Changes in Mortality Related to Human Immunodeficiency Virus Infection: Comparative Analysis of Inpatient Deaths in 1995 and in

Transcription

1 HIV/AIDS MAJOR ARTICLE Changes in Mortality Related to Human Immunodeficiency Virus Infection: Comparative Analysis of Inpatient Deaths in 1995 and in Mamta K. Jain, 1 Daniel J. Skiest, 1 Jeff W. Cloud, 2 Charu L. Jain, 2 Dennis Burns, 3 and Ruth E. Berggren 4 1 Division of Infectious Diseases, 2 Department of Internal Medicine, and 3 Department of Pathology, University of Texas Southwestern Medical Center, Dallas; and 4 Infectious Diseases Section, Tulane University Health Sciences Center, New Orleans, Louisiana We conducted a retrospective chart review of human immunodeficiency virus (HIV) infected patients who died in 1995 and in We found an increase in the proportion of patients who died from an illness that was not related to acquired immunodeficiency syndrome (AIDS). Although there was a decrease in the prevalence of AIDS-defining illnesses, 185% of patients died with CD4 counts of!200 cells/ml. The leading cause of death was Pneumocystis carinii pneumonia (PCP). Nonadherence to therapy and new diagnosis of HIV infection were the leading reasons why patients were not receiving antiretroviral therapy. The leading causes of non AIDS-related deaths in were non AIDS-defining infections and end-stage liver disease. At our hospital, PCP remains an important cause of death in the highly active antiretroviral therapy (HAART) era, possibly because 150% of HIV-infected patients who died were not receiving HAART. AIDSdefining illnesses continue to be a major cause of mortality in the HAART era in populations where access to care and adherence to HAART is limited. Since the introduction of HAART, mortality due to AIDS-defining illnesses has decreased [1 5]. The specific causes of mortality in the HIV-infected population continue to be investigated. A study from Cleveland, Ohio, that examined the cause of death in 255 HIVinfected persons showed that the percentage of patients who died of an AIDS-defining illness decreased in the HAART era, from 60% in 1995 to 30% in 1999 [6], Received 5 September 2002; accepted 19 December 2002; electronically published 2 April Presented in part: 39th Annual Meeting of Infectious Diseases Society of America, San Francisco, November Financial support: Unrestricted educational grant from Schering Plough (to M.J.) and the National Institutes of Health (grants NIMH R24 MH59656 [to D.S.] and K23 RR [to R.B.]). Reprints or correspondence: Dr. Mamta K. Jain, Div. of Infectious Diseases, University of Texas Southwestern Medical Center, 5323 Harry Hines Blvd., Dallas, TX (Mamta.Jain@UTSouthwestern.edu); or Dr. Ruth E. Berggren, Div. of Infectious Diseases, Tulane University Medical Center, 1430 Tulane Ave., SL- 71, New Orleans, LA (rberggre@tulane.edu). Clinical Infectious Diseases 2003; 36: by the Infectious Diseases Society of America. All rights reserved /2003/ $15.00 whereas a study from France suggested that 50% of deaths in the HAART era continue to be due to AIDSdefining illnesses [7]. Some studies of HIV-infected persons suggest that the proportion of non AIDS-related deaths has increased in the HAART era [6, 8 10], whereas other studies have found no such changes [11, 12]. An increasing proportion of deaths due to liver disease have been noted in some descriptive and cohort-based mortality studies but not in others [13 15]. We conducted a retrospective chart review of patients who died at Parkland Memorial Hospital in Dallas, Texas, to evaluate specific causes of death in 1995 and in , reflecting the deaths before introduction of HAART and several years after widespread use of HAART began. PATIENTS AND METHODS Clinical setting and patient selection. Parkland Memorial Hospital is a 990-bed county hospital and is the primary teaching institution for the University of Texas Southwestern Medical School. A retrospective chart re CID 2003:36 (15 April) HIV/AIDS

