Examining Depression as a Risk Factor for Cardiovascular Disease in People with HIV: A Systematic Review

Abstract

Background

People with human immunodeficiency virus (HIV) have an increased risk of cardiovascular disease (CVD) not fully accounted for by traditional or HIV-specific risk factors. Successful management of HIV does not eliminate this excess risk. Thus, there is a need to identify novel risk factors for CVD among people with HIV (PWH).

Purpose

Our objective was to systematically review the literature on one such candidate CVD risk factor in PWH—depression.

Methods

A systematic literature search of PubMed, PsycINFO, EMBASE, Web of Science, and CINAHL was performed to identify published English-language studies examining associations of depression with clinical CVD, subclinical CVD, and biological mechanisms (immune activation, systemic inflammation, altered coagulation) among PWH between the earliest date and June 22, 2021.

Results

Thirty-five articles were included. For clinical CVD (k = 8), findings suggests that depression is consistently associated with an increased risk of incident CVD. For subclinical CVD (k = 5), one longitudinal analysis reported a positive association, and four cross-sectional analyses reported null associations. For immune activation (k = 13), systemic inflammation (k = 17), and altered coagulation (k = 5), findings were mixed, and there was considerable heterogeneity in sample characteristics and methodological quality across studies.

Conclusions

Depression may be an independent risk factor for CVD among PWH. Additional research is needed to confirm depression’s association with clinical CVD and to determine whether depression is consistently and meaningfully associated with subclinical CVD and biological mechanisms of CVD in HIV. We propose a research agenda for this emerging area.

Depression may be related to cardiovascular disease risk among people with HIV and additional research is needed to better understand the underlying biological reasons for this association.

Introduction

Cardiovascular Disease in HIV

Antiretroviral therapy (ART) has transformed human immunodeficiency virus (HIV) from a fatal disease into a chronic, manageable condition. However, the decrease in acquired immunodeficiency syndrome (AIDS)-related mortality has been accompanied by an increase in the prevalence of noncommunicable chronic diseases among people with HIV (PWH), including cardiovascular disease (CVD), also known as HIV-CVD. In fact, CVD is a leading cause of death in PWH [1–3]. A 2018 meta-analysis reported that PWH have over two times greater risk of CVD than adults without HIV [4].

The major mechanism underlying CVD development is atherosclerosis [5], which is characterized by thickening and hardening of the arteries. In atherosclerosis, a proinflammatory cascade is initiated by the accumulation of low-density lipoprotein (LDL) in the artery walls. This cascade includes activation of endothelial cells; migration of molecules that adhere to platelets, monocytes, and lymphocytes to the intima of the artery wall; activation of macrophages that release proinflammatory cytokines; and the transformation of macrophages into foam cells after attempting to clear accumulating LDL. Macrophage activation and the subsequent systemic inflammation continues in a positive feedback loop. Over time, a fibrous, calcified cap develops over the foam cell-filled plaque. The continued proinflammatory cascade includes substances that degrade the fibrous cap, which can destabilize the plaque and increase the chance of rupture and thrombotic complications, such as myocardial infarction and stroke.

A conceptual framework presented by Beltran and colleagues [6] details the impact of HIV infection on LDL clearance, immune activation, inflammation, coagulation, and endothelial dysfunction. HIV infection has metabolic effects that decrease the availability of high-density lipoprotein (HDL), which reduces LDL clearance from the artery wall. HIV infection results in chronic activation of monocytes and macrophages, increased release of proinflammatory cytokines, and elevations in circulating substances that promote coagulation. HIV infection directly damages the endothelium and increases its permeability. HIV infection also indirectly promotes endothelial dysfunction through increasing immune activation and inflammation. Thus, the presence of a developing atherosclerotic plaque is not a necessary stimulus in PWH for the initiation and maintenance of the proinflammatory cascade characteristic of the atherosclerotic process. Though ART has been successful in reducing HIV viral loads and improving some immune function markers (e.g., CD4+ T-cell counts), it does not entirely eradicate reservoirs of viral replication [7] or fully normalize markers of immune or coagulatory function [8–12]. Consequently, successful virologic and immunologic management of HIV does not eliminate excess CVD risk. Thus, there is a need to identify novel risk factors for HIV-CVD and novel drivers of the biological mechanisms underlying HIV-CVD, which could serve as targets for future CVD prevention efforts in PWH.

Evidence from large prospective studies utilizing healthcare system- and international research-based cohorts suggests that HIV-CVD is likely due to the virus itself [13–15], cardiotoxicity of certain ART medications [16, 17], and nontraditional (e.g., renal disease and hepatitis C) [18–20] CVD risk factors. However, findings from large Veterans-based samples suggest that excess CVD risk remains among PWH after controlling for HIV viral load, ART use, and traditional and nontraditional CVD risk factors [21] and is observed among PWH with optimal CVD risk profiles [22]. Traditional CVD risk factors are important considerations in PWH, given the higher prevalence rates of smoking and dyslipidemia compared to the general population [23] and the increase in obesity prevalence among PWH in the absence of HIV-associated wasting [24]. In addition, nontraditional CVD risk factors influence CVD risk in PWH, with higher rates of illicit drug use, hepatitis C co-infection, and renal disease [25–27].

Based on current knowledge in the general population, depression is a candidate novel risk factor for HIV-CVD and a candidate novel driver of the biological mechanisms underlying HIV-CVD. First, general population meta-analytic evidence shows that adults with depression have a 20%–57% greater risk of clinical CVD than adults without depression [28–30]. For example, the most recent meta-analysis, utilizing a sample of 16 cohort studies of participants free of CVD at baseline, found that adults with depression had 20% greater risk of incident myocardial infarction or coronary death than those without depression [30]. Furthermore, this increased risk has been found for several other clinical CVD outcomes and appears to be independent from traditional CVD risk factors [29]. It is also worth noting that meta-analytic evidence indicates positive associations between depression and subclinical CVD, such as coronary artery calcification, carotid intima-media thickness (cIMT), pulse wave velocity, and brachial flow-mediated dilation (FMD) [31, 32]. Second, in general population samples, depression has been linked to some putative biological mechanisms underlying HIV-CVD, including immune activation [33, 34], systemic inflammation [35], and altered coagulation [36, 37]. More specifically, prior studies have observed positive associations between depression and biomarkers, such as activated monocytes and proinflammatory cytokines that are dysregulated in PWH and likely contributors to HIV-CVD. Third, evidence suggests that the successful treatment of depression may reduce CVD risk. In an 8-year follow-up study of the Improving Mood-Promoting Access to Collaborative Treatment (IMPACT) trial [38], it was found that depressed patients without pre-existing CVD treated with a 12-month collaborative care program for depression (intervention) had a 48% lower risk of a fatal/nonfatal myocardial infarction or stroke than patients receiving usual primary care for depression (comparator). Consequently, the authors concluded that collaborative care for depression may be a viable option as a primary prevention strategy for CVD. Given the general population evidence above, depression is a strong candidate for investigation as a contributor to the excess CVD risk among PWH.

There is currently limited understanding of associations between depression and CVD-related outcomes in PWH. One possibility is that there is no association between depression and HIV-CVD, given the potential overwhelming contribution of other factors like unmanaged viral load, immune dysregulation, traditional and nontraditional CVD risk factors overrepresented in PWH, and the potential cardiotoxicity of certain ART medications. However, we may extrapolate from general population research to hypothesize that the depression-CVD relationship may be applicable to and important in PWH. First, depression affects PWH at a high rate with a global prevalence of 31% [39], higher than the 13% observed among the general population [40]. Second, the biological pathways underlying depression and CVD identified in general population research (e.g., systemic inflammation) are among the central biological mechanisms in current conceptualizations of HIV-CVD [6, 41]. A specific examination of the potential role of depression in HIV-CVD is warranted, as general population research does not address the question of whether this relationship is independent from HIV-specific factors (e.g., viral load, CD4 count, and cardiotoxic ART) and the nontraditional CVD risk factors disproportionately affecting PWH. Accordingly, the objectives of this review are: (1) to systematically review the literature examining the associations of depression with clinical CVD, subclinical CVD, and the putative biological mechanisms of HIV-CVD (immune activation, systemic inflammation, and altered coagulation) in PWH and (2) to propose a future research agenda for this emerging area.

