Comparisons of Causes of Death and Mortality Rates Among HIV-Infected Persons

Analysis of the Pre-, Early, and Late HAART (Highly Active Antiretroviral Therapy) Eras

Abstract

Methods: 

Comparisons of death-related variables during the 3 eras were performed.

Results: 

The number of deaths declined over the study period, with 987 deaths in the pre-HAART era, 159 deaths in the early HAART era (1997-1999), and 78 deaths in the late HAART era (2000-2003) (P < 0.01). The annual death rate peaked in 1995 (10.3 per 100 patients) and then declined to <2 deaths per 100 persons in the late HAART era (P < 0.01). The proportion of deaths attributable to infection decreased, but infection remained the leading cause of death in our cohort, followed by cancer. Of those who died, there was an increasing proportion of non-HIV-related deaths (32% vs. 9%; P < 0.01), including cardiac disease (22% vs. 8%; P < 0.01) and trauma (8% vs. 2%; P = 0.01) in the post-HAART versus pre-HAART era. Despite the absence of intravenous drug use and the low prevalence of hepatitis C coinfection in our cohort, an increasing proportion of deaths in the HAART era were attributable to liver disease, although the numbers are small.

Conclusions: 

Despite increasing concerns regarding antiretroviral resistance, the death rate among HIV-infected persons in our cohort continues to decline. Our data show a lower death rate than that reported among many other US HIV-infected populations; this may be the result of open access to health care. A shift in the causes of death toward non-HIV-related causes suggests that a more comprehensive health care approach may be needed for optimal life expectancy; this may include enhanced screening for malignancy and heart disease as well as preventive measures for liver disease and accidents.

HIV mortality rates have dramatically declined since the availability of highly active antiretroviral therapy (HAART) and effective prophylaxis for opportunistic infections.^1-9 Given increasing antiretroviral-related complications and resistance, however, whether this decrease in deaths will be sustained is debated.^3,10,11

Some studies continue to describe high rates of death attributable to AIDS-defining conditions.^12-14 A recent study by Jain et al¹⁵ found that most deaths occurred among patients with a CD4 count of <200 cells/mL and that a leading cause of death remained Pneumocystis carinii (jiroveci) pneumonia (PCP). Other studies have shown an increasing proportion of deaths attributable to non-HIV-related conditions, especially that of liver failure.^3,16-20 In some cohorts, liver disease now accounts for greater than 50% of the deaths among patients with a CD4 count >200 cells/mL or an undetectable HIV viral load.^15,16

Divergent results regarding the cause of death are likely related to the underlying characteristics of the study populations, including injectable drug use, coinfection with hepatitis B and C, medication adherence, and the availability of antiretrovirals.¹⁵ In addition, patients with private insurance have been shown to receive more intensive drug regimens and to have lower mortality rates.⁴ A study among patients with open access to medical care as well as a low rate of drug use and hepatitis C coinfection may provide some insight regarding the effects of these barriers on overall mortality. We evaluated such a population, US military beneficiaries, to assess causes of death and mortality rates in this cohort during the years 1990 through 2003.

METHODS

Data collected during an HIV natural history study were retrospectively analyzed for causes of death and annual death rates. The original study is an ongoing, prospective, continuous enrollment, longitudinal cohort study conducted among HIV-infected Department of Defense (DoD) beneficiaries as part of the Triservice AIDS Clinical Consortium funded jointly by the US Military HIV Research Program and the National Institutes of Health. Military members in the study had data on the seroconversion window available; we defined the year of HIV infection as the midpoint between the last seronegative HIV test and the first seropositive HIV test. An HIV provider evaluated each participant biannually at one of the DoD HIV clinics. All clinical information, including new medical diagnoses, medication changes, and laboratory values (CD4 cell count, HIV viral load, and routine laboratory values), were recorded at each visit. These data were extracted from the medical charts by trained study coordinators and entered into a centralized computerized database. Baseline values for hepatitis coinfection were entered during the initial visit.

Participating centers included the Naval Medical Center San Diego, San Diego, CA; Walter Reed Army Medical Center, Washington, DC; National Naval Medical Center, Bethesda, MD; Naval Medical Center Portsmouth, Portsmouth, VA; Wilford Hall Air Force Medical Center and Brooke Army Medical Center, San Antonio, TX; and Tripler Army Medical Center, Honolulu, HI. The protocols and consent forms for the study were approved by the institutional review boards of each participating institution and by the DoD Triservice Human Subjects Research Review Board.

