Methods

By law, persons living in San Francisco, California, meeting the definition for AIDS as defined by the Centers for Disease Control and Prevention (CDC) criteria [3] are reported to the San Francisco Department of Public Health. The medical records are reviewed at the time of initial report and every 18 months thereafter, and reporting through this system has been evaluated to be 97% complete [4]. Mortality is ascertained through weekly review of local death certificates, reports from the State Office of AIDS, and annual matches with the National Death Index

All deaths among persons aged >12 years who were given an AIDS diagnosis in San Francisco between 1 January 1994 through 31 December 1998 were included in the analysis. Cause-of-death information, as determined by death certificate review, was coded by use of the International Classification of Diseases–9th revision (ICD-9) [5]. We examined trends in both underlying and contributory causes of death that occurred in at least 1% of the deaths during the 5-year study period. HIV- or AIDS-related causes of death were defined as causes of death identified by ICD-9 codes 042–044, as well as for all AIDS-related opportunistic infections as defined by CDC criteria [3]. In addition, trends in syndromes associated with mitochondrial toxicity (lactic acidosis, acute hepatitis not otherwise specified, polyneuropathy, myopathy, myocarditis, cardiomyopathy, pancreatitis, and adverse drug reaction) were evaluated

We analyzed the temporal trends among AIDS deaths in 3 ways. First, we compared annual changes in the proportion of underlying and contributory causes of deaths combined due to each cause by use of the Mantel-Haenszel χ² test. Second, we calculated the trends in the mortality rates for HIV- or AIDS-related and non–HIV- or non–AIDS-related deaths by use of the χ² test for linear trend. The single underlying cause of death was used for the numerator, and the denominator included the number of persons living with AIDS and the number of persons with AIDS who died in each year. Third, we compared trends in the observed number of deaths due to selected non–HIV- or non–AIDS-related causes, with the expected number of deaths derived from a reference population adjusted by race and age. Because of sample size constraints, this analysis was limited to men. We used the underlying cause of death and the indirect standardization method in which the standard population was California men aged 20–79 years. We calculated the standardized mortality ratio by dividing the number of observed deaths by the number of expected deaths. All statistical tests were done with use of SAS software (SAS Institute) and EpiInfo (CDC)

Results

There were 5234 deaths among persons with AIDS in San Francisco from 1 January 1994 through 31 December 1998. The demographic and risk characteristics of the deaths were similar to those of persons living with AIDS in the same time period, occurring mostly among men (96%), whites (71%), and men who have sex with men (men who have sex with men as the only risk, 74%; men who have sex with men and who also inject drugs, 14%). The median age at time of death increased from 40 years for persons with AIDS dying in 1994 to 44 years for persons with AIDS dying in 1998. Twenty causes listed as the underlying or contributory cause of death occurred in at least 1% of deaths during the study period (table 1). There was a decline in the proportion of deaths associated with HIV- or AIDS-related causes between 1994 and 1998 and an increase in the proportion of deaths associated with non–HIV- or non–AIDS-related causes

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Figure 1

Trends in human immunodeficiency virus (HIV)– or AIDS-related and non–HIV- or non–AIDS-related mortality rates among patients with AIDS in San Francisco, 1994–1998. Numerator includes single underlying cause of death. Denominator includes sum of no. of persons living with AIDS and no. of persons with AIDS who died in each year

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Table 1

Temporal trends in proportions of selected causes of death (underlying and contributory) among persons with AIDS, San Francisco, 1994–1998

Among causes of death associated with AIDS-related opportunistic infections, there were statistically significant declines in the proportion associated with nontuberculous mycobacteria, wasting syndrome, cytomegalovirus infection, Pneumocystis carinii pneumonia, Kaposi sarcoma, progressive multifocal leukoencephalopathy, toxoplasmosis, and cryptosporidiosis (P<.01). There was a small increase of borderline significance for non-Hodgkin lymphoma (P<.08)

Of the non–HIV- or non–AIDS-related causes of death, the proportion of deaths associated with malignancies other than Kaposi sarcoma or non-Hodgkin lymphoma, viral hepatitis, chronic liver disease, coronary artery disease, obstructive lung disease, septicemia, and overdose increased significantly (P<.05). Of the causes of death that might represent mitochondrial toxicity, pancreatitis and cardiomyopathy were found to occur in ⩾1% of total deaths and were associated with an increasing proportion of deaths (P<.05 and P<.08, respectively; table 1)

In total, 352 deaths were associated with non–AIDS-defining malignancies. Of these, lung cancer was most frequent (37 [11%]), followed by Hodgkin disease (18 [5%]), hepatocellular cancer (15 [4%]), and anal cancer (10 [3%]). Deaths associated with viral hepatitis were equally attributed to hepatitis B virus (HBV) and hepatitis C virus (HCV) (80 [51%] and 85 [55%] of 156, respectively; 12 deaths were associated with both). HBV and HCV were listed as contributory causes of death for 4 (27%) and 3 (20%), respectively, of the 15 deaths associated with hepatocellular cancer. Injection drug use was significantly associated with deaths due to accidental overdose (54 [71%] of 76), viral hepatitis (62 [40%] of 156), and septicemia (119 [33%] of 358; P<.01)

Annual changes in mortality rates are shown in figure 1. Decreases in all-cause mortality as well as mortality due to HIV- or AIDS-related conditions were observed from 1995 through 1998 (P<.01). Death rates due to non–HIV- or non–AIDS-related causes remained relatively stable (P=.15)

