Cardiac Dysfunction and Mortality in HIV-Infected Children

Abstract

Background

Left ventricular (LV) dysfunction is common in children infected with the human immunodeficiency virus (HIV), but its clinical importance is unclear. Our objective was to determine whether abnormalities of LV structure and function independently predict all-cause mortality in HIV-infected children.

Methods and Results

Baseline echocardiograms were obtained on 193 children with vertically transmitted HIV infection (median age, 2.1 years). Children were followed up for a median of 5 years. Cox regression was used to identify measures of LV structure and function predictive of mortality after adjustment for other important demographic and baseline clinical risk factors. The time course of cardiac variables before mortality was also examined. The 5-year cumulative survival was 64%. Mortality was higher in children who, at baseline, had depressed LV fractional shortening (FS) or contractility; increased LV dimension, thickness, mass, or wall stress; or increased heart rate or blood pressure (P≤0.02 for each). Decreased LV FS (P<0.001) and increased wall thickness (P=0.004) were also predictive of increased mortality after adjustment for CD4 count (P<0.001), clinical center (P<0.001), and encephalopathy (P<0.001). FS showed abnormalities for up to 3 years before death, whereas wall thickness identified a population at risk only 18 to 24 months before death.

Conclusions

Depressed LV FS and increased wall thickness are risk factors for mortality in HIV-infected children independent of depressed CD4 cell count and neurological disease. FS may be useful as a long-term predictor and wall thickness as a short-term predictor of mortality.

Children infected with the human immunodeficiency virus (HIV) may develop a wide range of cardiovascular abnormalities, some of which are known to be associated with poor survival.^1–3 In addition, we recently reported that among HIV-infected children, baseline echocardiographic abnormalities are common, persistent, and often progressive.¹

To determine the clinical value of baseline echocardiographic findings as predictors of mortality, we studied children with vertically transmitted HIV infection participating in a National Heart, Lung, and Blood Institute study, Pediatric Pulmonary and Cardiac Complications of Vertically Transmitted HIV Infection (P²C² HIV). We examined 9 echocardiographic measures of left ventricular (LV) structure and function at enrollment to determine whether any abnormalities predicted mortality after adjustment for demographic variables and other risk factors. We also constructed longitudinal profiles of the echocardiographic measurements to determine how early the predictors could distinguish between survivors and nonsurvivors.

Methods

We have previously described the P²C² HIV study, a natural history study of cardiac and pulmonary complications of vertically transmitted HIV infection at 5 clinical centers (10 hospitals) in different parts of the United States.⁴ Although 2 prospective cohorts of HIV-infected children were included in the study design, this report is limited to the older HIV-infected cohort (group 1). Briefly, 205 group 1 children >28 days old with documented maternally transmitted HIV infection were enrolled between May 1990 and April 1993. Recruitment of patients occurred in pediatric clinics (hospital-based and non-hospital-based) and inpatient wards, resulting in a sequential enrollment of all consenting patients. The protocol was approved by the institutional review board at each center. Informed consent was obtained from the patient, the parent, or the legal guardian. Patient interval histories were obtained during visits to the clinic or retrospectively from medical records. The other prospective cohort consisted of 600 infants born to HIV-infected mothers enrolled during pregnancy or before 28 days postpartum. They are not included in the present article because the covariates measured at birth are very different from those measured in a cohort of patients who enroll at later ages and are mostly symptomatic.

All children underwent protocol-directed echocardiographic testing every 4 months regardless of clinical status using Hewlett Packard 500, 1000, 1500, and Acuson 128 XP equipment. All echocardiograms were centrally remeasured by 1 of 3 technicians unaware of the clinical status of the patient. For each echocardiographic study, children <4 years old were sedated if necessary. 2D echocardiography and Doppler studies with stress-velocity analysis were performed for each child.¹ Afterload was measured as meridional end-systolic LV wall stress. LV mass was calculated from the M-mode measurements by the method of Devereux et al.⁵