Data were abstracted from the study database in January of 2004 for all participants who died during the study period (1990-2003). The years of the study were divided into pre-HAART (1990-1996), early HAART (1997-1999), and late HAART (2000-2003). Information collected from the computer database included demographics, year of HIV infection, year of death, CD4 cell count and HIV viral load (Roche Amplicor, Branchburg, NJ) at baseline and death, use of antiretroviral and prophylactic medications, hepatitis B (core antibody and surface antigen) and C (antibody test) infection, history of opportunistic infection, laboratory values (cholesterol, triglycerides, alanine transferase, and aspartate transferase), and cause of death. The primary and secondary causes of death were determined by review of death certificates, discussion with the primary physician at the time of death, and/or searching the Social Security Death Index (SSDI) database, and these data were prospectively entered into the database.

The terms used in the study included death due to an AIDS-defining illness, which was defined as death attributable to a category C disease listed by the Centers for Disease Control and Prevention,²¹ and death due to an HIV-infection associated disease, defined as an immunodeficiency-related disease (eg, bacterial pneumonia caused by Streptococcus pneumoniae, chronic diarrhea). A cause of death not attributable to an AIDS-defining or HIV-related illness was characterized as “non-HIV-related” (eg, trauma, suicide, liver disease). HAART was defined as the use of ≥3 full-dose antiretroviral drugs. The most recent evaluation before death was used to determine the medications and laboratory values at death for each patient; these ranged from the time of death to 6 months before death.

Descriptive statistics were performed. Death rates were calculated using the number of participants in the study for each year. Comparisons of the proportions of deaths in the pre-HAART and post-HAART eras were performed using the Fisher exact test for categoric variables and rank-sum tests for continuous variables. Comparisons of the variables of interest in the pre-HAART, early HAART, and late HAART eras were performed using the Fisher exact test for categoric variables and the Kruskal-Wallis test for continuous variables. Rates over time used rank-sum tests and the test of equality of regression slopes (Stata 7.0 software; Stata Corporation, College Station, TX). P values were 2-tailed, and values of <0.05 were considered statistically significant.

RESULTS

A total of 1224 (28.9%) deaths occurred among our cohort of 4241 participants from January 1990 through December 2003. Of these, 987 deaths occurred in the pre-HAART era and 237 in the HAART era; the latter included 159 deaths during the early HAART era (1997-1999) and 78 in the late HAART era (2000-2003). The annual death rate peaked in 1995 (10.3 per 100 patients) and then steadily declined (Fig. 1A) to the current rate during the late HAART era of <2 deaths per 100 persons (a test of regression slopes yields P < 0.01; Table 1). The early HAART era had a higher death rate than the late HAART era (P = 0.03). Overall, there was an 80% decrease in deaths from 1990 to 2003. The median time of survival after HIV seroconversion during 1994 through 1996 was 8 years compared with 10 years for the early HAART era and 12 years during the late HAART era.

TABLE 1: Annual Death Rates, 1990 Through 2003

FIGURE 1: A, Death rates during the pre- versus post-HAART eras. B, Percentage of deaths due to cancer during the pre- versus post-HAART eras. C, Percentage of deaths due to infection during the pre- versus post-HAART eras.

Demographic data demonstrated a slight increase in age at the time of death among HIV-infected persons from 37 to 40 years (P < 0.01; Table 2). The sex or race of patients who died did not substantially change over the study period. Patients were less likely to die of an AIDS-defining illness over the study period (Table 3). Eighty percent of pre-HAART deaths were attributable to AIDS-defining conditions compared with 65% in the early post-HAART era and 56% in the late post-HAART era (P < 0.01). Likewise, of those who died, the proportions of death attributable to causes unrelated to the HIV infection increased over time: 9%, 28%, and 32% in the 3 eras, respectively (P < 0.01).

TABLE 2: Comparisons of Characteristics of Patients Dying in the Pre-HAART, Early HAART, and Late HAART Eras

TABLE 3: Cause of Death Categories*

The proportion of deaths attributable to opportunistic infections as the primary cause of death diminished from 59% during 1990 through 1996 to 24% during 2000 through 2003 (P = 0.03; see Table 2). Furthermore, the number of opportunistic infections that contributed to the cause of death dramatically declined; 17% of those dying in the pre-HAART period had 2 or more opportunistic infections contributing to their death compared with 1.7% in the early HAART era and 0% in the late HAART era. PCP significantly decreased as a cause of death (P = 0.01); cytomegalovirus (CMV) and Mycobacterium avium complex (MAC) infections also declined, but the rates of death attributable to these causes did not meet statistical significance.

