Sex modulates the association between inflammation and coronary atherosclerosis among older Ugandan adults with and without HIV

Abstract

Objective:

Inflammation is key in the pathogenesis of atherosclerotic coronary artery disease (CAD). Distinct sex-specific inflammatory mechanisms may contribute to CAD in sub-Saharan Africa (SSA), where environmental and biological determinants of systemic inflammation may differ from those in high-income settings.

Approach and Results:

We investigated sex differences in inflammatory markers and CAD in a 2-year prospective cohort of Ugandan adults enriched for cardiometabolic risk factors (RFs) and HIV. Seven plasma biomarkers were quantified at the baseline visit among 125 females and 75 males (50% with HIV) ≥45 years old at enrollment with ≥1 major cardiovascular RF. In year 2, coronary CT angiography (CCTA) was performed in 82 females and 50 males returning for follow-up (52% with HIV). In sex-specific models adjusted for cardiovascular RFs and HIV, TNF-α RII and sCD163 predicted subsequent CAD in females, while only fibrinogen was predictive in males (p < 0.05). IL-6 and sCD14 were inversely associated with CAD in males (p < 0.05). Sex modified the associations of TNF-α RII, sCD14, and sCD163 with CAD (p < 0.05 for interaction). In multivariable multiple imputation models applied to missing year 2 CCTA data to test associations between serum biomarkers in the baseline cohort (n = 200) and subsequent CAD, higher sCD163 was predictive in females only (p < 0.05).

Conclusions:

The positive link between inflammation and subclinical CAD was stronger among females than males in Uganda. Mechanisms by which sex modulates the relationship between inflammation and CAD should be further investigated in SSA.

Introduction

Inflammation and immune activation play a central role in the development and progression of atherosclerosis.¹ Epidemiologic trends over decades have demonstrated differences in the incidence of atherosclerotic coronary artery disease (CAD) by sex,² but the differential contributions of inflammation underlying these patterns remain under-investigated in both high-income and resource-limited settings. Several serum inflammatory markers (e.g., high-sensitivity C-reactive protein,³ interleukin-6⁴) are known to have independent predictive value for atherosclerotic cardiovascular events in high-income settings. However, limited data are available in sub-Saharan Africa (SSA) on the epidemiology and contribution of inflammation to CAD. Notably, CAD may be an emerging noncommunicable disease in the region given the increasing burden of cardiometabolic risk factors and the high prevalence of HIV,⁵ the latter of which appears to be a risk factor for CAD in high-income settings.⁶

Studies from high-income countries have found sex differences in circulating protein markers implicated in cardiovascular disease (CVD), with pathways relevant to inflammation and adiposity pathways upregulated in women and pathways involved in fibrosis and platelet homeostasis upregulated in men.⁷ However, such mechanistic sex differences in biomarkers have not previously been investigated in an African population, where distinct biological determinants of systemic inflammation may play a role in atherosclerosis. Identifying such novel sex- and geography-specific pathways may have implications for precision-based targeted screening and management of CVD in resource-limited settings and globally.

In this study, we sought to assess relationships between sex, serum biomarkers of systemic inflammation, and subsequent CAD in a mixed 2-year prospective cohort of older persons living with HIV (PLWH) and HIV-uninfected controls in Uganda. Throughout our analysis, the term “sex” is used to represent sex assigned at birth. We hypothesized that the association of circulating inflammatory and immune markers with subsequent CAD would differ by sex.

Methods

Study population.

We leveraged data from the Ugandan STudy of HIV effects on the Myocardium and Atherosclerosis (mUTIMA), a prospective study of 100 PLWH and 100 age- and sex-matched HIV-uninfected controls ≥45 years of age in Kampala, Uganda. PLWH were recruited from the Joint Clinical Research Center (JCRC) near Kampala, Uganda between April 2015 and May 2017, were on antiretroviral therapy (ART) for >6 months with a stable ART regimen for >12 weeks and had an HIV viral load <400 copies/mL. Control participants were recruited from local clinics. All study participants had at least one major cardiovascular risk factor (hypertension, diabetes, smoking, low high-density lipoprotein level [<60 mg/dL], hypercholesterolemia [total cholesterol ≥200 mg/dL], or family history of early CAD). Exclusion criteria were known history of CAD, peripheral arterial disease, ischemic stroke, active chronic inflammatory condition other than HIV, pregnancy, use of chemotherapy or immunomodulating agents, or estimated glomerular filtration rate <30 mL/min/1.73 m². The JCRC Community Advisory Board was consulted for the design, conduct, and dissemination of the study, which was reviewed and approved by the Institutional Review Boards of University Hospitals Cleveland Medical Center, the JCRC, and by the Uganda National Council for Science and Technology. All study participants provided written informed consent.

