Progression of cardiac structure and function in people with human immunodeficiency virus

Abstract

Objective:

People living with HIV (PLWH) are at increased risk for cardiac dysfunction. It is unknown how their global longitudinal cardiac function, cardiac structure, and other indices of function progress over time. We aimed to characterize the longitudinal trend in cardiac structure and function in PLWH.

Design:

Retrospective study of PLWH with clinically obtained echocardiograms at an academic medical center.

Methods:

We reviewed archived transthoracic echocardiograms (TTEs) performed between 2001 and 2012 on PLWH. The primary outcome measures were progression of global longitudinal strain (GLS, left and right ventricles), LV mass, E/e′ ratio, LV end-systolic, and -diastolic volumes using hierarchical mixed model analysis as a function of CD4+ T cell count and HIV RNA suppression. Models were adjusted for clinical and demographic characteristics.

Results:

We analyzed 469 TTEs from 150 individuals (median age 46 years, 58% male). Median CD4+ T cell counts at nadir and proximal to first echocardiogram were 85 and 222 cells/mm³, respectively. Over a median of 5 years, LV mass index increased regardless of nadir or proximal CD4+ T cell count or viral suppression status. PLWH with viral suppression at baseline had more normal GLS throughout the follow-up period. There were no significant trends in LV end-systolic volume index or E/e′.

Conclusions:

In PLWH, HIV viral suppression is associated with early gains in echocardiographic indices of cardiac function that persist for up to >5 years. HIV disease control impacts routine echocardiographic measures with known impacts on long-term prognosis.

1 |. INTRODUCTION

In the era of effective antiretroviral therapy (ART), HIV infection is now a chronic disease^1,2 and cardiovascular disease (CVD) is among the leading causes of death among people living with HIV (PLWH) in the United States.³ The risk of clinical heart failure (HF) and antecedent structural and functional abnormalities are well-described in this population.^4,5 Cross-sectional studies have also shown that PLWH with diastolic dysfunction have multiple alterations in cardiac structure and function, including myocardial fibrosis and left atrial abnormalities, and worse quality of life.⁶

A growing body of literature also points to the presence of myocardial dysfunction in PLWH without overt clinical HF.⁷ Serum levels of biomarkers associated with cardiac stress and cardiac dysfunction are elevated in PLWH compared to those without HIV infection,⁸ as are molecular patterns indicating myocardial inflammation.^9,10 Compared to uninfected persons, diastolic function is also more often abnormal among PLWH.¹¹ Although left ventricular ejection fraction (LVEF) is a commonly reported measure of left ventricular function, myocardial strain imaging from speckle tracking echocardiography (STE) may detect abnormalities prior to overt HF.¹² Global longitudinal strain (GLS) is the most common application of STE, showing strong correlation to prognosis and disease severity in a variety of primarily cardiac and non-cardiac clinical conditions.^13–18 Myocardial strain imaging by STE has not been extensively investigated in PLWH without reduced EF (EF > 40%).^15,19,20

Our group has shown a high prevalence of abnormal GLS in an HIV clinical cohort in a tertiary referral center, and demonstrated that low proximal and nadir CD4+ T cell count are independently associated with GLS in patients with normal LVEF in a cross-sectional study.⁷ However, the long-term trajectory of GLS and other echocardiographic parameters in PLWH has not been characterized. Accordingly, the aims of this study were to characterize the longitudinal trend in cardiac structure and function by echocardiography in a cohort of PLWH at a tertiary referral hospital, and determine HIV disease characteristics significantly associated with GLS and other measures to identify patients who may benefit from early interventions to preserve normal cardiac function.

2 |. METHODS

2.1 |. Study design and sources of data

All participants were selected from a query of the Duke HIV Clinical Database which includes all PLWH cared for at Duke since 1988. Results of this query were linked with records from the Duke Echocardiography Lab Database (DELD), a digital archive of all echocardiograms performed at Duke University Medical Center since 1995.²¹ We conducted a retrospective review of all transthoracic echocardiograms (TTEs) performed in PLWH at Duke between 2001 and 2012, as previously described.²²

All PLWH with a verifiable date of HIV diagnosis in the electronic health record and available CD4+ T cell count data were included in the study. HIV diagnosis dates were ascertained by manual chart review of clinician notes and HIV serologies. Only patients with HIV who had more than one echocardiogram after HIV diagnosis of sufficient quality for interpretation were included in this study cohort. We included all echocardiograms performed during the study period for all patients, except echocardiograms which were performed less than 2 months apart in which case we analyzed only the first echocardiogram. The first echocardiogram after HIV diagnosis defined the baseline time point for each patient.

