Human immunodeficiency virus (HIV)–infected patients are disproportionately affected by cardiovascular disease and sudden cardiac death (SCD). Whether left ventricular (LV) dysfunction predicts SCD in those with HIV is unknown. We sought to determine the impact of LV dysfunction on SCD in patients with HIV. We previously characterized all SCDs and acquired immunodeficiency syndrome (AIDS) deaths in 2,860 consecutive patients in a public HIV clinic from 2000 to 2009. Transthoracic echocardiograms (TTEs) performed during the study period were identified. The effect of ejection fraction (EF), diastolic dysfunction, pulmonary artery pressure, and LV mass on SCD and AIDS death were evaluated: 423 patients had at least 1 TTE; 13 SCDs and 55 AIDS deaths had at least 1 TTE. In the propensity-adjusted analysis, EF 30% to 39% and EF <30% predicted SCD (hazard ratio [HR] 9.5, 95% confidence interval [CI] 1.7 to 53.3, p = 0.01 and HR 38.5, 95% CI 7.6 to 195.0, p <0.001, respectively) but not AIDS death. Diastolic dysfunction also predicted SCD (HR 14.8, 95% CI 4.0 to 55.4, p <0.001) but not AIDS death, even after adjusting for EF. The association between EF <40% and SCD was greater in subjects with detectable versus undetectable HIV RNA (adjusted HR 11.7, 95% CI 2.9 to 47.2, p = 0.001 vs HR 2.7, 95% CI 0.3 to 27.6, p = 0.41; p = 0.07 for interaction). In conclusion, LV systolic dysfunction and diastolic dysfunction predict SCD but not AIDS death in a large HIV cohort, with greater effect in those with detectable HIV RNA. Further investigation is needed to thoroughly evaluate the effect of low EF and HIV factors on SCD incidence and the potential benefit of implantable cardioverter-defibrillator therapy in this high-risk population.
We recently determined that sudden cardiac deaths (SCDs) composed the majority of cardiac deaths over a 10-year period in a large, urban, human immunodeficiency virus (HIV)–positive cohort, at an adjusted rate 4.5-fold higher than expected. In the general population, left ventricular (LV) systolic dysfunction is strongly associated with an increased risk of SCD, but this association has not been evaluated in the setting of HIV infection. Because most deaths in large HIV cohorts are still acquired immunodeficiency syndrome (AIDS) related, whether LV dysfunction carries the same prognostic importance for HIV-infected subjects is unknown. In addition, HIV-infected subjects may be at risk for ventricular arrhythmias by mechanisms independent of LV systolic dysfunction, including QT interval prolongation, inflammation, and direct viral effects on cardiomyocyte depolarization and repolarization. We therefore sought to evaluate any potential association between premortem LV dysfunction and SCD and AIDS-related death in a large urban cohort of HIV-infected patients.
Methods
We previously identified records of 2,860 consecutive patients monitored at a public HIV clinic at San Francisco General Hospital from April 1, 2000 to August 31, 2009. This clinic serves approximately 20% of HIV-infected patients in San Francisco. For this analysis, we included all patients aged ≥18 years with documented HIV infection who had at least 1 transthoracic echocardiogram (TTE) during this period. The study was approved by the Institutional Review Board of the University of California, San Francisco.
We previously identified and classified all deaths in this cohort. Briefly, SCDs were defined as meeting 2 published criteria: (1) primary International Classification of Diseases, tenth revision, code for all cardiac causes and (2) circumstances of death meeting World Health Organization criteria for SCD (death within 1 hour of symptom onset if witnessed or within 24 hours of being observed alive and symptom-free if unwitnessed) or unexpected out-of-hospital death. Hospice, overdose, violence, suicide, cancer, or opportunistic infection deaths were excluded. All unexpected deaths classified as SCD were confirmed as not meeting criteria for AIDS death. AIDS death required 2 of 3 published criteria: (1) primary International Classification of Diseases, tenth revision, code for HIV disease–related illness, (2) circumstances of death involving HIV-related infection or illness, or (3) most recent CD4 cell count of <50 cells/mm 3 .
