Pericardial Effusion in AIDS
Incidence and Survival
Background Although pericardial effusion is known to be common among patients infected with HIV, the incidence of pericardial effusion and its relation to survival have never been described.
Methods and Results To evaluate the incidence of pericardial effusion and its relation to mortality in HIV-positive subjects, 601 echocardiograms were performed on 231 subjects recruited over a 5-year period (inception cohort: 59 subjects with asymptomatic HIV, 62 subjects with AIDS-related complex, and 74 subjects with AIDS; 21 HIV-negative healthy gay men; and 15 subjects with non-HIV end-stage medical illness). Echocardiograms were performed every 3 to 6 months (82% had follow-up studies). Sixteen subjects were diagnosed with effusions (prevalence of effusion for AIDS subjects entering the study was 5%). Thirteen subjects developed effusions during follow-up; 12 of these were subjects with AIDS (incidence, 11%/y). The majority of effusions (80%) were small and asymptomatic. The survival of AIDS subjects with effusions was significantly shorter (36% at 6 months) than survival for AIDS subjects without effusions (93% at 6 months). This shortened survival remained significant (relative risk, 2.2, P=.01) after adjustment for lead time bias and was independent of CD4 count and albumin level.
Conclusions There is a high incidence of pericardial effusion in patients with AIDS, and the presence of an effusion is associated with shortened survival. The development of an effusion in the setting of HIV infection suggests end-stage HIV disease (AIDS).
A variety of cardiac abnormalities have been observed in patients infected with HIV, including myocardial, valvular, and pericardial disease.1 2 3 4 5 6 Several studies describe an increased prevalence of pericardial effusion in patients with HIV infection, particularly in those with end-stage disease.3 7 Conversely, reports from Africa have noted a high rate of HIV infection in patients with pericardial effusions.8 Several recent prospective studies in HIV-infected patients9 10 11 12 13 have confirmed the increased prevalence of effusion, but the limited follow-up of these studies has not allowed determination of the incidence and implications of pericardial effusion in HIV infection.
The pathogenesis of pericardial effusion in HIV infection is also unclear. Numerous case reports have described Kaposi’s sarcoma,14 mycobacteria,15 16 cytomegalovirus,17 prosthetic valve endocarditis,18 bacterial pericarditis,19 and lymphoma20 as causes of HIV-related pericardial effusion. However, the acutely ill inpatient population described in these studies may not represent the majority of HIV-infected individuals with pericardial effusions, many of whom are asymptomatic.21
AIDS patients with pericardial effusions have demonstrated a lower CD4 count than AIDS patients without effusions9 ; however, the relation of pericardial effusion to mortality has not been described.
To determine the prevalence, incidence, and relation to mortality of pericardial effusion in HIV-infected subjects, we analyzed data from an ongoing long-term prospective study of cardiac disease in HIV-infected patients, the Prospective Evaluation of Cardiac Involvement in AIDS (PRECIA).
This investigation evaluated subjects enrolled in PRECIA. The study is a 5-year prospective echocardiographic evaluation of homosexual men at various stages of HIV infection.
Five groups of subjects were recruited. Group 1 was made up of asymptomatic HIV-positive homosexual men with no illness (Centers for Disease Control and Prevention [CDC] group II). HIV-positivity was defined as a positive ELISA with Western blot confirmation. Group 2 included HIV-positive homosexual men with persistent generalized lymphadenopathy (formerly AIDS-related complex [ARC]) defined by CDC group III. Group 3 comprised HIV-positive homosexual men diagnosed with AIDS within the previous 12 months. AIDS was defined by CDC group IV. Group 4 included HIV-negative homosexual men, and group 5 was made up of HIV-negative men with chronic and/or terminal illness with the degree of inanition seen in AIDS subjects.