Cardiac, immune, and inflammatory biomarkers.

As previously described,^8,9 we measured 7 batched soluble protein biomarkers from cryopreserved plasma at −80°C: markers associated with atherosclerotic cardiovascular (ASCVD) events, coronary plaque burden and progression (high-sensitivity C-reactive protein [hsCRP],³ soluble intracellular and vascular adhesion molecule [sICAM and sVCAM, respectively],^10,11 interleukin-6 [IL-6],⁴ soluble tumor necrosis factor α receptor II [TNF-α RII],¹² fibrinogen¹³); and markers of monocyte activation, CAD, and mortality in both PLWH and the general population (soluble CD14 [sCD14]^14,15, soluble CD163 [sCD163]^16,17). sICAM, sVCAM, TNF-α RII, sCD163, and sCD14 were measured by ELISA (R&D Systems, Minneapolis, MN); IL-6 was measured by electrochemiluminescence (Meso Scale Diagnostics, Rockville, MD); hsCRP and fibrinogen were measured by nephelometry (Siemens, Munich, Germany). Lower limits of detection for assays were as follows: 3.12 ng/mL for sICAM; 6.25 ng/mL for sVCAM; 7.8 pg/mL for TNF-α RII; 1.6 ng/mL for sCD163; 250 pg/mL for sCD14; 0.05 pg/mL for IL-6; 0.16 ug/mL for hsCRP; and 56 mg/dL for fibrinogen.

Coronary CT angiography.

After collection of demographic and baseline biomarker data at the initial visit, participants returned for coronary CT angiography (CCTA) in year 2. CCTA was performed on a 128-slice Siemens Somatom Definition scanner at Nsambya St. Francis Hospital in Kampala. Participants with estimated glomerular filtration rate >60 mL/min/1.73 m² were eligible to receive intravenous contrast; image acquisition and analysis protocols were developed in accordance with Society of Cardiovascular Computed Tomography Guidelines.¹⁸ Two hours prior to the scan, 100 mg metoprolol was administered orally, with another 50 mg dose 30 minutes prior to the scan if the heart rate remained >60 beats per minute. Data acquisition for some participants was limited by technical issues, including inability to adequately lower heart rate. CT scans were read by a local radiologist for clinically significant findings, then batched and read offline by a single expert reader (MSB) for research. Using an 18-segment model, the number of evaluable segments per participant was determined and scans of poor technical quality (i.e., anatomically absent segments, non-evaluable due to artifact) were excluded from analyses. Because the average number of non-evaluable segments per participant was modest and given the low prevalence of CAD in this population, absent segments were assumed to be normal for our analyses. CAD was classified as a binary outcome defined as presence of detectable plaque in any evaluable coronary segments.

Statistical analyses.

We summarized baseline characteristics as frequency (percentage) for categorical variables and mean (standard deviation) for continuous variables. Baseline serum biomarker levels are expressed as median (inter-quartile range). Baseline characteristics were compared by sex using two-proportion Z-tests, Chi-squared tests, and two-sample t-tests for binary, categorical, and continuous variables, respectively. Median serum biomarker levels were compared using the Wilcoxon rank sum test.