2 view of all HIV-infected adults who died at Parkland Memorial Hospital in 1995 and was conducted. International Classification of Diseases, Ninth Revision, codes were used to identify those with documented HIV infection. Data collection. Data from medical records were collected regarding each patient s demographic characteristics, HIV risk factors, liver disease risk factors, CD4 cell count, and HIV load, as well as antiretroviral therapy and Pneumocystis carinii pneumonia (PCP) prophylaxis received. We determined the cause of death by a chart review on the basis of clinical, laboratory, and microbiologic data. Autopsy reports, when available, were also used to determine the cause of death. An attempt was made to assign one primary cause of death to each case. For patients who had several possible causes of death, 3 investigators (M.J., D.S., and D.B.) reviewed the records, including available autopsy reports, to determine the probable primary cause of death. Other possible causes of death were considered to be contributing causes of death. Definitions. In cases in which no autopsy was performed, we used the following definitions to determine the primary cause of death. Death due to end-stage liver disease (ESLD) was defined as death of a patient with underlying liver disease and 1 of the following conditions: coagulopathy, bleeding esophageal varices, hepatic encephalopathy, hepatorenal syndrome, or spontaneous bacterial peritonitis. Death due to definite PCP was defined as death of a patient with respiratory failure and a positive result of a PCP direct fluorescence antibody test of sputum. Death due to probable PCP was defined as death of a patient with respiratory failure and 2 ofthe following conditions: hypoxia, compatible chest radiograph findings demonstrating an interstitial infiltrate, or an elevated lactate dehydrogenase level without microbiologic confirmation of another infectious process. Death due to bacterial pneumonia was defined as death of a patient with either compatible chest radiograph findings (patchy alveolar infiltrates or consolidation) and a blood culture result positive for a bacterial pathogen or compatible chest radiograph findings and 2 of the 3 following conditions: a sputum culture result revealing a pathogen known to cause bacterial pneumonia, fever (temperature of 138 C), or an elevated level of polymorphonuclear leukocytes as determined by a differential WBC count of peripheral blood. Death due to malignancy was defined as death of a patient with a biopsy-proven malignancy and without another probable cause of death. Death due to sepsis was defined as death of a patient with a blood culture positive for a bacterial pathogen but without bacterial pneumonia and/or with evident refractory hypotension that required vasopressor therapy in the absence of another cause of refractory hypotension. Death due to cytomegalovirus (CMV) infection was defined as death of a patient with biopsy-proven or culture-proven end-stage organ disease due to CMV. For patients whose death was more likely due to other causes, CMV infection was considered a contributing cause. Finally, a patient was considered to have died of Mycobacterium avium-intracellulare (MAI) infection if this infection was proven by culture and no other cause of death was likely. Other terms used in this study were defined as follows. Death due to an AIDS-defining illness was defined as death attributable to one of the Centers for Disease Control and Prevention (CDC) category C diseases [16]. Probable immunodeficiencyrelated diseases included CDC category B diseases or diseases that we considered to probably be CDC category C conditions but for which there was insufficient data for a conclusive determination. These diseases include undiagnosed CNS mass (i.e., probable toxoplasmosis or primary brain lymphoma) and bacterial pneumonia due to Streptococcus pneumoniae or Haemophilus influenzae (i.e., probable recurrent pneumonia). Death due to HIV-associated diseases was defined as death of a patient due to an AIDS-defining illness or a probable immunodeficiency-related disease. The outpatient HIV clinic database was searched to determine the number of visits made by each patient in the 12 months before death. A patient was considered to have an HIVinfection primary care provider if he or she had made 3 visits in the past year or if HIV-infection specific treatment had been initiated. Pharmacy data were reviewed to verify that the patient was receiving HAART (defined as 3 antiretroviral drugs belonging to 2 classes, the prescriptions for which were refilled in the 3 months before death) or had an undetectable HIV load (defined as!400 HIV RNA copies/ml) in the 2 months before death. Patients were considered to have been receiving PCP prophylaxis if pharmacy records showed prescription refills during the 2 