Traditional risk factors and D-dimer predict incident cardiovascular disease events in chronic HIV infection

Objective: Cardiovascular disease (CVD) contributes significantly to HIV-related morbidity and mortality. Chronic immune activation and inflammation are thought to augment the progression of atherosclerotic disease. In this retrospective, case–control study of HIV-infected individuals, we investigated the association of traditional cardiac risk factors, HIV-related disease, and inflammation with CVD events.

Methods: HIV-infected individuals who experienced an incident CVD event while enrolled in National Institutes of Health clinical protocols from 1995 to 2009 were matched 2: 1 to HIV-infected individuals without known CVD. Markers of inflammation and cell activation were measured in serum or plasma using ELISA-based assays and peripheral mononuclear cells by four-color flow cytometry.

Results: Fifty-two patients experienced an incident CVD event. Events were related to smoking, dyslipidemia, hyperglycemia, and family history as well as elevated D-dimer, soluble vascular cell adhesion molecule-1, tissue inhibitor of metalloproteinase-1, and soluble tissue factor, but not high-sensitivity C-reactive protein. No significant differences in antiviral therapy, CD4⁺ T-cell count, or CD38 and human leukocyte antigen-DR expression were identified between patients and controls. In multivariable analysis, smoking, family history, D-dimer, and glucose were independently related to CVD risk.

Conclusion: In this cohort, CVD risk was related to traditional CVD risk factors and markers of thrombosis and endothelial damage, but not to high-sensitivity C-reactive protein or markers of T-cell activation such as CD38/human leukocyte antigen-DR coexpression. D-dimer may help identify HIV-infected patients at elevated CVD risk.

Introduction

Non-infection-related morbidity and mortality has become increasingly significant in the care and management of HIV-infected patients treated with combination antiretroviral therapy (ART). An increased risk of myocardial infarction (MI) in HIV-infected patients was described in 2007 by Triant et al. ^[1]. In this large cohort, the relative risk of MI was 1.4-fold greater in men with HIV and three-fold greater in women with HIV. HIV infection is associated with known cardiovascular risk factors such as increased total cholesterol, decreased high-density lipoprotein (HDL) cholesterol, lipodystrophy, and the metabolic syndrome ^[1,2]. Furthermore, the Data Collection on Adverse Events of Anti-HIV Drugs (DAD) study group showed that the risk of MI and cardiac death increased with duration of ART ^[3]. Although much of the increased risk associated with protease inhibitor use is attributable to related dyslipidemia, the nucleoside reverse transcriptase inhibitors abacavir and, to a lesser degree, didanosine, have also been linked to increased cardiovascular disease (CVD) risk in some cohorts ^[4,5].

Contrary to expectations that reduced exposure to antiretrovirals would decrease CVD risk in HIV-infected patients, the Strategies for Management of Antiretroviral Therapy (SMART) study of CD4⁺ T-cell count-guided ART demonstrated that CVD events increased in the treatment interruption arm [hazard ratio 1.57 (95% confidence interval (CI) 1.00–2.46, P = 0.05)] over the continuous viral suppression arm ^[6,7]. This observation suggests that HIV infection itself is an important contributor to CVD. HIV infection is also associated with greater carotid artery intimal medial thickness ^[8] and reduced arterial elasticity ^[9,10], both surrogate markers of vascular disease.

In a subsequent analysis of the SMART data, elevated levels of D-dimer and the proinflammatory cytokine interleukin-6 (IL-6) were associated with HIV-RNA viremia and related to all-cause mortality, supporting a mechanism by which HIV infection contributes to a pro-inflammatory, pro-thrombotic state ^[11]. D-dimer, high-sensitivity C-reactive protein (hsCRP), intercellular adhesion molecule-1 (ICAM-1), and IL-6 are all associated with risk of CVD in non-HIV-infected individuals ^[12–14]. Much of HIV-related disease is attributed to HIV-induced immune activation and dysregulation ^[15]. T-cell markers of activation and exhaustion such as CD38, human leukocyte antigen (HLA)-DR, Ki-67, and programmed death (PD)-1 are all upregulated ^[16–18]. T-cell activation and chronic inflammation are known to participate in CVD, as is evidenced by the increased incidence of atherosclerotic heart disease in populations with autoimmune inflammatory conditions ^[19]. Innate immunity also likely plays a role in CVD; increased circulating CD14⁺ monocytes have been linked to CVD risk ^[20]. In HIV-infected individuals, the CD14⁺⁺CD16⁺ subset has been linked to the progression of CVD ^[21,22] and HIV ^[23].

