| |
Patterns of the hazard of death after AIDS through the evolution of ART: 1984-2004
|
| |
| |
AIDS: Volume 19(17) 18 November 2005
The purpose of this analysis was to characterize the effect of introduction and prolonged use of HAART on the changing pattern of survival after clinical AIDS among participants enrolled in the Multicenter AIDS Cohort Study (MACS) and the Women's Interagency HIV Study (WIHS), two cohort studies among men and women in the United States that use similar methods and the same data-coordinating center.
The study focused on the important subgroup of individuals with incident clinical AIDS as these patients are at greater risk for disease progression in both the natural history setting and while taking HAART. The individuals in this group may be at higher risk for the emergence of viral resistance because they are more likely to have received antiviral therapies prior to HAART initiation, have the highest rates of viral replication, and have significant comorbidity.
ABSTRACT
Methods: A total of 1504 men and 461 women were followed for all-cause mortality after an incident AIDS diagnosis. Relative hazards of death and relative times to death were determined in five therapy eras: no/monotherapy (July 1984-December 1989), monotherapy/combination therapy (January 1990-December 1994), HAART introduction (January 1995-June 1998), short-term stable HAART use (July 1998-June 2001), and moderate-term stable HAART use (July 2001-December 2003).
Results:
A total of 1057 (54%) study participants died. The time at which 25% of individuals died after an AIDS diagnosis increased significantly from 0.56 years [95% confidence interval (CI), 0.50-0.64] in the no/monotherapy era to 0.74 (95% CI, 0.67-0.82), 1.78 (95% CI, 1.29-2.44), 4.22 (95% CI, 2.94-6.05) and 5.08 years (95% CI, 2.39-10.79) in the four subsequent therapy eras, respectively.
Inferences on the beneficial effects of HAART were confirmed after adjustment by age, sex, type of AIDS diagnosis and CD4 cell count at diagnosis. The pattern of the hazard of death after AIDS changed from increasing in the pre-HAART era to being lower and non-increasing in the eras of HAART.
Conclusions: The sustained beneficial effect of HAART, even in individuals with clinical AIDS and extensive treatment histories, attenuates concerns about emergence of resistance but augurs that a substantial number of HIV-infected individuals may require care for very long periods.
Discussion
Showing the beneficial effect of HAART in individuals who have progressed to clinical AIDS is reassuring because many were treated in a stepwise fashion (Table 2), which is the ideal scenario for emergence of resistance. Benefits of HAART are expected to be even better in individuals who start therapy with more potent and tolerable combinations.
Deaths
1984-1989: person-years 685; deaths (no. (% person-years) 388 (57%); relative hazard 1; relative time: 1
1990-1994: person-years 912; deaths 445 (49%); relative hazard 0.65; relative time 1.42 yrs
1995-1998: person-years 796; deaths 109 (14%); relative hazard 0.21; relative time 3.57 yrs
1998-2001: person-years 1,156; deaths 71 (6%); relative hazard 0.08; relative time 7.82 yrs
2001-2003: person-years 992; deaths 44 (4%); relative hazard 0.06; relative time 10.65 yrs
Substantial differences in survival were seen between the last two HAART eras and the no/monotherapy era, with the time from AIDS to death (adjusted by age, sex, type of AIDS diagnosis and CD4 cell count at diagnosis) expanded by factors close to 8 and 11 in the periods July 1998 to June 2001 and July 2001 to December 2003, respectively.
Although the cumulative mortality rates after AIDS were lower in the monotherapy/combination therapy era compared with the no/monotherapy era, large declines in mortality rates were not observed until the three therapy eras during which HAART was primarily used. Mortality was lowest in the last therapy era (July 2001 to December 2003) where we estimate that a cumulative 30% (95% confidence interval, 22-38%) of individuals die 8 years after their initial AIDS diagnosis.
The hazards of death in the two earliest therapy eras increased monotonically over time. In contrast, the hazards of death in all three HAART eras decreased over time, with the greatest decline occurring in the first 6 months after AIDS. The hazards of death in the last two HAART eras were very low, with a slight and non-significant (P = 0.23) decrease between the short-term stable HAART era and the moderate-term stable HAART era.
The times at which 25% of participants died after AIDS (i.e., the first quartiles) in each of the therapy eras are depicted with closed triangles in Fig. 2. Relative to the no/monotherapy era where the first quartile was close to half of a year, the survival times were extended in the stable HAART eras with the first quartiles being close to 4 and 5 years, respectively. Furthermore, if the conditions of the latest era were to remain, the model predicts that 50% of the persons who had AIDS will survive more than 16.0 years after their initial AIDS diagnosis, compared with the observed median of 1.2 years in the no/monotherapy era.