Normative values for the echocardiographic measures according to age or body surface area were developed from 285 healthy children (these external control subjects were not part of the P²C² HIV study) measured at the same central interpretation unit in the same manner as the patients.¹ To adjust for growth, z scores were created for the HIV-infected children by taking each echocardiographic measure, subtracting the age-appropriate or body surface area-appropriate mean, and dividing by 1 SD. Therefore, a z score of 0 represents a measurement equal to the normal mean value for the child’s age or body surface area, whereas a z score of −2 represents a measurement 2 SD below average. Age correction was used for fractional shortening (FS), wall stresses, blood pressure, and heart rate; body surface area correction was used for LV dimension, end-diastolic posterior wall thickness, and mass.¹

The prognostic variables were measured at the time of the initial echocardiogram. Covariates were chosen because of prior work suggesting that they were important for cardiovascular morbidity and mortality.^1–3 Baseline covariates were sex, race, clinical center, CD4 cell count z score, height and weight z scores,⁶ CDC HIV-disease stage, encephalopathy, a diagnosis of Pneumocystis carinii pneumonia, chest radiographic findings, any zidovudine exposure, and age at initial echocardiogram. Baseline measures of cardiac function were FS, end-diastolic dimension, end-systolic dimension, heart rate, LV mass, end-systolic wall stress (afterload), wall thickness, contractility, and diastolic blood pressure. Each of the echocardiographic parameters was dichotomized at ≥2 SD except FS (≤−2 SD), contractility (≤−2 SD), and wall thickness (≤1 SD because too few children exceeded 2 SD). An abnormal chest radiograph was defined by the presence of nodular densities, reticular densities, parenchymal consolidation, or increased bronchovascular markings. These radiographs were read with a standardized tool⁷ at each center by a pediatric radiologist unaware of the child’s clinical status. CD4 lymphocyte counts were determined from the first available reading within 6 months of the initial echocardiogram at laboratories using AIDS Clinical Trials Group quality assurance protocols. z scores were determined for CD4 counts⁸ and for height and weight⁶ and were categorized as either ≤−2 or >−2. Serum was analyzed for HIV-1 RNA concentration by quantitative HIV-1 RNA polymerase chain reaction.⁹ The 1994 revised CDC classification system¹⁰ was used to classify each child according to the most severe clinical HIV stage up to the time of first echocardiogram.

Statistical Analyses

Cumulative survival was estimated with the Kaplan-Meier method. Log-rank tests were used to compare survival according to baseline clinical characteristics and baseline measures of cardiac function, with groups defined by dichotomized z scores. Relative risks were calculated to measure the degree of association between the baseline cardiac function z scores and survival by fitting the Cox proportional-hazards regression model separately for each baseline echocardiographic measurement. The Spearman rank-order correlation coefficient was used to determine the association between echocardiographic parameters. All tests were 2-sided and unadjusted for multiple comparisons. A value of P<0.05 indicated statistical significance.

Forward and backward stepwise selection were used to choose prognostic variables for a Cox proportional-hazards regression model, and both methods led to the same results. Only factors that were significant at P<0.05 in the univariable analyses were included in the multivariable analyses. The relative risk and its 95% CI were calculated for each factor in the presence of others in the final model.

Repeated-measures analyses were performed for each cardiac function measurement and z score to examine the amount of time before death that a predictor could discriminate children who died from those who were still alive at the end of follow-up. For each cardiac outcome, these analyses performed by SAS Proc Mixed provided separate estimates of the mean and 95% CIs according to vital status and time before the last echocardiogram or death.

Results

Patient Population

Of the 205 enrolled children, a central analysis of echocardiographic data was performed for 193 children, who constitute the study cohort. The remaining 12 children were not included because of inability to centrally remeasure the initial echocardiogram (5 children, of whom 4 died and 1 was lost to follow-up), 1 large atrial septal defect, pulmonary hypertension and wall motion abnormalities (2 children, 1 of whom died), and absence of an echocardiogram (4 children, of whom 2 died and 2 were lost to follow-up).