In our primary analyses, we used univariable and multivariable logistic regression models to assess relationships between log-transformed baseline biomarker levels and subsequent presence of CAD in the full cohort and in sex-specific models adjusted for traditional cardiovascular risk factors (age [categorical; reference 45–54, 55–64, ≥65 years], diabetes, hypertension, elevated serum total cholesterol [>200 mg/dL], past or current smoking, abdominal obesity [waist-to-hip ratio ≥0.85 for females, ≥0.9 for males]) and non-traditional risk enhancers (HIV, chronic kidney disease [CKD] defined as urine albumin ≥30 mg/dL). Models in the full cohort were adjusted for sex. Separate interaction models were used to assess whether sex modifies the association of each serum biomarker with subsequent CAD. We report odds ratios (OR), 95% confidence intervals (CI), and p-values estimated using the likelihood ratio method. In secondary analyses, we applied multiple imputation using the Mice package in RStudio version 4.0.5 (R Foundation) to missing year 2 CCTA data to test associations between tertiles of serum biomarkers in the baseline cohort (n = 200) and subsequent CAD. Parameter estimates from each imputation were pooled using Rubin’s rules to generate confidence intervals and p-values. All analyses were performed in RStudio version 4.0.5; p < 0.05 was considered statistically significant.

We performed sensitivity analyses in the non-sex-stratified cohort using multivariable logistic regression models adjusted for ASCVD risk category instead of individual risk factors (calculated using the American College of Cardiology [ACC]/American Heart Association [AHA] Pooled Cohort Equations with race classified as “other”), abdominal obesity, HIV, and CKD; these models were fitted to the follow-up cohort (n = 132) and the full baseline cohort with multiple imputation applied to missing year 2 CCTA data (n = 200).

Results

Characteristics of study cohort.

Table 1 summarizes baseline characteristics in the study cohort. One hundred thirty-two participants who returned for follow-up in year 2 and had complete CCTAs were included in the primary analysis. The mean age of study participants was 54 ± 7 and 56 ± 6 years among females and males, respectively. Hypertension and obesity were more common among females than males (93% vs 74% and 54% vs 16%, respectively). A higher proportion of females were in the low calculated ASCVD risk category based on the Pooled Cohort Equations (51% of females vs 12% of males), while a higher proportion of males were high risk (52% of males vs 13% of females). There were no significant differences in demographic or risk factor composition between the baseline cohort (n = 200) and the follow-up cohort (n = 132; Supplemental Table 1). In the follow-up cohort, 13 females (16%) and 10 males (20%) had detectable CAD by CCTA, for an overall prevalence of 17%. Of those with CAD, 7 (30%) were PLWH.

Table 1.

Demographic and baseline clinical characteristics of study participants in the mUTIMA cohort. Percentages/proportions are represented in parentheses. P-values are calculated using t-tests, Wilcoxon rank-sum tests, chi-squared tests, and two-proportion z-test for continuous means, continuous medians, categorical, and binary variables, respectively.

	Female (N = 82)	Male (N = 50)	P (Female vs. Male)
Age, years (mean ± SD)	54.3 ± 6.6	55.5 ± 6.4	0.32
Age group categories (n)			0.80
45 –54	40 (49)	27 (54)
55 –64	35 (43)	20 (40)
65 +	7 (9)	3 (6)
ASCVD risk group (n)			< 0.001*
Low (< 5%)	42 (51)	6 (12)
Borderline (5 – 7.5%)	17 (21)	9 (18)
Intermediate (7.5 – 10%)	12 (15)	9 (18)
High (> 10%)	11 (13)	26 (52)
Past Medical History (n)
Diabetes	21 (26)	19 (38)	0.14
Hypertension	76 (93)	37 (74)	< 0.01*
Current smoker	2 (2)	5 (10)	0.14
BMI			< 0.001*
Normal	12 (15)	19 (38)
Overweight	26 (32)	21 (42)
Obese	44 (54)	8 (16)
Underweight	0 (0)	2 (4)
Waist-to-hip ratio (mean ± SD)	0.87 ± 0.07	0.96 ± 0.10	< 0.001*
Abdominal obesity (high WHR) †	54 (66)	39 (78)	0.20
Chronic kidney disease	7 (9)	4 (8)	0.99
GFR (mean ± SD)	102 ± 19.8	106 ± 17.1	0.22
Abnormal urine ACR ‡	15 (18)	9 (18)	0.99
High total cholesterol	46 (56)	24 (48)	0.47
High low-density lipoprotein cholesterol	46 (56)	25 (50)	0.62
Low high-density lipoprotein cholesterol	48 (59)	34 (68)	0.37
HIV §	42 (51)	26 (52)	0.99
Time since diagnosis, years (median [1st – 3rd quartile])	11.8 (9.9 – 12.1)	11.6 (9.9 – 13.4)	0.52
Current CD4+ (cells/mm3)	624 (498 – 747)	416 (375 – 488)	< 0.001*
Nadir CD4+ (cells/mm3)	143 (81 – 202)	148 (65 – 220)	0.86
Antiretroviral therapy duration, years (median [1st – 3rd quartile])	10.9 (8.9 – 11.9)	11.1 (9.0 – 12.6)	0.49
Current protease inhibitor	12 (29)	3 (12)	0.18
Current integrase inhibitor	0 (0)	0 (0)	--
Current abacavir	2 (5)	1 (4)	0.99