Clinical data, including sociodemographic characteristics, behavioral risk data, CVD comorbidities, and laboratory data, were abstracted from medical charts and entered into an electronic database by trained staff. In addition, data on cocaine and other substance abuse, hepatitis C, and antiretroviral exposure (i.e., protease inhibitors, abacavir, zidovudine [AZT], stavudine, and didanosine) were also collected for all study participants at baseline.

We defined HIV viral suppression as HIV-1 RNA level of less than 400 copies/mL. Proximal CD4+ T cell count was defined as the most recent cell count prior to the baseline echocardiogram. Nadir CD4+ T cell count was defined as the lowest cell count in the electronic health record. We used a threshold of 200 cells/mm³ to create two categories of CD4+ T cell count groups based on established clinical relevance.²³

2.2 |. Echocardiographic measurements

All two-dimensional TTEs which had been acquired using standard clinical protocols were uploaded to an independent software package (2D Cardiac Performance Analysis version 4.5, Tom Tec Imaging Systems, Unterschleissheim, Germany) at frame rate ≥50/s for STE analyses. All analyses were interpreted by a single experienced reader (F.A.) blinded to the clinical, HIV specific markers, and immune system status. LV volumes and LVEF were calculated by the biplane method, according to the American Society of Echocardiography (ASE).²⁴ We used values indexed to body surface area for LV end-systolic volume, end-diastolic volume, and mass.

For STE, LV endocardial border was manually traced in end systole. In case of insufficient border tracking, manual correction of the endocardial tracing was attempted and if still unsatisfactory, or inadequate image quality (defined as poor visualization or poor tracking of >2 LV segments, segment dropout, missing view, or significant fore-shortening of the LV) the study was excluded from the final analysis. GLS was the average of the three apical peak longitudinal strain measurements (apical 4, 3, and 2 chamber views), based on prevailing recommendations.²⁵ Intra-observer variability for these measures has been previously reported with no significant systematic differences observed and intra-class coefficient of correlation of .70 (95% confidence interval [CI] .74, .97).⁷ The six main echocardiographic measures of interest were E/average e′ ratio, left ventricular mass index (LVMi), GLS (left and right ventricles), left ventricular end-systolic volume index (LV ESVi), and left ventricular end-diastolic volume index (LV EDVi).

2.3 |. Statistical analysis

A descriptive table of baseline characteristics for the study population included demographics, medical history, and HIV characteristics stratified by viral suppression status at baseline echocardiogram. Continuous variables are presented as medians and inter-quartile ranges (IQR). Categorical variables are presented as frequencies and percentages. All statistical analyses were calculated using SAS statistical software, Version 9.4 (Cary, NC).

For longitudinal data description, we categorized echocardiograms into the following time bins: baseline, 0–1 years, 1–2 years, 2–3 years, 3–4 years, 4–5 years, and ≥5 years after baseline echocardiogram, and summarized continuous echocardiographic measurements using medians and IQR within each bin in the overall cohort and stratified by sex. To visualize unadjusted longitudinal changes in echocardiographic measures, we plotted echocardiographic characteristics of interest over time using linear regression and penalized b-splines, as well as loess smoothing to visualize trends. These smoothed curves were overlayed on top of the raw data to determine any non-linear trends overtime. Based on exploratory observation there did not seem to be strong indication that a non-linear trend fit the data any better than a linear trend.

To assess the longitudinal relationship of each echocardiographic characteristic of interest with HIV characteristics of interest, we conducted hierarchical mixed model analyses, in which the longitudinal measurements of each echocardiographic characteristic were modeled as a linear function of time post-baseline, with the slope and intercept modeled as a function of the HIV characteristic of interest. The models included a random effect for patient to account for the repeated measurements within patient. We analyzed each of three HIV characteristics separately (i.e., proximal CD4+ T cell count, nadir CD4+ T cell count, and HIV RNA viral suppression), thus we had seven models for each HIV characteristics in which each of the seven echocardiographic characteristics (i.e., six main outcome measures and LVEF) described above served as response variables and the HIV characteristic served as the main covariate of interest. GLS values are discussed as absolute numbers such that lower absolute values indicate worse strain.

