Association Between Left Ventricular Scar and Ventricular Ectopy in People Living With and Without HIV

Abstract

Background

People living with HIV (PLWH) have greater risk for arrhythmic sudden death and heart failure than people without HIV (PWOH), though risk identifiers remain understudied. Higher ventricular ectopy (VE) burden reflects increased arrhythmic susceptibility and cardiomyopathy risk.

Objectives

The purpose of this study was to test if myocardial scar measured by late gadolinium-enhancement cardiovascular magnetic resonance (LGE-CMR) associates with VE by ambulatory electrocardiographic monitoring among PLWH and PWOH with risk factors for HIV, and if the association differs by HIV.

Methods

Participants from 3 cohorts of PLWH and PWOH underwent electrocardiographic monitoring (median wear time 8.3 days) and CMR. Using multivariable regression, we assessed: 1) associations between scar metrics and VE, adjusting for demographics, HIV serostatus, substance use, cardiovascular risk factors, and left ventricular (LV) function/structure; and 2) effect measure modification by HIV.

Results

Of 329 participants (median age 55 years, 30% women, 62% PLWH), 109 had LGE (62% PLWH). Ischemic or major nonischemic pattern LGE was associated with high VE burden (adjusted OR: 2.32, P = 0.004) and more PVCs/day (141% higher, P < 0.001). Among people with LGE, greater scar mass correlated with more PVCs/day (P = 0.028). Associations persisted after adjustment for LV function/structure and when excluding PLWH with HIV viremia and showed no effect measure modification by HIV.

Conclusions

Ischemic or major nonischemic pattern LGE and greater scar mass correlated with higher VE burden, independently of LV structure/function, HIV serostatus, and HIV viremia. The findings highlight specific scar characteristics common to PLWH and PWOH with risk factors for HIV that may portend higher risk for arrhythmias and heart failure.

Central Illustration

Early initiation of combination antiretroviral therapy (cART) has lengthened survival and reduced rates of AIDS-defining illnesses among people living with HIV (PLWH).¹ Nonetheless, the marked extension in longevity among PLWH has resulted in increased morbidity and mortality attributable to cardiovascular disease.² PLWH have greater risk for sudden arrhythmic cardiac death and heart failure, in part due to higher prevalence of traditional cardiovascular risk factors and unhealthy behaviors, metabolic dysregulation from cART, immune dysregulation, and/or chronic immune activation.³ However, subclinical markers identifying specific subgroups of PLWH at elevated cardiovascular risk remain understudied but could inform prevention and therapeutic strategies. Abnormal myocardial substrate, including greater myocardial scar burden, is a marker for increased risk for arrhythmias and heart failure among the general population.^4,5 Data are more sparse among PLWH. We previously reported that PLWH (vs people without HIV [PWOH] with risk factors for HIV acquisition) have greater prevalence of a nonischemic subepicardial scar pattern as well as an expanded interstitial space, assessed as extracellular volume fraction (ECV) by cardiovascular magnetic resonance imaging (CMR).⁶

While adjudicated cardiovascular event data collection is limited in community-based cohorts of PLWH limiting the power of outcome studies, subclinical markers such as high ventricular ectopy (VE) burden may reflect increased risk for cardiomyopathy and ventricular arrhythmic susceptibility.7, 8, 9 Whether VE burden is associated with scar characteristics has not been investigated, but addressing this question could identify a higher risk phenotype, inform our understanding of drivers of increased risk for clinical events, and further elucidate the connection between myocardial substrate and adverse prognosis specifically among PLWH. Thus, our objectives here were to assess if: 1) presence, extent, and pattern of myocardial scar, as measured by late gadolinium-enhancement (LGE)-CMR, is associated with VE, as measured by ambulatory electrocardiographic (ECG) patch monitoring (ZioXT, iRhythm Technologies, Inc) specifically among PLWH and people at high risk of HIV acquisition; and 2) if any observed association is stronger among PLWH compared to PWOH within this sample. This study included PLWH and PWOH who have risk factors for HIV acquisition concurrently enrolled in 3 large longitudinal cohort studies including the MACS (Multicenter AIDS Cohort Study), the ALIVE (AIDS Linked to the Intravenous Experience (ALIVE) study, and the WIHS (Women’s Interagency HIV Study).

Methods

Study cohort and procedures

Active participants of the ongoing, prospective MACS, ALIVE, and WIHS cohort studies were recruited for the SMAsH (Subclinical Myocardial Abnormalities in HIV) study for which the main goal was comparison of the prevalence of myocardial abnormalities among PLWH and PWOH with risk factors for HIV acquisition.⁶ Analyses assessing the association between the CMR-LGE metrics and VE burden were performed for the current study. MACS includes HIV seropositive and seronegative men who have sex with men in Baltimore/Washington DC, Chicago, Pittsburgh, and Los Angeles.¹⁰ The ALIVE study is a community-based cohort study that includes men and women with a history of injection drug use in Baltimore, MD.¹¹ The WIHS comprises HIV seropositive and seronegative women enrolled from 10 cities including Washington DC.¹² Participants in each cohort undergo semiannual visits for standardized interviews, physical examination, and biological specimen collection for laboratory measurements. Please see the Supplemental Appendix for detailed cohort descriptions.