^†WHR, waist-to-hip ratio; abdominal obesity defined as WHR ≥0.85 for females and ≥0.9 for males

^‡ACR, albumin-to-creatinine ratio; abnormal: ACR ≥30

^§HIV-specific variables are reported for females and males living with HIV only

Serum biomarkers and CAD.

Table 2 shows median serum levels for 7 biomarkers by sex. hsCRP, IL-6, sCD163, and fibrinogen were higher among males, while there were no sex differences in in levels of the remaining inflammatory markers. In multivariable models fitted to the full cohort and adjusted for sex, only baseline levels of fibrinogen (OR 8.6 [95% CI 1.3–81]) were predictive of subsequent CAD in both primary and sensitivity analyses (p < 0.05; Supplemental Table 2). In sex-stratified univariable models, sCD163 (OR 4.8 [95% CI 1.2–22]) in females and sCD14 and fibrinogen in males (0.02 [0.001 – 0.33]; 50 [2.7–2500], respectively) were predictors (p < 0.05 for all). In multivariable models adjusted for cardiometabolic risk factors, HIV, abdominal obesity, and CKD, TNF-α RII and sCD163 were predictive in females (42 [1.2–2900]; 5.2 [1.1–33], respectively), while only fibrinogen was predictive in males (11400 [5.6–5e11]; Table 3). By contrast, IL-6 and sCD14 were inversely associated with subsequent CAD in males (0.01 [1e-5–0.3]; 0.005 [4e-6–0.8], respectively; p < 0.05). In multivariable models in the full cohort assessing for formal interaction between sex and each serum biomarker, sex modified the association of TNF-α RII (p = 0.03), sCD14 (p = 0.005), and sCD163 (p = 0.01) with subsequent CAD.

Table 2.

Median (interquartile range) serum biomarker levels. P-values are based on Wilcoxon rank-sum tests.

Biomarker§	Female (N = 82)	Male (N = 50)	P (Female vs. Male)	PLWH (N = 68)	Controls (N = 64)	P (PLWH vs. Controls)
hsCRP (ug/mL)	0.78 (0.43–1.9)	2.8 (0.78–4.8)	< 0.001*	1.22 (0.43–4.9)	1.6 (0.66–3.3)	0.92
sICAM (ng/mL)	201 (128–259)	163 (128–245)	0.47	181 (128–255)	173 (127–252)	0.99
sVCAM (ng/mL)	573 (495–727)	622 (503–761)	0.44	602 (518–760)	599 (485–732)	0.93
IL-6 (pg/mL)	0.52 (0.36–0.76)	0.70 (0.46–1.0)	0.014*	0.53 (0.41–0.91)	0.63 (0.42–0.86)	0.79
TNF-α RII (pg/mL)	2379 (2052–3247)	2631 (2251–3138)	0.27	2575 (2107–3196)	2568 (2202–3133)	0.95
sCD14 (ng/uL)	1382 (1113–1653)	1524 (1320–1742)	0.67	1589 (1394–1848)	1366 (1138–1536)	<0.001*
sCD163 (ng/mL)	517 (387–736)	623 (489–887)	0.007*	572 (444–750)	598 (465–891)	0.21
Fibrinogen (mg/dL)	309 (241–359)	340 (270–417)	0.005*	310 (254–379)	342 (280–404)	0.09

^§hsCRP, high-sensitivity C-reactive protein; sICAM, soluble intercellular adhesion molecule; sVCAM, soluble vascular cell adhesion molecule; IL-6, interleukin 6; TNF-α RII, soluble tumor necrosis factor α receptor II; sCD14, soluble CD14; sCD163, soluble CD163

Table 3.