Methods

Our systematic review was conducted in accordance with the Preferred Reporting Items for Systematic reviews and Meta-Analysis (PRISMA) guidelines [42] and registered in PROSPERO (ID: CRD42018094612).

Search Strategy

The search was conducted by author BMP for all references between the earliest date available and June 22, 2021 in the following databases: PubMed, PsycINFO, EMBASE, Web of Science, and CINAHL. Distinct search terms in the domains of HIV, depression, and CVD-related outcome categories are detailed in Table 1. Search terms within each domain were combined using the OR search operator. After creating each domain, searches were performed in the following fashion using the AND search operator: HIV domain AND depression domain AND a CVD-related outcome domain. An example search syntax for PubMed is provided in Supplementary Table S1. Of note, clinical CVD, subclinical CVD, immune activation, systemic inflammation, and altered coagulation were all searched separately rather than simultaneously. In addition to our comprehensive search, we performed backward (examination of references of an included study) and forward (examination of articles citing an included study) referencing of articles selected for inclusion. Forward referencing was performed using PubMed.

Table 1.

Summary of Results of Included Studies Examining Clinical CVD (k = 8)

Author	Follow-up duration	Depression variable(s)	Clinical CVD variable(s)	Covariates	Results
Longitudinal
Castilho [43]	Median = 3.5 years	History of Depression Defined by depressive disorders (including bipolar disorders) abstracted and validated from EMR at baseline.	Incident CVD (n = 69) Incident CAD (n = 46) Defined by CAD events (MI, coronary artery atherosclerosis, ischemic cardiomyopathy), cerebrovascular events, or peripheral artery disease abstracted and validated from EMR during follow-up.	Age, sex, race, CD4/CD8 ratio, and CD4 count.	History of depression was associated with an increased risk of incident CVD (HR = 2.35, 95% CI: 1.44–3.85, p = .001). History of depression trended towards an increased risk of incident CAD (HR = 1.69, 95% CI: 0.93–3.10, p = .088).
French [44]	Length = 10 years	CES-D ≥ 16 Time-dependent variable measured at each semiannual visit.	CVD-related mortality (n = 33) Defined by death certificate obtained after active and passive ascertainment of participant death and search of the National Death Index-Plus performed annually during follow-up.	Age, race/ethnicity, income, smoking, heavy alcohol consumption, cocaine use, BMI, ART use, viral load, CD4 cell count, history of AIDS, and hepatitis B and C status.	Elevated depressive symptoms were associated with an increased risk of CVD-related mortality (HR = 4.61, 95% CI: 1.55–13.74, p < .05).
Khambaty [45]	Median = 5.8 years	MDD Dysthymic disorder Defined by ICD-9 codes (MDD: 296.2, 293.3; dysthymic disorder: 300.4) abstracted from EMR (diagnosis in medical record prior to the beginning of follow-up).	Incident AMI (n = 490) Defined by nonfatal AMI ([1] adjudicated events in discharge summary if at VA hospital or [2] ICD-9 code 410 in Medicare or VA fee-for-service data if not at non-VA hospital) or fatal AMI (ISCDRHP-10 AMI codes 121.0-.9 as first listed cause of death) during follow-up.	Age, sex, race/ethnicity, hypertension, diabetes, dyslipidemia (LDL-C, HDL-C, triglycerides), statin use, viral load, CD4 count, and ART regimen.	MDD was associated with an increased risk of incident AMI (HR = 1.30, 95% CI: 1.05–1.62, p < .05). Non-significant trend with the addition of hepatitis C, renal disease, history of alcohol or cocaine abuse/dependence, hemoglobin (HR = 1.25, 95% CI: 1.00–1.56, p = .053) and these variables plus antidepressant use (HR = 1.12, 95% CI: 0.87–1.42, p = .37), BMI (HR = 1.25, 95% CI: 1.00–1.56, p = .049) or smoking (HR = 1.22, 95% CI: 0.98–1.53, p = .08). Dysthymic disorder showed a nonsignificant trend with incident AMI (HR = 1.28, 95% CI: 0.96–1.71, p > .08).
Parruti [46]	Mean = 2.1 years	BDI-II ≥ 17 Measured once at baseline.	Incident vascular event (n = 19). Defined by TIA, stroke, acute coronary syndrome (unstable angina, non-ST elevation and ST elevation MI), and/or myocardial or other organ infarction actively assessed at 3-month follow-up visits and adjudicated by physician.	None.	No difference in frequency of incident vascular events between those with and without depression (p = .14).
White [47]	Median = 5.8 years	MDD Defined by ICD-9 codes (296.2, 293.3) abstracted from EMR (diagnosis in medical record prior to the beginning of follow-up).	Incident heart failure (n = 2,666) Defined by ICD-9 codes 428.xx, 429.3, 402.91, 425.x in VA or Medicare data during follow-up.	Age, sex, race/ethnicity, BMI, hypertension, diabetes, dyslipidemia, statin use, smoking status, alcohol/cocaine abuse/dependence, viral load, CD4 count, ART, hepatitis C, hemoglobin, renal function, atrial fibrillation, and atrial flutter.	MDD was associated with an increased risk of incident heart failure (HR = 1.30, 95% CI: 1.11–1.51, p NR).
Vadini [48]	Mean = 4.4 years	BDI-II ≥ 17 Measured once at baseline.	Vascular event (n = 54) Defined by AMI, angina, PCI, stroke, TIA, cardiac syncope, organ infarctions (intestinal or renal infarction), or peripheral artery disease actively assessed at 6-month follow-up visits and confirmed by clinical, laboratory, and imaging data.	Age, sex, HIV transmission risk factor, smoking, BMI, hypertension, diabetes, FRS CVD risk score, AIDS, and alexithymia.	Elevated depressive symptoms were not associated with vascular events (HR = 1.37, 95% CI: 0.72–2.58, p = .3).
Cross-sectional
Havlik [49]	N/A	CES-D	Heart condition (n = 123) Stroke (n = 29) Defined by participant self-report via endorsement on a list of medical illnesses experienced in the past year.	None.	Depressive symptoms were positively correlated with heart condition (r = .086, p = .05) but not stroke (r = .059, p NR).
Tymchuck [50]	N/A	PHQ-9 Comparison of 3 groups: minimal (0–4), mild (5–9), and moderate-to-severe symptoms (≥10).	Cardiovascular conditions (n = 37) Defined by ischemic heart disease, CHD, hypertension, MI, or deep vein thrombosis extracted from clinic database.	None.	No difference in frequency of cardiovascular conditions between the three depressive symptom groups (p NR).

AIDS acquired immunodeficiency syndrome; AMI acute myocardial infarction; ART antiretroviral therapy; BDI Beck’s Depression Inventory-II; BMI body mass index; CAD coronary artery disease; CES-D Center for Epidemiological Studies Depression Scale; CHD coronary heart disease; CI confidence interval; CVD cardiovascular disease; EMR electronic medical record; FRS Framingham Risk Score; HDL-C high-density lipoprotein cholesterol; HR hazard ratio; ICD-9 International Classification of Diseases, 9^th Revision; ISCDRHP-10 International Statistical Classification of Diseases and Related Health Problems, 10^th Revision; LDL-C low-density lipoprotein cholesterol; MI myocardial infarction; MDD major depressive disorder; NR not reported; PCI percutaneous coronary intervention; PHQ-9 Patient Health Questionnaire-9; TIA transient ischemic attack; VA Veterans Affairs.

Inclusion and Exclusion Criteria

Included articles were published empirical studies that examined the association between depression and at least one CVD-related variable in PWH. Articles were excluded if they: (1) were nonquantitative (e.g., qualitative studies), (2) were in a language other than English, (3) were an animal or human cell study, (4) did not include an analysis of PWH, or (5) did not examine the association between depression and at least one CVD-related variable. We did not have exclusion criteria related to participant age or to type of measurement of the depression variable.

Search Protocol, Data Extraction, and Quality Assessment

First, references identified through the database searches were exported into EndNote. Second, titles and abstracts of the articles identified by the comprehensive literature search were screened by author BMP to determine eligibility for inclusion using the criteria defined above. If a study’s inclusion status was unclear after the initial screening, the full text was acquired to determine final inclusion status. Third, full texts were obtained for articles deemed eligible for inclusion. The following data were extracted from each study by author BMP: basic study information (location, years of data collection, design, sample size, recruitment locations), sample characteristics (inclusion/exclusion criteria, mean/median age, % female, % non-White, % on ART, % with HIV- and CVD-related comorbidities), and information regarding the depression and CVD-related association (depression definitions, CVD-related variable definitions, event counts, statistical analysis information, and results).