Most of the children were black (86 children) or Hispanic (73 children), and only 22 children were asymptomatic before the initial echocardiogram. The median age at the first echocardiogram was 2.1 years and the median CD4 cell count was 690/mm³ (median z score, −1.92 SD; the normal CD4 cell count for a 2-year-old was 2298/mm³).⁸ Table 1 shows the cumulative survival, with 64 of the 193 study children dying. The overall 5-year survival was 64% (95% CI, 56.6% to 71.3%). The median length of follow-up for the 129 children alive at last contact was 60 months. Of the 27 lost to follow-up, additional data on vital status were obtained for 21 children.

TABLE 1.

Cumulative Survival Among 193 HIV-Infected Children According to Baseline Clinical Characteristics

		Deaths		Cumulative Survival (%)±SEM
Characteristic	n	n	%	1 y	5 y	P
All children	193	64	33.2	89.6±2.2	64.0±3.8
Sex
Male	89	31	34.8	92.1±2.9	62.9±5.5	0.70
Female	104	33	31.7	87.4±3.3	64.9±5.1
Race or ethnicity
White	28	6	21.4	96.4±3.5	76.4±8.7	0.20
Black	86	27	31.4	88.2±3.5	65.9±5.6
Hispanic	73	30	41.1	89.0±3.7	56.0±6.2
Other	6	1	16.7
Clinical center
1	28	14	50.0	78.6±7.8	52.2±9.7	0.02
2	38	8	21.1	92.1±4.4	78.1±6.9
3	40	19	47.5	87.2±5.4	45.2±9.1
4	34	11	32.4	91.2±4.9	66.7±8.2
5	53	12	22.6	94.3±3.2	72.3±7.2
Age at initial echocardiogram, y
0–1	40	15	37.5	92.2±4.3	58.2±8.4	0.70
1–2	48	12	25.0	85.4±5.1	74.5±6.4
2–4	47	17	36.2	89.4±4.5	59.8±7.8
4+	58	20	34.5	91.4±3.7	62.9±7.1
CD4 count z score within 6 mo of first echocardiogram
≤−2	87	50	57.5	80.5±4.3	36.8±6.0	<0.001
>−2	97	11	11.3	97.9±1.5	87.7±3.7
CDC symptom status before or at time of echocardiogram
Asymptomatic	22	5	22.7	100.0	77.3±10.2	<0.001
A	32	3	9.4	100.0	89.2±6.0
B	36	8	22.2	94.4±3.9	75.2±7.7
C	103	48	46.6	82.4±3.8	49.3±5.4
Baseline encephalopathy before or at time of echocardiogram
No	149	41	27.5	94.6±1.9	69.7±4.1	<0.001
Yes	44	23	52.3	72.1±6.8	44.2±8.1
Height z score
≤−2	53	26	49.1	75.0±6.0	46.9±7.5	<0.001
>−2	140	38	27.1	95.0±1.9	70.3±4.2
Weight z score
≤−2	39	25	64.1	74.4±7.0	33.2±8.1	<0.001
>−2	154	39	25.3	93.4±2.0	72.1±4.0
Zidovudine before echocardiogram
No	57	13	22.8	100.0	74.5±6.2	0.03
Yes	136	51	37.5	85.1±3.1	59.5±4.6
P carinii pneumonia before echocardiogram
No	164	49	29.9	90.2±2.3	67.7±3.9	0.02
Yes	29	15	51.7	86.2±6.4	42.6±10.6
Chest radiograph findings within 6 mo of first echocardiogram
Abnormal	61	24	39.3	93.3±3.2	53.6±7.4	0.28
Normal	121	36	29.8	88.4±2.9	70.0±4.4

Univariable Predictors of Mortality

Survival was not affected by race or ethnicity, sex, age category (0 to 1, 1 to 2, 2 to 4, >4 years), or chest radiograph findings (Table 1). However, when a Cox model was used with the actual noncategorized age, the association between age and mortality was significant (P=0.02). Survival was lower for children who were short for age (P<0.001) or underweight for age (P<0.001) at baseline. The presence of encephalopathy and increasing severity of CDC symptoms at baseline were both associated with higher mortality (P<0.001). Survival also differed among clinical centers (P=0.02). Other factors associated with lower cumulative survival included suppressed CD4 cell counts (P<0.001), a history of zidovudine therapy (P=0.03), and Pneumocystis carinii pneumonia diagnosed before the initial echocardiogram (P=0.02).