Identification of markers of increased CVD risk in HIV-infected patients is important and may inform the implementation of preventive measures by early identification of high-risk patients who may be candidates for intervention. Additionally, in the HIV-infected population, current methods of CVD risk stratification such as the Framingham Risk Calculator may underestimate the actual CVD risk ^[24]. In this retrospective, case–control study of HIV-infected adults enrolled in National Institutes of Health (NIH) intramural clinical protocols, we hypothesized that inflammation and immune activation, which have both been associated with HIV infection and HIV-related mortality, may contribute to an increased risk of CVD in HIV-infected patients.

Methods

Case identification

HIV-infected patients who experienced an incident CVD event while participating in National Institute of Allergy and Infectious Diseases (NIAID), NIH intramural clinical protocols between January 1995 and March 2009 were identified from electronic medical records. Seven categories of CVD events were defined: acute myocardial infarction (AMI) – acute clinical coronary syndrome associated with elevated cardiac enzymes or EKG documenting acute ischemia or with evidence of recent blockage by invasive imaging (N = 25); silent MI (SMI) – the appearance of EKG changes indicating history of ischemia with new Q waves in two contiguous leads, dated by discovery of EKG changes (N = 4); coronary revascularization (CRv) – angioplasty with stent placement or coronary artery bypass grafting without preceding AMI or SMI (N = 14); acute coronary syndrome (ACS) – clinical signs and symptoms of ACS, including chest pain at rest lasting longer than 30 min associated with changes on EKG or elevated cardiac enzymes, but not considered to be sufficient for AMI by clinical judgment at the time (N = 2); cerebrovascular accident (CVA) – focal neurologic deficit lasting longer than 24 h and determined to be of ischemic origin (N = 4); lower extremity revascularization – procedure to restore arterial blood supply (N = 1); or sudden cardiovascular death – death related to CVD as determined by treating facility or autopsy (N = 2). (End points modeled after those defined by the PROactive study ^[25]).

Events were confirmed through detailed chart review, including records submitted from outside healthcare facilities when available. Participants were excluded if they had experienced a CVD event prior to enrollment in NIAID/NIH clinical protocols (N = 17) or the acquisition of HIV (N = 1). If a patient experienced multiple events while enrolled, only the first was considered.

Matching

Case patients were matched 2: 1 to HIV-infected individuals by sex, age (+/− 2.5 years), and enrollment date (+/− 3 years) in NIAID intramural protocols. Controls were considered eligible if they had not experienced any of the above-defined CVD events and had continuing participation in NIAID protocol(s) that included sample storage at the time of the matching event.

Clinical data

Hypertension was defined as the use of antihypertensive medications at the event or a documented blood pressure reading more than 140 mmHg systolic or more than 90 mmHg diastolic on two or more occasions prior to the event (PTE). Dyslipidemia was defined as the use of lipid-lowering therapy or the presence of any of the following: total cholesterol more than 240 mg/dl, low-density lipoprotein (LDL) more than 160 mg/dl, HDL less than 40 mg/dl, or triglycerides more than 250 mg/dl at the visit most closely preceding the event date ^[26]. Diagnosis of diabetes was confirmed by a random glucose level more than 200 mg/dl or fasting glucose more than 126 mg/dl on two occasions PTE or use of antidiabetic medications. Family history of premature MI was defined as MI in a male less than 55 years or female less than 65 years first-degree relative as documented PTE. History of smoking was positive if a patient reported having smoked more than 100 cigarettes; smoking was considered to be current if the patient reported on-going use or quitting less than 1 month PTE. Patient deaths were confirmed through the online social security death index (SSDI) database when possible. Cumulative exposure to ARVs, ARV regimen at event, and use of risk-related medications were determined by prescribing information and history in the medical record. Standard clinical and laboratory values were collected from patient records at approximately 4 months and 2 years PTE.