Subjects were excluded if they were younger than 18 years of age, showed hemodynamic instability (systolic pressure <90 mm Hg), or were in the terminal stage of disease (estimated survival <4 months). Subjects in groups 1 through 3 were recruited only in the outpatient setting, and an attempt was made to enroll consecutive subjects. The period of recruitment was 4.5 years, beginning in 1988. All patients of any of the three AIDS primary care physicians at the University of California at San Francisco and San Francisco General Hospital were eligible. Other groups were identified by advertisement in local bulletins. To reduce selection bias, descriptions of the study did not mention the heart as the organ of interest to investigators. Both inpatients and outpatients were recruited for group 5. All subjects provided informed consent in a procedure approved by the Committee on Human Research at the University of California at San Francisco.
A total of 248 subjects volunteered for the study. Of these, 231 had at least one echocardiogram and were included in the investigation.
After informed consent was given, a history was obtained and a physical examination was performed. An ECG, chest roentgenography, and laboratory data, including routine chemistry, hematology, and T4 cell counts, were obtained at baseline. All subjects underwent transthoracic echocardiography according to the following schedule: groups 1 through 3 and 5, at baseline and then every 4 months; group 4, at baseline and then annually. Annual follow-up included history, physical, and standard laboratory data. During the 5 years of the study, 41 of 231 subjects (18%) did not have follow-up echocardiograms. An attempt was made to contact all subjects at the conclusion of the study to determine survival. If a subject demonstrated HIV disease progression (asymptomatic HIV infection to ARC to AIDS), he was assigned to more than one group. The subject was considered to be in the initial group until disease progression was demonstrated, at which point he was added to the appropriate second group (ARC or AIDS). The study entry time for the second group was at the diagnosis of disease progression.
A diagnosis of tuberculosis was based on a positive culture. Clinical congestive cardiomyopathy was considered present if an experienced cardiologist diagnosed congestive heart failure. Pneumocystis infection was considered present if sputum smears were positive. A diagnosis of cytomegalovirus infection required histological demonstration of infection in the setting of clinical disease. Diagnoses of Kaposi’s sarcoma and lymphoma were based on histology.
Two-dimensional echocardiography was performed with commercial ultrasound scanners (Acuson and Hewlett Packard). Standard views, including the left lateral decubitus and supine positions, were obtained. An experienced echocardiographer who did not know the subjects’ diagnostic groups reviewed all echocardiograms. A pericardial effusion was diagnosed if an echo-free space between the visceral and parietal pericardium persisted throughout the cardiac cycle. The size of the pericardial effusion was defined as follows: small when the maximum pericardial space at end diastole was <10 mm, moderate when the space was ≥10 mm but <20 mm, and large when the pericardial space was ≥20 mm between the pericardial layers.
Data are expressed as mean±SD or proportions. Comparisons between proportions were evaluated by the χ2 test with the Yates correction. Survival and incidence are displayed with Kaplan-Meier and actuarial survival curves. Differences in mortality for the entire survival curve were determined with the log-rank test. The independent effects of several variables on survival were evaluated by use of stepwise logistic regression (sas, SAS Institute Inc). The 95% CIs were obtained for two-tailed tests. Differences were considered significant at P<.05.
Table 1⇓ gives a description of the 231 subjects who entered the study. A total of 601 studies were performed over 26±17 months. Follow-up studies (range, 1 to 11 studies per subject) were obtained in 190 subjects (82%). The analysis included all 231 subjects with at least one echocardiogram.
Fifty-nine subjects entered the study as asymptomatic HIV (group 1). Five of these subjects progressed to ARC and 13 progressed to AIDS during the study. Sixty-two subjects entered the study as ARC (group 2), and 5 asymptomatic HIV subjects developed ARC, for a total of 67. Progression to AIDS occurred in 12 ARC subjects. Seventy-four subjects entered the study with a diagnosis of AIDS (group 3), and 25 asymptomatic HIV and ARC subjects developed AIDS, for a total of 99. Twenty-one uninfected homosexual men entered the study (group 4), none of whom developed HIV infection. Fifteen subjects were enrolled with chronic debilitating illnesses (group 5).
Prevalence and Incidence
A pericardial effusion was diagnosed in 16 subjects during the study. Three subjects had an effusion on study entry (2 subjects with AIDS and 1 subject with end-stage medical illness), and 3 more subjects developed an effusion at the time of progression to AIDS (group 3). Therefore, 5 effusions were present on the subjects’ first echocardiograms as AIDS subjects (prevalence, 5% [5 of 99]), and 1 was present as a group 5 subject (end-stage medically ill; prevalence, 7% [1 of 17]; P=NS, group 3 versus group 5).