From March 2015 to February 2018, as previously described, active cohort participants from the Baltimore and Chicago MACS sites, ALIVE, and the WIHS Washington DC site were invited to undergo a CMR with LGE examination and complete a nurse-administered standardized questionnaire and blood draw.⁶ After August 2015, participants were invited to wear an ambulatory ECG device (ZioXT, iRhythm Technologies, Inc), a single-use, skin-adhesive patch for up to 14 days of continuous, single-channel ECG monitoring with details of the device described previously.¹³ Institutional Review Boards for all study sites (Johns Hopkins University, Georgetown University, and Northwestern University) approved the study and all participants provided written informed consent.

Inclusion criteria comprised active cohort participants between 40 and 70 years of age. Exclusion criteria included estimated glomerular filtration rate <45 mL/min/1.73 m², weight >350 lbs, known claustrophobia or contrast allergy, and contraindications to CMR.⁶ We oversampled for PLWH aiming for ∼60% of our total target size of 400 participants. We screened 603 participants meeting age and glomerular filtration rate cutoffs and enrolled 468 of whom 402 completed CMR-LGE. Participants with <24 hours of analyzable ZioXT wear time were excluded (n = 24) and analyses of total scar mass were performed only among participants with LGE. The final study cohort for this analysis included 329 participants with both analyzable CMR-LGE and ambulatory ECG data. See Supplemental Figure 1 for the participant flow diagram.

Definitions of VE burden

Premature ventricular contractions (PVCs)/day were calculated from ZioXT output as ([1·number of isolated PVCs] + [2·number of couplets] + [3·number of triplets] + [(4+maximum beats per nonsustained ventricular tachycardia [NSVT] episode)/(2·number of NSVT episodes)])/total analyzable days of ZioXT monitor wear. (Note that for NSVT episodes comprising ≥4 beats, ZioXT output only reports the minimum, that is, 4 beats, or maximum number of beats/episode. We thus estimated a mean number of beats for NSVT episodes with ≥4 beats). High burden of VE was defined as any NSVT episode comprising ≥4 PVCs or a PVC frequency that was within the highest quartile of PVCs/day. Alternate definitions of high VE burden as explored in the literature were also calculated (see Table 1) but were too infrequent to analyze further.^7,14

Table 1.

Participant Demographic and Clinical Characteristics Stratified by Ventricular Ectopy Burden (N = 329)

	Low VE Burdena (n = 220)	High VE Burdena (n = 109)
Study
MACS Baltimore	32.3% (71)	31.2% (34)
MACS Chicago	13.6% (30)	11.0% (12)
WIHS	20.0% (44)	20.2% (22)
ALIVE	34.1% (75)	37.6% (41)
Demographics
Age	55 (51−58)	56 (52−60)
Female	29.5% (65)	31.2% (34)
Race/ethnicity
Black	67.3% (148)	70.6% (77)
White	27.3% (60)	23.9% (26)
Hispanic/other	5.5% (12)	5.5% (6)
Education (≥ high school diploma)	74.1% (163)	69.7% (76)
Substance use
Never smoker	17.3% (38)	17.4% (19)
Current smoker	47.3% (104)	44.0% (48)
Former smoker	30.0% (66)	38.5% (42)
Pack-years of smokingb	0.69 (0.00−2.54)	0.46 (0.00−2.10)
Hazardous alcohol use	11.4% (25)	13.8% (15)
Stimulant useb	41.4% (91)	29.4% (32)
CVD-related clinical
CVD historyc	1.8% (4)	13.8% (15)
Body mass index, kg/m2	26.7 (23.7−30.9)	27.0 (23.3−30.8)
Hypertensiond	50.5% (111)	64.2% (70)
Blood pressure lowering therapy use	34.1% (75)	44.0% (48)
Dyslipidemiae	60.5% (133)	62.4% (68)
Lipid-lowering therapy use	23.6% (52)	29.4% (32)
Diabetesf	9.1% (20)	19.3% (21)
Diabetes medication use	7.3% (16)	13.8% (15)
HIV-related clinical
HIV diagnosis	64.1% (141)	56.9% (62)
Among PLWH
HIV viral load detectable, >50 RNA copies/mL	25.5% (36)	30.6% (19)
CD4+ cell count, cells/μL	604 (380−794)	630 (473−886)
CD4+ nadir cell count, cells/μL	261 (131−364)	278 (161−454)
History of AIDS diagnosisg	25.0% (24)	19.4% (7)
cART use	86.5% (122)	91.8% (56)
Duration of cART, yearsg	11.9 (4.9−15.8)	13.2 (9.3−15.6)
Protease inhibitor-based	37.6% (53)	31.1% (19)
Nonnucleoside reverse transcriptase inhibitor-based	23.4% (33)	29.5% (18)
Integrase strand inhibitor-based	25.5% (36)	26.2% (16)
Other cART	0.0% (0)	4.9% (3)
Ambulatory ECG monitoring (ZioXT)
Analyzable wear time, days	7.7 (3.5−13.4)	9.1 (5.6−13.6)
Any premature ventricular contraction	95.0% (209)	100.0% (109)
Any NSVT episode, run of ≥4 ectopic beats	0.0% (0)	42.0% (46)
Premature ventricular contractions per day, n	25 (6−67)	418 (192−740)
Percent premature ventricular contractions, %	0.02 (0.01−0.06)	0.36 (0.17−0.63)
Alternate VE burden definitions
Prevalence of >30 PVCs/hour or a single run ≥20 beats	0.0% (0)	29.4% (32)
Prevalence of percent PVCs ≥0.7%	0.0% (0)	22% (24)
Prevalence of total percent PVCs ≥1% or NSVT episodes/day in highest quartile	0.0% (0)	48.6% (53)

