Author + information
- Received December 26, 2002
- Accepted January 16, 2003
- Published online May 7, 2003.
- Inas Al-Attar, MD*,†,‡,
- E.John Orav, PhD∥,
- Vernat Exil, MD§,
- Sarah A Vlach, MD*,¶ and
- Steven E Lipshultz, MD*,‡,¶,#,* ()
- ↵*Reprint requests and correspondence:
Dr. Steven E. Lipshultz, Division of Pediatric Cardiology, University of Rochester Medical Center, 601 Elmwood Avenue, Box 631, Rochester, New York, USA 14642.
Objectives The aim of this study was to determine the prevalence of cardiovascular dysfunction and its predictors in children with acquired immunodeficiency syndrome (AIDS).
Background Cardiovascular manifestations are common among children with AIDS but may be clinically occult.
Methods We reviewed the medical records, echocardiograms, electrocardiograms, and Holter monitor studies of 68 children with AIDS. We tested clinical and demographic characteristics at the time of AIDS diagnosis for their ability to predict serious cardiac events, death, and cardiac death.
Results The median time from AIDS diagnosis to death or end of follow-up was 1.0 year (range, 1 week to 7.9 years). Nineteen patients (28%) experienced serious cardiac events after AIDS diagnosis. Of 43 patients who died, 15 (35%) had cardiac dysfunction. Multivariable analyses revealed that recurrent bacterial infections, wasting, encephalopathy, male gender, and an earlier year of AIDS diagnosis were predictors of serious cardiac events (relative risk [RR] = 9.3, 6.9, 4.7, 4.1, and 0.76, respectively, p < 0.05). Wasting, encephalopathy, a low age-adjusted CD4 count, a low age-adjusted immunoglobulin G (IgG) level, and an earlier year of AIDS diagnosis increased the risk of all-cause mortality (RR = 8.9, 5.1, 2.7, 0.82, and 0.8, respectively, p ≤ 0.02). Male gender, a low age-adjusted CD4 count, and a low age-adjusted IgG level increased the risk for cardiac death (RR = 16.9, 4.2, and 0.68, respectively, p ≤ 0.05).
Conclusions Serious cardiac events and cardiac death are common among children with AIDS. Factors such as recurrent bacterial infections, wasting, encephalopathy, male gender, low CD4 and IgG levels, and an earlier year at AIDS diagnosis may identify high-risk patients.
Cardiac manifestations of human immunodeficiency virus (HIV) infection in children are protean and common (1,2). They include asymptomatic and symptomatic cardiac dysfunction, hemodynamic abnormalities, dysrhythmias, cardiac arrest, and cardiac death (1–11). These abnormalities occur frequently as the prevalence of HIV infection among children is high and as advances in therapy and management improve life expectancy. More than 90% of the estimated 1,700 new pediatric HIV infections that occur each day around the world occur in developing countries (12). The ability to identify patients at particularly high risk for cardiovascular abnormalities would improve care because care of HIV-infected patients relies on accurate diagnosis and prompt management of underlying illnesses, including symptomatic cardiac diseases (13).
In previous work, we focused on different stages of HIV disease (1,5,6,10). Those studies were not restricted to predictors at the time of acquired immunodeficiency syndrome (AIDS) diagnosis, as performed in this report.
In studies of HIV-infected children, adverse cardiac events were more frequent and severe in children with advanced HIV disease (1,5,8,10). The risk of most adverse cardiac outcomes was higher in children with AIDS-defining conditions (except for those with lymphoid interstitial pneumonitis [LIP]) than in HIV-infected children without AIDS (1). We therefore studied AIDS patients to identify predictors at the time of the first AIDS-defining illness that could identify patients at high risk for subsequent cardiac morbidity and mortality. We focused on predictors that are commonly recorded and involve safe, simple, and inexpensive regimens so that the study results would be relevant in all settings in the U.S. and also in countries with less developed health systems.
The eligible population consisted of 87 children infected with HIV who met the Centers for Disease Control and Prevention (CDC) criteria for AIDS (14)and who were followed at Children’s Hospital, Boston from January 1984 through June 1994. Thirteen patients were excluded because they were enrolled in a National Institutes of Health study of heart and lung complications of pediatric HIV infection that restricted the use of study data; four were excluded because they had been diagnosed with AIDS at another institution; and two were excluded because their medical records were unavailable. All of the remaining 68 patients (78% of those eligible) were included.