Eight subjects developed an effusion after an initial negative echocardiogram as AIDS subjects, 1 as an asymptomatic HIV subject, and 1 as an ARC subject. The asymptomatic HIV subject would now be classified as AIDS by the more recent CD4 criteria (<200 cells/cm3).22 Thus, of the 16 effusions diagnosed, 15 were in HIV-infected patients (13 AIDS patients, 1 ARC patient, and 1 asymptomatic HIV patient). The latter two patients developed AIDS during the course of the investigation and were included as AIDS patients from the time of AIDS diagnosis.
The incidence of pericardial effusion in surviving AIDS subjects was 11%/y over a 2-year period (Fig 1⇓). No effusions were diagnosed in the healthy HIV-negative subjects.
Of the 15 effusions in HIV-infected patients, 12 (80%) were small, 2 (13%) were moderate, and 1 (6%) was large. All 12 small and 1 moderate effusions (87%) were asymptomatic. The 1 symptomatic moderate and 1 large effusion developed signs and symptoms of tamponade and demonstrated right ventricular collapse on echocardiography (incidence of tamponade, 9%/y for AIDS subjects with effusion and 1%/y for all AIDS subjects). Both subjects were treated with pericardiocentesis, and the subject with the large effusion underwent partial pericardectomy.
Of the 15 HIV-infected subjects with effusions, 8 died and 1 withdrew before the scheduled 4-month follow-up study. Of the remaining 6 with follow-up studies, 4 effusions were no longer present. Of these 4 subjects, 1 previously underwent partial pericardectomy, and another demonstrated effusion recurrence on a second follow-up study.
Of the 15 HIV-infected subjects with pericardial effusions, a cause was suggested in only 4 patients. Two subjects had non-Hodgkin’s lymphoma (not known to involve the heart), and 2 had depressed left ventricular function (ejection fraction <50%). Only 2 subjects underwent pericardiocentesis, and only 1 of these was cultured for bacteria and fungi (negative). No autopsy data were available. The 1 group 5 subject (HIV negative, ill) with an effusion had hepatocellular carcinoma and severe liver disease, which probably contributed to the development of the effusion.
Only 2 of 231 study patients developed clinical pericarditis (1 asymptomatic HIV patient with no effusion and 1 AIDS patient with effusion). When AIDS subjects with effusions were compared with AIDS subjects without effusions, there were no significant differences in intravenous drug use, age, weight, height, race, history of Kaposi’s sarcoma, tuberculosis, or use of AZT (Table 2⇓). Symptomatic left ventricular systolic dysfunction was diagnosed in 1 patient with an effusion and in 7 AIDS patients without effusions. Cardiac disease was suggested by physical examination in 3 other subjects with effusions (S4, n=2; aortic insufficiency murmur, n=1). An additional patient with an effusion developed atrial flutter.
The causes of death for 6 subjects with effusions were known: dilated cardiomyopathy (n=1), non-Hodgkin’s lymphoma (n=2), Pneumocystis carinii pneumonia (n=1), other pneumonia (n=1), and wasting syndrome (n=1). In only 1 subject (large effusion, dilated cardiomyopathy) did the pericardial effusion probably contribute to death.
Effusion and CD4 and Albumin Levels
The average CD4 count for AIDS subjects at the diagnosis of pericardial effusion was 59±41 cells/cm3; range, 4 to 180 cells/cm3 (Table 2⇑). This was lower than the CD4 count for noneffusion AIDS subjects (146±20 cells/cm3, P=.06) on study entry. A weaker trend was observed for the CD4 percent of total T cells (10±3% for patients with effusions versus 15±1% for patients without effusions, P=.12). There was no obvious difference in total T-cell count (868±214 cells/cm3 for patients with effusions versus 956±87 cells/cm3 for patients without effusions, P=.7).
