-CAD than leukocyte-CAD association [35]. However, the precise role of different leukocyte subtypes in predicting CAD events remains unresolved.
The leukocyte-CAD association was in part attenuated by smoking, a factor that is well-recorded to increase leukocytes, but less so by alcohol and black ethnicity, both of which may decrease leukocytes, or other factors with an established inflammatory link such as detectable HIV viremia or abdominal obesity.
Our study has limitations. Our population was 87% male, 94% white, and relatively young; therefore, results should only cautiously be extrapolated to other PWH. Leukocyte subtypes were available only in the Zurich participants, and insufficient information was available to analyze possible associations of leukocytes with chronic inflammatory conditions or corticosteroid therapy. Inflammatory markers such as high sensitivity C-reactive protein and interleukin-6 are not routinely measured in the SHCS. A potential link between inflammatory biomarkers and leukocytes would therefore be an important avenue for future investigation. Finally, we did not compare the leukocyte-CAD association in our PWH with a All variables were associated with the CAD-odds ratio and had P < .01.
control population without HIV. However, the effect size of leukocytes on CAD risk that we report in PWH is very similar to effect sizes reported in the general population [14,15].
In conclusion, we show how a high leukocyte count, most often in the normal range, may identify PWH at independently increased risk for CAD events. This increased risk persists after adjustment for traditional and HIV-related risk factors. Our findings expand on how inflammation (that may not yet be captured by current CAD risk assessment methods) may contribute to high leukocytes and CAD events in PWH.