The relationship between mortality and baseline impaired cardiac function is shown in Table 2 and Figure 1. Mortality was higher in children with a depressed baseline FS (P<0.001). A similar pattern was noted for contractility (P=0.01). Mortality was also significantly higher for children with increased end-diastolic dimension (P<0.001), end-systolic dimension (P<0.001), wall stress (P=0.002), heart rate (P=0.006), LV mass (P<0.001), end-diastolic posterior wall thickness (P=0.02 for ≥1 SD), or diastolic blood pressure (P=0.003).

TABLE 2.

Cumulative Survival Among 193 HIV-Infected Children According to Baseline Echocardiographic Measurements

			Deaths		Cumulative Survival (%)±SEM
Measurement	z score	n	n	%	1y	5y	P
LV mass	<2	165	43	26.1	92.7±2.0	70.6±3.9	<0.001
	>2	24	20	83.3	66.7±9.6	20.0±8.4
End-diastolic dimension	<2	176	50	28.4	92.0±2.1	69.0±3.8	<0.001
	>2	17	14	82.4	64.7±11.6	14.1±9.0
End-systolic dimension	<2	159	42	26.4	93.0±2.0	71.0±3.9	<0.001
	>2	34	22	64.7	73.5±7.6	31.6±8.8
FS	<−2	56	32	57.1	73.0±6.0	38.1±7.3	<0.001
	>−2	137	32	23.4	96.3±1.6	74.4±4.1
Heart rate	<2	152	44	29.0	92.7±2.1	68.3±4.1	0.006
	>2	41	20	48.8	78.0±6.5	47.9±8.5
Contractility	<−2	33	16	48.5	75.8±7.5	50.0±9.1	0.01
	>−2	144	45	31.3	93.0±2.1	65.5±4.4
End-diastolic posterior wall thickness	<1	159	47	29.6	91.1±2.3	68.4±3.9	0.02
	>1	34	17	50.0	82.4±6.5	44.5±9.5
Afterload	<2	141	42	29.8	92.9±2.2	67.4±4.3	0.002
	>2	36	19	52.8	77.8±6.9	42.9±9.3
Diastolic blood pressure	<2	173	53	30.6	91.3±2.2	67.3±3.8	0.003
	>2	17	11	64.7	70.6±11.1	33.1±11.8

Figure 1.

Cumulative survival for 193 HIV-infected children according to baseline clinical characteristics and baseline echocardiographic measurements. A, CD4 cell count z score; B, encephalopathy; C, FS z score; D, LV mass z score; E, wall thickness z score; F, enddiastolic dimension z score; G, end-systolic dimension z score; and H, contractility (stress-velocity index) z score.

Figure 1 illustrates the impact on cumulative survival of each of the clinical and echocardiographic measures that are also significant in the multivariable models described below. Alternative analyses using Cox proportional-hazards models separately for each echocardiographic measure as a continuous z score produced similar results. An increased risk of death per 1 SD change in z score occurred with increased LV mass (relative risk, 1.84; P<0.001), increased end-diastolic dimension (relative risk, 1.69; P<0.001), increased end-systolic dimension (relative risk, 1.68; P<0.001), decreased FS (relative risk, 1.39; P<0.001), increased heart rate (relative risk, 1.34; P=0.001), increased wall thickness (relative risk, 1.29; P=0.009), depressed contractility (relative risk, 1.30; P<0.001), and increased afterload (relative risk, 1.25; P<0.001). Survival was not affected by baseline diastolic blood-pressure z scores when analyzed as a continuous variable (P=0.31).

Statistically significant associations were found between FS and contractility (ρ=0.56), end-diastolic dimension (ρ=−0.31), end-systolic dimension (ρ=−0.61), afterload (ρ=−0.63), and LV mass (ρ=−0.31) at baseline. Wall thickness and LV mass were also closely correlated at baseline (ρ=0.45).