Echocardiograms, electrocardiogram (ECGs), or 24-h continuous ambulatory electrocardiographic monitoring studies (Holter studies) were performed.
Of 309 echocardiograms from 61 patients (median, 5 studies per patient; range 1 to 18 studies per patient), 278 included a complete assessment of left ventricular fractional shortening. The remaining 31 studies were limited to a qualitative or subjective assessment of left ventricular systolic performance. During all echocardiographic studies, a Dinamap automated vital signs monitor (Critikon Inc., Tampa, Florida) was used to monitor heart rate and blood pressure (median, 5 min) while the child was in the relaxed state or, for children under age 2, while the child was sedated with chloral hydrate. Left ventricular fractional shortening was calculated as previously reported (3,15–17). Sixty-two patients had 262 ECG studies (median, 3 studies per patient; range, 1 to 16 studies per patient). Thirty-six patients had 24-h electrocardiographic Holter monitoring (80 studies; median, 2 studies per patient; range, 1 to 10 studies per patient). The ECGs were evaluated using standard criteria (18,19).
Medical record review
Inpatient and outpatient medical records were reviewed by unmasked pediatric physicians who recorded information about the potential clinical predictors at the time of AIDS diagnosis and who looked for documented abnormalities in heart rate, blood pressure, and heart rhythm. Potential clinical predictors for which data were collected included demographics, mode of acquisition of HIV, AIDS classification, co-infection with Epstein-Barr virus (EBV) and cytomegalovirus (CMV), CD4 cell count, serum immunoglobulin levels, antiretroviral therapy, and history of cardiac abnormalities. Approval for this medical record review was obtained from the Committee on Human Investigation at Children’s Hospital, Boston.
Definition of clinical predictors
The AIDS-defining illnesses were classified according to the 1987 CDC classification system for children that was being used at the time the study children were diagnosed with AIDS (14). These illnesses were encephalopathy (P-2B), LIP (P-2C), opportunistic infections (P-2D1), severe recurrent bacterial infections (P-2D2), secondary cancer (P-2E1), and wasting.
Ebv and CMV status
Coinfection with EBV was determined by the presence of serum immunoglobulin G (IgG) antibody to EBV capsid antigen measured by immunofluorescence (20,21). Co-infection with CMV was determined by the presence of serum IgG antibody to CMV measured by enzyme-linked immunosorbent assay (22), or by culturing the virus in urine (by the spin-enhanced technique or by growth in human diploid fibroblast cells) (23). Seropositive test results for EBV and CMV were not considered evidence of infection if they were obtained in infants under 12 months old because persistent maternal antibody may have confounded a serologic diagnosis. Similarly, seropositive test results for EBV and CMV were excluded if the tests were performed within three months after intravenous immunoglobulin therapy.
T-cell lymphocyte count
CD4 and CD8 cell counts were performed within three months after AIDS diagnosis using standard flow cytometry (FACScan, Becton Dickinson, San Jose, California) after whole-blood lysis. Normal age-related changes in absolute CD4 counts were adjusted by computing a z-score for each subject using the means and standard deviations for age in normal healthy children. The standards used were those generated by Mofenson et al. (24)from CD4 lymphocyte counts reported in normal children in nine published reports. The z-score is the difference between the child’s CD4 count and the count predicted from the reference standard for age, divided by the standard deviation.
Serum levels of IgG, immunoglobulin A, and immunoglobulin M (IgM) were determined within three months after AIDS diagnosis by nephelometry and electroimmunoassay. No child in this database was receiving intravenous IgG at the time of baseline data collection. The immunoglobulin levels were standardized for age by creating z-scores from the published means and standard deviations from 201 normal children (25). We used polynomial regression against age to smooth the mean immunoglobulin levels as reported by Buckley et al. (25). Similar polynomial regression models were used to smooth the upper and lower confidence bounds (25). Z-scores for each child in our cohort were then calculated by subtracting the smoothed age-adjusted mean and dividing by the smoothed age-adjusted standard deviation.
Hemoglobin, anemia, and albumin
Anemia was defined as a hemoglobin level below 8 g/dl. Serum albumin levels were used as continuous measures.