Measurement of soluble biomarkers

Cryopreserved stored samples from approximately 4 months (4.5 ± 0.1 months) and 2 years (21.6 ± 0.2 months) prior to the matching event were tested. Cytomegalovirus (CMV) DNA was quantitated by real-time PCR as described by Yun et al. ^[27]. Plasma samples were analyzed by ELFA (enzyme-linked fluorescent assay) on a Vitek Immuno Diagnostic Assay System (VIDAS) instrument for D-dimer (bioMerieux Inc., Durham, North Carolina, USA) and by ELISA for High motility group box (HMGB)-1 (Shino-Test Corp., Tokyo, Japan), soluble CD14, and soluble tissue factor (R&D Systems, Minneapolis, Minnesota, USA). Serum samples were analyzed by ELFA on the VIDAS for N-terminal prohormone brain natriuretic peptide (NT-pro BNP; bioMerieux Inc.) and ELISA using a 9-plex Pro-inflammatory kit, a 9-plex Chemokine kit, a 4-plex Vascular injury II kit, and four single-plex kits (Meso Scale Discovery, Gaithersburg, Maryland, USA) for the detection of the following cytokines: granulocyte/macrophage colony-stimulating factor (GM-CSF), interferon-γ (IFN-γ), IL-1β, IL-10, IL-12p70, IL-2, IL-6, IL-8, tumor necrosis factor alpha (TNF-α), eotaxin, eotaxin-3, interferon-γ-induced protein (IP)-10, monocyte chemotactic protein (MCP)-1, MCP-4, macrophage-derived chemokine (MDC), macrophage inflammatory protein (MIP)-1β, thymus and activation regulated chemokine (TARC), serum amyloid A, vascular cellular adhesion molecule-1 (VCAM-1), CRP, ICAM-1, myeloperoxidase (MPO), tissue inhibitor of metalloproteinase-1 (TIMP-1), tumor necrosis factor receptor-II (TNFR-II), and adiponectin. All of the listed tests were performed according to the manufacturers' instructions. Reference values from pooled and individual healthy donors as provided by the manufacturers (mean and SD), and the NIAID clinical laboratories as available are as follows: D-dimer less than 0.5 μg/ml; CRP less than 3.0 mg/l; GM-CSF, 1.1 ± 0.4 pg/ml; IFN-γ, 3.1 ± 3.8 pg/ml; IL-1β, 1.3 ± 0.9 pg/ml; IL-10, 8.2 ± 24.4 pg/ml; IL-12p70, 17.8 ± 59.1 pg/ml; IL-2, 1.0 ± 0.3 pg/ml; IL-6, 3.2 ± 9.3 pg/ml; IL-8, 20.6 ± 14.0 pg/ml; TNF-α, 8.2 ± 2.0 pg/ml; eotaxin, 1516 ± 226 pg/ml; eotaxin-3, 12.3 ± 5.3 pg/ml; interferon γ-induced protein-10 (IP-10), 86 ± 29 pg/ml; MCP-1, 502 ± 57 pg/ml; MCP-4, 1043 ± 137 pg/ml; MDC, 3230 ± 790 pg/ml; MIP-1β, 151 ± 19 pg/ml; TARC, 753 ± 111 pg/ml; serum amyloid A, 1485 ± 189 pg/ml; VCAM-1, 419 ± 127 pg/ml; ICAM-1, 250 ± 36 pg/ml; MPO, 383 ± 84 pg/ml; TIMP-1, 308 ± 43 ng/ml; TNFR-II, 1490 ± 354 pg/ml; and adiponectin, 8460 ± 189 ng/ml.

Flow cytometry

To investigate contribution of cellular immune activation, T-cell and monocyte markers of activation were measured by multi-color flow cytometry in cryopreserved peripheral blood mononuclear cells (PBMCs) from time points matching those used for the soluble markers. The fluorochrome-conjugated antibodies used were as follows: anti-CD3 Pacific Blue, anti-CD3 APC-Cy7, anti-CD25 PE-Cy7, anti-CD36 FITC, anti-CD38 APC, anti-CD56 PB, anti-CD142 PE, anti-Ki-67 FITC, anti-HLA-DR PE-Cy7, anti-CCR5 PE and anti-HLA-DR APC-Cy7 from BD-Biosciences (San Jose, California USA), PD-1 PerCP-Cy5.5 from BioLegend (San Diego, California, USA), anti-CD3 PE, anti-CD4 Qdot605, anti-CD8 PB, anti-CD14 Qdot605, anti-CD16 PerCP, anti-CD20 PB and Live/Dead Fixable Blue Dead Cell Stain Kit with UV excitation from Invitrogen (Carlsbad, California, USA), and anti-CD11c PE-Cy7, anti-CD39 FITC, and anti-FoxP3 PE-Cy5 from eBioscience (San Diego, California, USA). Samples were acquired on a Life Sciences Research (LSR)-II flow cytometer (BD) (Franklin Lakes, New Jersey, USA) and data were analyzed using FlowJo software version 8.8.6 (Treestar Inc., Ashland, Oregon, USA).