Multivariable Predictors of Mortality

Data were available for all covariates in 184 children (9 children who did not have a CD4 cell count within 6 months of the baseline echocardiogram could not be included in the analyses). A multivariable Cox model using stepwise selection was used to identify a subset of covariates as independent risk factors for survival. CD4 count z score, clinical center, encephalopathy, and age at initial echocardiography remained significantly associated with survival. Factors that did not remain significant included zidovudine exposure, P carinii pneumonia, and continuous-weight z score.

After these significant nonechocardiographic covariates had been included, the Cox model was refitted separately for each echocardiographic z score. FS (P<0.001), contractility (P<0.001), end-systolic dimension (P<0.001), LV wall thickness (P=0.008), and LV mass (P=0.007) remained significant. Heart rate, afterload, and end-diastolic dimension lost significance after adjustment for nonechocardiographic covariates.

In the final model in Table 3, both decreased FS z score (P<0.001) and increased wall thickness z score (P=0.004) were independent prognostic risk factors of mortality after adjustment for CD4 count z score (P<0.001), encephalopathy (P<0.001), and clinical center (center 1 versus center 5, P<0.001; center 3 versus center 5, P=0.03). The adjusted relative risk for FS was 1.31 per 1 SD drop, and for wall thickness was 1.35 per 1 SD increase. Age, LV mass, contractility, and dimension lost significance and were not retained in the model. Because LV mass and wall thickness were highly correlated, the model was also fitted with LV mass instead of wall thickness; both FS (P<0.001) and LV mass (P=0.04) were independently associated with survival after adjustment for CD4 count, encephalopathy, and clinical center. Likewise, the model was also fitted with LV contractility instead of FS; both wall thickness (P=0.02) and contractility (P=0.006) were independently associated with survival after adjustment for CD4 count, encephalopathy, and clinical center.

TABLE 3.

Multivariable Analysis of Factors Associated With Survival for Children Infected With HIV

Effect	Estimated β	Standard Error	Relative Risk=exp (β)	95% CI	P
FS z score (per 1 SD decrease)	−0.2704	0.0639	1.31	1.16–1.49	<0.001
Wall thickness z score (per 1 SD increase)	0.2987	0.1029	1.35	1.10–1.65	0.004
CD4 count z score (per 1 SD decrease)	−1.5029	0.2405	4.49	2.81–7.20	<0.001
Encephalopathy (yes/no)	1.2312	0.3059	3.43	1.88–6.24	<0.001
Clinical center 1 (1/5)	1.7995	0.3870	6.05	2.83–12.91	<0.001
Clinical center 3 (3/5)	0.7126	0.3354	2.04	1.06–3.94	0.03

High FS or contractility and low LV mass were found not to be risk factors for death. However, we did find an additive relationship between LV wall thickness and dimension. Cumulative 5-year survival for the 18 children with elevated (>1 SD) ventricular dimension and elevated (>0.5 SD) wall thickness was 22.5%. In contrast, survival among the 100 children with normal dimension and thickness was 76.2% (P<0.001). Five-year survival was 64.2% among 40 children with only increased thickness and 51.7% among 35 children with only increased dimension.

In a subset of 157 patients for whom HIV RNA copy number was available, decreased FS and increased wall thickness z scores still remained significant predictors of mortality after adjustment for HIV RNA copy number (analyzed on a logarithmic base 10 scale) (P=0.04), CD4 cell count z score (P<0.001), encephalopathy (P<0.001), and clinical center (center 1 versus center 5, P<0.001; center 3 versus center 5, P=0.02). The relative risk was 1.39 per 1 SD increase (P=0.004) for wall thickness and 1.41 per 1 SD decrease (P<0.001) for FS. Replacing wall thickness z score with LV mass z score provided similar findings.