Sex-specific odds of subclinical CAD in n = 132 study participants who returned for follow-up CCTA in year 2. Results are odds ratios (OR) and 95% confidence intervals (CI) per one-unit difference in log-transformed baseline biomarker level. Interaction models include both females and males, with an interaction term between sex and log-transformed biomarker.

Log-transformed biomarker§	Female: Univariable		Female: Multivariable†		Male: Univariable		Male: Multivariable†		Sex Interaction‡
	OR (95% CI)	P	OR (95% CI)	P	OR (95% CI)	P	OR (95% CI)	P	P
hsCRP	0.9 (0.56–1.47)	0.67	0.89 (0.49–1.62)	0.70	0.83 (0.42–1.55)	0.56	0.36 (0.05–1.3)	0.18	0.49
sICAM	1.19 (0.74–2.5)	0.55	1.38 (0.8–3.18)	0.32	1.18 (0.46–5.07)	0.78	1.38 (0.11–22.0)	0.80	0.65
sVCAM	4.88 (0.92–29.7)	0.07	3.58 (0.53–34.0)	0.21	0.85 (0.1–6.59)	0.88	0.09 (0.0002–8.14)	0.33	0.27
IL-6	0.57 (0.18–1.46)	0.29	0.66 (0.18–1.74)	0.45	0.46 (0.1–1.34)	0.25	0.01 (1e-5–0.29)	0.0010*	0.096
TNF-α RII	9.42 (0.69–156)	0.098	41.8 (1.17–2929)	0.047*	0.55 (0.04–6.95)	0.65	0.05 (0.0002–3.32)	0.19	0.026*
sCD14	8.42 (0.56–174)	0.14	18.07 (0.53–1204)	0.14	0.02 (0.001–0.33)	0.016*	0.005 (4e-6–0.81)	0.028*	0.005*
sCD163	4.79 (1.18–22.1)	0.033*	5.16 (1.07–32.7)	0.034*	0.43 (0.07–2.19)	0.33	0.13 (0.004–1.54)	0.15	0.012*
Fibrinogen	2.18 (0.45–15.4)	0.39	1.94 (0.16–28.5)	0.61	50 (2.66–2501)	0.022*	11418 (5.67–5e11)	0.0096*	0.14

^§hsCRP, high-sensitivity C-reactive protein; sICAM, soluble intercellular adhesion molecule; sVCAM, soluble vascular cell adhesion molecule; IL-6, interleukin 6; TNF-α RII, soluble tumor necrosis factor α receptor II; sCD14, soluble CD14; sCD163, soluble CD163

^†Multivariable models adjusted for age, diabetes, hypertension, high total cholesterol, past or current smoking, abdominal obesity, HIV, and chronic kidney disease

^‡Interaction models adjusted for all covariates in multivariable models

In our secondary analyses utilizing all baseline serum biomarker data and applying multiple imputation to missing year 2 CCTA data, higher baseline serum sCD163 predicted year 2 CAD in females only (p < 0.05; Figure 1, Supplemental Table 3). Additionally, there was a trend toward increased odds of CAD at higher levels of baseline TNF-α RII among females (3^rd vs 1^st tertile, 4.0 [0.79–20; p = 0.09]). None of the inflammatory markers were predictive of CAD in males.

Figure 1.