Values are % (n) or median (IQR).

ALIVE = AIDS Linked to the Intravenous Experience; cART = combination antiretroviral therapy; CVD = cardiovascular disease; ECG = electrocardiogram; MACS = Multicenter AIDS Cohort Study; NSVT = nonsustained ventricular tachycardia; PVC = premature ventricular complex; VE = ventricular ectopy; WIHS = Women’s Interagency HIV Study.

^aHigh VE burden defined as any nonsustained VT episode with ≥4 consecutive ectopic beats or highest quartile of premature ventricular contractions/day.

^bAcross 5 years prior to cardiovascular magnetic resonance imaging.

^cDefined as prior myocardial infarction, coronary angioplasty, stent, coronary artery bypass grafting, other heart surgery, or heart failure confirmed by medical records.

^dDefined as use of antihypertensive medications or systolic blood pressure ≥140 mm Hg or diastolic blood pressure ≥90 mm Hg averaged over the preceding 5 years when available or at the time of CMR study visit.

^eDefined as use of lipid-lowering medications or fasting total cholesterol level ≥200 mg/dL or low-density lipoprotein (LDL) cholesterol level ≥130 mg/dL or high-density lipoprotein (HDL) cholesterol level <40 mg/dL or serum triglyceride level ≥150 mg/dL closest to the time of CMR study visit.

^fDefined as use of hypoglycemic meds or fasting serum glucose levels ≥126 mg/dL closest to the time of CMR study visit. Hemoglobin A1C level <6.5% was used to exclude diabetes if fasting glucose levels were not available.

^gData unavailable for ALIVE participants.

CMR acquisition, metrics, and analyses

CMR studies were acquired and analyzed as described.⁶ Using a standardized protocol, CMR studies were acquired at Johns Hopkins Hospital (Baltimore, Maryland, USA) or Northwestern Memorial Hospital (Chicago, Illinois, USA) with 1.5 T scanners (Siemens Avanto or Aera). Short- and long-axis cines (30 phases/cardiac cycle) were acquired with a steady-state free precession sequence. Two-dimensional LGE short- and long-axis images were acquired 15 minutes after intravenous administration of 0.2 mmol/kg of gadobutrol (GadavistVR, Bayer) using an inversion-recovery fast gradient-echo sequence. T1 mapping with a modified Look-Locker inversion recovery sequence was performed precontrast and postcontrast.¹⁵ T2 mapping was performed in a subset of participants using a T2-prepped balanced steady-state free precession sequence with motion correction.¹⁵

CMR analyses were blinded to participant characteristics and HIV serostatus. Analyses of biventricular structure, function, and LGE were conducted with Segment v2.0 with biventricular volumes and mass quantified by standard methods. All LGE images were reviewed for LGE presence in 2 planes and regions with hyper-enhanced signal intensity were then quantified as total scar mass in grams. Myocardial scar pattern was classified based on LGE location. LGE distributed along a coronary artery territory and involving the subendocardium with or without transmural involvement was categorized as an ischemic pattern and otherwise categorized as nonischemic. Nonischemic LGE patterns were categorized as “major” if involving the midwall or subepicardium as previously described.^6,16 Minor pattern included small foci of LGE including those involving the right ventricular insertion points.^6,16

T1 and T2 mapping images obtained from the mid-ventricular short axis slice were postprocessed with MRmap version 1.2 (Charite University Medicine, Berlin, Germany). Extracellular volume fraction (ECV, %) was calculated as: (ECV = 100 × partition coefficient × [1 − hematocrit]) for which partition coefficient was determined by the slope of the linear relationship of 1/T1myocardium vs 1/T1blood precontrast and postcontrast.¹⁷

The primary CMR outcomes of interest comprised LGE presence (vs absence), LGE pattern (presence of ischemic or major nonischemic pattern vs absence), and total scar extent. Both ischemic and major nonischemic LGE patterns were combined in the primary analysis given their known prognostic significance among the general population.¹⁶ Secondary outcomes included native T1, ECV, and native T2 values.