months before death for trimethoprimsulfamethoxazole, pentamidine, dapsone, or atovaquone. Because pharmacy data were not available for 1995, PCP prophylaxis data were determined by chart review. Data from review of charts from misclassified 18% of patients, compared with data from pharmacy records. Statistical analysis. SAS, version 8.0 (SAS Institute), was used to analyze the data. Statistical comparisons were made by use of x 2 likelihood ratio or Fisher s exact test, where appropriate, and Student s t test for comparison of mean values. All analyses were 2-tailed, and P.05 was considered significant. RESULTS A total of 210 deaths of HIV-infected persons were identified: 119 occurred in 1995, 44 in 1999, and 47 in Charts for 112 (94%) of 119 patients identified as being HIV infected who died in 1995 (period 1) and 88 (97%) of 91 in (period 2) were available for review. The demographic characteristics, risk factors for HIV acquisition, and HIV care HIV/AIDS CID 2003:36 (15 April) 1031

3 related profile of the study population are presented in table 1. The 2 groups were similar with respect to age and sex, but HIV risk factors were different in period 1 than they were in period 2, with men who have sex with men being more common in period 1. In period 1, 54 (48%) of 112 patients were of black or Hispanic origin, compared with 53 (60%) of 88 in period 2 ( P p.09). No patients from period 1 received HAART, which was unavailable in Despite widespread availability after 1995, only 48% of patients from period 2 received HAART, and 13 patients had HIV RNA levels of!400 copies/ml. The median CD4 cell counts (measured within 8 months of the time of death) were similar in both groups. CD4 cell counts were not available for 14 (13%) of 112 patients in period 1 and 4 (5%) of 88 patients in period 2. A comparison of the leading causes of death in each group Table 1. Demographic characteristics, risk groups for HIV infection, and HIV care related profiles of patients in a comparative analysis of inpatient deaths in 1995 (period 1) and (period 2). Variable is shown in table 2. A trend toward a decrease in AIDS-defining illnesses as cause of death occurred. A significant decrease in deaths due to HIV-associated diseases occurred between period 1 and period 2. The proportion of patients who died as a result of PCP (probable or definite) did not change between the 2 periods. There were no deaths due to CMV, MAI, or toxoplasmosis in period 2. Because the proportion of deaths attributable to HIV-associated illnesses decreased from period 1 to period 2, non AIDS-related illnesses increased. The leading non AIDS-related causes of death in both time periods were infections and ESLD. Four patients in period 2 died of intracerebral hemorrhages (2 parenchymal and 2 subarachnoid). Other causes of death due to non-aids-related illnesses did not differ between the 2 time periods. Bacterial pneumonia and sepsis were the most common Period 1 (n p 112) Period 2 (n p 88) P Male sex 100 (89) 71 (81).09 Age, mean years (range) 38 (23 74) 39 (22 72).46 Race White 57 (51) 35 (40).12 Black 34 (30) 34 (39).22 Hispanic 20 (18) 19 (22).51 Other 1 (!1) a HIV infection risk group b Men who have sex with men 68 (61) 33 (38)!.01 Heterosexual 34 (30) 30 (34).57 Injection drug user 25 (22) 28 (32).13 Transfusion recipient 5 (4.5) 5 (5.7).75 a Unknown 14 (13) 16 (18).26 HIV infection care related profile History of AIDS-defining illness c 85 (76) 54 (61).03 Receipt of PCP prophylaxis 77 (69) 44 (50)!.01 Receipt of HAART d 0 42 (84)!.001 HIV care provider 77 (69) 50 (57).08 CD4 cell count of!200 cells/ml e 87 (89) 74 (88).89 CD4 cell count, median cells/ml (range) e 16 (0 912) 33 (1 743).38 HIV load, median log copies/ml (range) ND 5.12 ( ) ND HIV load of!400 copies/ml ND 13 (15) ND Autopsy performed 19 (17) 9 (10).17 NOTE. Data are no. (%) of patients, unless otherwise indicated. P values were determined by x 2 test, for proportions, or Student s t test, for continuous variables, unless otherwise indicated. ND, not done; PCP, Pneumocystis carinii pneumonia. a Determined by Fisher s exact test. b One patient may have belonged to 11 risk group for HIV infection c Defined by Centers for Disease Control and Prevention category C criteria. d Defined as 3 drugs in 2 classes. e Values were missing CID 2003:36 (15 April) HIV/AIDS