A second series of “baseline adjustment” models were fit that additionally included the following baseline adjustment characteristics selected a priori: age, sex, race (Black vs. non-Black), hypertension, diabetes, dyslipidemia, CAD, and smoking status (former vs. never, current vs. never). For each model we evaluated whether the HIV characteristic of interest was associated with the overall level of the echocardiographic measurements, whether there was evidence of an overall trend over time, and whether there was evidence that the time trend differed based on the HIV characteristic. Model-based least square means of the echocardiographic characteristic were estimated for each level of the HIV characteristic of interest at baseline, 1, 2, and 5 years, and reported with 95% CIs. This study of retrospective data during this time frame for waiver of informed consent was reviewed and approved by the Duke Institutional Review Board.

3 |. RESULTS

3.1 |. Cohort and individual characteristics

Among 5215 patients identified from the Duke HIV Clinical Database, there were 766 with a TTE in DELD, and 498 patients with at least one TTE after HIV diagnosis. After final exclusions for patients with less than two TTEs post-HIV diagnosis and imaging quality, there were 150 individuals included in the final sample (Figure 1).

FIGURE 1.

Flow chart showing cohort development. TTE, transthoracic echocardiogram

Distribution of baseline characteristics including demographics, medical history, and HIV clinical characteristics are presented in Table 1, stratified by viral suppression status at baseline echocardiogram. Of the 150 patients, 74 (49%) were not virally suppressed, 37 (25%) were virally suppressed, and 39 (26%) did not have sufficient data to determine viral suppression at time of baseline echocardiogram.

TABLE 1.