Statistical methods

The prevalence of clinical risk factors was compared by χ² analysis with Fisher's exact test. Continuous variables were compared by Student's t-test. Nonnormally distributed variables were log₁₀-transformed for analyses to approximate a normal distribution. Multivariable analysis was performed using a forward stepwise regression model including all significant variables as defined by univariate logistic regression (P < 0.05). All statistical analyses were performed with JMP software (JMP; Version 7. SAS Institute Inc., Cary, North Carolina, USA; 1989–2007).

Results

Fifty-two patients who experienced incident CVD events were identified from a pool of 1709 HIV-infected individuals enrolled in NIAID/NIH clinical protocols between 1995 and 2009 (5.1 events/1000 patient years). One hundred and four HIV-infected individuals without a history of CVD events were matched as described above to serve as controls. Cardiac events represented 90% of the case population. Both cases of sudden cardiac death were determined to be due to AMI, one by the receiving emergency care facility and the second by autopsy. All cases of SMI with Q wave changes in successive EKGs had been confirmed by invasive imaging techniques.

Clinical risk factors

There were no significant differences in race, nadir CD4⁺ count, peak plasma HIV-RNA, mode of acquisition, or duration of HIV infection between cases and controls (Table 1).

Table 1

The prevalence of dyslipidemia was greater in cases than controls (87.0 vs. 71.9%, P = 0.05). The prevalence of hypertension in this population was relatively high, but no greater in cases than controls (72.5 vs. 67.3%, P = 0.57). Cases were significantly more likely to be active smokers at the time of the event (49.0 vs. 25.0%, P = 0.004). A positive family history of premature MI was also significantly more likely in the cases than controls (29.8 vs. 10.9%, P = 0.003). There were no differences in the prevalence of diabetes or in mean BMI between groups (Table 1), or in history of substance abuse or known cocaine use (data not shown).

There were no statistically significant differences between cases and controls in cumulative months of exposure to all antiretroviral agents or classes of antiretroviral agents (Table 1). In addition, there was no difference in the prevalence of protease inhibitor or nonnucleoside reverse transcriptase inhibitor (NNRTI) use at the time of the event, nor were there differences in the participants' recent or past exposure to IL-2, didanosine, or abacavir (data not shown).

Total and LDL cholesterol were both elevated in the cases; this difference was significant at 4 months PTE (P = 0.002 and P = 0.04, respectively). HDL and triglycerides were not significantly different at either time point. Serum glucose was significantly higher in the cases at 2 years PTE (115 vs. 102, P = 0.03). The number of circulating CD14⁺ monocytes was significantly higher in the cases 4 months PTE (P = 0.04). The prothrombin time (PT) was significantly lower in cases than controls at 4 months PTE. Plasma HIV-RNA was significantly lower in the cases than the controls at 4 months PTE (2540 vs. 13860 copies/ml, P = 0.04). No differences were seen in CD4⁺ or CD8⁺ T-cell count at either time point (Table 2). The prevalence of CMV viremia (at 4 months PTE) was 8% in controls and 6% in cases; neither prevalence nor CMV viral load was significantly different between the groups (data not shown).

Table 2

No differences in prevalence of febrile, neoplastic, or inflammatory diseases were observed between cases and controls, nor did the analysis change significantly when these patients were excluded (data not shown).

Biomarkers

Two biomarkers were elevated in the cases at both tested time points PTE: D-dimer (P = 0.003 at 4 months and P = 0.04 at 2 years), and VCAM-1 (P = 0.02 at 4 months and P = 0.03 at 2 years). Soluble tissue factor and TIMP-1 were elevated in the cases at 4 months PTE (P = 0.02 in both). Serum amyloid A and MPO were elevated in the patients compared to controls at 2 years PTE (P = 0.03 and P = 0.005, respectively; Table 3 and Fig. 1). IL-6, TNF-α, MDC, and NT-proBNP tended to be higher in the cases than controls, but these differences were not significant (Table 3). Serum levels of hsCRP, adiponectin, IL-1β, IL-2, IL-8, IL-10, IL-12p70, TARC, eotaxin, MIP-1β, GM-CSF, IFN-γ, MCP-1, MCP-4, sICAM-1, and TNF-RII did not differ significantly between patients and controls.