Covariate measurement

The analysis used primary data collected at the SMAsH study CMR visit and secondary longitudinal data collected prospectively within the 3 parent cohorts. HIV serostatus was measured within the parent studies by a positive antibody test and confirmed by Western blot. Date of birth, sex, race/ethnicity, and education level were self-reported via the parent studies. Body mass index (BMI) was calculated from height and weight measured at the CMR visit. Recent hazardous alcohol use as defined by the AUDIT assessment was measured by nurse-administered questionnaire at the SMAsH visit, as were the current use of antihypertension medication and substance use. In order to better leverage exposure to risk factors over the 5 years prior to the CMR, we combined data from both sources to generate an average systolic blood pressure, classify diabetes, cumulate pack-years of cigarette smoking, and capture any reported stimulant use during that timeframe. We further supplemented parent cohort data with the collection of self-reported cardiovascular events, and medications, lipid profile, serum creatinine, and hematocrit at the time of CMR. Among PLWH, measures of HIV disease activity including plasma HIV RNA concentrations (viral load), current and nadir CD4+ T-lymphocyte cell counts/μL (CD4), history of acquired immunodeficiency syndrome (AIDS)-defining clinical events, and cART use were obtained from the parent cohort data.

Statistical methods

We summarized participant demographic, clinical, CMR, and ZioXT data by median (IQR) or proportion (count). We performed complete case analyses estimating cross-sectional odds ratios for high VE burden using multivariable logistic regression, comparing those with vs without dichotomous scarring measures or per 1 g increment in total scar mass. We also estimated the difference in PVCs/day by the same scarring parameters using multivariable linear regression. We log-transformed PVCs/day due to a heavy right-skew and reported e^β, which is equivalent to the ratio of geometric mean PVCs/day per increment in scarring parameter and can be arithmetically transformed to be interpreted as a % difference = (ratio−1)·100. Sequential models were fit to account for both potential confounders as well as causal mediators. Models were adjusted for age, sex, race/ethnicity, education level, HIV serostatus, BMI, systolic blood pressure, use of antihypertensive medications, diabetes, pack-years of smoking over the 5 years preceding CMR, current hazardous alcohol use, and stimulant use over the 5 years preceding CMR (Model 1). Total analyzable duration of ZioXT monitoring was included in the models assessing high VE burden but not those assessing PVCs/day, as the latter was calculated with wear time in the denominator. Model 2 further adjusted for left ventricular (LV) ejection fraction (EF), LV mass indexed to body surface area (LVMi), and LV end-diastolic volume indexed to body surface area (LVEDVi).

To assess heterogeneity of the associations between myocardial scarring and VE by HIV serostatus, we performed stratified multivariable analyses by HIV serostatus and also tested a multiplicative interaction term for each scarring measure (HIV × scar) in primary models. Post hoc sensitivity analyses included alternative approaches to calculating PVCs/day, exclusion of PLWH with detectable HIV viral load, and further adjustment of PVCs/day models by total analyzable ZioXT wear time. Sensitivity analyses using alternative methods for calculating PVCs/day—that is, excluding NSVT episodes and using maximum recorded beats per VT episode instead of a calculated mean as defined—yielded nearly identical results to those of primary analyses, as did further adjustment for duration of analyzable ZioXT wear time in models regressing PVCs/day.

All analyses were conducted using R version 4.0 and a type I error rate of 0.05 was used to assess statistical significance.

Results

Characteristics of the study cohort

Table 1 shows unadjusted baseline characteristics by VE burden for the final analytic sample of 329 participants. The median age was 55 years (IQR: 51-59 years); 30% were women; 68% self-identified as Black; and 62% were PLWH, of whom 73% were virally suppressed (<50 HIV RNA copies per ml) and 88% were receiving cART. Overall, 31% had a history of injection drug use. Analyzable ZioXT data were available for a median of 8.3 days (IQR: 3.8-13.5 days) per participant. CMR metrics by VE burden are summarized in Table 2. Most participants had normal LV systolic function as well as normal LV volumes, and mass. Only 3 participants had LVEF <50%, all of whom also had high VE burden. LGE was present among 33% of participants (n = 109), of whom 8% (n = 9) had an ischemic pattern and 67% (n = 73) had a major nonischemic pattern. Among those with LGE, median total scar mass was 4.0 g [IQR: 2.8-7.0 g]. Prevalence of HIV seropositivity among those with LGE was 62% (n = 68). Supplemental Table 1 shows unadjusted baseline characteristics by parent cohort.

Table 2.