Methodological quality of included studies was assessed by author BMP using the Study Quality Assessment Tools of the National Institutes of Health (https://www.nhlbi.nih.gov/health-topics/study-quality-assessment-tools). Studies were classified as good, fair, or poor based on criteria for the respective article study design (i.e., controlled intervention studies or observational cohort and cross-sectional studies).

Results

Study Characteristics

As is shown in Supplementary Fig. S1, the systematic search identified 6,836 citations. A total of 35 studies met inclusion criteria—33 from the systematic search results and two from backward/forward referencing. Several studies examined multiple depression and/or CVD-related variables, resulting in more examined associations than studies. Of the included studies, eight examined clinical CVD [43–50], five examined subclinical CVD [46, 48, 51–53], 13 examined immune activation [51, 54–65], 17 examined systemic inflammation [51, 54, 58, 64–77], and five examined coagulation [51, 64, 74, 76, 77].

Supplementary Table S2 presents the study designs and sample characteristics of the included studies. The majority of studies were cross-sectional, conducted in the United States, and included adult participants. As shown in Tables 1–5, studies varied in their definitions of the depression and CVD-related variables and often used more than one definition.

Table 5.

Summary of Results of Included Studies Examining Coagulation (k = 5).

Author	Follow-up duration	Depression variable(s)	Coagulation variable(s)	Covariates	Results
Randomized controlled trial
Gupta [51]	24 weeks	Intervention (internet cognitive-behavioral therapy) vs. usual care	12 week change in D-dimer 24 week change in D-dimer 12 week change in fibrinogen 24 week change in fibrinogen	None.	No difference in mean change in D-dimer or fibrinogen at 12 (p = .933 and p NR, respectively) or 24 weeks (p = .749 and p NR, respectively).
Cross-sectional
Rivera-Rivera [74]	N/A	PHQ-9 ≥ 10	Platelet count	None	Depressed group had significantly lower levels of platelets than non-depressed group (p = .048).
Saylor [76]	N/A	CES-D ≥ 16	D-dimer	None.	No group differences in D-dimer between those with and without depression (p NR).
Stewart [64]	N/A	PHQ-9 total depressive symptoms PHQ-9 somatic depressive symptoms PHQ-9 cognitive/affective depressive symptoms	D-dimer	Age, sex, race/ethnicity, CVD, diabetes, hypertension, LDL-C, HDL-C, triglycerides, statin use, hepatitis C, renal function, hemoglobin, BMI, smoking, alcohol use, and cocaine abuse/dependence.	Positive trend between total depressive symptoms and D-dimer (exp(b) = 1.05, 95% CI: 1.00–1.11, p = .05–.10) and significant positive association between somatic depressive symptoms and D-dimer (exp(b) = 1.06, 95% CI: 1.00–1.11, p < .05). Associations were attenuated and non-significant in models controlling for antidepressant use, CD4 cell count, HIV RNA level, and ART use. No association observed between cognitive/affective symptoms and D-dimer (exp(b) = 1.04, 95% CI: 0.99–1.10, p NR). Results were similar when controlling for antidepressant use, CD4 cell count, HIV RNA level, and ART use.
Zuñiga [77]	N/A	PHQ-9 ≥ 10	D-dimer Fibrinogen von Willebrand factor	None.	Using a machine learning approach (random forest regression), none of the examined coagulation markers exhibited high enough feature importance (42%–71%) in predicting depression (i.e., mean frequency of selection by recursive feature elimination; selection required ≥90%).

BMI body mass index; CEDS-D Center for Epidemiological Studies Depression Scale; CI confidence interval; CVD cardiovascular disease; exp(b) exponentiated beta regression coefficient; HDL-C high-density lipoprotein cholesterol; HIV human immunodeficiency virus; LDL-C low-density lipoprotein cholesterol; NR not reported; PHQ-9 Patient Health Questionnaire-9.

Using the NIH assessment tools, six studies were deemed to be of good quality [45, 47, 52, 64, 72, 73], 13 of fair quality [44, 48, 51, 53–55, 57, 60, 67–69, 71, 74], and 16 of poor quality [43, 46, 49, 50, 56, 58, 59, 61-63, 65, 66, 70, 75–77]. The factors most commonly resulting in lower quality were lack of adjustment for confounding variables [43, 46, 49, 50, 53, 56–63, 65–71, 74–77] and small sample size (n < 100) with little to no comment on sample size justification or power [51, 53, 57, 58, 61, 62, 65–67, 69, 70, 74, 75, 77]. The included studies overwhelmingly used cross-sectional designs (k = 24), which prevented determination of the directionality of observed associations in these investigations. In addition, only five of these cross-sectional studies excluded prevalent CVD (if it was not the outcome) [57, 59, 60, 65, 72] and 16 did not report CVD status [54, 56, 58, 61–63, 66–71, 73–76]. Thus, clinical CVD could be operating as a confounder of observed associations in these investigations. Finally, four studies utilized eligibility criteria that excluded some participants with depression (e.g., presence of known psychiatric disorders, use of antidepressant medication) [53, 59, 60, 66], which may have restricted the range of the depression variable in these studies and reduced the generalizability of findings to the target population of PWH with depression.

Depression and Clinical CVD

Eight studies contributed eight longitudinal and three cross-sectional associations (see Table 1). For the eight longitudinal examinations, results are as follows: four statistically significant positive associations between history of a depressive disorder and CVD [43], major depressive disorder (MDD) and acute myocardial infarction [45], MDD and heart failure [47], and elevated depressive symptoms (Center for Epidemiological Studies Depression Scale [CES-D] ≥ 16) and CVD-related mortality [44]; two trending positive associations between history of a depressive disorder and coronary artery disease [43] and between dysthymic disorder and acute myocardial infarction [45]; and two null associations between elevated depressive symptoms (Beck Depression Inventory-II [BDI-II] ≥ 17) and vascular events [46, 48].

For the three cross-sectional examinations, results are as follows: one statistically significant positive association between depressive symptoms (CES-D) and prevalent CVD (i.e., “heart condition”) [49], and two null associations between depressive symptom groups (Patient Health Questionnaire-9 [PHQ-9] scores) and prevalent CVD [50] and between depressive symptoms (CES-D) and prevalent stroke [49].

Depression and Subclinical CVD

Five studies contributed one randomized controlled trial (RCT), one longitudinal association, and five cross-sectional associations (see Table 2). The sole RCT [51] was a pilot study that tested the effects of depression treatment (internet cognitive-behavioral therapy vs. usual care) on brachial FMD, finding no difference in mean change between treatment groups at 12 or 24 weeks. Correlational analyses were also null between changes in depressive symptoms (PHQ-9 and 20-item Hopkins Symptom Checklist [SCL-20]) and changes in brachial FMD at 12 and 24 weeks.

Table 2.

Summary of Results of Included Studies Examining Subclinical CVD (k = 5).