Baseline characteristics were abstracted from the clinic’s electronic medical record. We recorded the following variables: age, gender, race, CD4 cell count, HIV viral load, antiretroviral medication use, cardiac medication use, coronary artery disease (CAD), hypertension (HTN), diabetes mellitus, smoking, disorders of lipid metabolism, chronic kidney disease, and illicit drug use.
We searched all cohort patients for any TTE evaluation at San Francisco General Hospital during the study period. On TTE, LV systolic function, diastolic function, pulmonary artery systolic pressure (PASP), and LV mass were analyzed. LV function was independently assessed by 2 study investigators and categorized as normal (ejection fraction [EF] >50%), mildly reduced (EF 40% to 50%), moderately reduced (EF 30% to 39%), or severely reduced (EF <30%) by visual inspection. LV diastolic function was classified as normal, impaired relaxation (stage I), pseudonormal (stage II), or restrictive (stage III) using American Society of Echocardiography criteria. Significant diastolic dysfunction was classified as stage II or III. PASP (mm Hg) was measured using tricuspid regurgitation jet velocity plus the estimated right atrial pressure by inferior vena cava diameter and response to inspiration. Pulmonary HTN was defined as PASP >30 mm Hg. LV hypertrophy was classified as none, mild, moderate, or severe by standard criteria.
Baseline characteristics were compared using t or chi-square tests as appropriate. We used Cox proportional hazards models to estimate the association of TTE parameters with SCD and AIDS-related death, on the scale of time since the first TTE. To account for multiple TTEs, each parameter was treated as a time-dependent covariate, updated at the time of each later TTE. Because the SCD outcome was uncommon, we used propensity scores rather than conventional multivariate adjustment to control for potential confounding variables. Covariates in the logistic model used to estimate the scores were specified a priori. We also assessed modification of the TTE effects by 2 factors specified a priori: HIV RNA treated as either a continuous or dichotomous variable (detectable vs undetectable) and nadir and most recent CD4 count (<200 vs >200 cells/mm 3 ), using interaction terms. Because several baseline characteristics differed between those who did and did not receive a TTE, we assessed for potential selection bias, first fitting a logistic model for the probability of inclusion in the subsample with at least 1 TTE as a function of baseline characteristics, then refitting the Cox models for SCD and AIDS-related death, weighted by the inverse of these probabilities. In additional sensitivity analyses, we used Fine-Gray models to estimate TTE effects on SCD, treating other deaths as competing risks.
Results
The baseline characteristics of this cohort have been previously described. Four hundred twenty-three of the total 2,860 patients (15%) of the cohort had echocardiographic evaluation. One hundred thirty-six patients had >1 echocardiogram (range 2 to 5) and 654 total studies were evaluated. Of the 423 patients with at least 1 TTE, 13 had SCD and 55 died of AIDS-related causes. The median time from most recent TTE to death was 566 days for SCDs and 366 days for AIDS deaths. Compared with subjects who did not have TTEs, subjects with TTEs were slightly older and more likely to be female and African-American. In addition, they had higher rates of traditional risk factors including HTN, diabetes mellitus, chronic kidney disease, and more advanced HIV disease, including lower CD4 counts and higher HIV RNA levels.
Overall demographics between SCDs and AIDS deaths were similar, with AIDS deaths having higher HIV RNA levels and lower CD4 counts. Prevalence of HTN, CAD, and antiretroviral therapy use was not significantly different between SCDs and AIDS deaths ( Table 1 ). A measurable tricuspid regurgitation jet permitted the measurement of PASP in 314 patients (74%) whereas diastolic dysfunction was assessed in 299 (not assessable in 124 because of atrial fibrillation or flutter or mitral stenosis). Compared with AIDS deaths, SCDs were more likely to have systolic dysfunction and diastolic dysfunction but had similar levels of pulmonary HTN and LV hypertrophy ( Table 2 ).