4 Table 2. Primary cause of death, by year, in HIV-positive patients in a comparative analysis of inpatient deaths in 1995 (period 1) and (period 2). No. (%) of patients Cause of death Period 1 (n p 112) Period 2 (n p 88) HIV-associated illness a 69 (62) 41 (47).03 AIDS-defining illness All 57 (51) 33 (38).058 Pneumocystis carinii pneumonia 21 (19) 15 (17).76 Definite 7 (6) 10 (11).20 Probable 14 (13) 5 (6).09 Kaposi sarcoma 7 (6) 3 (3).52 b Cytomegalovirus infection 7 (6) 0.02 b Mycobacterium avium-intracellulare infection 5 (4) 0.07 b AIDS-associated dementia/progressive 2(2) multifocal leukodystrophy 4 (4).70 b Cryptococcus neoformans infection 3 (3) 1 (1).63 b Tuberculosis 3 (3) 2 (2) 1.0 b Non-Hodgkin lymphoma 2 (2) 6 (7).14 b CNS lymphoma 2 (2) 2 (2) 1.0 b Toxoplasmosis 2 (2) 0.50 b Histoplasmosis 1 (!1) 2 (2).58 b Probable immunodeficiency related c 12 (11) 8 (9).70 Non-AIDS related All 38 (34) 42 (48).048 Non AIDS-defining infection d 15 (13) 16 (18).35 End-stage liver disease 11 (10) 11 (13).55 CNS hemorrhage 0 4 (5).04 b Other e 12 (11) 11 (13) ND NOTE. Ten patients were not categorized (5 patients in each time period) because cause of death was unspecified pneumonia, unspecified meningitis, or unknown. Insufficient data precluded placement into AIDS-defining illness vs. non AIDS-related category. P values were determined by x 2 test, for proportions, or Student s t test, for continuous variables, unless otherwise indicated. a HIV-associated illnesses include AIDS-defining illnesses and probable immunodeficiency-related illnesses. b Determined by Fisher s exact test. c Includes chronic diarrhea, disseminated herpes, unspecified CNS mass, and bacterial pneumonia caused by Streptococcus pneumoniae or Haemophilus influenzae. d Includes sepsis, bacterial pneumonia not considered to be HIV associated, Aspergillus brain abscess, pseudomembranous colitis, and endocarditis. Not included in infections are unspecified pneumonia and unspecified meningitis, because not enough data were available to determine whether the infectious process was an AIDS-defining illness or not. e Other conditions not included in this table are coronary artery disease, non AIDS-associated malignancies, hemolytic anemia, pericardial hemorrhage, pontine infarct, pulmonary embolism, overdose/suicide, heatstroke, gastrointestinal hemorrhage, pancreatitis, and critical care neuropathy. P types of infections seen during both time periods. Twelve of 17 patients who died of bacterial pneumonia had positive sputum or blood culture results. No significant differences were seen in types of bacterial pneumonia between the 2 groups. Seven patients died of S. pneumoniae infection. One death occurred due to each of the following organisms: Klebsiella pneumoniae, Escherichia coli/enterobacteriaceae, Pseudomonas aeruginosa, H. influenzae, and Rhodococcus equi. Fourteen of 18 patients who died of nonpneumonia sepsis had positive blood culture results. Similar organisms were found in septic patients from both periods. They included the following: P. aeruginosa ( n p 6), with 1 patient having concomitant infection with K. pneumoniae; Staphylococcus aureus ( n p 5); and E. coli, Campylobacter jejuni, orserratia marcescens (1 each). Contributing causes of death. Contributing causes of HIV/AIDS CID 2003:36 (15 April) 1033

5 death were examined for both periods. In period 1, 7 patients (6%) were considered to have pneumonia that did not meet the case definition of PCP contributing to death, compared with none in period 2 ( P p.02). No significant difference was seen in other contributing causes of death. Analysis stratified by receipt of HAART and CD4 cell count. Analysis of deaths stratified by patients receipt of HAART, presented in table 3, is limited to period 2, when HAART was available. Forty-two (48%) of 88 patients in period 2 were receiving HAART. The reasons why patients were not receiving HAART were the following: 18 (39%) of 46 patients were not adherent to regimens, 12 patients (26%) had a new diagnosis of HIV infection within 6 months of death, 5 patients (11%) could not tolerate HAART because of underlying liver disease, 5 patients (11%) had unavailable data, 4 patients (9%) were unable to tolerate HAART because of side effects or progression of AIDS, and 2 patients (4%) were lost to follow-up (one of these patients had HIV infection diagnosed within 9 months of death). Patients who received HAART had a median CD4 cell count (43 cells/ml) similar to those who did not