Participant Cardiac Structural, Functional, and Tissue Characteristics Measured via Cardiovascular Magnetic Resonance Imaging Stratified by Ventricular Ectopy Burden

	Low VE Burdena		High VE Burdena
LGE presence	220	30.0% (66)	109	39.4% (43)
LGE pattern among those with LGE
Major nonischemic	66	63.6% (42)	43	72.1% (31)
Ischemic	66	1.5% (1)	43	18.6% (8)
Major nonischemic or ischemic	66	65.2% (43)	43	90.7% (39)
Total scar mass, g	66	3.25 (2.31−5.17)	43	5.88 (3.62−9.94)
LGE pattern among those with nonischemic LGE
Minor foci	65	35.4% (23)	35	11.4% (4)
Subepicardial	65	30.8% (20)	35	62.9% (22)
Mid-myocardial	65	33.8% (22)	35	25.7% (9)
Major LGE (subepicardial or mid-myocardial)	65	64.6% (42)	35	88.6% (31)
Native T1, ms	212	998 (973−1,028)	107	1,000 (973−1,022)
Native T2, ms	72	50.5 (47.8−53.9)	45	52.2 (50.9−54.3)
Extracellular volume fraction, %	187	28.4 (26.1−30.1)	100	29.0 (26.4−30.7)
High extracellular volume fraction, ≥30% among women, ≥28% among men	187	47.1% (88)	100	56.0% (56)
Left ventricular mass index, g/m2	218	61.0 (55.2−67.3)	108	61.9 (55.6−69.4)
Left ventricular end-diastolic volume index, mL/m2	218	65.5 (58.5−74.0)	108	67.8 (57.2−79.7)
Left ventricular end-systolic volume index, mL/m2	218	17.8 (14.5−21.6)	108	20.0 (15.1−26.3)
Left ventricular ejection fraction, %	218	72.5 (69.3−75.5)	108	70.4 (65.8−76.4)
Left ventricular ejection fraction <50%	218	0.0% (0)	108	2.8% (3)

Values are n, % (n), or median (IQR).

LGE = late gadolinium enhancement; VE = ventricular ectopy.

^aHigh VE burden defined as any nonsustained ventricular tachycardia episode with ≥4 consecutive ectopic beats or highest quartile of premature ventricular contractions/day.

Scar characteristics and VE burden

Associations between scar presence and pattern and both: 1) high vs low VE burden; and 2) mean PVCs/day are shown in Figure 1. There was a nonsignificant association between LGE presence (vs absence) and odds of higher VE burden after adjustment for demographics and clinical characteristics (Model 1) and with further adjustment for CMR LVEF, LVMi, and LVEDVi (Model 2). LGE presence was significantly associated with a higher number of PVCs/day among all participants in Model 1 (1.88 times higher geometric mean PVCs/day, 95% CI: 1.21-2.92; P = 0.005) and Model 2 (1.64 times higher geometric mean PVCs/day; CI: 1.05-2.56; P = 0.029). Parameter estimates were similar when stratified by HIV serostatus.

Figure 1.

Cross-Sectional Associations Between Myocardial Scar (LGE) and Ventricular Ectopy Among All Participants and by HIV Serostatus

LGE = late gadolinium enhancement; PLWH = people living with HIV; PWOH = people without HIV; VE = ventricular ectopy.

Among all participants, ischemic or major nonischemic pattern LGE presence (vs absence) was associated with 2.32 times higher odds of high VE burden (95% CI: 1.31-4.11; P = 0.004) following adjustment for demographic and clinical characteristics (Model 1) as well as a higher number of PVCs/day (2.41 times higher geometric mean PVCs/day; 95% CI: 1.50-3.87; P < 0.001). Upon further adjustment for LVEF, LVMi, and LVEDVi (Model 2), both parameter estimates were mildly attenuated but remained statistically significant. In models regressing high vs low VE burden on ischemic or major nonischemic pattern LGE and demographic and clinical characteristics, no significant independent associations were observed for any other covariate (Supplemental Table 2a). In models regressing incremental PVCs/day on the same variables, independent associations were observed for older age (P = 0.02), higher BMI (P = 0.03), diabetes (P = 0.02), and lower LVEF (P = 0.04) (Supplemental Table 2b). Of note, stimulant use was not independently associated with high VE burden or number of PVCs/day in the multivariable models. There was no evidence of effect measure modification of any association by HIV serostatus, assessed via multiplicative interaction terms (all P > 0.10).

Total scar mass and VE burden

Table 3 shows cross-sectional associations between total scar mass and VE among participants with LGE. Within this subsample, total scar mass was associated with significantly higher mean PVCs/day (1.07 times higher geometric mean PVCs/day per 1 g increment; 95% CI: 1.01-1.14; P = 0.028) and 1.10 times higher odds of high VE burden (95% CI: 1.00-1.21; P = 0.06), following adjustment for demographics and traditional CVD risk factors (Model 1). These associations were mildly attenuated and no longer statistically significant following further adjustment for LVEF, LVMi, and LVEDVi (Model 2).