Author	Follow-up duration	Depression variable(s)	Subclinical CVD variable(s)	Covariates	Results
Randomized controlled trial
Gupta [51]	24 weeks	Intervention (internet cognitive-behavioral therapy) vs. usual care PHQ-9 SCL-20 Questionnaires measured at baseline, 12, and 24 weeks.	12-week change in brachial FMD 24-week change in brachial FMD	None in primary outcome analyses.	No difference in mean change in FMD at 12 (p = .323) or 24 weeks (p = .719). In a subgroup analysis of participants completing 6-8/8 intervention sessions, a nonsignificant trend for larger decrease (i.e., worsening) in FMD in the treatment group than usual care at 12 (p = .051) but not 24 weeks (p = .679). Changes in depressive symptoms (as measured by either PHQ-9 or SCL-20) were not correlated with changes in 12 or 24 week changes in FMD (all p > .2).
Longitudinal
Levy [52]	Median = 6.6 years	“High persistence of high depressive symptoms” Defined by CES-D ≥ 16 at ≥45% of semi-annual visits over a median of 13 semiannual visits during follow-up.	New carotid plaque formation (n = 61) Defined as an increase in number of focal plaques (cIMT > 1.5 mm) between baseline and end of follow-up ultrasound assessment.	Duration of follow-up, plaque at baseline, age, race/ethnicity, education, income, smoking status, alcohol use, crack/cocaine use since last visit, history of injection drug use, history of hepatitis C, menopausal status, BMI, SBP, total cholesterol, HDL-C, lipid-lowering therapy, antihypertensive medication, diabetes, nadir CD4 cell count, history of AIDS, CD4 cell count, HIV viral load, and HAART.	High persistence of high depressive symptoms was associated with an increased odds of new carotid plaque formation (OR = 1.96, 95% CI: 1.06–3.64, p = .033).
Cross-sectional
Levy [52]	N/A	CES-D ≥ 16	Carotid plaque (n = 71) Defined by cIMT > 1.5 mm.	Age, race/ethnicity, income, education, smoking status, alcohol use, crack/cocaine use since last visit, history of injection drug use, history of hepatitis, menopausal status, BMI, SBP, total cholesterol, HDL-C, lipid-lowering therapy, antihypertensive, diabetes, nadir CD4 cell count, history of AIDS, CD4 cell count, HIV viral load, and HAART.	Elevated depressive symptoms were associated with an increased odds of prevalent carotid plaque (OR = 1.97, 95% CI: 1.06–3.67, p = .032). Nonsignificant trend with addition of psychiatric medication (OR = 1.87, 95% CI: 0.99–3.55, p = .055).
Parruti [46]	N/A	BDI-II ≥ 17	cIMT Carotid plaque (n = 59) Carotid plaque defined by cIMT ≥ 1.5 mm.	None.	No difference in mean cIMT (p = .22) or frequency of carotid plaque (p = .2) between those with and without depression.
Vadini [48]	N/A	BDI-II ≥ 17	Carotid plaque (n = 175) Defined by cIMT ≥ 1.5 mm.	Age, gender, employment, HIV transmission risk factor, smoking status, total cholesterol, hypertension, diabetes, AIDS, duration HIV infection, and alexithymia.	Elevated depressive symptoms were not associated with prevalent carotid plaque (OR = 0.84, 95% CI: 0.46–1.55, p = .5).
Yaldizli [53]	N/A	German General Depression Scale	cIMT	None.	Depressive symptoms were not correlated with cIMT (p NR).

AIDS, acquired immunodeficiency virus; BDI, Beck’s Depression Inventory-II; BMI, body mass index; CES-D, Center for Epidemiological Studies Depression Scale; CI, confidence interval; cIMT, carotid intima-media thickness; CVD, cardiovascular disease; FMD, flow-mediated dilation; HAART, highly active retroviral therapy; HDL-C, high-density lipoprotein cholesterol; HIV, human immunodeficiency virus; NR, not reported; OR, odds ratio; PHQ-9, Patient Health Questionnaire-9; SBP, systolic blood pressure; SCL-20, Hopkins Symptom Checklist-20.

The sole longitudinal examination observed a statistically significant positive association between persistently elevated depressive symptoms (CES-D ≥ 16 at ≥ 45% of semiannual visits) and new carotid plaque formation [52].

For the five cross-sectional examinations, results are as follows: one statistically significant positive association between elevated depressive symptoms (CES-D ≥ 16) and prevalent carotid plaque [52] and four null associations between depressive symptoms (BDI-II ≥ 17 and German Depression Scale) and prevalent carotid plaque [46, 48] or cIMT [46, 53].

Depression and Putative Biological Mechanisms of CVD in PWH

Immune activation.

Thirteen studies contributed one RCT, nine longitudinal associations, and 36 cross-sectional associations (see Table 3). Of note, all the longitudinal studies (k = 3) examined associations between markers of immune activation at baseline and future depressive symptoms. The RCT by Gupta and colleagues [51] (N = 54) that tested the effects of internet cognitive behavioral therapy observed a larger mean reduction in soluble CD14 (sCD14) in the intervention group (versus usual care) at 12 but not 24 weeks and a larger mean reduction in sCD163 at 24 but not 12 weeks. No difference was found for soluble vascular cell adhesion molecule-1 (sVCAM-1).

Table 3.

Summary of Results of Included Studies Examining Immune Activation (k = 13).

Author	Follow-up duration	Depression variable(s)	Immune activation variable(s)	Covariates	Results
Randomized controlled trial
Gupta [51]	24 weeks	Intervention (internet cognitive-behavioral therapy) vs. usual care	12 week change in sCD14 24 week change in sCD14 12 week change in sCD163 24 week change in sCD163 12 week change in sVCAM-1 24 week change in sVCAM-1	None.	Observed a larger mean reduction in sCD14 in the intervention group than usual care at 12 (p = .014) but not 24 weeks (p = .108). Observed a larger mean reduction of sCD163 in the intervention group than usual care at 24 (p = .024) but not 12 weeks (p = .622). No group differences observed in sVCAM-1 at 12 or 24 weeks (ps NR).
Longitudinal
Lu [54]	Median = 4 visits (duration in time metric NR)	CES-D Measured at each semiannual visit. Multiple definitions examined: clinically relevant depression (>20), mild depression (12–20), and no depression (<12); CES-D ≥ 16; inclusion of antidepressant medication in prior 2 definitions (current use at time of measurement).	EIP characterized by immune activation markers sTNF-RII, sIL-2Ralpha, sCD27, B-cell activating factor, IP-10, sIL-6R, sCD14, and sGP130. Biomarkers measured at “selected study visits”.	Age, black race, college education, current smoking, obesity, hepatitis C infection, alcohol intake, and cocaine use.	A 1-SD higher EIP was associated with increased odds of clinical (OR = 1.09, 95% CI: 1.03–1.16, p NR) and mild (OR = 1.09, 95% CI: 1.03–1.16, p NR) depression. Similar results were observed for clinical depression with antidepressant inclusion (OR = 1.11, 95% CI: 1.05–1.18, p NR). No moderation was observed by ART use, HIV suppression status, or pre- vs. post- cART era. Similar results were also observed using a different CES-D cut-off without (OR = 1.08, 95% CI: 1.02–1.14, p NR) and with (OR = 1.09, 95% CI: 1.03–1.16, p NR) antidepressant inclusion. Similar results were also observed when the EIP was defined by highest vs. lowest tertiles for clinical depression with (OR = 1.21, 95% CI: 1.05–1.39, p NR) and without (OR = 1.21, 95% CI: 1.06–1.37, p NR) the inclusion of antidepressants.
McGuire [55]	Mean = 3.2 years	CES-D Measured at 6-, 12-, and 24 months.	CD8+/CD38+ Measured at baseline.	Sex, race, recent psychiatric care, recent illicit drug use, and recent alcohol use.	Higher levels of CD8+/CD38+ were associated with lower depressive symptoms at 12 months (β = −.954, 95% CI: −1.854 to .054, p = .038). Results were similar when examining outcomes at 6 and 24 months.
Veenhuis [56]	Median = 1 year	CES-D Measured at follow-up.	% CD14+CD16+ of subset of MNCs with TLR2+ Measured at baseline.	None.	No significant correlation between baseline % of CD14+CD16+ of MNCs subset and depressive symptoms at follow-up (rs = .23, p NR).
Cross-sectional
Evans [57]	N/A	Current MDD (structured clinical interview) HDRS-17 HDRS-11 (excludes somatic symptoms)	CD8+/CD38+/DR+	ART use and viral load.	MDD was not correlated with CD8+/CD38+/DR+ (r = .18, p = .22). Depressive symptoms, as measured by the HDRS-17 and HDRS-11 (excludes somatic symptoms), were positively correlated with CD8+/CD38+/DR+ (r = .30, p = .03; r = .28, p = .05, respectively).
Gold [58]	N/A	BDI-II POMS-Depression-Dejection	CSF neopterin	None.	Depressive symptoms, as assessed by either measure, were not correlated with CSF neopterin (ps NR).
Greeson [59]	N/A	BDI	CD3+/CD8+/CD38+ CD3+/CD8+/HLADR+	None.	Depressive symptoms were not correlated with CD3+/CD8+/CD38+ (r = .06) or CD3+/CD8+/HLADR+ (r = .05, ps>.05).
Hellmuth [60]	N/A	PHQ-9 ≥ 10 HADS-D ≥ 8	Plasma neopterin CSF neopterin	Viral load, CD4 count, and time on ART (or time since diagnosis).	Depression, as defined by PHQ-9 and HADS-D, was positively associated with plasma neopterin (p = .001 and .007, respectively), controlling for covariates. PHQ-9 was correlated with plasma neopterin (r = .29, p = .045). Depressed group defined by either measure had higher plasma (p = .011) but not CSF neopterin (p NR). Depressed group defined by HADS-D had higher CSF neopterin (p = .028).
Kemeny, 1994 [61]	N/A	POMS-Depression-Dejection	CD8+/CD38+	None.	Depression was correlated with CD8+/CD38+ among non-bereaved participants (r = .54, p < .01) but not among bereaved participants (r = −0.41, p NR).
Kemeny, 1995 [62]	N/A	POMS-Depression-Dejection	Neopterin CD8+/HLA-DR+ CD8-/HLA-DR+	In separate partial correlation analyses: alcohol, cigarettes, exercise, sleep hours, HIV-related symptoms, or sexual behavior.	Among bereaved participants, no significant correlations were observed between depression and neopterin, CD8+/HLA-DR+ or CD8-/HLA-DR+ (ps NR). Among nonbereaved participants, depression was correlated with CD8-/HLA-DR+ (r = .56, p < .05). This association remained significant when controlling for covariates in separate partial correlation analyses (r range = .54–.60, p < .05) and was unaffected by exclusion of participants with current drug use (r = .51, p < .10) or CNS impairment (r = .57, p < .05).
Schroecksnadel [63]	N/A	BDI ≥ 10	Plasma neopterin Urinary neopterin	None.	Depressed group had higher plasma and urinary neopterin than nondepressed group (p < .01). Both plasma (OR = 1.021, 95% CI: 1.004–1.024, p = .018) and urinary neopterin (OR = 1.002, 95% CI: 1.001–1.004, p = .003) were associated with depression. In the total sample, depression was positively correlated with neopterin (r = .295, p < .001). Among those not on an antidepressant, depression was positively correlated with neopterin (r = .331, p < .001) but not among those on an antidepressant.
Stewart [64]	N/A	PHQ-9 total depressive symptoms PHQ-9 somatic depressive symptoms PHQ-9 cognitive/affective depressive symptoms	sCD14	Age, sex, race/ethnicity, CVD, diabetes, hypertension, LDL-C, HDL-C, triglycerides, statin use, hepatitis C, renal function, hemoglobin, BMI, smoking, alcohol use, and cocaine abuse/dependence.	Non-significant positive trend for association between total depressive symptoms and sCD14 (exp(b) = 1.01, 95% CI: 1.00–1.03, p = .05–.10). No longer a trend with the addition of antidepressant medication, CD4 cell count, HIV RNA level, and ART use. A 1-SD increase in somatic depressive symptoms was associated with an increase in sCD14 (exp(b) = 1.02, 95% CI: 1.00–1.03, p < .05). Similar results were observed with the addition of antidepressant medication (p < .05), but this association became a nonsignificant trend with addition of CD4 cell count, HIV RNA level, and ART use (p = .5–.10). No significant association between cognitive/affective depressive symptoms and sCD14 (exp(b) = 1.01, 95% CI: 0.99–1.02, p NR). Similar results were observed with the addition of antidepressant medication, CD4 cell count, HIV RNA level, and ART use (ps NR).
Warriner [65]	N/A	BDI Cognitive/Affective (Cog-Aff) score	Neopterin	None.	Cog-aff symptoms were positively correlated with neopterin (r = .28, p < .05). Among those on an antidepressant, cog-aff symptoms were correlated with neopterin (r = .83, p < .001) but not among those not on an antidepressant (r = .03, p > .05). Cog-aff depression group (Cog-aff score >10) had higher neopterin than no depression group among those on an antidepressant (F = 45.7, df = 1,11, p < .001) but not among those not on an antidepressant (p > .05).
Veenhuis [56]	N/A	CES-D	% CD14+CD16+ of total MNCs %CD14+CD16+ of subset of MNCs with TLR2+	None.	No significant correlations between % of CD14+CD16+ defined as proportion of total MNCs (r = .01) or as proportion of subset of MNCs (r = .02, p NR) and depressive symptoms.