Serum albumin levels for AIDS subjects were available at entry into the study for all subjects and within 1 month (n=8) and 4 months (n=10) of diagnosis of effusion. Within 4 months of effusion diagnosis, AIDS subjects had a significantly lower albumin level (4.00±0.1 g/dL) than AIDS subjects without an effusion on study entry (4.25±0.06 g/dL, P=.02).
Survival and Pericardial Effusion
Survival for all groups was related to the stage of HIV infection (2- year survival, 33±5.4% for patients with AIDS, 77±6.2% for patients with ARC, 88±4.6% for asymptomatic HIV-positive patients, and 95±4.9% for HIV-negative subjects). The survival of group 5 subjects (severe non-HIV medical illness) was less than that for AIDS subjects (50% versus 65% at 2 years); however, this did not reach statistical significance.
For the 13 AIDS subjects who developed effusions, the survival from onset of the effusion was markedly decreased (36±11% at 6 months; Fig. 2⇓). This was significantly shorter than survival in AIDS subjects without effusions from study entry (93±3% at 6 months, P<.01). This comparison is subject to lead time bias. To control for this bias, each AIDS patient with an effusion was matched by time in study to 6 AIDS patients without effusions. The time from study entry to effusion development for each subject with an effusion was subtracted from the time in the study for the matched patients. With this analysis, the difference in survival between AIDS patients with and without effusions remained significant for the entire study period (relative risk, 2.2; 95% CI, 1.2 to 4.0; P=.01).
Univariate predictors of survival (P≤.2) included T4 count, albumin level, history of tuberculosis, and effusion development. Other variables tested that were not associated with survival included age, history of tuberculosis, cytomegalovirus, pneumocystis or other pneumonia, and Kaposi’s sarcoma. The four variables associated with survival (effusion development, T4 count, albumin level, and tuberculosis) were entered stepwise in multivariable logistic regression models of 12- and 18-month survival for AIDS subjects. The final model included three variables: T4, albumin level, and effusion development (Table 3⇓). The addition of tuberculosis (P>.9 for adjusted association with survival) did not improve the model and increased the lack of fit; therefore, it was excluded. To control for lead time bias, survival was measured from study entry for all subjects. Only the effusion development was associated with decreased 12-month survival. When survival was measured at 18 months, both effusion development and CD4 cell count were independently associated with decreased survival.
This study demonstrated a high prevalence and incidence of pericardial effusion in AIDS subjects. The prevalence of pericardial effusion was clearly related to the stage of HIV infection because only 1 non-AIDS HIV-infected subject developed an effusion, and this subject would now be classified as an AIDS subject if more recent CD4 criteria were used.22 This study is in agreement with previous echocardiographic1 7 9 11 and autopsy2 4 investigations that have noted an increased prevalence of pericardial effusion in patients with AIDS.
Several previous prospective studies9 10 11 12 13 have followed AIDS patients with serial echocardiograms; however, short follow-up time limited the determination of incidence. In this study, the incidence of pericardial effusion increased as HIV infection progressed (0%/y in asymptomatic HIV-infected subjects versus 11%/y for subjects with AIDS).
The large majority of effusions were small (80%) and asymptomatic (87%). Only 1 patient developed tamponade, and only 2 patients were treated with pericardial drainage. Recent prospective investigations9 11 have also found large, hemodynamically significant effusions to be less common than small effusions. These findings are in contrast to previous retrospective reports of hospitalized AIDS patients23 that suggested a high incidence of large effusions requiring pericardiocentesis. One explanation for this discrepancy is the patient populations evaluated. In our study, only outpatients were recruited, and preterminal echocardiograms frequently were not performed. If echocardiography had been performed during the terminal period of HIV disease, a greater number of large effusions may have been identified.
A previous investigation of AIDS-related cardiac abnormalities has suggested that effusions frequently resolve over time.9 Only 2 of 15 subjects (13%) with intact pericardium demonstrated persistent resolution. This rate of resolution is less than that in the study by Blanchard et al,9 which noted resolution in 42% of subjects (5 of 12) with pericardial effusion.