receive HAART (30 cells/ml). Patients receiving HAART were more frequently receiving PCP prophylaxis and less frequently had a CD4 cell count of!200 cells/ml. A significantly higher proportion of patients who were receiving HAART died of non AIDS-related illnesses. No patients receiving HAART died of the following AIDS-defining illnesses: Cryptococcus infection, histoplasmosis, or Kaposi sarcoma. No significant differences existed between Table 3. Cause of death Causes of death in in HIV-positive patients, by HAART use. No. (%) of patients Not receiving HAART (n p 46) Receiving HAART (n p 42) HIV-associated illness 26 (57) 15 (36).05 AIDS-defining illness All 21 (46) 12 (29).10 Pneumocystis carinii pneumonia 10 (22) 5 (12).22 Definite 8 (17) 2 (5).09 a Probable 2 (4) 3 (7).67 a Kaposi sarcoma 3 (7) 0.24 a Non-Hodgkin lymphoma 2 (4) 4 (10).42 a Histoplasmosis 2 (4) 0.50 a CNS lymphoma 1 (2) 1 (2) 1.0 a Cryptococcus neoformans infection 1 (2) a AIDS-related dementia 1 (2) 1 (2) 1.0 a Tuberculosis 1 (2) 1 (2) 1.0 a Probable immunodeficiency related 5 (11) 3 (7).72 a Non AIDS related All 17 (37) 25 (60).03 Non AIDS-defining infections b 8 (17) 8 (19).84 End-stage liver disease 4 (9) 7 (17).26 Other c 5 (11) 10 (24) ND HIV care profile Receipt of P. carinii pneumonia prophylaxis 11 (24) 32 (76)!.01 a CD4 cell count of!200 cells/ml d 44 (98) 30 (77)!.01 a P NOTE. Five patients were not categorized (3 not receiving HAART and 2 receiving HAART) because of insufficient data. P values were determined by x 2 test, for proportions, or Student s t test, for continuous variables, unless otherwise indicated. a Determined by Fisher s exact test. b Includes sepsis, bacterial pneumonia not considered to be HIV associated, pseudomembranous colitis, and endocarditis. c Other conditions not included in this table are coronary artery disease, CNS hemorrhage, Hodgkin disease, hemolytic anemia, lung cancer, pontine infarct, pericardial hemorrhage, and critical care neuropathy. d Values were missing CID 2003:36 (15 April) HIV/AIDS

6 the 2 groups with respect to other causes of death. The leading causes of death in patients receiving HAART were bacterial infections and ESLD, whereas, in those patients who were not receiving HAART, the leading cause of death was PCP. Stratification of patients by CD4 cell count of 200 cells/ ml or!200 cells/ml, regardless of period, revealed 161 patients with a CD4 cell count of!200 cells/ml and 21 patients with a CD4 cell count of 200 cells/ml. As shown in figure 1, 9 (43%) of 21 patients with a CD4 cell count of 200 cells/ml died of ESLD, compared with 13 (8%) of 161 who had a CD4 cell count of!200 cells/ml ( P!.001). PCP was the primary cause of death in 33 patients (20%) who had a CD4 cell count of!200 cells/ml, whereas no PCP-related deaths occurred in patients who had a CD4 cell count of 200 cells/ml ( P p.02). Liver disease. A similar proportion of patients in period 1 were hepatitis C virus (HCV) antibody positive, were hepatitis B surface antigen (HBsAg) positive, had a history of heavy alcohol intake, had underlying liver disease, or had elevated serum alanine aminotransferase levels at admission, compared with patients from period 2 (data not shown). However, when stratified by CD4 cell count, as shown in table 4, each of the following was more common in patients with a CD4 cell count of 200 cells/ml: HCV or HBsAg positivity, history of heavy alcohol intake, and evidence of liver disease at admission. A significantly higher proportion of ESLD deaths occurred in patients with higher CD4 cell counts regardless of year of death (data not shown). Further analysis of all patients who died of ESLD ( n p 22) showed that, in 13 (59%) of 22 patients, HCV contributed to liver disease; in 9 (41%) of 22, liver disease was related to hepatitis B virus; and, in 12 (55%) of 22, alcohol use contributed to liver disease. Eleven (58%) of 19 patients with ESLD had both HCV or HBsAg positivity and concomitant heavy alcohol intake. DISCUSSION Among HIV-infected persons hospitalized at Parkland Memorial Hospital, HAART has affected cause-specific HIVrelated mortality only modestly. Nationwide, the leading causes of death before 1996 were PCP, MAI, CMV, and bacterial pneumonia [17]. In developed countries, HAART and prophylaxis for opportunistic infections have markedly reduced opportunistic infection related deaths [2, 6, 9]. Yet, at our safety-net hospital, which