ART antiretroviral therapy; BDI Beck’s Depression Inventory; BMI body mass index; cART combined antiretroviral therapy; CEDS-D Center for Epidemiological Studies Depression Scale; CI confidence interval; CNS central nervous system; CSF cerebrospinal fluid; CVD cardiovascular disease; EIP exploratory factor analysis-identified inflammatory process; exp(b) = exponentiated beta regression coefficient; HADS-D Hospital Anxiety and Depression Scale; HDL-C high-density lipoprotein cholesterol; HDRS Hamilton Depression Rating Scale; HIV human immunodeficiency virus; IP-10 interferon gamma-induced protein 10; LDL-C low-density lipoprotein cholesterol; MDD major depressive disorder; MNC mononuclear cells: NR not reported; OR odds ratio; PHW-9 Patient Health Questionnaire-9; POMS Profile of Mood States; sCD14 soluble CD14; sCD27 soluble CD27; sCD163 soluble CD163; SD standard deviation; sGP130 soluble GP130; sIL-2Ralpha soluble interleukin-2 receptor alpha; sIL-6R soluble interleukin-6 receptor; sTNF-RII soluble tumor necrosis factor receptor II; sVCAM-1 soluble vascular cell adhesion molecule-1; TLR2 toll-like receptor 2.

For the nine longitudinal examinations, results are as follows: seven positive associations between a higher composite measure of immune activation and future depression (multiple definitions used based on CES-D cut points and inclusion of antidepressant medication) [54]; one null association between %CD14+CD16+ and future depressive symptoms (CES-D) [56]; and one negative association between CD8+/CD38+ and future depressive symptoms (CES-D) [55].

For the 36 cross-sectional examinations, the results are as follows: 19 positive associations between depressive symptoms and neopterin [60, 63, 65, 70], CD8+/CD38+/DR+ [57], CD8-/HLA-DR+ [62], CD8+CD38+ [61], and sCD14 [64]; one trending positive association between depressive symptoms and sCD14 [64]; and 16 null associations between current MDD and CD8+/CD38+/DR+ [57] and between depressive symptoms and neopterin [58, 60, 62, 63, 65], CD8+/CD38+/DR+ [57], CD3+/CD8+/CD38+ [59], CD3+/CD8+/HLA-DR+ [59], CD8-/HLA-DR+ [62], CD8+/HLA-DR+ [62], CD8+CD38+ [61], %CD14+CD16+ [56], and sCD14 [64].

Systemic inflammation.

Seventeen studies contributed one RCT, eight longitudinal associations, and 88 cross-sectional associations (see Table 4). In the sole RCT [51] that tested the effects of internet cognitive behavioral therapy, no difference in mean change in C-reactive protein (CRP) or IL-6 was observed between the intervention and usual care groups at 12 or 24 weeks.

Table 4.

Summary of Results of Included Studies Examining Systemic Inflammation (k = 17).