Baseline characteristics according to degree of viral suppression

Characteristic	All patients 150	Unknown viral suppression 39	No viral suppression 74	Yes viral suppression 37	*P-value
Age in years, median (IQR) [nobs]	46 (38–52) [148]	47 (38–52) [38]	43 (36–49) [73]	50 (45–55) [37]	<.01
Age at HIV diagnosis, median (IQR) [nobs]	36 (29–43) [146]	37 (32–44) [37]	35 (28–43) [72]	39 (32–43) [37]	.40
Female	61/147 (41.5%)	14/38 (36.8%)	34/72 (47.2%)	13/37 (35.1%)	.29
Race/ethnicity					.06
Black	114/146 (78.1%)	30/37 (81.1%)	58/72 (80.6%)	26/37 (70.3%)
White/non-Hispanic	25/146 (17.1%)	5/37 (13.5%)	9/72 (12.5%)	11/37 (29.7%)
Hispanic	5/146 (3.4%)	1/37 (2.7%)	4/72 (5.6%)	0/37 (.0%)
Other	2/146 (1.4%)	1/37 (2.7%)	1/72 (1.4%)	0/37 (.0%)
History of CAD	14/149 (9.4%)	2/39 (5.1%)	4/73 (5.5%)	8/37 (21.6%)	<.05
Hypertension	65/149 (43.6%)	18/39 (46.2%)	24/73 (32.9%)	23/37 (62.2%)	<.01
Hyperlipidemia	27/149 (18.1%)	2/39 (5.1%)	8/73 (11.0%)	17/37 (45.9%)	<.0001
Diabetes mellitus	30/146 (20.5%)	8/37 (21.6%)	14/72 (19.4%)	8/37 (21.6%)	.79
BMI, median (IQR)	26.5 (21.8–31.6) [98]	26.2 (24.1–30.4) [19]	27.5 (19.8–32.0) [44]	26.4 (22.2–33.7) [35]	.47
Any cocaine use	56/146 (38.4%)	13/37 (35.1%)	31/72 (43.1%)	12/37 (32.4%)	.54
Any tobacco use					.10
No tobacco use	56/146 (38.4%)	16/37 (43.2%)	31/72 (43.1%)	9/37 (24.3%)
Current tobacco use	64/146 (43.8%)	13/37 (35.1%)	28/72 (38.9%)	23/37 (62.2%)
Former tobacco use	24/146 (16.4%)	8/37 (21.6%)	11/72 (15.3%)	5/37 (13.5%)
Duration of HIV infection at time of ECHO (year), median (IQR) [nobs]	8 (4–12) [148]	8 (4–11) [38]	7 (4–11) [73]	10 (4–15) [37]	.03
Mean CD4 count (cells/mm3), median (IQR) [nobs]	151 (21–376) [26]	164 (12–337) [4]	98 (17–167) [16]	437 (202–499) [6]	.01
Proximal CD4 count at time of ECHO (cells/mm3), median (IQR) [nobs]	222 (65–488) [139]	203 (92–418) [32]	112 (31–275) [73]	446 (278–696) [34]	<.0001
Proximal CD4 group					<.0001
CD4 < 200	65/139 (46.8%)	15/32 (46.9%)	46/73 (63.0%)	4/34 (11.8%)
CD4 ≥ 200	74/139 (53.2%)	17/32 (53.1%)	27/73 (37.0%)	30/34 (88.2%)
CD4 nadir at time of ECHO (cells/mm3), median (IQR) [nobs]	85 (19–216) [150]	72 (10–193) [39]	66 (7–181) [74]	185 (112–295) [37]	<.0001
Nadir CD4 group					<.01
CD4 < 200	107/150 (71.3%)	30/39 (76.9%)	58/74 (78.4%)	19/37 (51.4%)
CD4 ≥ 200	43/150 (28.7%)	9/39 (23.1%)	16/74 (21.6%)	18/37 (48.6%)
Plasma HIV RNA < 400 for 6 months or more prior to ECHO	37/111 (33.3%)	–	0/74 (.0%)	37/37 (100.0%)	<.0001
Cumulative viremia copy years log10, median (IQR) [nobs]	3.72 (1.54–5.21) 35	2.59 (.33–9999.00) [10]	5.05 (4.93–5.47) [12]	1.86 (.53–2.77) [13]	<.001
Years of known uncontrolled viremia, median (IQR) [nobs]	1.15 (.08–4.88) [94]	1.38 (.18–5.29) [26]	1.48 (.04–4.58) [42]	.33 (.04–2.96) [26]	.66
Hepatitis C diagnosis	36/149 (24.2%)	10/39 (25.6%)	15/73 (20.5%)	11/37 (29.7%)	.28
ACEI or ARB use	40/149 (26.8%)	9/39 (23.1%)	19/73 (26.0%)	12/37 (32.4%)	.48
Beta blocker use	35/149 (23.5%)	9/39 (23.1%)	15/73 (20.5%)	11/37 (29.7%)	.28
Antiretroviral use	111/149 (74.5%)	27/39 (69.2%)	50/73 (68.5%)	34/37 (91.9%)	<.01
Abacavir (Ziagen) use	25/149 (16.8%)	4/39 (10.3%)	11/73 (15.1%)	10/37 (27.0%)	.13
Didanosine/stavudine/zalcitabine use	45/149 (30.2%)	12/39 (30.8%)	24/73 (32.9%)	9/37 (24.3%)	.36
Zidovudine (AZT, Retovir) use	51/149 (34.2%)	13/39 (33.3%)	24/73 (32.9%)	14/37 (37.8%)	.61
Protease inhibitor use	77/149 (51.7%)	18/39 (46.2%)	38/73 (52.1%)	21/37 (56.8%)	.64

BMI, body mass index; CAD, coronary artery disease; IQR, inter-quartile range; nobs, number of observations available for analysis.

^*P-values are for the comparison of patients with responses of yes versus no for viral suppression.

The median [IQR] age at HIV diagnosis was 36 [29–43] years and median age at time of baseline echocardiogram was 46 [38–52] years. The majority of patients were male (58%) and Black (78%). Nearly half (44%) had hypertension, 18% had hyperlipidemia, and 21% had diabetes. Those with viral suppression were older (50 vs. 43 years, P < .01), and more often had histories of CAD (22% vs. 6%, P < .05), hypertension (62% vs. 33%, P = .003), and hyperlipidemia (46% vs. 11%, P < .0001) than those who were not virally suppressed. The CD4+ T cell counts proximal to first echocardiogram and at nadir were greater in those with viral suppression than those without viral suppression (446 vs. 112 cells/mm³, P < .001; and 185 vs. 66 cells/mm³, P < .0001, respectively). More patients with viral suppression were currently on ART than those without viral suppression (92% vs. 69%, P = .006). There were no significant differences in prevalence of hepatitis C diagnosis, cardiac medications, AZT, or protease inhibitor use between those with and without viral suppression.