Height and weight
Height and weight were adjusted for age and gender using the National Center for Health Statistics norms (26)and were expressed as z-scores. Wasting was defined according to the 1987 CDC classification system for children (14).
Patients were categorized as having received antiretroviral therapy if they had been on zidovudine (ZDV), dideoxyinosine (DDI), or dideoxycytosine (DDC) for at least one month before AIDS was diagnosed. Protease inhibitors were not used.
Cardiac events before aids diagnosis
Serious cardiac events and any cardiac events that occurred at least one month before AIDS diagnosis were considered as potential predictors. Serious cardiac events included transient and chronic congestive heart failure, hypotension, severe dysrhythmia, cardiac tamponade, cerebrovascular accident associated with hemodynamic instability, and cardiac arrest. The category of “any cardiac event” included any of the serious cardiac events, plus tachycardia, bradycardia, and hypertension.
Definition of clinical end points
Data on rhythm abnormalities were extracted from echocardiographic, ECG, and Holter monitor studies, as well as from medical records. For all the end points except death, the cardiac event considered was the first one occurring at least seven days after the diagnosis of AIDS. If a cardiac end point occurred within seven days after the diagnosis of AIDS, the patient’s data were removed from the analyses of cardiac end points.
Serious ECG abnormalities were defined as any of the following: supraventricular tachycardia, atrial fibrillation or flutter, third-degree atrioventricular block, torsade de pointes, ventricular tachycardia or fibrillation, myocardial infarct pattern, ST- and T-wave changes consistent with cardiac ischemia or injury, or a prolonged QTc interval, defined as 0.48 s or longer.
Transient and chronic congestive heart failure were defined by depressed left ventricular systolic function on echocardiography (fractional shortening of ≤25%), in addition to clinical signs and symptoms of heart failure necessitating anticongestive therapy for less than 30 days (transient) or more than 30 days (chronic). Signs and symptoms of heart failure in children include pulmonary or peripheral edema, tachypnea, tachycardia, poor feeding, increased liver size, and deterioration in exercise tolerance. Echocardiographic evidence of cardiac dysfunction was required for all diagnoses of congestive heart failure because other illnesses associated with HIV infection may mimic the signs and symptoms of congestive heart failure.
Cardiac tamponade was defined by echocardiographic evidence of a marked accumulation of pericardial fluid that was consistent with tamponade, that resulted in clinical symptoms, and that required pericardiocentesis.
Cerebrovascular accident secondary to hemodynamic abnormalities was defined as either a cerebral hemorrhage or a cerebral infarct associated with severe hypertension or hypotension.
Cardiorespiratory arrest was defined as the sudden onset of hemodynamic abnormalities severe enough to require chest compressions and pharmacologic intervention.
Death with cardiac dysfunction (cardiac death) was defined as either sudden death presumed to be secondary to dysrhythmia or as death associated with marked left ventricular dysfunction within the preceding six months. Marked left ventricular systolic dysfunction was defined by low contractility (a stress-velocity index >2 SD below normal) and depressed fractional shortening (≤25%) as measured by echocardiography (15–17).
The predictor analyses were performed for three clinical outcomes: serious cardiac events, all-cause mortality, and death with cardiac dysfunction. Serious cardiac events were as follows: transient or chronic congestive heart failure, marked dysrhythmia, cardiac tamponade, cerebrovascular accident associated with hemodynamic instability, or cardiac arrest. Follow-up for clinical events ended in the summer of 1994 when the principal investigator left the study institution.
Cox proportional hazards regression analysis was used to identify predictors of each of the three study outcomes. All of the previously described demographic, clinical, and immunologic characteristics, measured at the time of AIDS diagnosis, were considered in univariate analyses as potential risk factors. Risk factors with a significance level of 0.10 or less were then included in the final multivariable models. All of the final multivariable models also included CD4 counts and an indicator variable for wasting because of their confounding effects on the other predictors. A two-sided p value of 0.05 or less in the final model was considered indicative of statistical significance.
Nineteen patients with missing CD4 counts were coded with an indicator variable for missing data. In this way, all of the subjects could be used for multivariable modeling, but only those children with actual CD4 counts were used to estimate the impact of CD4 on the end points.