Table 3

Fig. 1

Cell surface markers

There were no significant differences in the expression of CD38/HLA-DR, CCR5, or PD-1 on CD4⁺ and CD8⁺ T cells or in the relative percentages of T-regulatory cells as measured by CD25/FoxP3 coexpression, though cases tended to have lower PD-1 expression on CD8⁺ T cells at 4 months PTE (18.8 vs. 22.4%, P = 0.15) and lower coexpression of CD11c and CD36 on monocytes at 2 years PTE (40.5 vs. 46.0%, P = 0.06; Table 3). There were no significant differences in the expression of CD16, CD142 (tissue factor), CD11c, coexpression of CD11c with CD16, or in the relative sizes of monocyte CD14⁺/CD16⁺ subsets of patients and controls (data not shown).

Multivariate analysis

After determining variables for entry into multivariate analysis by univariate regression (P < 0.05), a stepwise multivariate model was constructed for each of the two time points. At 2 years PTE, family history of premature MI (P = 0.03), plasma D-dimer (P = 0.006), and serum glucose (P = 0.001) contributed independently to CVD event risk. At 4 months PTE, plasma D-dimer (P = 0.02), family history (P = 0.006), current smoking (P = 0.004), and total cholesterol (P = 0.0005) contributed independently to CVD event risk.

Discussion

In this study, traditional risk factors and markers of inflammation and cell activation were evaluated in HIV-infected individuals prior to an incident CVD event to investigate the potential shared and unique CVD risk factors that exist within the HIV-infected population. Although the strongest contributors were traditional CVD risk factors such as smoking and high cholesterol, markers of innate immune activation, endothelial cell dysfunction, and thrombosis were also related to CVD events.

The two clinical factors most strongly associated with future CVD events were family history of premature MI and active smoking at the time of the event. The overall prevalence of a history of smoking was relatively high in both groups, but at the time of the event was nearly twice as high in patients as controls, demonstrating the known influence of active smoking on CVD risk. It also highlights the importance of smoking cessation counseling in the HIV clinical setting.

The high frequency of dyslipidemia in both patients and controls was likely due to a combination of genetic predisposition, HIV infection, and ARV therapy. Although there were no differences in HDL or triglycerides, which are frequently abnormal and directly related to HIV infection, patients had higher total and LDL cholesterol compared to controls. The prevalence of diabetes was 13% [compared to 7.8% in the US population (http://diabetes.niddk.nih.gov/)] and the average BMI was above the normal range but nonobese (25.6 kg/m²). One indication of the relation of hyperglycemia with CVD risk in this study, however, was that patients had significantly higher blood glucose at 2 years PTE. This effect was maintained in multivariable analysis, highlighting a long-term risk of hyperglycemia. It was not possible to determine the prevalence of the metabolic syndrome, as data on abdominal girth were not available.

Large multicenter cohorts such as the DAD study group ^[3] have identified an increased risk of MI with increased years of antiretroviral exposure compared to HIV-infected individuals who are antiretroviral-naive. Our data did not support a connection between specific antiretroviral drugs, drug classes, or duration of antiretroviral exposure and incidence of CVD, though this may be due to the small number of patients and inclusion of only two antiretroviral-naive individuals (both controls).

Plasma HIV-RNA viremia, which is known to cause immune activation, was not associated with increased CVD risk in our cohort. Instead, prior to the event, the patients were more likely to have lower plasma HIV-RNA, which correlated with a trend toward lower PD-1 expression in CD8⁺ T cells, a marker of immune dysregulation that is strongly related to viremia ^[28]. The difference in plasma HIV-RNA between the groups was not associated with significant differences in duration of therapy or treatment regimen.

HIV also causes activation and dysregulation in the innate immune system ^[29,30] that may negatively influence cardiovascular health. Activated CD14⁺ monocytes are associated with increased CVD risk in the general population ^[20,31]. Additionally, monocyte-associated tissue factor (CD142) is upregulated in HIV infection and related to viremia, potentially indicating an induced hypercoagulable state ^[32,33]. Although the number of CD14⁺ cells was significantly higher in the cases, this was not associated with elevated expression of other markers of monocyte activation such as HLA-DR and CD142. The detection of these monocyte markers may have been limited by cryopreservation. Importantly, increased monocyte activation in the cases was demonstrated by significantly higher levels of soluble tissue factor and MPO and a trend toward higher TNF-α, MDC, and IL-6. Soluble CD14 (sCD14), a marker of monocyte activation by lipopolysaccharide (LPS) that is elevated in HIV-infected individuals and is suspected to contribute to chronic inflammation and cardiovascular risk ^[32,34], was not different between cases and controls.