provides a Ryan White funded, state-of-theart HIV outpatient clinic and a National Institutes of Health sponsored AIDS clinical trials unit, the HAART era brought a less-marked change. The proportion of deaths due to PCP was the same in as it was in This observation differs from numerous published studies documenting a reduction in mortality from PCP [2, 8, 18 20]. Several factors may explain why PCP remains a leading cause of death in our patient population. First, despite the availability of HAART, the majority (89%) of patients in died of AIDS (as defined by a CD4 cell count of!200 cells/ml). Second, almost one-half of patients who died in were not receiving HAART because of nonadherence to medications or new diagnosis of HIV infection. Nonadherence to HAART also appeared to be associated with nonadherence to PCP prophylaxis Figure 1. Leading cause of death, by CD4 cell count. All comparisons were made by Fisher s exact test. Definitions of causes of death are as follows: sepsis, nonpneumonia causes of sepsis; malignancies, all AIDS-defining or non AIDS-related neoplastic causes of death; and pneumonia, all bacterial pneumonias, including those due to Streptococcus pneumoniae and Haemophilus influenzae. ESLD, end-stage liver disease; PCP, Pneumocystis carinii pneumonia. HIV/AIDS CID 2003:36 (15 April) 1035

7 Table 4. Risk factors for and indicators of liver disease, stratified by CD4 cell count. Risk factor or indicator!200 Cells/mL (n p 161) CD4 cell count 200 Cells/mL (n p 21) HCV antibody positive or HBsAg positive a,b 53 (33) 13 (62).01 HCV antibody positive a 41 (28) 9 (45).12 HBsAg positive a 16 (10) 4 (20).26 c Heavy alcohol intake d 36 (22) 10 (48).02 History of liver disease e 27 (17) 10 (48).01 c ALT level of 140 IU/L 49 (30) 10 (48).12 ALT level, median IU/L (range) a 26 (2 1040) 31 (8 1598).29 AST level, median IU/L (range) a 56 (9 1449) 66 ( ).32 P NOTE. Data are no. (%) of patients, unless otherwise indicated. P values were determined by x 2 test, for proportions, or Student s t test, for continuous variables, unless otherwise indicated. Ag, antigen; ALT, serum alanine aminotransferase; AST, aspartate aminotransferase; HBsAg, hepatitis B surface antigen; HCV, hepatitis C virus. a Values were missing. b Combined HCV antibody positivity and HBsAg positivity. c Determined by Fisher s exact test. d Defined by either a history of alcoholism or consumption of 16 drinks per day. e Defined by clinical diagnosis of liver disease with 1 of the following: cirrhosis, hepatomegaly, fatty liver, ascites, low albumin level, jaundice, coagulopathy, or spider angiomas. in this resource-limited patient population. Longitudinal data show the risk of death is significantly decreased in patients receiving primary PCP prophylaxis [3]. A significant increase in the proportion of non AIDS-related deaths did occur between 1995 and Unlike other studies that show an increase in the proportion of deaths due to liver disease [10, 15], we did not find a significant change in the proportion of deaths due to this cause. Like other studies [6, 12], we did not see an increase in the proportion of deaths due to non AIDS-defining infections. However, these 2 etiologies were the most common non AIDS-related causes of death in both time periods. We did see an increase in intracerebral hemorrhages not related to an underlying CNS lesion. Changes in diagnostic evaluation may account for these differences. Comparison of patients who were receiving HAART to those who were not revealed a decrease in HIV-associated deaths and an increase in non AIDS-related deaths. Although we did not see a significant decrease in AIDS-defining illnesses in our patients receiving HAART, other studies have shown that use of HAART does decrease the number of deaths due to AIDSdefining illnesses [1, 3, 4]. HAART has also been associated with a decrease in disease progression and death in patients with CD4 cell counts of!200 cells/ml [21, 22]. Seventy-seven percent of our patients receiving HAART had CD4 cell counts of!200 cells/ml, but 60% of these deaths were due to non AIDS-related diseases. Our small sample size may have precluded detection of significant differences in specific causes of death. Despite this limitation, some interesting trends were noted. Patients not receiving HAART more frequently died of PCP, which was also the leading cause of death in patients in the pre-haart era. However, the