Because EBV and CMV status were not available for many children, separate analyses were performed by adding these predictors to the final multivariable models. These results may not be generalizable to the entire study cohort and are considered secondary findings.
Results are presented as two-sided p values, relative risks, and 95% confidence intervals. To illustrate the differences in time to cardiac death between children with and without selected identified risk factors, we used the SAS statistical program to estimate survival curves based on the Cox regression coefficients. A baseline survival curve is estimated by maximal likelihood from the entire cohort, and then modified according to the proportional hazards assumption to the degree dictated by the regression coefficients (27).
Of the 68 patients, 48 (71%) acquired HIV infection by vertical transmission, and 20 (29%) by transfusion of blood products (10 of these patients had hemophilia). All of the 19 patients excluded from the study acquired HIV by vertical transmission; the only significant difference between the vertically infected study group and the excluded children was that males were more common in the vertically infected study group (Table 1).
At the time of the first AIDS-defining condition, seven patients were anemic and 22 were on antiretroviral therapy (12 on ZDV, 7 on DDI, and 3 on DDC). Other characteristics are listed in Table 2.
Cardiac events occurred in 56 (82%) patients after the diagnosis of AIDS (Table 3). Nineteen (28%) patients experienced a serious cardiac event after AIDS diagnosis (Table 3). Among the 50 patients with data available before AIDS diagnosis, 22 (44%) had a cardiac event, including 5 (10%) who had serious cardiac events.
Forty-three (63%) of the patients died during the study period. Fifteen deaths (35%) were associated with marked cardiac dysfunction. The estimated median survival after AIDS diagnosis was 1.6 years. Median time to death associated with cardiac dysfunction was 5.4 years. Fifteen of the 43 deaths were among transfusion patients (15/20 or 75% mortality); the remaining 28 deaths were among vertically infected study patients (28/48 or 58% mortality). The 58% mortality in vertically infected study patients was not different from the 53% mortality in excluded vertically infected patients.
Clinical predictors of adverse cardiac events
Factors univariately linked to an increased relative risk of overall death were an earlier year of AIDS diagnosis, encephalopathy, and wasting (Table 4). The diagnosis of LIP and higher CD4, IgG, or IgM z-scores were predictors of longer survival. Factors linked with an increased relative risk of death with cardiac dysfunction were male gender, transfusion-acquired HIV infection, and the occurrence of a severe cardiac event before AIDS diagnosis. High IgG and IgM z-scores were protective for death from cardiac dysfunction. Recurrent severe bacterial infections (CDC class P-2D2) and the occurrence of any heart abnormality before AIDS diagnosis predicted the occurrence of serious cardiac events after AIDS diagnosis.
For death and cardiac death, both IgG and IgM were found to be significant on univariate testing, but both were highly collinear. Therefore, only IgG was included in the final models. Similarly, for predicting serious cardiac events, encephalopathy and opportunistic infections were found to be exclusive diagnoses in this cohort and only encephalopathy was entered into the model.
In multivariable analyses (Table 5), an earlier year at AIDS diagnosis, encephalopathy, a low IgG z-score, wasting, and a low age-adjusted CD4 count significantly predicted overall mortality. Male gender, a low IgG z-score, and a low age-adjusted CD4 count predicted death with cardiac dysfunction (Fig. 1). When IgM was substituted for IgG, IgM was found to be significantly related to cardiac death but not to overall mortality.
Significant multivariable predictors of serious cardiac events were an earlier year at AIDS diagnosis, male gender, encephalopathy, recurrent bacterial infections, and wasting (Table 5). Opportunistic infections were not significant when substituted for encephalopathy.
The effects of the two significant predictors, CD4 z-score and IgG z-score, on cardiac death are shown in Figure 1. For example, the median time to cardiac death after AIDS diagnosis for an average patient in this cohort (with a CD4 count approximately 2 SD below normal and an IgG level 3 SD above normal) was about 3.4 years (Fig. 1, curve B). In contrast, median time to cardiac death increases to 5.4 years in patients whose CD4 counts are only 1 SD below normal and whose IgG is 4 SD above normal (Fig. 1, curve A). Conversely, median time to cardiac death falls to approximately 2.1 years for a patient with a CD4 count 3 SD below normal and an IgG level 2 SD above normal (Fig. 1, curve C). Male gender was a very strong predictor for death with cardiac dysfunction; 14 of 15 patients who had this outcome were males.