D-dimer, a soluble product of fibrinogen breakdown, is a marker of thrombosis that is elevated in patients with premature CVD ^[35] and was associated with mortality in the SMART study ^[11]. In our cohort, D-dimer was significantly elevated in the cases at both time points and independently related to CVD risk in multivariate analysis. In contrast, there was no difference in hsCRP between cases and controls, though hsCRP is related to CVD event risk and associated risk factors such as the metabolic syndrome in the non-HIV population ^[36–38]. The capacity for CRP to serve as a discriminatory marker of CVD risk in the HIV-infected population may be confounded by multiple, distinct factors influencing CRP, including significant underlying disease.

Five markers of endothelial activation and damage were measured: sVCAM-1, sICAM-1, TIMP-1, and eotaxins 1 and 3. TIMP-1 is an inhibitor of matrix metalloproteinases postulated to indicate vascular damage or oxidative stress ^[39] and eotaxin-3 paradoxically is lower in patients with elevated CVD risk ^[40]. In our study, TIMP-1 was significantly elevated in the patients 4 months PTE, and eotaxin-3 tended to be lower in the patients 2 years PTE. VCAM-1 is expressed by vascular endothelial cells in response to pro-inflammatory cytokines TNF-α and IL-1 as well as shearing or wall stress ^[41]. In the HIV-infected population, VCAM-1 expression is elevated in early HIV infection and with viremia, reported to return to baseline after 2 years of consistent ART therapy ^[42] and is associated with the metabolic syndrome ^[43]. Here VCAM-1 was significantly elevated in the cases over controls at both time points, indicating its potential use in risk stratification in the HIV-infected population. Subgroup analysis excluding the five cases with noncardiac CVD-related events did not significantly change this analysis.

In conclusion, in this cohort of HIV-infected patients who experienced cardiovascular events, the factors most related to a participant's status as a case or control included active smoking, family history of premature MI, elevated total cholesterol, glucose, plasma D-dimer, and soluble tissue factor. Unexpectedly, in this cohort of patients with long-term HIV infection who are well controlled on HAART, HIV-related characteristics such as ARV therapy or regimen and CD4⁺ T-cell count were not related. Additionally, although markers of innate immune activation, thrombosis, and endothelial cell damage, including D-dimer, soluble tissue factor, VCAM-1, MPO, and serum amyloid A were associated with CVD events, these elevations were not accompanied by higher expression of markers of T-cell activation that are characteristic of HIV infection. It is possible that other important markers such as NT-proBNP, IL-6, MDC, and GM-CSF did not reach statistical significance due to the small sample size. Our findings support an aggressive approach in identifying significant family history and targeting traditional cardiac risk factors for therapeutic intervention such as management of dyslipidemia, smoking cessation, as well as the potential addition of biomarkers such as D-dimer in further stratification of high-risk patients.

Acknowledgements

The present research was supported in part by the Intramural Program of the National Institutes of Health (NIH), National Institute of Allergy and Infectious Diseases (NIAID), and Critical Care Medicine Department. Additionally, this project has been funded in part with federal funds from the National Cancer Institute, NIH, under Contract No. HHSN261200800001E. The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products, or organizations imply endorsement by the US Government.

The authors would like to thank all study participants and the staff of outpatient clinic 8 at NIAID of the NIH clinical center. Thanks also to William Thompson for technical assistance and Dr Douglas Rosing for his review of study participant EKGs.

I.S. and C.H. designed the study, identified case participants, and assisted in data analysis and in writing the article. E.F. identified patients and controls, gathered clinical data, and assisted in performance of laboratory studies, data analysis, and in writing the article. J.G. and A.R. assisted in the performance of laboratory studies and data analysis. A.R., Y.B., and V.N. assisted in the performance of laboratory studies. L.D. assisted in the matching of control participants. C.R. assisted in the identification of participants and patient samples. All authors assisted in reviewing the article and approved the final version of the article.

E.S.F. was a 2008–2009 participant in the Clinical Research Training Program, a public–private partnership supported jointly by the NIH and Pfizer Inc. via a grant to the Foundation for NIH from Pfizer Inc.

These data have been presented in part at Keystone HIV Immunobiology, Keystone, CO, April 2009, abstract number 220, and at the 17th Conference on Retroviruses and Opportunistic Infections, San Francisco, CA, February 2010, abstract number 713.