Author	Follow-up duration	Depression variable(s)	Inflammation variable(s)	Covariates	Results
Randomized controlled trial
Gupta [51]	24 weeks	Intervention (internet cognitive-behavioral therapy) vs. usual care	12 week change in CRP 24 week change in CRP 12 week change in IL-6 24 week change in IL-6	None.	No difference in mean change in CRP or IL-6 at 12 (p = .902 and p = .537, respectively) or 24 weeks (p = .332 and p = .269, respectively).
Longitudinal
Lu [54]	Median = 4 visits (duration in time metric NR)	CES-D Measured at each semiannual visit. Multiple definitions examined: clinically relevant depression (>20), mild depression (12–20), and no depression (<12); CES-D ≥ 16; inclusion of antidepressant medication in prior 2 definitions.	CRP EIP characterized by inflammation markers IL-6, IL-8, TNF-alpha, and MIP-1beta. Biomarkers measured at “selected study visits”.	Age, black race, college education, current smoking, obesity, hepatitis C infection, alcohol intake, and cocaine use.	A 1-SD higher CRP was associated with an increased odds of clinical depression (OR = 1.04, 95% CI: 1.00–1.09, p NR). A 1-SD higher EIP was associated with an increased odds of mild (OR = 1.11, 95% CI: 1.05–1.18, p NR) but not clinical depression without (OR = 1.04, 95% CI: 0.99–1.10, p NR) and with (OR = 1.02, 95% CI: 0.97–1.08, p NR) inclusion of antidepressants. No moderation was observed by ART use (yes/no), HIV suppression status, or pre- vs. post-cART era. A positive association was observed with the use of a different CES-D cut-off (OR = 1.07, 95% CI: 1.01–1.13, p NR) but this became nonsignificant with inclusion of antidepressants (OR = 1.04, 95% CI: 0.99–1.09, p NR). A nonsignificant trend was observed when EIP was defined by highest vs. lowest tertile without the inclusion of antidepressants (OR = 1.08, 95% CI: 0.96–1.22, p NR), which became significant with inclusion of antidepressants (OR = 1.12, 95% CI: 1.00–1.26, p NR).
Cross-sectional
Bekhbat [66]	N/A	CES-D ≥ 16	LPS-induced IL-6 LPS-induced TNF-alpha	Groups frequency matched on marijuana use.	Depressed group exhibited higher LPS-induced TNF-alpha than nondepressed group (p = .023). No group difference was observed for LPS-induced IL-6 (p NR).
Cassol [67]	N/A	BDI > 14 (NNTC/CHARTER) CES-D>16 (ALIVE)	IL-1beta IL-6 IL-8 IL-10 IL-12 MCP-1 MIGIP-10 IFN-alpha IFN-gamma	Groups matched on age, gender, race, stage of disease (current and nadir CD4), viral load, and ART use.	Depressed group exhibited significantly higher levels of IP-10 (p = .011) than nondepressed group. A positive trend was observed for MIG (p = .058) and IFN-gamma (p = .084). No group differences were observed for IL-1beta, IL-6, IL-8, IL-10, IL-12, MCP-1, or IFN-alpha (p NR).
Derry [68]	N/A	CES-D-10	CRP Composite score including IL-6, TNF-alpha, and IFN-gamma	Age, sex, race, BMI, VACS score, smoking status, and anti-inflammatory medications.	No significant correlation between depressive symptoms and CRP (r = .02, p = .82). A positive association was observed between depressive symptoms and higher levels of the composite inflammation score (β = .22, p = .008). In unadjusted models, no moderation was observed by sex or age (did not test in adjusted model).
Fumaz [69]	N/A	HADS-D	IL-6	Age, sex, infection route, years since HIV diagnosis, years on ART, current ART, other medications, hepatitis B or C status, CD4 cell count, CD8 cell count, CD4/CD8 ratio, exercise, sleep quality, healthy diet, sexual risk intercourse, drug intake, ART adherence, and psychological stress.	Depressive symptoms were positively correlated with IL-6 (r = .41, p = .000). Depressive symptoms were positively associated with IL-6 in univariate (b = .64, SE = .20, p < .01) but not multivariate analyses (p NR).
Ghosh [70]	N/A	CES-D > 16	IL-1alpha IL-1beta IL-6 IL-8 MIP-3alpha MCP-1IP-10 TNF-alpha	CD4 count, viral load, and chronic sexual abuse.	Depression was significantly associated with lower MIP-3alpha (b = −1.00, p < .01). No association was observed between depression and IL-1alpha, IL-1beta, IL-6, IL-8, MCP-1, IP-10, or TNF-alpha (p > .18).
Gold [58]	N/A	BDI-II POMS-Depression-Dejection	MCP-1 IP-10	None.	Depressive symptoms, as assessed by either measure, were not correlated with MCP-1 or IP-10 (p NR).
Musinguzi [71]	N/A	Current MDD (structured clinical interview)	IL-6 TNF-alpha CRP	Age, sex, highest educational attainment, study site, IL-6, CRP, and TNF-alpha.	CRP was not associated with MDD (p = .60). IL-6 was non-linearly (100/[IL-6 + 1]), positively associated with MDD (OR = 0.98, 95% CI: 0.97–0.99, p < .001). Compared to no TNF-alpha, low levels of TNF-alpha were associated with a decreased odds of MDD (50–<500 pg/mL: OR = 0.31, 95% CI: 0.10–0.94), whereas high levels of TNF-alpha were associated with an increased odds of MDD (500–<1,000 pg/mL: OR = 3.98, 95% CI: 1.29–12.33; ≥1,000 pg/mL: OR = 7.24, 95% CI: 1.65–31.82).
Norcini Pala [72]	N/A	PHQ-9 Defined by latent class analysis subtypes: “severe depressive symptoms”, “moderate/severe somatic symptoms” and “absent depressive symptoms”	IL-6 TNF-alpha	Age, gender, viral load, CD4 cell count, ART use, monocyte count.	The somatic group exhibited a trend for higher TNF-alpha and IL-6 than the no depression group (p > .05). The somatic group exhibited a significantly higher IL-6 (p < .05) and a trend for higher TNF-alpha than the severe group (p > .05). No group differences were observed between the severe and no depression groups for TNF-alpha or IL-6 (p > .05).
Poudel-Tandukar [73]	N/A	BDI	CRP	Age, marital status, education, occupation, alcohol intake, smoking status, physical activity, BMI, history of any disease in the past 12 months, SBP, cholesterol, triglycerides, CD4 cell count, duration of ART use, type of ART.	CRP was significantly associated with depressive symptoms in men (b = 1.06, SE = 0.43, p = .015) and women (b = 1.29, SE = 0.62, p = .038). CRP > 3 mg/L group exhibited a significantly higher odds of depression (BDI ≥ 20) than CRP ≤ 3 mg/L group in men (OR = 3.68, 95% CI: 1.56–8.64, p = .002) but not women (OR = 1.66, 95% CI: 0.56–4.88, p = .350).
Rivera-Rivera [74]	N/A	PHQ-9	IL-1alpha IL-1beta IL-15 MIP-1alpha MIP-1beta MCP-1 IP-10 IFN-gamma	None.	Depression was correlated with IL-15 (r = .47, p = .011) and IP-10 (r = .35, p = .05). Depressed group (PHQ-9 ≥ 10) had significantly higher levels of IL-15 (p = .044) and IP-10 (p = .025) than non-depressed group. No group differences were observed for IL-1alpha, IL-1beta, MIP-1alpha, MIP-1beta, MCP-1, or IFN-gamma (p NR).
Rubin [75]	N/A	Remitted MDD (structured clinical interview)	IL-1beta IL-6 IL-8 CRP MMP-9 TNF-alpha IP-10 MCP-1 MIG	None.	Among women with HIV, the remitted MDD group exhibited higher levels of IL-8, CRP, IL-6, and TNF-alpha than the no depression group (p < .05). No group differences observed for MMP-9, IL-1beta, IP-10, MCP-1, or MIG (p NR). Among men with HIV, the remitted MDD group exhibited higher levels of IL-8, IL-1beta, and TNF-alpha but lower levels of IP-10 than the no depression group (p < .05). No group differences observed for CRP, MMP-9, IL-6, MCP-1, or MIG (p NR).
Saylor [76]	N/A	CES-D ≥ 16	IL-6	None.	Depressed group exhibited higher mean levels of IL-6 than non-depressed group (p = .03) but this was a non-significant trend when median levels of IL-6 were compared (p = .06).
Stewart [64]	N/A	PHQ-9 total depressive symptoms PHQ-9 somatic depressive symptoms PHQ-9 cognitive/affective depressive symptoms	IL-6	Age, sex, race/ethnicity, CVD, diabetes, hypertension, LDL-C, HDL-C, triglycerides, statin use, hepatitis C, renal function, hemoglobin, BMI, smoking, alcohol use, and cocaine abuse/dependence.	No significant associations observed between total (exp(b) = 1.00, 95% CI: 0.96–1.04, p NR), somatic (exp(b) = 1.01, 95% CI: 0.98–1.05, p NR), or cognitive/affective depressive symptoms (exp(b) = 0.99, 95% CI: 0.95–1.03, p NR) and IL-6. Similar results were observed with the addition of antidepressant medication, CD4 cell count, HIV RNA level, and ART use.
Warriner [65]	N/A	BDI Total and Cognitive/Affective (Cog-Aff) score	IL-6 mRNA expression TNF-alpha mRNA expression TNF-alpha serum	None.	A significant difference in IL-6 mRNA expression was observed between depression groups (F = 7.7, df = 3,24, p < .001). The severe depression group (BDI > 29) exhibited higher IL-6 mRNA expression than the moderate (BDI 17–29), mild (BDI 10–16), and minimal depression groups (BDI < 10; pairwise comparison p < .01). Cog-aff depressive symptoms were positively associated with IL-6 mRNA expression (r = .40, p < .05), even after adjustment for subjective fatigue and cognitive complaints in partial correlations (r = .39, p < .05). No association was observed between cog-aff depressive symptoms and either measure of TNF-alpha.
Zuñiga [77]	N/A	PHQ-9 ≥ 10	IL-1beta IL-6 sIL-1RII sIL-6R TNF-alpha sTNF-RI sTNF-RII CRP adiponectin	None.	Using a machine learning approach (random forest regression), it was observed that TNF-alpha had a feature importance of 97% (i.e., mean frequency of selection by recursive feature elimination; selection required ≥90%) in predicting depression. The other inflammatory markers exhibited low feature importance (30-62%).