| Characteristic | All Subjects With TTE (n = 423) | SCD (n = 13) | AIDS Death (n = 55) | p Value ∗ |
|---|---|---|---|---|
| Mean age (yrs) | 42.3 ± 9.4 | 43.6 ± 10.8 | 41.2 ± 8.6 | 0.39 |
| Women | 80 (19) | 1 (8) | 10 (18) | 0.36 |
| Body mass index (kg/m 2 ) | 25.0 ± 5.2 | 27.0 ± 6.9 | 24.4 ± 5.3 | 0.52 |
| Race/ethnicity | 0.38 | |||
| African-American | 147 (35) | 8 (62) | 23 (42) | |
| Asian American | 14 (3) | 0 | 1 (2) | |
| European American | 174 (41) | 5 (38) | 21 (38) | |
| Hispanic American | 83 (20) | 0 | 10 (18) | |
| Other | 5 (1) | 0 | 1 (2) | |
| CAD | 24 (6) | 1 (8) | 1 (2) | 0.26 |
| MI | 13 (3) | 0 | 1 (2) | 0.62 |
| HTN | 101 (24) | 4 (31) | 9 (16) | 0.21 |
| Smoking | 113 (27) | 2 (15) | 11 (20) | 0.52 |
| Diabetes mellitus | 38 (9) | 1 (8) | 5 (9) | 0.67 |
| Chronic kidney disease | 46 (11) | 2 (15) | 8 (15) | 0.62 |
| Illicit drug use | 128 (30) | 2 (15) | 14 (25) | 0.44 |
| Medications | ||||
| ACE inhibitor | 66 (16) | 1 (8) | 11 (20) | 0.30 |
| β Blocker | 64 (15) | 1 (8) | 7 (13) | 0.61 |
| Statin | 51 (12) | 1 (8) | 2 (4) | 0.52 |
| NRTI | 260 (61) | 6 (46) | 39 (71) | 0.09 |
| NNRTI | 138 (32) | 3 (23) | 17 (31) | 0.58 |
| PI | 229 (54) | 5 (38) | 32 (58) | 0.20 |
| CD4 count (cells/mm 3 ) † | 274 (89–458) | 310 (268–534) | 111 (35–278) | 0.003 |
| HIV RNA (log copies/ml) † | 3.7 (1.9–4.8) | 3.6 (2.3–4.0) | 4.5 (1.9–5.3) | 0.09 |
∗ p Value for comparison between SCDs and AIDS deaths.
| Characteristic | All Subjects (n = 423) | SCD (n = 13) | AIDS Death (n = 55) | p Value ∗ |
|---|---|---|---|---|
| Left ventricular ejection fraction (%) | ||||
| >50 | 339/423 (80) | 3/13 (23) | 45/55 (82) | <0.0001 |
| 40–50 | 39/423 (9) | 4/13 (31) | 5/55 (9) | 0.04 |
| 30–39 | 29/423 (7) | 3/13 (23) | 4/55 (7) | 0.09 |
| <30 | 16/423 (4) | 3/13 (23) | 1/55 (2) | 0.003 |
| Diastolic dysfunction † | 29/299 (10) | 5/8 (63) | 1/38 (3) | <0.0001 |
| PASP (mm Hg) | 27.0 (21–33) | 29.5 (22–42) | 27.0 (23–33) | 0.28 |
| Pulmonary HTN ‡ | 100/314 (32) | 4/13 (31) | 13/55 (24) | 0.59 |
| LV hypertrophy | 127/420 (30) | 3/13 (23) | 20/53 (38) | 0.32 |
∗ p Value for comparison between SCDs and AIDS deaths.
† Stage II and III diastolic dysfunction.
LV dysfunction predicted SCD but not AIDS death. In the Cox model for SCD, employing a propensity score adjusting for age, race, gender, diabetes mellitus, CAD, most recent CD4 count, most recent HIV viral load, cardiac medication use by class, and antiretroviral medication use by class, the hazard ratio (HR) for EF <40% was 13.7 (95% confidence interval [CI] 4.1 to 46.3, p <0.001). Lower EF was associated with progressively greater SCD risk, with HRs of 9.5 (95% CI 1.7 to 53.3, p = 0.01) for EF 30% to 39% and 38.5 (95% CI 7.6 to 195.0, p <0.001) for EF <30% ( Figure 1 ). Diastolic dysfunction was also associated with SCD (HR 14.8, CI 4.0 to 55.4, p <0.001), even after adjusting for the effect of LV systolic dysfunction (HR 8.0, 95% CI 2.5 to 25.8, p <0.001).