No definitive cause was determined for any pericardial effusion in this study, but causal agents were suggested by associated disease in 4 subjects (lymphoma in 2 and cardiomyopathy in 2). No subject was diagnosed with an infectious cause. However, because pericardiocentesis was performed in only 1 subject, infectious or malignant causes cannot definitively be ruled out.
This study demonstrated a markedly shortened survival once a pericardial effusion was diagnosed. In AIDS subjects with effusions, the mortality at 6 months from onset of the effusion (62%) was almost ninefold greater than the 6-month mortality for subjects without effusions (7%). The difference in survival remained significant after adjustment for lead time bias. A multivariable analysis, including CD4 count and serum albumin as determinants of survival, also demonstrated that the development of a pericardial effusion was independently associated with increased mortality. These findings indicate that a pericardial effusion in the setting of HIV infection is a marker of shortened survival.
Several potential explanations exist for this shortened survival after the development of a pericardial effusion. First, an effusion may be a marker for undiagnosed opportunistic infections. The development of an effusion may also be a function of the immune status, as reflected in the CD4 count. Indeed, several reports6 9 have described a significantly reduced CD4 count in patients with effusions compared with patients without effusions. In the present study, subjects with effusions had lower CD4 counts compared with AIDS subjects without effusions. Hypoalbuminemia is also a marker of end-stage HIV infection, and predictive of shortened survival independent of CD4 count.24 However, when survival was controlled for CD4 count and albumin level, the development of a pericardial effusion remained a strong predictor of 12- and 18-month survival.
Several hypotheses for the increased incidence of pericardial effusion in HIV infection have been advanced. End-stage HIV infection is associated with an increased risk of malignancies and infections known to involve the pericardium. Numerous case reports have described HIV-related pericardial effusion in the setting of lymphoma,25 Kaposi’s sarcoma,14 and infection with bacteria,19 tuberculosis,8 15 other mycobacterium,16 18 cytomegalovirus,17 nocardia,26 and cryptococcus.27
Our study suggests that these are uncommon causes of effusions in the majority of AIDS subjects. We found little relation between Kaposi’s sarcoma, tuberculosis, and cytomegalovirus infection in AIDS subjects and the development of a pericardial effusion although the small number of effusions in this investigation limits the power of this observation. Dilated cardiomyopathy was infrequently identified in HIV-infected patients and was not associated with pericardial effusion. This finding is in accordance with an autopsy study2 that found cardiomyopathy in only 2 of 38 AIDS patients with pericardial effusion. Hypoalbuminemia, which is associated with ascites and pleural and pericardial effusions, is a potential cause of pericardial effusion in end-stage HIV infection. In this investigation, subjects with effusions had lower albumin levels at the time of diagnosis than subjects without effusions on study entry. This difference, though statistically significant, was small (4.00 versus 4.25 g/dL) and is unlikely to be the explanation for the high prevalence of effusion in AIDS.
Other studies have described a similar lack of diagnosis for pericardial effusion in the setting of HIV infection.2 11 18 28 An autopsy review of 115 patients with AIDS was unable to determine an obvious diagnosis in all 35 patients with pericardial effusions.2 In the only cytological study to date, Zakowski and Ianuale28 examined effusions in 14 AIDS patients and biopsies in 10 of these 14 patients. No infectious causes were found, and only 2 patients (13%) demonstrated malignancy (lymphoma in patients with a previous lymphoma diagnosis). The remainder (87%) were idiopathic. In a previous study,29 10 of 25 consecutive patients with HIV infection and pericardial effusion underwent pericardiocentesis or surgical pericardiotomy. Fluid (all exudative) and tissue samples did not provide a diagnosis in any patient.
Another explanation for the increased incidence of pericardial effusion is an end-stage HIV capillary leak syndrome. Evidence for this theory comes from autopsy studies demonstrating frequent coexistence of pericardial effusions, serous pleural effusions, and ascites.2 30 Cytokines such as interleukin-2 and tumor necrosis factor are known to be elevated in end-stage HIV infection31 and are associated with capillary leak syndromes.32 Unexplained pericardial effusion in AIDS patients may therefore be secondary to and a marker of cytokine activation in end-stage HIV infection.