3.2 |. Echocardiographic data quality

The final analysis cohort consisted of 469 echocardiograms from 150 individuals over a period of over 10 years. The maximum number of echocardiograms per patient was nine (n = 1) and median time from HIV diagnosis to baseline echocardiogram was 8.3 years. The number of individuals with two, three, four, and five or more echocardiograms during the study period was 67, 38, 21, and 24, respectively. Of the 319 follow-up echocardiograms, most (n = 87) were performed between 0 and 365 days after the index echocardiogram. In successive years 1–2, 2–3, 3–4, 4–5, and ≥5, there were 53, 42, 36, 25, and 76 echocardiograms performed and included in the analysis spanning the 11-year time period (Supplementary Table S1).

We assessed the degree to which each echocardiogram had sufficient data and image quality for analysis. The most frequently missing echocardiographic measure at the index echocardiogram was E/e′ (43%), which became less missing on subsequent echocardiograms. For all measures, the degree of missingness on available echocardiograms progressively decreased over time (Supplementary Figure S1).

3.3 |. Progression of echocardiographic measures

To examine unadjusted trends in echocardiogram characteristics overtime, we plotted the main echocardiographic measures for the overall group according to days from the baseline echocardiogram to the last date of follow-up (Figure 2). E/e′, LV-end systolic volume, LV-end diastolic volume, and LV GLS did not show gross changes over time for the overall cohort. LV mass increased over time while RV GLS appeared to numerically decrease (worsen) over time. We next assessed the longitudinal trends in echocardiographic measures as a function of proximal and nadir CD4+ T cell count category and HIV RNA viral suppression to identify time trends in the echocardiographic measure, and group differences. Results were generally consistent between our unadjusted (see Supplementary Table S4) and adjusted models (discussed below).

FIGURE 2.

Progression of echocardiographic parameters over time. Unadjusted values of echocardiographic measures overtime for the overall cohort for the six main echocardiographic measures: E/e′ ratio, left ventricular mass, LV global longitudinal strain, RV global longitudinal strain, LV end-systolic volume, and LV end-diastolic volume. LV, left ventricular; RV, right ventricular

3.4 |. Echocardiographic trends in relation to CD4+ T cell count

In analyses adjusted for age, sex, race, hypertension, diabetes, dyslipidemia, CAD, and smoking status, LVMi increased over time irrespective of proximal CD4+ T cell count group (P for time = .01) or nadir CD4+ T cell count group (P for time = .01, Table 2). LV EDVi decreased significantly over time regardless of CD4+ T cell count group (P < .01 for both proximal and nadir CD4+ T cell count groups, Supplementary Table S2). We found no statistically significant associations between CD4+ T cell count, LV GLS, LV ESVi, and E/e′ in our adjusted analyses.

TABLE 2.

Adjusted LV mass index, LV ESV index, and LV GLS over time by CD4+ and viral suppression category

HIV disease measure	Baseline 95% CI	1 Year 95% CI	2 Year 95% CI	5 Year 95% CI	Time P-value	Interaction P-value	Group P-value
LV mass index
Proximal CD4					.01	.65	.11
<200	115.3 (104.8, 125.7)	116.5 (106.5, 126.6)	117.8 (107.9, 127.7)	121.7 (110.6, 132.9)
≥200	103.9 (94.7, 113.0)	105.7 (97.0, 114.4)	107.5 (99.0, 115.9)	113.0 (103.6, 122.3)
Nadir CD4					.01	.46	.98
<200	109.4 (101.4, 117.4)	111.5 (103.9, 119.2)	113.6 (106.1, 121.1)	120.0 (111.5, 128.3)
≥200	109.2 (96.5, 121.9)	110.4 (98.4, 122.5)	111.6 (99.9, 123.3)	115.2 (102.4, 128.0)
Viral suppression					<.001	.06	.55
No	106.8 (97.0, 116.6)	108.5 (99.1, 117.8)	110.1 (101.0, 119.2)	115.2 (105.1, 125.2)
Yes	101.7 (88.8, 114.6)	106.1 (93.9, 118.3)	110.4 (98.4, 122.4)	123.7 (109.9, 137.6)
LV ESVi
Proximal CD4					.08	.57	.63
<200	32.9 (26.9, 39.0)	32.1 (26.4, 37.8)	31.3 (25.7, 36.9)	28.9 (22.6, 35.2)
≥200	30.9 (25.8, 36.1)	30.5 (25.6, 35.4)	30.1 (25.4, 34.9)	28.9 (23.7, 34.0)
Nadir CD4					.06	.76	.78
<200	31.6 (27.3, 36.0)	31.1 (27.0, 35.3)	30.6 (26.6, 34.7)	29.1 (24.6, 33.6)
≥200	30.5(23.7, 37.2)	29.8 (23.4, 36.2)	29.1 (22.8, 35.3)	26.9 (20.2, 33.7)
Viral suppression					.69	.10	.16
No	32.1 (26.5, 37.7)	31.6 (26.2, 36.9)	31.1 (25.9, 36.3)	29.6 (23.9, 35.3)
Yes	25.3 (18.0, 32.6)	26.1 (19.2, 33.0)	26.9 (20.1, 33.6)	29.3 (21.6, 37.0)
LV GLS
Proximal CD4					.45	.77	.16
<200	−13.6 (−14.8, −12.5)	−13.6 (−14.7, −12.5)	−13.5 (−14.5, −12.4)	−13.2 (−14.5, −11.8)
≥200	−14.9 (−16.1, −13.7)	−14.8 (−15.9, −13.7)	−14.8 (−15.9, −13.7)	−14.7 (−15.9, −13.4)
Nadir CD4					.30	.80	.25
<200	−14.2 (−15.1, −13.3)	−14.2 (−15.0, −13.3)	−14.1 (−14.9, −13.2)	−13.9 (−14.9, −12.8)
≥200	−15.3 (−16.8, −13.7)	−15.1 (−16.6, −13.7)	−15.0 (−16.4, −13.6)	−14.7 (−16.3, −13.0)
Viral suppression					.30	.73	<.01
No	−13.7 (−14.8, −12.6)	−13.6 (−14.6, −12.6)	−13.6 (−14.5, −12.6)	−13.3 (−14.6, −12.0)
Yes	−16.8(−18.6, −15.0)	−16.6 (−18.4, −14.9)	−16.5 (−18.2, −14.8)	−16.0 (−18.1, −13.8)