Liver disease as the primary cause of death increased from 0.2% (1 of 640 patients) in the pre-HAART era to 3.7% (3 of 83 patients) in the post-HAART era (P < 0.01). Although there was no difference in the percentage of patients with hepatitis B (chronic) infection, hepatitis C was more common among those dying in the post-HAART era (12% vs. 7%; P < 0.05).

All causes of death during the pre-HAART and post-HAART eras are listed in Table 4. Combined through the study period, the cause of death was reported as infection in 67% and as cancer in 22%. Ten percent of patients are reported to have died of cardiac disease, but most of these deaths were recorded as cardiac arrest; 2.2% died of congestive heart failure or cardiomyopathy. Of those who died, post-HAART versus pre-HAART deaths were significantly more likely related to cardiac disease (22% vs. 8%; P < 0.01), diabetes mellitus (3% vs. 0%; P = 0.01), anal cancer (3% vs. 0%; P = 0.01), trauma (8% vs. 2%; P = 0.01), hepatitis B (4% vs. 1%; P = 0.04), or hepatitis C (4% vs. 0%; P = 0.01).

TABLE 4: All Causes of Death During Pre-HAART and Post-HAART Eras*

The proportion of deaths that were attributable to cancer, including non-Hodgkin lymphoma (NHL) and Kaposi sarcoma, was not significantly different comparing the pre-HAART and post-HAART eras; however, trends over time demonstrated an increased proportion of deaths attributable to cancer between 1990 and 1997, after which time there was an overall decline (P < 0.01; see Fig. 1B). Deaths occurring in the post-HAART era were significantly less likely than in the pre-HAART era to be related to an infection (44% vs. 80%; P < 0.01), specifically bacterial causes (19% vs. 32%; P = 0.02) and PCP (0% vs. 13%; P < 0.01). The proportion of deaths related to infections during the years 1990 through 2003 declined 5.6% per year (P < 0.01; see Fig. 1C). Despite the reduction in deaths associated with infection and cancer, these conditions remain the most common cause of death among HIV-infected patients in our cohort.

Laboratory data showed that 93% of those dying in the pre-HAART era had a CD4 count less than 200 cells/mL at the time of death compared with 77% in the early HAART era and 61% in the late HAART era (P < 0.01). The mean CD4 count at death was 122 cells/mL during the pre-HAART era compared with 316 cells/mL in the late post-HAART era (P < 0.01). Those dying in the late HAART era were 2.8 times more likely to have an undetectable HIV viral load (<400 copies/mL) at death compared with those in the early HAART era (15% vs. 6%; P = 0.08). The cholesterol level was elevated (>200 mg/dL) among 10% of those dying in the pre-HAART era compared with 29% of those dying in the post-HAART era (P < 0.01). There were no differences in triglyceride values at the time of death over the period of this study.

Of those dying with a CD4 cell count >200 cells/mL, 38% had a non-HIV-related primary cause of death. The most common causes of death in this group were trauma (60%), suicide (20%), cardiac disease (13%), and other (7%). Patients with a CD4 cell count at death of >200 cells/mL (compared with ≤200 cells/mL) were more likely to have concurrent hepatitis C (odds ratio [OR] = 2.95; P < 0.01) or an undetectable HIV viral load (OR = 6.53; P < 0.01); this group was less likely to have a history of MAC (OR = 0.06; P < 0.01), PCP (OR = 0.19; P < 0.01), or CMV (OR = 0.46; P < 0.01).

DISCUSSION

These data show that deaths from HIV infection have declined in our cohort similar to the trend seen across the United States since the introduction of HAART. With improvement in HIV treatment, proportionally fewer patients are currently dying of the “typical” HIV-related illnesses; of those who die, the causes are increasingly attributable to cardiac disease, trauma, and liver disease.

AIDS-defining causes of death declined to 56%, whereas non-HIV-related conditions have increased to 32% of the causes of deaths among our study cohort. This is similar to other studies, which have shown that AIDS-defining illnesses as a cause of death are rapidly falling.^3,12 In 2000 through 2003, 39% of the patients in our cohort who died had a CD4 count >200 cells/mL (overall mean CD4 cell count = 316 cells/mL) compared with 7% in the pre-HAART era, corroborating that proportionally more patients are dying with relative immune preservation.