This study has several potential limitations, including the lack of pathological and autopsy diagnoses. Because the majority of pericardial effusions were small and asymptomatic, pericardiocentesis was rarely performed. In addition, very few subjects underwent autopsy. Therefore, the cause of pericardial effusion and the severity of pericardial disease at the time of death are unknown. As noted above, this may account for the limited number of large pericardial effusions identified.
The small total number of effusions identified limited the determination of risk factors for effusion development. Given the rate of effusion development observed, a study three times the size of this investigation would be required to detect a 20% difference in patient characteristics. Only men were included in this investigation; therefore, the association between pericardial effusion and HIV infection in women is unclear.
The diagnosis of an asymptomatic pericardial effusion may signal end-stage HIV disease despite a relatively preserved CD4 count. Because these effusions are frequently small and rarely progressive, an exhaustive search for a pericardial diagnosis is usually not indicated. Large symptomatic pericardial effusions do occur, however, and may need aggressive evaluation and therapy. If future studies confirm these findings, a pericardial effusion in the setting of HIV infection may be considered a diagnostic criterion for end-stage HIV infection (AIDS).
In conclusion, we reported the results of a 5-year prospective evaluation of pericardial effusion in HIV-infected men. A high prevalence and incidence of pericardial effusion is noted and is related to the stage of HIV infection. Most effusions in this cohort of AIDS outpatients were small and asymptomatic but were associated with a markedly shortened survival independent of albumin level and CD4 count.
This investigation was supported in part by NIH grant HL-41495. We are grateful to Gunnard Modin for his assistance in statistical matters.
- Received January 18, 1995.
- Revision received June 5, 1995.
- Accepted July 17, 1995.
- Copyright © 1995 by American Heart Association
Corallo S, Mutinelli MR, Moroni M, Lazzarin A, Celano V, Repossini A, Baroldi G. Echocardiography detects myocardial damage in AIDS: prospective study in 102 patients. Eur Heart J. 1988;9:887-892.
DeCastro S, Migliau G, Silvestri A, D’Amati G, Giannantoni P, Cartoni D, Kol A, Vullo V, Cirelli A. Heart involvement in AIDS: a prospective study during various stages of the disease. Eur Heart J. 1992;13:1452-1459.
Karve MM, Murali MR, Shah HM, Phelps KR. Rapid evolution of cardiac tamponade due to bacterial pericarditis in two patients with HIV-1 infection. Chest. 1992;101:1461-1463.
Kelsey RC, Saker A, Morgan M. Cardiac lymphoma in a patient with AIDS. Ann Intern Med. 1991;115:370-371.
Guenter P, Muurahainen N, Simons G, Kosok A, Cohan GR, Rudenstein R, Turner JL. Relationships among nutritional status, disease progression, and survival in HIV infection. J Acquir Immune Defic Syndr. 1993;6:1130-1138.
Holtz H, Lavery D, Kapila R. Actinomycetales infection in AIDS. Ann Intern Med. 1985;102:203-205.
Schuster M, Valentine F, Holzman R. Cryptococcal pericarditis in an intravenous drug abuser. J Infect Dis. 1985;152:842. Letter.
Galli F, Cheitlin M. Pericardial disease in AIDS: frequency of tamponade and therapeutic and diagnostic use of pericardiocentesis. J Am Coll Cardiol. 1992;19:266A. Abstract.
Shepard C, Alexander S, Burstein S. Effusions of body cavity fluids in patients with the acquired immunodeficiency syndrome or human immunodeficiency virus infection. Acta Cytol. 1991;35:632-633.
Hack CE, Ogilvie AC, Eisele B, Eerenberg AJ, Wagstaff J, Thijs LG. C1-inhibitor substitution therapy in septic shock and in the vascular leak syndrome induced by high doses of interleukin-2. Intensive Care Med. 1993;19(suppl 1):S19-S28.
Calabrese LH, Proffitt MR, Yen LB, Hobbs RE, Ratliff NB. Congestive cardiomyopathy and illness related to the acquired immunodeficiency syndrome (AIDS) associated with isolation of retrovirus from myocardium. Ann Intern Med. 1987;107:691-692.