Table 3.

Cross-Sectional Associations Between Total Scar Mass and Ventricular Ectopy Among Those With LGE Detected on Cardiovascular Magnetic Resonance Imaging (N = 109)

	Estimate per 1 g Increment in Total Scar Mass (95% CI)
	Odds Ratio for High VE Burdena	P Valuea	Ratio of Geometric Mean PVCs per db	P Valueb
Model 1: demographic and clinicalc	1.10 (1.00-1.21)	0.060	1.07 (1.01-1.14)	0.028
Model 2: +LV structure and functiond	1.09 (0.98-1.21)	0.111	1.05 (0.99-1.12)	0.116

LGE = late gadolinium enhancement; LV = left ventricular; PVC = premature ventricular contraction; VE = ventricular ectopy.

^aModeled using multivariable logistic regression.

^bModeled using multivariable linear regression with log-transformed outcome (PVCs/d). The exponentiated parameter estimate is presented, equivalent to the ratio of geometric mean PVCs per day for every 1 g increment in total scar mass. This can be interpreted as a percent difference, eg, adjusting for demographic and clinical characteristics, there was, on average, 7% higher PVCs per day for every 1 g increment in total scar mass.

^cAdjusted for age, sex, race/ethnicity, education, HIV serostatus, body mass index, pack-years of smoking over 5 years preceding exam, current hazardous alcohol use, stimulant use over 5 years preceding exam, systolic blood pressure, antihypertensive medication, and diabetes. Total analyzable time on electrocardiogram monitoring was included in models assessing high VE burden but not those assessing PVCs/d.

^dFurther adjusted for left ventricular (LV) ejection fraction, LV mass index, and LV end-diastolic volume index.

Scar characteristics and VE burden among virally suppressed PLWH

Subgroup analyses of associations between myocardial scar measures and VE among PLWH excluding those with detectable HIV viral load yielded similar results to primary analyses (Supplemental Table 3). Among participants with nonischemic LGE, both a subepicardial pattern (adjusted OR: 8.30; 95% CI: 2.37-29.05; P = 0.001) and a major LGE pattern (adjusted OR: 8.75; 95% CI: 2.08-36.74, P = 0.003) were associated with higher VE burden compared to other patterns.

T1, ECV, and T2 and VE burden

Results from analyses of secondary myocardial tissue characteristics are shown in Supplemental Table 4. There was no significant relationship between native T1 or ECV values and VE. An SD increment of native T2 values was associated with 1.76 times greater odds of a high VE burden, on average (95% CI: 1.10-2.80; P = 0.018) and 1.39 times higher geometric mean PVCs/day (95% CI: 1.01-1.92; P = 0.047). With further adjustment for LV structure and function, these estimates were only mildly attenuated.

Discussion

In this study of community-dwelling PLWH and PWOH who have risk factors for HIV acquisition, LGE presence and an ischemic or major nonischemic LGE pattern were associated with higher VE burden as measured by higher frequency of PVCs/day, independent of traditional CVD risk factors (Central Illustration). Among participants with LGE, total scar mass also correlated with higher VE burden. These results did not differ significantly by HIV serostatus, persisted following exclusion of PLWH with detectable HIV viral load and were largely independent of LV structure and function. By leveraging both CMR and ambulatory ECG patch monitoring, the present study provides novel correlations between VE burden reflecting ventricular irritability and abnormal myocardial substrate, assessed as myocardial scar extent and patterns that may identify a cohort at higher risk for subsequent ventricular arrhythmic and heart failure events.

Central Illustration.

CMR-LGE Presence and Specific Patterns Are Associated With Higher Burden of Ventricular Ectopy Among People Living With HIV or Who Have Risk Factors for HIV

ALIVE = AIDS Linked to the Intravenous Experience; CVD = cardiovascular disease; LV = left ventricular; MACS = Multicenter AIDS Cohort Study; PLWH = people living with HIV; PWOH = people without HIV; WIHS = Women’s Interagency HIV Study.

Myocardial scar and scar patterns among PLWH and PWOH who have risk factors for HIV: Comparison with prior studies

Among the general population, the prognostic significance of myocardial scar assessed by CMR has been well-described among high-risk subgroups (eg, those with reduced LVEF).^4,5,18 Furthermore, among older cohorts of asymptomatic individuals, the prevalence of clinically unrecognized myocardial scar with either ischemic or nonischemic pattern is common and ranges from 6.2% up to 19.8%.18, 19, 20 Both ischemic and major nonischemic scar patterns have been associated with increased risk for subsequent major adverse cardiovascular events among a community-based sample of elderly people without prevalent HIV.^16,20,21 Among PLWH, the prevalence of LGE reported in the literature varies widely at 8% to 83%, for which our current findings of 33% are intermediate; these differences likely reflect differing cohort characteristics and definitions of scar.^6,22, 23, 24 Our findings confirm those of prior studies which reported predominantly nonischemic rather than ischemic scar patterns among PLWH as well as nonischemic midwall and subepicardial involvement.^6,22, 23, 24