patients who died in while receiving HAART more frequently died of ESLD. Non AIDS-defining infections in patients receiving HAART were similar to those in patients who were not receiving HAART. Larger cohort studies are needed to determine the impact of HAART on deaths due to liver disease and other causes. Many of the patients not receiving HAART were of ethnic minorities. Nine (75%) of 12 patients with newly diagnosed HIV infection and 12 (67%) of 18 nonadherent patients were black or Hispanic. Deaths among HIV-infected minorities increased at our hospital from 48% in 1995 to 61% in National surveys show that the incidence of AIDS (per 100,000 persons) in 2000 was 3 times higher among Hispanic persons (30.4) and 9 times higher among black persons (74.2) than it was among white persons (7.9) [23]. Minorities may have limited access to medical care for economic, social, and cultural reasons. Lower socioeconomic status has been associated with a decreased duration of survival in patients with AIDS [24]. Efforts targeted at improving early diagnosis and adherence to medical therapy may decrease the mortality rate in this population. Our data demonstrate that the leading cause of death in patients who died with a CD4 cell count of 200 cells/ml was ESLD. The finding that 10% of our patients died with a CD4 cell count of 200 cells/ml in both 1995 and differs from the findings of a study from Boston in which the proportion of patients who died with a CD4 cell count of 200 cells/ml increased from 2.8% in 1996 to 18.2% in [15]. A possible explanation for the increasing deaths due to ESLD from the Boston study may be the higher proportion of Boston pa CID 2003:36 (15 April) HIV/AIDS

8 tients dying with CD4 cell counts of 200 cells/ml. Because the proportion of patients who died with a CD4 cell count of 200 cells/ml did not change over time in our study, we did not observe an increase in the proportion of deaths due to ESLD. Our study is limited to inpatient deaths, and, therefore, we cannot calculate mortality rates or evaluate a causal association between HAART and HIV-associated deaths. As in any retrospective study, we cannot exclude bias, but we used consistent definitions to minimize potential bias. Although the case definition of probable PCP may have misclassified a few cases in 1995, the increase in definite PCP in suggests that PCP remains an important cause of death in our population. Larger prospective cohort studies in the HAART era are needed to define the impact of HAART on the changing pattern of deaths. In our study, a trend toward a decrease in deaths due to AIDS-defining illness did occur in the HAART era compared with the pre-haart era. However, PCP still remains an important cause of death. The leading non AIDS-related deaths were due to non AIDS-defining infections and ESLD in both time periods. In the subset of patients with CD4 cell counts of 200 cells/ml, ESLD was the leading cause of death. Strategies for the prevention and treatment of alcohol addiction, hepatitis B virus infection, and HCV infection may have an important impact on deaths due to ESLD in HIV-infected patients receiving HAART. The majority of patients who died in the HAART era had a CD4 cell count!200 cells/ml, and almost one-half of these patients who died were not receiving HAART at the time of death. Strategies to improve early diagnosis and maximize treatment adherence may lead to improved survival in high-risk populations. Acknowledgments Special thanks to Lori Fischbach and Rajeev Jain, for reviewing the article in manuscript, and to Jill Hester, for data collection. References 1. Palella FJ Jr, Delaney KM, Moorman AC, et al. Declining morbidity and mortality among patients with advanced human immunodeficiency virus infection. HIV Outpatient Study Investigators. N Engl J Med 1998; 338: Brodt HR, Kamps BS, Gute P, Knupp B, Staszewski S, Helm EB. Changing incidence of AIDS-defining illnesses in the era of antiretroviral combination therapy. AIDS 1997; 11: McNaghten AD, Hanson DL, Jones JL, Dworkin MS, Ward JW. Effects of antiretroviral therapy and opportunistic illness primary chemoprophylaxis on survival after AIDS diagnosis. Adult/Adolescent Spectrum of Disease Group. AIDS 1999; 13: Hogg RS, O shaughnessy MV, Gataric N, et al. Decline in deaths from AIDS due to new antiretrovirals. Lancet 1997; 349: Chiasson MA, Berenson L, Li W, et al. Declining HIV/AIDS