Association between tertiles of baseline serum biomarkers (n = 200) and subsequent CAD in overall cohort and stratified by sex in multiple imputation models adjusted for age, diabetes, hypertension, high total cholesterol, past or current smoking, abdominal obesity, HIV, and CKD. Non-stratified models are additionally adjusted for sex. Given the large standard errors in a small dataset, odds ratios are presented on a logarithmic scale. Horizontal dotted line represents reference odds of 1 in the first tertile; filled circles indicate p < 0.05. Numerical odds ratios and 95% confidence intervals can be found in Supplemental Table 3. Abbreviations: hsCRP, high-sensitivity C-reactive protein; sICAM, soluble intercellular adhesion molecule; sVCAM, soluble vascular cell adhesion molecule; IL-6, interleukin 6; TNF-alpha RII, soluble tumor necrosis factor α receptor II; sCD14, soluble CD14; sCD163, soluble CD163.

Discussion

While sex differences in cardiovascular events are well-established,² few studies have explored potential inflammatory mechanisms and objective subclinical disease assessments directly, as we have done here. Our study represents the first effort to examine potential sex-specific associations between systemic inflammation and subsequent atherosclerotic CAD in SSA, with a key strength being its prospective design. We leverage a well-phenotyped cohort that was initially recruited to examine the effects of HIV on early structural heart disease and now examine inflammation and CAD among residents of this understudied geographic region. Our analyses furnish novel hypotheses suggesting sex and region-of-residence influence on the manner in which inflammation predisposes to atherosclerotic CAD. Specifically, we found a low overall prevalence of CAD, as well as distinct sex-specific relationships of several key inflammatory markers (sCD14, sCD163, and IL-6) with subsequent CAD.

Notably, among the Ugandan population studied (which was enriched for cardiometabolic risk factors and HIV), the prevalence of subclinical CAD (17%) is low, compared with 24–53% in low- to intermediate-risk CCTA cohorts from high-income settings.^19–22 In addition, while PLWH represented approximately half our cohort, they comprised only 30% of those with CAD, indicating that our findings were not driven solely by residual HIV-mediated immune dysregulation, which has been well-described.²³ These observations point to potential unique mechanistic pathways underlying atherosclerosis in this setting that further differ by sex, as cardiometabolic risk factors did not appear to confer the same CAD risk in this cohort as that described in high-income populations from which most ASCVD risk frameworks are derived.⁸

Striking results with respect to sex included (1) the differential effects of sCD163 and sCD14 on subsequent CAD, and (2) the inverse association of circulating sCD14 and IL-6 with subsequent CAD among males in our cohort. Few studies have examined the interaction of sCD163, a marker of macrophage activation, with sex in the context of CVD. In human atherosclerotic lesions and in mouse models, CD163+ macrophages were associated with plaque progression by a non-lipid-driven mechanism.²⁴ sCD163 predicts non-calcified coronary plaque in PLWH, which is known to harbor a component of residual immune dysregulation despite viral suppression,^17,23 and accelerated carotid atherosclerosis in patients with rheumatologic conditions such as lupus.²⁵ Taken together, the current literature suggests that sCD163 may be a marker for atherosclerosis and CVD in populations with existing systemic inflammation and/or immune dysregulation. Our study excluded participants with known rheumatologic or inflammatory conditions, and our models were adjusted for HIV status, with only 30% of those with CAD being PLWH. Despite this, there was a strong link between sCD163 and subsequent CAD among females in both our primary and secondary analyses, suggesting that novel attributes of female sex may interact with complex innate immune mechanisms in this environment to increase the risk of atherosclerosis. Indeed, a recent sub-analysis of the Center for AIDS Research Network of Clinical Systems (CNICS) cohort in the US noted that females with treated HIV infection had higher levels of baseline immune activation than their male counterparts; this immune activation was also more strongly associated with incident myocardial infarction in females.²⁶ Further work is needed to delineate the differential contribution of these inflammatory pathways to the development and progression of atherosclerosis in females and males in global populations.

sCD14 is receptor for lipopolysaccharide and is a marker of gut microbial translocation and monocyte activation;²⁷ in the African population, however, unique endemic diseases, such as enteric parasites, may play key roles in its circulating levels and its differential expression on monocytes.^28,29 In Black South Africans with CKD, sCD14 was positively associated with subclinical carotid atherosclerosis only among those with elevated serum levels of endotoxin.³⁰ To our knowledge, sex differences in circulating sCD14 and its association with subsequent CVD have not been investigated in depth. Our exploratory analyses suggest that its role in potential anti-atherogenic pathways, and ways that this relationship is modulated by sex, should be further investigated.