Including CMV and EBV status in the multivariable models substantially restricted sample size, and therefore these variables were incorporated only in the final models. Positive CMV and EBV status were associated with increased risks for most of our end points. However, statistical significance was reached only for the impact of EBV positivity on overall mortality (hazard ratio = 4.6; 95% confidence interval: [1.5, 14.0]; p = 0.01, after adjusting for all of the factors in Table 5).
Correlations with year of AIDS diagnosis
In more recent years, vertically infected patients experienced their first AIDS-defining condition at an older age (r = 0.34, p = 0.02). AIDS diagnosis in more recent years also correlated with a longer follow-up before AIDS diagnosis (r = 0.31, p = 0.01) and a lower CD4 z score at the time of AIDS diagnosis (r = −0.32, p = 0.03). However, regression analysis indicated that patients diagnosed with AIDS in more recent years survived longer and were less likely to experience serious cardiac events (Table 5).
Serious cardiac events occurred frequently (28%) in pediatric patients with AIDS. Of the 43 patients who died during the study period, 15 (35%) died with cardiac dysfunction.
Several factors were able to predict mortality, death with cardiac dysfunction, and serious cardiac events. After adjustment in multivariable models, encephalopathy and wasting significantly predicted mortality and serious cardiac events, whereas opportunistic infections as the first AIDS-defining illness did not. This is unfortunate, as there have been many advances in the management and prophylaxis of opportunistic infections but not in the treatment of encephalopathy and wasting. A more immunocompromised state, marked by a lower age-adjusted CD4 count, was a significant prognostic sign of poor survival and death with cardiac dysfunction, corroborating results from other studies (1,8). Also in multivariable models, a high IgG z score predicted longer survival, and EBV co-infection predicted poorer survival. Hypergammaglobulinemia in HIV may indicate an immune system reaction that is not yet complete and is perhaps a marker of less severe immunosuppression. Serum IgG level has been identified as an immune factor that independently predicts survival in HIV-infected children (28). Gender was a strong predictor of death with cardiac dysfunction: 14 of 15 of these deaths were in males. In another study in patients without HIV infection, women with heart failure due to nonischemic causes had significantly better survival than men with or without coronary disease as their primary cause of heart failure, supporting the gender-related differences in cardiac outcomes observed in this study (29). Adverse cytokine and endotoxin cardiac effects in children with recurrent bacterial infections may relate to increased adverse cardiac events (30). Many studies in primates, rodents, and cardiac biopsy have addressed theories of AIDS cardiomyopathy (31,32). Wasting also significantly predicted serious cardiac events. The association may be the result of micronutrient deficiencies that affect the heart (33–36). Encephalopathy has been previously associated with adverse cardiac outcomes in HIV-infected children (1,5)and may relate, in part to associated autonomic dysfunction.
Clinical and laboratory characteristics of patients at the time of AIDS diagnosis at our hospital have changed since the first decade of the epidemic. In more recent years, patients have been older, have had lower CD4 counts at the time of their first AIDS-defining illness, and have survived longer. They have also been less likely to experience serious cardiac events and less likely to die with cardiac dysfunction. These changes may reflect advances in the management of HIV-infected patients, such as prophylaxis against secondary infections, antiretroviral therapy, nutritional interventions, and closer cardiac monitoring. The duration of follow-up before AIDS diagnosis has also been longer in recent years, as it has been in other studies (37). The World Health Organization recently reported that only 5% of the estimated five to six million people with advanced HIV disease in developing countries are being treated with highly active antiretrovirals, falling to 1% in sub-Saharan Africa where nearly 1 in 11 adults are infected and women, mostly of childbearing age, account for 58% of infections (38). There are currently 900,000 Americans living with HIV, and this number has increased by 50,000 since 1998, largely because advances in treatment have controlled the infection in many people, allowing some to live longer. However, about half of these infected Americans have not been given a diagnosis or treated or both. Worldwide, 42 million people are living with HIV, including 3.2 million children under age 15; more than 25 million have died of AIDS; 14,000 new cases of HIV infection occur every day; and 100 million people over the next eight years could become HIV-infected. The United Nations has estimated that AIDS will kill 68 million people over the next 20 years, more than the number claimed by AIDS in the past 20 years, and prevention can directly impact this statistic. This suggests that the results of our study are relevant to the majority of HIV-infected children globally.