Deaths during the post-HAART versus pre-HAART era are proportionally more often attributable to cardiac disease, trauma, and liver disease. We also noted increasing cholesterol levels in the post-HAART era; this finding may be related to the induction of protease inhibitor therapy and may be an important factor in the increasing rate of cardiac disease. Other studies have also noted an increased number of deaths attributable to “non-HIV”-related diseases, including liver disease, drug overdose, non-AIDS-defining malignancy, cardiac disease, obstructive lung disease, suicide, homicide, and trauma.^14-16,22-24

Despite having a patient population without significant intravenous drug use and relatively few hepatitis C coinfections, we noted a rising rate of deaths attributable to liver disease, although our numbers are small. Other study cohorts have also shown more liver-related deaths^16,17; Bica et al¹⁶ noted an increase from 12% in 1991 to 50% in 1998 through 1999 in a population consisting mainly of intravenous drug users. The impact of viral hepatitis in populations with concurrent hepatitis B and C coinfection is likely to become more evident, because progression to significant liver disease usually takes >10 years and opportunistic infections may have masked the mortality of hepatitis coinfection in the pre-HAART era.²⁵ Concurrent viral hepatitis also increases the risk of hepatotoxicity during HIV therapy (especially with nevirapine and ritonavir); hence, it may increase morbidity and mortality in this setting.²⁶ Liver disease may also be related to the use of HIV medications; the precise impact of long-term antiretroviral drug use on the rising number of deaths attributable to liver disease remains uncertain. Whether the finding of an increased proportion of deaths attributable to liver disease in our cohort represents evidence addressing this question remains to be determined.

Infections, especially PCP, as a cause of death decreased dramatically from 1990 to 2003. In our cohort, no deaths were attributed to PCP during the HAART era; this differs from some studies, which have shown that PCP remains an important cause of death.¹⁵ Despite marked improvements in the prevention of many infections, infections remain the leading cause of death among HIV-infected persons.

Cancer rates increased until 1997 but declined thereafter in our cohort. HAART seems to have had a smaller impact on the occurrence of neoplastic conditions than on that of infections^3,24,28; therefore, if the number of infection-related deaths continues to decline, cancer may become a more prominent cause of death.²⁰ Although the rates of Kaposi sarcoma and other cancers have declined, trends in deaths attributed to NHL are less clear^3,29-31; this may be a result of the fact that NHL is caused by chronic B-cell stimulation and tends to occur at a less advanced stage of immunosuppression, suggesting that prevention may require early antiretroviral therapy and more complete immune reconstitution.³⁰ In addition, non-AIDS-defining cancers as the cause of death are becoming increasingly important.^24,32

Our study cohort is unique in several regards, including open access to medical care, free medications, and enrollment soon after the time of HIV seroconversion. The lower death rate (<2 deaths per 100 persons) compared with that reported in some other study cohorts (4 to 6 per 100 persons)^3,11,14 suggests the hypothesis that improved access to health care, including free medication, can result in reduced mortality among those with HIV infection. Other factors such as drug use, homelessness, and stage of disease would need to be accounted for to test this hypothesis. In a study of the US Medicaid system, using a mathematic model, the authors demonstrated that increasing access to free HIV medications by expanding the Medicaid program to include treatment of HIV could result in marked improvements in survival and a decreased incidence of AIDS among those who would be eligible for such benefits.³³ Another study demonstrated that access to a primary health care provider was directly related to receipt of antiretroviral treatment and PCP prophylaxis, which have been linked to survival and AIDS.³⁴ Such information is important in devising strategies to improve outcomes in HIV-infected populations.

Limitations of the study include a study population that differs from the typical US HIV patient. Our ethnically balanced cohort has open and equal access to health care, including free medications, and thus may not be representative of some groups with more limited resources. Also, most of our subjects have stable employment, income, and housing, circumstances that are different from other cohorts. We examined only deaths during the study period and did not use data among living patients; the goal of our analysis focused on examining causes of death rather than why certain persons died. The potential for delays in reporting deaths may create uncertainties regarding the frequency of recent deaths; however, most deaths are quickly reported because of close communication between HIV physicians and study data collectors who work within the same clinics. Information on death certificates may be unreliable because of confidentiality issues (sensitivity) and lack of diagnostic procedures during the end of life (specificity); these potential biases are unlikely to have changed over time.³⁵

This study demonstrates declining death rates among HIV-infected persons in our cohort with improved HIV control (fewer AIDS-defining deaths and increased CD4 cell count) at the time of death. Continued monitoring of the causes of death among HIV-infected persons will be necessary to improve screening and therapy for non-HIV-related conditions.

ACKNOWLEDGMENTS

The authors express their gratitude to Mark Milazzo, the data manager for this study cohort. In addition, they thank all the clinicians and study coordinators who devoted time in the collection of the data contained within this report.