mortality in New York City. J Acquir Immune Defic Syndr 1999; 21: Valdez H, Chowdhry TK, Asaad R, et al. Changing spectrum of mortality due to human immunodeficiency virus: analysis of 260 deaths during Clin Infect Dis 2001; 32: Lewden CH, Héripret L, Bonnet L, et al. Causes of death in HIVinfected adults in the era of highly active antiretroviral therapy (HAART): the French Survey Mortalite 2000 [abstract 753-W]. In: Programs and abstracts of the 9th Conference on Retroviruses and Opportunistic Infections (Seattle). Alexandria, VA: Foundation for Retrovirology and Human Health, 2002: Morgello S, Mahboob R, Yakoushina T, Khan S, Hague K. Autopsy findings in a human immunodeficiency virus infected population over 2 decades: influences of gender, ethnicity, risk factors, and time. Arch Pathol Lab Med 2002; 126: Sansone GR, Frengley JD. Impact of HAART on causes of death of persons with late-stage AIDS. J Urban Health 2000; 77: Wolfe MI, Hanson DL, Selik DL, Swerdlow R. Deaths from non AIDSrelated diseases have increased as a proportion of deaths of HIVinfected persons since the advent of HAART [abstract 14]. In: Programs and abstracts of the 9th Conference on Retroviruses and Opportunistic Infections (Seattle). Alexandria, VA: Foundation for Retrovirology and Human Health, 2002: Jambroes MD, Dieleman J, van de Wiel J, et al. Changing mortality in an HIV-infected population on HAART [abstract 754-W]. In: Programs and abstracts of the 9th Conference on Retroviruses and Opportunistic Infections (Seattle). Alexandria, VA: Foundation for Retrovirology and Human Health, 2002: Folch EH, Herández I, Vetter T, et al. AIDS-related mortality: improved survival with HAART but causes of death unchanged [abstract 756- W]. In: Programs and abstracts of the 9th Conference on Retroviruses and Opportunistic Infections (Seattle). Alexandria, VA: Foundation for Retrovirology and Human Health, 2002: Rodríguez BV, Valdez H, Salata H, et al. Changes in causes of death in a cohort of HIV-infected individuals: analysis of the last 6 years [abstract 755-W]. In: Programs and abstracts of the 9th Conference on Retroviruses and Opportunistic Infections (Seattle). Alexandria, VA: Foundation for Retrovirology and Human Health, 2002: Cacoub P, Geffray L, Rosenthal E, Perronne C, Veyssier P, Raguin G. Mortality among human immunodeficiency virus infected patients with cirrhosis or hepatocellular carcinoma due to hepatitis C virus in French Departments of Internal Medicine/Infectious Diseases, in 1995 and Clin Infect Dis 2001; 32: Bica I, McGovern B, Dhar R, et al. Increasing mortality due to endstage liver disease in patients with human immunodeficiency virus infection. Clin Infect Dis 2001; 32: Centers for Disease Control and Prevention Revised classification system for HIV infection and expanded surveillance case definition for AIDS among adolescents and adults. MMWR Recomm Rep 1992; 41: Chan IS, Neaton JD, Saravolatz LD, Crane LR, Osterberger J. Frequencies of opportunistic diseases prior to death among HIV-infected persons: community programs for clinical research on AIDS. AIDS 1995; 9: Hofman P, Saint-Paul MC, Battaglione V, Michiels JF, Loubiere R. Autopsy findings in the acquired immunodeficiency syndrome (AIDS): a report of 395 cases from the south of France. Pathol Res Pract 1999; 195: Masliah E, DeTeresa RM, Mallory ME, Hansen LA. Changes in pathological findings at autopsy in AIDS cases for the last 15 years. AIDS 2000; 14: Selik RM, Karon JM, Ward JW. Effect of the human immunodeficiency virus epidemic on mortality from opportunistic infections in the United States in J Infect Dis 1997; 176: Hogg RS, Yip B, Chan KJ, et al. Rates of disease progression by baseline CD4 cell count and viral load after initiating triple-drug therapy. JAMA 2001; 286: HIV/AIDS CID 2003:36 (15 April) 1037

9 22. Egger M, Hirschel B, Francioli P, et al. Impact of new antiretroviral combination therapies in HIV infected patients in Switzerland: prospective multicentre study. Swiss HIV Cohort Study. BMJ 1997; 315: Centers for Disease Control and Prevention. US HIV and AIDS cases reported through December AIDS Surveill Rep 2001; 12: Rapiti E, Porta D, Forastiere F, Fusco D, Perucci CA. Socioeconomic status and survival of persons with AIDS before and after the introduction of highly active antiretroviral therapy. Lazio AIDS Surveillance Collaborative Group. Epidemiology 2000; 11: CID 2003:36 (15 April) HIV/AIDS