Author discussion:
Many of the MACS and WIHS participants initiated therapy in a stepwise fashion, as they first received monotherapy consisting of one nucleoside reverse transcriptase inhibitor, then combination therapy (multiple nucleoside reverse transcriptase inhibitors), and finally HAART, a pattern that is more likely to foster virological resistance. As expected, poor adherence and the inability to sustain long-term use has resulted in resistance to HAART, the development of virological failure, and the lack of immune reconstitution. Toxicity to HAART has been well described, and liver failure is the most common cause of non-AIDS death in the WIHS cohort. Nevertheless, our data demonstrated a nearly 11-fold increase in survival after AIDS diagnosis despite the difficulties with adherence, the development of resistance, the high rate of regimen switching, and adverse events. These improvements in survival were sustained, if not improved, in the last two calendar periods. In spite of these reassuring results, cohort studies must continue to monitor individuals using HAART, since long-term use of the therapies may result in new and deleterious toxicities.
Authors:
Schneider, Michael Fa; Gange, Stephen Ja; Williams, Carolyn Mb; Anastos, Kathrync; Greenblatt, Ruth Md; Kingsley, Lawrencee; Detels, Rogerf; Munoz, Alvaroa
From the aDepartment of Epidemiology, Johns Hopkins Bloomberg School of Public Health
bDivision of AIDS, National Institute of Allergy and Infectious Diseases, National Institutes of Health, Bethesda, Maryland
cDepartments of Medicine and Epidemiology and Population Health, Montefiore Medical Center, Bronx, New York
dDepartments of Medicine and Epidemiology, University of California, San Francisco, California
eGraduate School of Public Health, University of Pittsburgh, Pittsburgh, Pennsylvania
fDepartment of Epidemiology, University of California Los Angeles School of Public Health, Los Angeles, California, USA.
DISCUSSION
Concern that the success of HAART will wane is based on the common occurrence of treatment non-adherence and discontinuation, the frequency by which viral resistance emerges, and the development of clinically significant toxicities. Adherence to therapy by MACS and WIHS participants has been shown to be heterogeneous [31,32]. Here we have shown that adherence to therapy was not 100%, with more than 20% of person-visits reporting less than 95% adherence.
Many of the MACS and WIHS participants initiated therapy in a stepwise fashion, as they first received monotherapy consisting of one nucleoside reverse transcriptase inhibitor, then combination therapy (multiple nucleoside reverse transcriptase inhibitors), and finally HAART, a pattern that is more likely to foster virological resistance. As expected, poor adherence and the inability to sustain long-term use has resulted in resistance to HAART, the development of virological failure, and the lack of immune reconstitution [33-35]. Toxicity to HAART has been well described, and liver failure is the most common cause of non-AIDS death in the WIHS cohort [24]. Nevertheless, our data demonstrated a nearly 11-fold increase in survival after AIDS diagnosis (Table 3) despite the difficulties with adherence (Table 2), the development of resistance, the high rate of regimen switching, and adverse events. These improvements in survival were sustained, if not improved, in the last two calendar periods (Fig. 2a). In spite of these reassuring results, cohort studies must continue to monitor individuals using HAART, since long-term use of the therapies may result in new and deleterious toxicities.
The high effectiveness of HAART shown here was based on considering all causes of death as the endpoint of interest. HAART has not only extended survival times after AIDS but also decreased the percentages of deaths caused by AIDS, with the expected rise of non-AIDS deaths (i.e., competing risks) resulting from increasing age. In our multivariate analysis, we included age at AIDS diagnosis, thus providing age-specific mortality. Our results are conservative; that is, restricting the analysis only to deaths from AIDS would result in better survival in the HAART era than those reported in Table 3.
Not only did the Weibull regression models fit the data (i.e., Kaplan-Meier curves) well and provide the estimated factor by which survival times after AIDS were expanded in the periods in which HAART was used, but they also proved to be useful for showing that HAART has changed the pattern of the hazard of death after AIDS. In the first two therapy eras, longer survival after AIDS was associated with a higher hazard of death. In contrast, the hazard of death decreased over time during the HAART era, indicating a lower risk of death among treated individuals surviving longer with AIDS. This is consistent with increasing immunosuppression in the early eras when there were no effective therapies, whereas the CD4 cell count of treated individuals was higher in the later eras owing to the reconstitution made feasible by HAART (Tables 2 and 3). However, only long-term studies of this and other similar cohorts will provide data to assess whether the hazard of death may start increasing after some time (i.e., a bath-tub-shaped curve).