ALIVE AIDS linked to the intravenous experience; ART antiretroviral therapy; BDI Beck’s Depression Inventory; BMI body mass index; CES-D Center for Epidemiological Studies Depression Scale; CHARTER CNS HIV Antiretroviral Therapy Effects; CI confidence interval; CRP C-reactive protein; EIP Exploratory factor analysis-identified inflammatory process; exp(b) exponentiated beta regression coefficient; HADS-D Hospital Anxiety and Depression Scale; HDL-C high-density lipoprotein cholesterol; HIV human immunodeficiency virus; IL interleukin; IFN interferon; IP-10 interferon gamma-induced protein 10; LDL-C low-density lipoprotein cholesterol; LPS lipopolysaccharide; MCP-1 monocyte chemoattractant protein-1; MDD major depressive disorder; MIG monokine induced by gamma interferon; MIP macrophage inflammatory protein; MMP-9 matrix metalloproteinase-9; mRNA messenger ribonucleic acid; NNTC National NeuroAIDS Tissue Consortium; NR not reported; OR odds ratio; PHQ-9 Patient Health Questionnaire-9; POMS Profile of Mood States; SBP systolic blood pressure; SD standard deviation; SE standard error; sIL-1RII soluble interleukin-1 receptor I; sIL-6R soluble interleukin-6 receptor; sTNF-RI soluble tumor necrosis factor receptor I; sTNF-II soluble tumor necrosis factor receptor II; TNF-alpha tumor necrosis factor alpha; VACS Veterans Aging Cohort Study.

The only longitudinal study examined associations between baseline inflammation (CRP or a composite score of systemic inflammation) and future depressive symptoms (multiple definitions used based on CES-D cut points and inclusion of antidepressant medication) [54]. Four positive associations were reported between CRP and clinical depressive symptoms (CES-D > 20, no antidepressants included in definition), composite inflammation and mild depressive symptoms (CES-D 12-20, no antidepressants included in definition), composite inflammation and clinical depressive symptoms (CES-D ≥ 16, no antidepressants included in definition), and between highest tertile of inflammation and clinical depressive symptoms (CES-D ≥ 16, antidepressants included in definition). One trending positive association was found between highest tertile of inflammation and clinical depressive symptoms (CES-D > 20, no antidepressants included in definition). Three null associations were observed between composite inflammation and clinical depressive symptoms (CES-D > 20, with and without inclusion of antidepressants in the definition) and clinical depressive symptoms (CES-D ≥ 16, with antidepressants included in definition).

The cross-sectional results are as follows. There are 27 significant positive associations between depressive symptoms and tumor necrosis factor-alpha (TNF-alpha) [66, 77], interleukin-6 (IL-6) [65, 69, 72, 76], IL-15 [74], CRP [73], interferon gamma-induced protein-10 (IP-10) [67, 74], and a composite measure of inflammation [68]; between remitted MDD and CRP, IL-1beta, IL-6, IL-8, and TNF-alpha [75]; and between current MDD and IL-6 and TNF-alpha [71]. There are six trending positive associations between depressive symptoms and TNF-alpha [72], interferon gamma (IFN-gamma) [67], IL-6 [72, 76], and monokine induced by interferon-gamma (MIG) [67]. There are 53 null associations between depressive symptoms and TNF-alpha [65, 70, 72], interferon alpha (IFN-alpha) [67], IFN-gamma [74], IL-6 [64, 66, 67, 69, 70, 72, 77], IL-1alpha [70, 74], IL-1beta [67, 70, 74, 77], IL-10 [67], IL-12 [67], IL-8 [67, 70], monocyte chemoattractant protein-1 (MCP-1) [58, 67, 70, 74], macrophage inflammatory protein 1-alpha (MIP-1alpha) [74], macrophage inflammatory protein 1-beta (MIP-1beta) [74], IP-10 [58, 70], CRP [68, 73, 77], soluble IL-1 receptor type II (sIL-1RII) [77], soluble IL-6 receptor (sIL-6R) [77], soluble tumor necrosis factor receptor I (sTNF-RI) [77], soluble tumor necrosis factor receptor II (sTNF-RII) [77], and adiponectin [77]; between remitted MDD and CRP, IL-6, matrix metalloproteinase-9 (MMP-9), MCP-1, MIG, IP-10, and IL-1beta [75]; and between current MDD and CRP [71]. Finally, there are two negative associations between depressive symptoms and macrophage inflammatory protein 3-alpha (MIP-3alpha) [70] and between remitted MDD and IP-10 [75].

Altered coagulation.

Five studies contributed one RCT and seven cross-sectional associations (see Table 5). In the RCT [51] that tested the effects of internet cognitive behavioral therapy, no difference in mean changes in D-dimer or fibrinogen was observed between the intervention and usual care groups at 12 or 24 weeks.

The cross-sectional results are as follows: one significant positive association between somatic depressive symptoms and D-dimer [64]; one trending positive association between total depressive symptoms and D-dimer [64]; four null associations between total depressive symptoms and D-dimer [76, 77], fibrinogen [77], and von Willebrand factor [77] and between cognitive/affective depressive symptoms and D-dimer [64]; and one negative association between elevated depressive symptoms and platelet counts [74].

Discussion

Our first objective was to systematically review the literature examining the associations of depression with clinical CVD, subclinical CVD, and putative biological mechanisms of HIV-CVD. For clinical CVD, the existent literature in PWH is small in size (k = 8) but it suggests the following conclusions. First, the majority of the literature supports a positive, longitudinal association between depression and clinical CVD among PWH. It is worth noting that the two longitudinal studies reporting null results may be underpowered [46, 48]. Second, this positive association was observed in investigations controlling for traditional CVD risk factors (e.g., BMI, diabetes, dyslipidemia, and smoking) [44, 45, 47]. Although these three studies included many traditional CVD risk factors as covariates, only one included all of them in the same model [47]. Third, this positive association was found in studies controlling for HIV-specific factors [44, 45, 47] and in a study using a virologically suppressed sample [43], suggesting that depression may also independently confer CVD risk beyond that of uncontrolled viral load, low CD4 count, and ART. Fourth, this positive association was detected in three of four studies examining incident CVD outcomes [45, 48, 51]. Based on the results of this review, depression appears to be a risk marker for clinical CVD in PWH and may be an independent risk factor, increasing its potential to serve as a CVD prevention target among PWH.

For subclinical CVD, minimal research has been conducted, with only five studies identified in PWH. The sole RCT of depression treatment [51] in this review did not observe differences in change in brachial FMD between the intervention and usual care groups at 12 or 24 weeks. The only longitudinal study found an association between persistently high depressive symptoms and increased odds of new carotid plaque formation in its longitudinal analysis and of prevalent plaque in its cross-sectional analysis [52]. The other contributing cross-sectional studies report null associations between depressive symptoms and either carotid plaque or cIMT [46, 48, 53]. Thus, it is currently unclear whether depression is associated with subclinical CVD in PWH, particularly what types of subclinical CVD.