To our knowledge, only 1 study by de Leuw et al has reported associations between cardiovascular prognosis and CMR metrics among PLWH.²⁵ It included a smaller cohort of 156 PLWH without a comparison group. A composite outcome was assessed, comprising not only arrhythmic or heart failure (HF) but also ischemic events.²⁵ Compared to our participants, the de Leuw study included White participants living in Europe who had a lower prevalence of any scar by LGE (22% vs 33%), more frequent ischemic scars among those with LGE (37% vs 8%), and higher prevalence of cardiovascular disease (12% with coronary artery disease and 19% with prior HF vs 6% overall rate in our study).²⁵ Variability in findings between the 2 studies as discussed below may thus partly be attributable to differences in the endpoint examined and importantly, cohort composition. The relationship between scar patterns and size and clinical outcomes was not assessed in the European study.

In the de Leuw study, the proportion of patients with LGE presence and an ischemic pattern of LGE was higher among those with vs without events but after multivariable adjustment, LGE was no longer associated with the composite outcome. Native T1, reflecting expanded interstitial space, and LV mass were the only variables independently associated with adverse outcomes.²⁵ In contrast, we found no associations between VE burden and native T1 (or ECV, another metric of expanded interstitial space) or LV mass in secondary analyses. Notably, slight variations across CMR vendors may affect the T2 sensitivity of native T1 mapping sequences and could explain the differences in the T1 results between the studies.^25,26 Both our study and that of de Leuw found that higher T2 values were related to greater odds of both adverse outcomes and higher VE burden, though associations were less strong compared to other CMR metrics. These observations support the hypothesis that myocardial edema due potentially to chronic inflammation plays a role in subclinical myocardial tissue remodeling and nonischemic fibrosis that may be precursors to symptomatic disease both among PLWH and PWOH with and without risk factors for HIV acquisition.^25,27,28

Significance of VE burden

Among community-dwelling populations of PWOH, VE has been reported as a surrogate marker of risk for adverse outcomes.⁸ Studies show an association between increased frequency of PVCs and greater odds of a subsequent decrease in LVEF and incident HF independent of coronary heart disease.^7,29 Presence of PVCs also identifies a subset of otherwise low-risk individuals, including those without overt structural and functional LV abnormalities, as having higher risk for cardiovascular disease-related death including sudden cardiac death.^30,31 However, the mechanisms for the observed association between PVC presence/burden and adverse outcome remain to be fully elucidated.

The relationship between PVCs and subclinical myocardial abnormalities as the present study highlights here has not been previously reported. Even small regions of myocardial scar could lead to increased ventricular irritability and thus explain the adverse prognosis associated with high VE burden, which requires further confirmation. Further investigation is also needed of factors that may worsen VE burden, either independently or in conjunction with myocardial scar characteristics. A prior study of VE burden in a community cohort of PWOH identified higher blood pressure, lack of physical activity, and smoking as modifiable risk factors.³² We identified higher BMI and prevalent diabetes as additional risk predictors of VE burden in our study population of PLWH and PWOH who have risk factors for HIV acquisition which may further inform our understanding of contributing factors.

In interpreting and comparing our study results with prior published literature and guiding future investigations, another important consideration is how VE burden was measured. The duration of monitoring for arrhythmias is quite heterogeneous in the literature, most often comprising 24- to 48-hour Holter monitoring but also including ectopy detected on a 10-second 12-lead electrocardiogram or 2-minute rhythm strip. Extended continuous monitoring as performed here increases the detection rate of VE and more importantly, affords the opportunity for more consistent assessment of arrhythmic burden.³³ There is, however, no universally accepted duration of monitoring, nor are there accepted definitions of high VE burden. As the greater convenience of continuously recording electrocardiographic devices has expanded usage, criteria are needed and could be informed by the present and subsequent studies.

Study limitations and strengths

Limitations of the present study include the cross-sectional design, which precludes ascertainment of causation. Because the cohorts included in this study comprised conveniently sampled men and women from the community, the results may not be representative of PLWH or people in the general population living without HIV in the United States. While we adjusted for key covariates, the potential for residual confounding cannot be excluded. T1 and T2 mapping results were not available in all patients. Assessment of morphologic features and other characteristics of PVCs is beyond the scope of this analysis. Furthermore, our sample size allows for the detection of small to moderate effect sizes, so elevated type II error should be considered when interpreting P values. This is of particular concern in subgroup analyses where there are fewer subjects per variable in the model. There are several notable strengths of this study. The participants were more racially diverse than in most other studies, and thus, results are potentially more generalizable. In contrast to other studies which often enrolled healthy controls or did not include a comparison group, our cohort comprised sociodemographically similar community-dwelling PLWH and PWOH with risk factors for HIV acquisition. This affords us the opportunity to investigate the role of HIV-specific factors in increasing risk among PLWH as compared to HIV-uninfected persons with otherwise similar clinical profiles. In so doing, we found no significant effect measure modification by HIV serostatus and similar parameter estimates when excluding PLWH with detectable viral load. These findings might suggest that the presence of the scar patterns described and increasing scar extent contribute to VE burden to a similar degree among PLWH and PWOH who have higher risk for HIV acquisition when accounting for traditional cardiovascular risk factors and measured risk factors for HIV acquisition. However, the presence of ischemic or major nonischemic pattern LGE among PLWH with high VE burden may represent a unique, high-risk subclinical phenotype that warrants further study.