Timing of Mortality

Figure 2 shows the model-based means and 95% CIs for measurements and z scores according to time before the last echocardiogram or death for the 64 children who died and the 129 children who remained alive at last contact. Figure 2, A and E, indicates that within 18 months of death, the mean FS z score was <−2.0, with a mean FS <31%. On the basis of the separation in curves that starts at 36 months, depressed FS may be a useful marker for increased mortality for up to 3 years before death. In contrast, LV contractility (Figure 2D) did not differ between survivors and nonsurvivors until 2 years before death. LV mass (Figure 2, C and G) and end-systolic dimension (Figure 2H) show a difference between survivors and nonsurvivors >2 years before death, suggesting that these may be long-term prognostic indicators. LV wall thickness (Figure 2, B and F) shows a difference only 18 to 24 months before the final echocardiogram. At the time of death, echocardiographic measurements differed between survivors and nonsurvivors for FS (mean z score, −1.11 and −2.32; mean FS, 32.3% and 29.8%), LV mass (mean z score, 0.29 and 1.75; mean LV mass, 60.7 and 73.8 g), wall thickness (mean z score, −0.14 and 0.46; mean wall thickness, 0.63 and 0.67 cm), end-diastolic dimension (mean z score, 0.20 and 1.31; mean end-diastolic dimension, 3.62 and 3.81 cm), end-systolic dimension (mean z score, 0.70 and 2.25; mean end systolic dimension, 2.44 and 2.74 cm), and contractility (mean z score, −0.94 and −1.88).

Figure 2.

Longitudinal change in echocardiographic measurements for survivors and nonsurvivors (64 dead and 129 alive). Time trend lines represent mean and 95% CIs according to time before last echocardiogram or death (months). A, FS z score; B, wall thickness z score; C, LV mass z score; D, contractility (stress-velocity index) z score; E, FS (%); F, wall thickness (cm, adjusted for body surface area); G, LV mass (g, adjusted for body surface area); and H, end-systolic dimension z score.

Discussion

Baseline echocardiographic abnormalities in HIV-infected children were associated with cumulative all-cause mortality. In multivariable analyses, baseline depressed LV systolic performance and increased LV wall thickness made statistically significant contributions to the prediction of mortality after adjustment for immunodeficiency, HIV viral load, encephalopathy, and clinical center. We show that echocardiographic measurements (z scores) of LV structure and performance provide noninvasive, independent markers of disease and death in HIV-infected children that may be clinically useful.

Both contractility and FS were predictive of survival in univariable analyses. In multivariable analyses, however, FS was a more important predictor. This is not surprising, considering that FS represents the end products of multiple processes, including preload, afterload, heart rate, and contractility, all of which may be disturbed in these patients.

Differences in survival between clinical centers did not appear to be attributable to disease differences at baseline, because the center differences remained significant in multivariable models after disease variables were included. Therapeutic differences during follow-up or unidentified patient features we have not accounted for may be responsible. However, these survival differences among clinical centers may suggest differences in patient populations and not differences in the use of antiretroviral therapies, because >90% of the cohort received antiretroviral medications, and only a small subset took protease inhibitors.⁹

The study has some limitations. All-cause mortality was analyzed instead of cardiac death because of the low autopsy rate (19 of the 71 children). However, a report by the P²C² HIV multidisciplinary mortality review committee found that among 93 group 1 and group 2 children who had an HIV-related death, 11 (11.8%) had chronic cardiac disease as the underlying cause of death, and 48 (51.6%) had evidence of chronic cardiac disease.¹¹ The children in the sample were those who came in for clinical visits and whose parents or guardians consented to cardiopulmonary function testing, and so the children may not be representative of the more general population of HIV-infected children (eg, increased disease severity). Echocardiographic measurements from children with impaired growth rates similar to those of the HIV-infected children would have provided more appropriate control data than the healthy control children used in this study.

In summary, echocardiographic measures of LV structure and function are independent and potentially useful long-term and short-term predictors of overall mortality in HIV-infected children. The regular use of serial echocardiograms in this population may identify children at risk who may benefit from more careful examination and potentially effective interventions¹² to alter the course of the disease. Future studies may determine whether treatment of baseline echocardiographic abnormalities associated with increased mortality is beneficial.