For putative biological mechanisms of HIV-CVD, there is a larger literature of mixed, but promising, evidence that depression may be related to immune activation (k = 13) and systemic inflammation (k = 17) in PWH. Most of this literature consists of cross-sectional studies that utilized small convenience samples of PWH and lacked attention to potential confounders, either through study eligibility criteria or statistical adjustment. However, one RCT of depression treatment was identified. This RCT found larger reductions in immune activation markers sCD14 at 12 weeks and sCD163 at 24 weeks in the intervention group versus the usual care group, providing initial evidence that improvements in depression may result in subsequent decreases in immune activation [51]. This RCT detected no treatment group differences for changes in inflammation markers at 12 or 24 weeks. The extant longitudinal studies focused on immune function’s potential role in the pathophysiology of depression, examining the direction of biological mechanism to future depression [54–56]. No longitudinal studies (with the exception of the RCT) were found examining associations of depression with subsequent changes in immune activation or systemic inflammation over time. There was no clear pattern of results across the cross-sectional studies for immune activation or systemic inflammation, with studies of various methodological qualities reporting contrasting results. Potential contributors to these mixed results include substantial heterogeneity among the studies in sample characteristics and methodology (as summarized in Supplementary Table S2). For example, studies widely differed on the study eligibility criteria, sample percent on ART, and gender breakdown. Of particular importance is the heterogeneity in the study inclusion criterion of virologic suppression and/or statistical adjustment for viral load. This may be an important source of confounding in the literature, as HIV disease progression could contribute to increased depressive symptoms, immune activation, and systemic inflammation.

For altered coagulation, minimal research has been conducted, with only five studies identified in PWH. The sole RCT of depression treatment [51] did not observe differences in changes in D-dimer or fibrinogen between the intervention and usual care groups at 12 or 24 weeks. The few cross-sectional studies provided mixed evidence and displayed significant between study heterogeneity in sample characteristics and methodological quality. For example, only one study performed a multivariable adjusted analysis [64], one utilized a machine learning method [77], and all differed considerably in the percentage of participants on ART.

Although a full discussion of the role of antidepressant use in the depression-CVD association is outside this review’s scope, there are important points worth noting. General population research is mixed regarding whether antidepressants reduce CVD risk, though adjusting for them often attenuates the depression-CVD association [78]. General population research on antidepressant class shows some evidence that selective serotonin reuptake inhibitors (SSRIs) are generally safe and perhaps beneficial for cardiovascular health due to modest anti-inflammatory and anticoagulatory effects [79, 80]. In contrast, serotonin norepinephrine reuptake inhibitors (SNRIs) have been associated with increases in heart rate and blood pressure, monoamine oxidase inhibitors (MAOIs) with hypotensive and hypertensive events, and tricyclic (TCAs) and tetracyclic (TeCAs) antidepressants with increased risk of arrhythmias [80]. To the best of our knowledge, only one study has reported associations between antidepressant use and biological mechanisms of HIV-CVD, observing that SSRIs related to lower and TCAs related to higher immune activation and inflammation markers [64]. Of the studies included in this review, two used current/recent antidepressant use as an exclusion criterion [59, 66], both of which observed positive associations between depressive symptoms and inflammation markers. One study incorporated antidepressant use into the depression variable definition [54], which largely did not affect the positive associations observed between immune activation and future depressive symptoms. Five other studies addressed potential confounding by antidepressant use through statistical adjustment [45, 52, 64], which typically resulted in slightly attenuated associations, or through stratification [63, 65], with studies conflicting on whether there was a positive association among those with or without antidepressant use. Overall, these initial findings suggest that the depression-CVD association in PWH is not accounted for by antidepressant use alone, providing further support that other mechanisms are at work. These findings also highlight the need for future research examining separate antidepressant classes, as their effects on CVD-related outcomes (beneficial vs. nil vs. harmful) and their roles in the depression-CVD association (mechanism vs. moderator) in PWH may vary by medication class.

Our second objective was to propose a future research agenda. We recommend the following for future research endeavors to elucidate depression’s role in HIV-CVD:

• 1) conduct prospective analyses in large, representative samples of PWH to confirm depression’s association with incident clinical CVD and to determine whether depression is consistently and meaningfully associated with future changes in subclinical CVD and putative biological mechanisms of HIV-CVD;

• 2) address baseline CVD as a potential confounder of associations of depression with CVD-related outcomes in PWH by excluding those with CVD at baseline or adjusting for this factor in statistical models;

• 3) adjust statistical models for traditional CVD risk factors, nontraditional CVD risk factors, and HIV-specific factors (e.g., viral load, CD4+ count, and ART use) to determine whether depression is independently associated with CVD-related outcomes;

• 4) examine overlap among depressive symptoms (including the somatic and cognitive/affective clusters), HIV symptoms, and ART side effects to facilitate teasing apart their relative contributions to CVD-related outcomes and to refine depression treatment targets;

• 5) assess whether associations of depression with CVD-related outcomes are independent, mediated, or confounded by contextual factors unique to HIV summarized below;

• 6) test whether sociodemographic or the unique contextual factors summarized below are moderators of associations of depression with CVD-related outcomes, which could identify subgroups in whom these associations are particularly strong and for whom intervention is especially warranted;

• 7) conduct randomized controlled trials to determine whether successful treatment of depression (e.g., via antidepressant medications and/or psychotherapy) improves CVD-related outcomes;

• 8) evaluate the possible roles of antidepressant use (e.g., mechanism, confounder, or moderator) in the depression-CVD association in PWH, ideally examining antidepressant classes (e.g., SSRIs, SNRIs, and TCAs) separately given their potential differential effects.

The aforementioned unique factors for consideration in future research are the medical and environmental/social contexts of PWH. Though potentially relevant to the depression-CVD association in the general population, these unique contextual factors likely play a greater role in the HIV population. Regarding the unique medical context, some groups of PWH exhibit high rates of hepatitis C co-infection and illicit drug use (e.g., cocaine) [81]. These comorbidities show strong associations with depression [82–84] and contribute to an increased risk of clinical CVD among PWH [18, 26]. Concerning the unique environmental/social context, some PWH are at greater risk for experiencing negative social determinants of health, such as inequitable access and provision of medical care [85, 86], individual- and structural-level HIV-related stigma, and discrimination [87]. HIV status may intersect with other traditionally marginalized groups (e.g., racial/ethnic minority, low socioeconomic status, and LGBTQ identities), calling for researchers and clinicians to attend to the unique intersectionality found among PWH that may influence the depression-CVD association. Another facet of the unique environmental/social context is the high prevalence of trauma experienced by PWH. It is well documented that some PWH have a higher burden of traumatic experiences than the general population, with the highest burden among those with more than one marginalized identity [88]. Given its frequent comorbidity with depression [89] and its potentially positive association with CVD in populations without HIV [90], trauma is an important factor to incorporate in future research.

Although the implications of our review are largely related to research, there are some clinical implications. First, this review should raise awareness among HIV providers of the possible role of depression in promoting CVD in PWH and that PWH with depression represent a subgroup at elevated CVD risk. Second, this awareness could lead to increased screening, identification, and treatment of depression in PWH, ideally where they receive their primary HIV care. The use of depression treatment as a complement to traditional CVD prevention approaches remains an under-addressed research question. However, the identification of this subgroup at elevated CVD risk creates an opportunity for providers to (a) address depression, which is worthy of treatment in its own right, and (b) closely monitor PWH with depression to determine the need for early intervention on established CVD risk factors.

Limitations of this systematic review should be considered. First, our review included only English-language, peer-reviewed, published studies potentially leading to a bias in our sample of studies. Second, the study identification, data extraction, and quality assessment were conducted by a single author, which precluded our ability to assess reliability. Third, our review investigated a limited set of potential pathways underlying the association between depression and HIV-CVD. Specifically, we examined biological mechanisms identified in general population depression-CVD research that overlapped with putative biological mechanisms of HIV-CVD. Thus, candidate behavioral mechanisms (e.g., smoking, diet quality, and physical activity) were not examined.

Future efforts addressing the aforementioned research priorities will undoubtedly lead to the creation of a comprehensive conceptual framework. The long-term goals of this work are (a) to determine if depression is a causal and modifiable risk factor for CVD among PWH and (b) to evaluate whether evidence-based depression treatment may serve as a novel CVD prevention strategy in PWH that could complement current prevention approaches. It is our hope that the present review will raise researcher and clinician awareness of the possible cardiotoxic effects of depression in PWH and will motivate future research efforts in this area of high potential clinical significance.

Conflicts of Interest and Sources of Funding

Compliance with Ethical Standards

Authors’ Statement of Conflict of Interest and Adherence to Ethical Standards Authors Brittanny M. Polanka, Samir K. Gupta, Kaku A. So-Armah, Matthew S. Freiberg, Tamika C.B. Zapolski, Adam T. Hirsh and Jesse C. Stewart declare that they have no conflict of interest.

Authors’ Contributions BMP, SKG, JCS conceived the review. BMP was the primary writer and conducted the search and data extraction. All authors contributed to the development of the search strategy and draft review/editing.