The numbers of observed and expected deaths due to selected non–HIV- or non–AIDS-related causes are shown in table 2. For the pooled years 1994–1998, the observed number of deaths exceeded the expected number for every cause examined. The ratio of observed to expected number of deaths (the standardized mortality ratio) was highest for viral hepatitis, septicemia, and pancreatitis

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Table 2

Trends in observed and expected numbers of deaths and standardized mortality ratio (SMR) due to underlying non–human immunodeficiency virus (HIV)- or –AIDS-related causes among San Francisco men with AIDS aged 20–79 years, 1994–1998

Discussion

As others have reported, we found a decline in the rate of HIV- or AIDS-related deaths from 1995 through 1998 [1, 2]. The relatively stable rate seen between 1997 and 1998 likely reflects a plateau and subsequent leveling off of the beneficial effects of HAART in this population. Also consistent with other studies, we found an increasing proportion of deaths from 1994 through 1998 associated with non–AIDS-defining malignancies and other chronic diseases [1, 2]. Furthermore, the observed numbers for these causes of death in our study population exceeded the expected numbers derived from an age- and race-adjusted reference population in both pre- and post-HAART eras

The increased prominence of deaths associated with overdoses, viral hepatitis and end-stage liver disease, obstructive lung disease, and septicemia suggests that risk behaviors that may be more prevalent in HIV-infected persons are playing an important role in mortality trends. Coinfection with HIV and the hepatitis viruses is common and may accelerate the progression of liver disease. Prevalence rates of HBV-induced cirrhosis are higher in the presence of HIV infection, and HIV- and HCV-coinfected patients have more rapid progression to fibrosis and liver failure and an increased risk of death [6]. Although recent reports have suggested that HAART may contribute to hepatotoxicity irrespective of presence of coinfection with HBV or HCV [7], we found no evidence for an increase in observed compared with expected numbers of deaths due to liver disease after the first 2 years of widespread use of HAART. The importance of deaths associated with septicemia was not surprising, given the association between HIV and injection drug use, a behavior that conveys an increased risk of death due to these and other causes independent of HIV serostatus [8]. Obstructive lung disease, chronic bronchitis, and bronchiectasis have been reported to occur at increased frequencies among HIV-infected persons, and infection with HIV may accelerate the onset of smoking-induced emphysema [9]. Smoking prevalence is high among HIV-infected persons [10], as well as among men who have sex with men [11], who contributed >70% of deaths in our study. The importance of tobacco exposure as a mediator is supported by our finding that lung cancer was the most common cause of death from non–AIDS-defining malignancies, followed by Hodgkin disease, liver cancer, and anal cancer. Almost half of the hepatocellular cancer deaths were associated with HBV and HCV. Other risk factors, such as tobacco, alcohol, or coinfection with sexually transmitted viruses, may also contribute to the pathogenesis of these diverse cancers. The role that HIV plays in inducing these varied cancers is unclear but may be related to defects in immunosurveillance with advancing immunosuppression, chronic B cell stimulation, deregulation of cytokine pathways, and individual host genetic susceptibility—pathogenetic mechanisms that might occur in the early stages of immune deficiency, when antiretroviral therapy is commonly not given and also may have little, if any, modulating effect

Risk factors such as tobacco or noninjection drug use may have also played a role in the increasingly higher proportion of deaths associated with coronary artery disease. However, primary HIV infection may also contribute to higher rates of baseline coronary heart disease [12]. More recently, frequent observations of fat redistribution, lipid disorders, and insulin resistance have focused attention on the potential for HAART to increase cardiovascular morbidity. These concerns are supported by findings of higher rates of coronary events in HIV-infected persons with longer duration of protease inhibitor use after adjustment for other cardiac risk factors [13]. Our data are insufficient to demonstrate an effect of HAART on cardiac deaths in the short term, but we note that the highest number of deaths occurred in 1998. Whether that single increase indicates a trend will require additional years of data to determine

Finally, we found an increasing proportionate mortality associated with pancreatitis and, to a lesser extent, cardiomyopathy. We were unable to control for alcohol consumption, which may have contributed to these deaths, but the trend toward increasing standardized mortality ratios after 1996 is intriguing. Dilated cardiomyopathy in the setting of HIV infection associated with chronic zidovudine use has been observed in children [14]. An association with long-term didanosine and lamivudine use and pancreatitis is well described [15], and hypertriglyceridemia, an adverse effect of protease inhibitors, is a strong risk factor for pancreatitis. Because of the lack of reported case series, further studies with longer follow-up are required before firm conclusions can be drawn about these potential life-threatening complications of HAART

Our study has several limitations. The quality of our data depends on how accurately and thoroughly the causes of death were reported on death certificates. Underreporting of deaths due to HIV- or AIDS-related causes (e.g., stemming from concerns for patient confidentiality) or overreporting of deaths due to HIV- or AIDS-related causes (e.g., when the cause of death is stated to be end-stage AIDS, a broad, nonspecific descriptor that includes a mix of difficult-to-diagnose or unidentified disease processes) can occur. Our data are limited to a population in which the majority of HIV infections are sexually acquired, and, thus, we may have not detected mortality trends that occur predominantly among injection drug users. By including only persons with AIDS in our analysis, we may have missed important mortality trends among persons at earlier stages of HIV infection. Finally, our study follows patients only through 1998, which may not allow enough time to fully assess mortality trends resulting from the widespread use of HAART

In conclusion, although the mortality rate due to many HIV- or AIDS-related conditions has declined, the mortality rate attributed to non–AIDS-defining malignancies, chronic diseases, and complications of injection drug use has remained stable, although these causes now make up a higher proportion of overall deaths among persons with AIDS. The small but increasing trend for deaths due to pancreatitis suggest that adverse effects from HAART may be contributing to increased mortality