CI, confidence interval; ESVi, end-systolic volume index; GLS, global longitudinal strain; LV, left ventricular.

3.5 |. Echocardiographic trends in relation to viral suppression

There were no significant associations observed between viral suppression status and LV ESVi, RV GLS, LV EDVi, or E/e′. There was a borderline statistically significant interaction between viral suppression and LVMi over time such that, LVMi increased to a greater degree in those with viral suppression in both unadjusted models and models adjusted models (P for interaction = .06, Table 2). Viral suppression was associated with higher LV GLS at baseline and throughout follow-up in those with viral suppression (P for group comparison < .01, Table 2). This difference in GLS based on viral suppression at baseline persisted out to >5 years of follow-up (P for viral suppression effect = .005, Figure 3). Median LVEF remained in the mid-range or normal range over the follow-up period for the majority of participants with the lower limit of IQR being >45% at most time points of our analyses stratified by nadir CD4+, proximal CD4+ T cell count, and viral suppression status (Supplementary Table S3). LVEF was higher at all time points in those with viral suppression at baseline (P = .03, Supplementary Table S3).

FIGURE 3.

Plot of estimated regression estimates for LV GLS, RV strain, E/e′, LV ESV, LV EDV, and LV mass by viral suppression. EDV, end-diastolic volume; ESV, end-systolic volume; GLS, global longitudinal strain; LV, left ventricular; RV, right ventricular

4 |. DISCUSSION

We analyzed electronic health records and archived digital echocardiograms in PLWH at single academic medical center. Our objective was to determine progression of echocardiographic markers that measure cardiac structure and function over a 10-year period. Our principal findings demonstrate the natural history of these measures, notably with an early separation in GLS based on level of viral suppression at baseline. There was an increase in LVMi over time, a decrease in LV EDVi, and no significant trend in estimated LV filling pressures (E/e′). In relation to markers of HIV disease control, we confirm prior research suggesting that early viral suppression forecasts better cardiac function and further show that these benefits persist for over a median of 5 years. Routinely acquired echocardiographic measures with known impacts on long-term clinical status are impacted by modifiable HIV disease status indicators.

We specifically addressed cardiac structure and function in a cohort without reduced LV function at baseline as measured by LVEF >40%.²⁰ GLS using STE has emerged as a reliable indicator of pre-clinical cardiac dysfunction in a number of systemic disease states.^17,18 Our previous work demonstrated that the absence of viral suppression was associated with worse LV GLS in a cross-sectional study.⁷ We now further demonstrate that viral suppression at baseline is a modifiable factor to preserve GLS over time. Baseline viral suppression differentiated GLS reliably in this group and these changes persisted over time out to >5 years.