The findings of this study are limited by its retrospective nature. Although all patients were examined at least every three months, the intervals between examinations were not equal, which may have resulted in underreporting of events or in spurious associations. However, although the practice of monitoring for cardiac events has increased at our institution in recent years, we found that serious cardiac events occurred more frequently in the earlier years. Using the proportional hazards model presumes that risks remain constant over time, which may not be the case in reality. However, our results represent average associations over time.
We studied many possible predictive variables, which may have led to a few false positive results, but we argue that most of our findings are biologically plausible. Other potentially predictive variables such as HIV viral load were not available during this study interval. Although most of the excluded patients from the study were patients enrolled on a protocol, their characteristics, except for gender, were similar to the vertically infected patients in the study group.
In conclusion, we found that adverse cardiac events are frequent in AIDS patients but appeared to be more frequent in earlier years when the AIDS population was monitored less frequently and treated less aggressively. The predictors identified in our study, at the very least, identify children who remain at inherent risk of death or serious cardiac events. In addition, these predictors may help in the management of AIDS patients by identifying patients at high risk. Identifying these predictors highlights areas where interventions may be most productive (13,39).
☆ Supported in part by grants HL53392, HL59837, HL48012, and HL72705 from the National Institutes of Health, Bethesda, Maryland.
- acquired immunodeficiency syndrome
- Centers for Disease Control and Prevention
- Epstein-Barr virus
- human immunodeficiency virus
- immunoglobulin G
- immunoglobulin M
- lymphoid interstitial pneumonitis
- relative risk
- Received December 26, 2002.
- Accepted January 16, 2003.
- American College of Cardiology Foundation
- Starc TJ, Lipshultz SE, Kaplan S, et al, for the P2C2HIV study group. Cardiac complications in children with human immunodeficiency virus infection. Pediatrics 1999;104:e14 (http://www.pediatrics.org/cgi/content/full/104/2/e14)
- Pediatric Pulmonary and Cardiovascular Complications of Vertically Transmitted HIV Infection Study Group,
- Lipshultz S.E.,
- Easley K.A.,
- Orav E.J.,
- et al.
- the P2C2Study Group,
- Langston C.,
- Cooper E.R.,
- Goldfarb J.,
- et al.
- Pediatric Pulmonary and Cardiovascular Complications of Vertically Transmitted HIV Infection Study Group,
- Lipshultz S.E.,
- Easley K.A.,
- Orav E.J.,
- et al.
- Fisher S.D.,
- Lipshultz S.E.
- Colan S.D.,
- Borow K.M.,
- Neumann A.
- Colan S.D.,
- Parness I.A.,
- Spevak P.J.,
- Sanders S.P.
- Garson A. Jr.
- Lennette E.T.
- Fleischer G.,
- Henle W.,
- Henle G.,
- et al.
- Gleaves C.A.,
- Smith T.F.,
- Shuster E.A.,
- Pearson G.R.
- Buckley R.H.,
- Dees S.D.,
- O’Fallon W.M.
- Dibley M.J.,
- Goldsby J.B.,
- Staehling N.W.,
- Trowbridge F.L.
- Collett D.
- Adams K.F.,
- Dunlap S.H.,
- Sueta C.A.,
- et al.
- ↵Fisher SD, Bowles NE, Towbin JA, Lipshultz SE. Cytokines and HIV-associated cardiovascular disease. AIDS. In Press
- Lipshultz S.E.
- Miller T.L.,
- Easley K.A.,
- Zhang W.,
- et al.
- Kavanaugh-McHugh A.L.,
- Ruff A.,
- Perlman E.,
- Hutton N.,
- Modlin J.,
- Rowe S.
- Miller T.L.
- Hoffman M.A.Lipshultz S.E.,
- Miller T.L.
- the New York City Pediatric Spectrum of HIV Disease Clinical Consortium,
- Obiri G.U.,
- Thomas P.A.,
- Caldwell B.
- Lipshultz S.E.,
- Orav E.J.,
- Sanders S.P.,
- Colan S.D.