In order to determine whether the shape of the hazard of death was different in the two cohorts in the eras of stable HAART, we tested the possible heterogeneity by gender of the Weibull model for the period July 1998 to December 2003. We found that there was no difference in the two cohorts for that period (ƒÔ2 = 2.1; degrees of freedom = 2; P = 0.35). This similarity, in spite of differential use of HAART (Fig. 1), reflects the influence of other concurrent factors, including duration of infection, prior exposure to antiretroviral therapy, and type of AIDS diagnosis (Table 1). These issues underscore the importance of the adjustments made in the multivariate models presented in Table 3.
The use of survival analysis methods that incorporated staggered entries allowed us to capitalize on the extensive follow-up of the MACS and the WIHS. The use of staggered entry analysis methods avoids the survival bias present in analyses that inappropriately include individuals who survive from previous periods as if they were at risk for death the entire time in the period of interest. Using these methods, we repeated the analysis allowing MACS participants in the first two eras to contribute to subsequent periods (i.e., a full period analysis). From this full period analysis, the RH values of the last four periods relative to the first (0.85, 0.33, 0.09, 0.08, respctively) were similar to those shown in Table 3 and yielded the same inferences.
Our results have important public-health implications. Population effectiveness analyses are ecological in the sense that they reflect many factors, including access, practice style, efficacy, adherence, discontinuation, regimen switching, and the occurrence of viral resistance to therapy components. They represent the net reduction on disease burden at the population level. Cohort studies offer an opportunity to quantify the effect of interventions that more closely represent the effect in the whole population than those reported in clinical trials. However, cohort studies may not fully represent the population at large. Previously, we have used similar methods for the assessment of effectiveness of therapies on the time from HIV infection to AIDS and death [13]. In that context, our findings have been confirmed by several other cohort studies [18,36,37]. Replications of our findings on survival after AIDS in other cohort studies will add to the generalization of our results.
A strength of our analysis is the inclusion of only incident AIDS cases, allowing us to anchor the analysis at the date of AIDS diagnosis. Many of the individuals never received therapy and those who did received it at different times and for different reasons. An advantage of the period analysis presented here is that in a given period there is a vast heterogeneity of therapy usage at the individual level. Therefore, our analysis allows for the dynamics of therapy usage by HIV-infected individuals in the population and is not subject to the biases of patients seeking care at a late stage of disease progression.
The prognosis of patients who are currently alive and being treated with HAART, including factors for treatment failure, have been widely studied [10-12] and complement the findings presented here. Our findings project that the demand for HIV care in areas where therapies are available will increase over time.
Study population
The MACS was initiated in 1983 to study the natural history of HIV-1 infection among homosexual and bisexual men in the United States. The study design and characteristics of the participants, including race and socio-economic status, have been described previously [22]. A total of 2788 participants were either HIV positive at enrollment (79%) or were documented to acquire HIV infection during follow-up (21%). The WIHS is a multicenter prospective cohort study of the natural history of HIV-1 infection in women. Methods and baseline cohort characteristics have been described previously [23]. A total of 2074 participants were either HIV positive at enrollment (99%) or were documented to have acquired HIV infection during follow-up (1%). Studies were approved by the Committees of Human Research of participating institutions.
Outcome variable
The origin for survival times was the diagnosis of an incident AIDS-defining condition. At each semi-annual visit, participants were asked to report dates of AIDS diagnoses. When the precise date was not reported, the date of AIDS diagnosis was defined by the midpoint between the last AIDS-free date and the first date with AIDS. Individuals who missed visits in between these two dates, and where the resulting lag time was more than 1 year, were not included in the analysis. Participant deaths were ascertained continuously in both cohorts using a combination of active (e.g., death certificate abstraction upon notification of a participant death) and passive (e.g., national death registry searches) surveillance methods. Detailed evaluations of the causes of death in the WIHS have been published previously [24]. Here, changes in all-cause mortality after an incident AIDS diagnosis are reported. AIDS was defined as a clinical illness consistent with 1993 Centers for Disease Control clinical surveillance conditions [25] but excluding the immunological-based criterion of low CD4 cell count. Analyses incorporated data through December 2003 with participants alive at the end of follow-up contributing censored observations.
Exposure variables
Distinct calendar periods that serve as proxies for different therapy eras were the primary exposures and were used to measure the effectiveness of antiretroviral therapy and HAART at the population level. HAART, combination therapy, and monotherapy have been described previously for the MACS and WIHS cohorts [26].
The five calendar periods used were July 1984 through December 1989, January 1990 through December 1994, January 1995 through June 1998, July 1998 through June 2001, and July 2001 through December 2003.
The impacts of age, sex, type of AIDS diagnosis, and CD4 cell count at diagnosis on survival after AIDS were evaluated. Clinical AIDS diagnoses were classified into seven categories as described by Munoz et al. [27].
|
|
| |
| |
|
|
|