To our knowledge, an inverse association of IL-6 with subsequent CAD in males, as we observed in our study, has not previously been reported. IL-6 is involved in established inflammatory pathways with significant evidence for a causal effect on the development of CAD.⁴ However, there are also indications that IL-6 has potent anti-inflammatory effects in certain contexts; infusion of recombinant human IL-6 at low doses into young males induced endogenous production of IL-1 receptor agonist, IL-10, and cortisol, all potent anti-inflammatory compounds.³¹ While our findings are strictly exploratory given our small sample and the low detected levels of IL-6 (<1 pg/mL), should our findings be validated in larger studies, such a physiologic mechanism may point to a potential “protective” pathway against the development of atherosclerosis among males in our population. If such a pathway exists, it merits additional investigation as a target for pharmacotherapeutic interventions. Future studies should examine circulating anti-inflammatory cytokines alongside IL-6 in other populations, including females.

Our study represents emerging, hypothesis-generating data pointing to differences in inflammatory mechanisms underlying ASCVD in SSA compared with high-income settings. While there are no comparable cross-country cohort studies in the general population examining the relationship between the biomarkers and imaging outcomes we have described herein, a recent cross-sectional analysis in the US-based REPRIEVE HIV cohort showed that IL-6 had a weak but positive association with CAD, in contrast to our study. Additionally, this same study found no association of subsequent CAD with sCD14 or sCD163, while the latter was predictive of CAD among females in our cohort.³² Remarkably, in our study, sCD14 was inversely associated with CAD among males, but demonstrated no association among females. By contrast, several studies implicate sCD14 in pro-atherogenic pathways. In the US-based REGARDS cohort, sCD14 was strongly predictive of incident ischemic stroke and CAD among Black, but not White, participants,³³ and predicted heart failure and any CVD among women in the Framingham cohort.⁷ In the Multicenter AIDS Cohort Study, higher sCD14 levels were positively associated with subclinical obstructive CAD among males with and without HIV, in direct contrast to the results of our sex-stratified models.³⁴ Our exploratory findings point to geographic differences in the relationship between circulating serum markers of inflammation and CAD, particularly as we noted significant differences despite our relatively small sample of Ugandan adults. Future international comparative cohort studies from SSA and high-income settings are needed to elucidate unique environmental or life-course modulators of the relationship between inflammation and atherosclerosis.

Limitations

Our study has several limitations, including a relatively small sample with a large number lost to follow-up, and resultant wide confidence intervals for many of our estimates; despite this, the consistency of associations by sex are hypothesis-generating, pointing to underlying geographic and sex variation in inflammation leading to CAD. In addition, we did not examine interactions between individual biomarkers due to the risk of model overfitting in a small study sample, though this will be key in defining relationships between single markers in the context of larger inflammatory pathways. Additionally, we did not adjust for pre- or post-menopausal status in this cohort given the small sample size.

We acknowledge that our analyses are exploratory, and additional studies in diverse populations are needed to substantiate these findings. As such, future studies should recruit larger cohorts that include participants from under-studied regions such as SSA, South Asia, and the Pacific Islands, to understand differences in drivers and phenotypes of CVD in global populations and thereby improve sex- and geography-based precision in CAD management.

Conclusions

In this 2-year prospective cohort study of older Ugandan adults enriched for cardiometabolic risk factors and HIV, the overall prevalence of subclinical CAD was markedly low. The association between inflammatory markers of CVD and subsequent CAD was stronger among females than males; indeed, among males, markers that predict ASCVD in high-income settings were inversely associated with CAD in Uganda, which may represent risk modulation by environmental exposures and their interaction with sex. Systems biology and high-throughput proteomics approaches should be applied in global populations, including in SSA, to better understand the interaction between sex, environment, and risk factors in the development of CVD. Clinical implications of sexual dimorphism in these pathways remain relevant globally for risk prediction, prognosis, and targeted therapies.