Consistent viral suppression is generally thought to be beneficial to cardiac health as it decreases the level of circulating virus and, as a result, decreases the inflammatory response to the virus. In the Strategies for Management of Antiretroviral Therapy (SMART) study, patients assigned to a drug conservation strategy experienced a greater risk of opportunistic diseases or death, compared to patients with uninterrupted ART.²⁷ The higher risk of opportunistic disease or death in patients with interrupted ART was associated with living longer with uncontrolled HIV replication even at higher CD4+ T cell counts.²⁸ Inflammation, as assessed by circulating biomarkers, is one important aspect of how cardiac tissue responds to chronic HIV infection.²⁹ Imaging studies have further shown that PLWH have more myocardial fibrosis and steatosis – which are closely linked to myocardial dysfunction – than uninfected controls.^30,31 Although we were not able to demonstrate the myocardial pathologies at play, it is clear that as far as viral suppression is concerned for cardiac function, the sooner the better.

Left ventricular GLS was impacted by viral suppression status more so than RV GLS. There are a number of potential reasons for this finding. RV function is mostly longitudinal and has more longitudinal muscle fibers than the LV.³² Right ventricular GLS was also measured in one view whereas LV strain was measured in a combination of views. There is no consensus on the optimal methods of assessment on RV GLS. We used vendor-independent software to assess both RV and LV GLS which mitigates the effect of widespread variability in RV GLS measures which have been reported.³¹ Nonetheless, we cannot exclude that an alternate approach to measuring RV GLS could have yielded different results. The observed differences in LV GLS could portend potential long-term consequences, which warrants further examination of clinical endpoints in this group.

We did not find any significant trends in echocardiographic measures associated with abnormal LV filling as measured by E/e′. A wealth of literature has previously linked HIV infection and disease characteristics to abnormal diastolic function^11,33,34 and both low CD4+ T cell count and high viral load increase the risk of HF with preserved ejection fraction.⁴ We previously reported a 37% prevalence of left ventricular hypertrophy (LVH) and 8% prevalence of diastolic dysfunction from this cohort.²² In that cross-sectional analysis, LVH was more common in PLWH with low CD4+ T cell counts and less common in persons with viral suppression. Lower nadir and proximal CD4+ T cell counts were also significantly associated with diastolic dysfunction. Most of the parameters used to detect diastolic dysfunction by echocardiogram are sensitive to altered hemodynamics or clinical condition.³⁵ Therefore, while we found no significant trends over time in E/e′, most of the available echocardiograms has missing parameters for the other measures to reliably comment on trends in the presence/absence of diastolic dysfunction. Owing to the impact of clinical condition, medications, and other factors, we agree with findings from the Characterization of Heart Function on Antiretroviral Therapy (CHART) study that the longitudinal change in diastolic function warrants prospective study.⁶

There are a number of limitations worth highlighting. We have reported the experience of a single academic tertiary medical center. The echocardiograms were performed for a variety of reasons which we were not able to confirm over the entire time period. Many may have had a clinical indication for echocardiography. Thus, these data may not represent the experience of PLWH who are not being sent for echocardiography. Few individuals had high nadir CD4+ T cell counts and few had CAD. Similarly, routine serial viral load measures overtime were not available for every individual. Missingness of echo measures would contribute to imprecise estimates of change over time, but most of our measures showed a low degree of missingness. Missingness over time was likely related variation in clinical decision making regarding requesting an echocardiogram. STE measurements are influenced by loading conditions and due to lack of consistently available data, were not able to take into account blood pressure at the time of the echocardiogram. Lastly, we defined baseline as time of first echocardiogram, which does not consider the intensity, type, and response to ART in the time interval between HIV diagnosis and TTE, and may be confounded by indication for TTE. It is also noteworthy to highlight some unique characteristics of this study. Leveraging the longest existing archive of digital echocardiograms in the country,³⁶ we were able to remeasure all of our key outcome measures by a single experienced echocardiographer. We were also able to match individuals to their clinical data by use of an electronic data warehouse.⁷ Lastly, we analyzed progression of a number of commonly obtained measures, many of which are recommended as part of routine practice which will enable reproducibility in other settings.

5 |. CONCLUSIONS

The impacts of HIV on cardiac structure and function persist for many years. LV GLS, mass, and volumes are associated with markers of HIV disease control early in the disease course. Progression of these markers was strongly related to baseline viral load and CD4+ T cell count. Early viral suppression is associated with better cardiac structure and function over time.