Conclusions

The present study highlights prevalent subclinical abnormalities in myocardial substrate associated with higher ventricular irritability among a subset of individuals living with or without HIV as well as the absence of modification by HIV serostatus. Specific CMR features, namely ischemic or major nonischemic LGE patterns and larger scar extent, when present among people with high VE burden, may portend an elevated risk for subsequent LV dysfunction, clinical HF, and/or clinically significant ventricular arrhythmic events that merits additional exploration.

PERSPECTIVES.

COMPETENCY IN MEDICAL KNOWLEDGE: PLWH have higher risk for subsequent heart failure and ventricular arrhythmias, though risk markers remain elusive. Among people in the general population without HIV, higher VE burden is a surrogate marker for adverse cardiovascular outcomes including heart failure and sudden arrhythmic cardiac death and reflects ventricular irritability. Myocardial scar could provide an abnormal substrate that increases VE burden. In this cohort of PLWH and PWOH with risk factors for HIV acquisition, the present study found that the presence and larger extent of myocardial scar as well as an ischemic or major nonischemic pattern of scar assessed by CMR-LGE were associated with a higher burden of VE. The strength of the relationship persisted among well-treated PLWH with low viral load and was similar between PLWH and PWOH with risk factors for HIV.

TRANSLATIONAL OUTLOOK 1: The results of the current study are based on cross-sectional analyses of a cohort study. Whether the CMR findings alone or in combination with higher VE burden identify a subgroup of PLWH with higher incidence of heart failure and ventricular arrhythmic events requires longitudinal follow-up.

TRANSLATIONAL OUTLOOK 2: Serial assessment of the evolution of CMR patterns and scar extents in conjunction with temporal changes in VE burden could inform a mechanistic understanding of the relative and causative roles of each factor in subsequent LV remodeling and cardiovascular prognosis in PLWH and PWOH who have risk factors for HIV.

Funding support and author disclosures

Data in this paper was collected by the Multicenter AIDS Cohort Study (MACS) and the Womens Interagency HIV Study, now the MACS/WIHS Combined Cohort Study (MWCCS). The contents of this publication are solely the responsibility of the authors and do not represent the official views of the National Institutes of Health (NIH). MWCCS sites contributing to this study include (Principal Investigators): Baltimore CRS (Todd Brown and Joseph Margolick), U01-HL146201; Data Analysis and Coordination Center (Gypsyamber D'Souza, Stephen Gange and Elizabeth Golub), U01-HL146193; Chicago-Northwestern CRS (Steven Wolinsky), U01-HL146240; and the Metropolitan Washington CRS (Seble Kassaye and Daniel Merenstein), U01-HL146205. The MWCCS is funded primarily by the National Heart, Lung, and Blood Institute (NHLBI), with additional co-funding from the Eunice Kennedy Shriver National Institute Of Child Health & Human Development (NICHD), National Human Genome Research Institute (NHGRI), National Institute on Aging (NIA), National Institute of Dental & Craniofacial Research (NIDCR), National Institute of Allergy and Infectious Diseases (NIAID), National Institute of Neurological Disorders and Stroke (NINDS), National Institute of Mental Health (NIMH), National Institute on Drug Abuse (NIDA), National Institute of Nursing Research (NINR), National Cancer Institute (NCI), National Institute on Alcohol Abuse and Alcoholism (NIAAA), National Institute on Deafness and Other Communication Disorders (NIDCD), National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK). MWCCS data collection is also supported by UL1-TR000004 (UCSF CTSA), P30-AI-050409 (Atlanta CFAR), P30-AI-050410 (UNC CFAR), and P30-AI-027767 (UAB CFAR). ALIVE data collection was supported by the NIH/NIDA: U01-DA036297 (Shruti Mehta and Gregory Kirk). This study was funded by the NIH/NHLBI: RO1 HL126552 (Drs Wu and Post); U01-HL146201, U01-HL146193, U01-HL146240, and U01-HL146205 (MACS, WIHS data collection); U01-DA036297 (ALIVE data collection); and T32-HL007227 (TEP training support). Dr Palella is a consultant and/or on the Speakers Bureau for Gilead Sciences, Janssen Pharmaceuticals, Merck & Co Inc and ViiV. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose.