We characterized factors and temporal trends associated with discontinuation of highly active antiretroviral therapy (HAART) among 936 HIV-infected women enrolled in the Women's Interagency HIV Study. A multivariate analysis of post-HAART initiation exposures found that high HIV RNA levels (relative hazard [RH] = 1.36, P < 0.001) and high depressive symptom scores (RH = 1.53, P = 0.012) were associated with HAART discontinuation. The adjusted hazard of discontinuation was higher in the 2 most recent calendar periods compared with the first (RH = 1.61, P = 0.026; RH = 1.56, P = 0.074, respectively). The increasing risk of HAART discontinuation in recent calendar periods and changes in the clinical factors associated with discontinuation reflect ongoing and dynamic shifts in the approach to HAART utilization.
From the *Department of Epidemiology, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD; †University of Texas School of Public Health, Houston, TX; ‡Montefiore Medical Center, Bronx, NY; §Cook County Hospital, Chicago, IL; ∥University of Southern California School of Medicine, Los Angeles, CA; ¶Maimonides Medical Center, Brooklyn, NY; #Georgetown University Medical Center, Washington, DC; and **Departments of Medicine and Epidemiology, University of California, San Francisco, CA.
Received for publication December 6, 2002; accepted June 23, 2004.
The primary authors of this article are Linda Ahdieh-Grant and Patrick M. Tarwater.
The Women's Interagency HIV Study is funded by the National Institute of Allergy and Infectious Diseases, with supplemental funding from the National Cancer Institute, the National Institute of Child Health and Human Development, the National Institute on Drug Abuse, and the National Institute of Craniofacial and Dental Research (U01-AI-35004, U01-AI-31834, U01-AI-34994, U01-AI-34989, U01-HD-32632, U01-AI-34993, U01-AI-42590, M01-RR00079, and M01-RR00083).
Reprints: Linda Ahdieh-Grant, Department of Epidemiology, Johns Hopkins Bloomberg School of Public Health, 615 North Wolfe Street, Room E-7650, Baltimore, MD 21205 (e-mail: email@example.com).
Since its introduction in late 1995, the success of highly active antiretroviral therapy (HAART) for extending AIDS-free survival and lowering AIDS-related mortality rates has been well-documented1-5 and widely acknowledged. In recent years, the challenges of adherence to complex regimens and failures of sustaining immunologic and virologic responses have come to the forefront of discussions regarding antiretro-viral treatment, and it is now generally recognized that long-term adherence to antiretroviral therapies (ARTs) is closely related to the occurrence of adverse effects. The risks of potentially severe adverse effects influence decisions regarding initiation of HAART in asymptomatic patients and regimen switching among patients on therapy. Adverse effects may also influence decisions to discontinue HAART.
We have previously shown that the prevalence of HAART discontinuation increased from 3% in late 1996 to 9% in mid-1999 in the Women's Interagency HIV Study (WIHS).6 To characterize temporal trends in discontinuation more appropriately, the current report measures the incidence of HAART discontinuation across calendar periods while accounting for the duration of HAART use after initiation. Furthermore, we describe the extent to which immunologic, virologic, and sociodemographic exposures, treated as time-varying factors, are associated with the risk of HAART discontinuation.
The study was based in the WIHS, a multisite cohort study of the natural history of HIV infection in women. Participants were recruited from October 1994 through November 1995, during which time 2059 HIV-seropositive and 569 HIV-seronegative women were enrolled at 6 sites. WIHS participants were interviewed every 6 months after an initial baseline visit. Blood specimens were collected for ascertainment of CD4 cell counts and plasma HIV RNA levels. At each WIHS visit, participants were shown photograph medication cards and were asked the names of antiretroviral medications used since their previous visit. More detailed information regarding the WIHS study design and data collection has been previously reported.7
The definition of HAART was guided by the International AIDS Society-USA (IAS-USA) Panel8 and the Department of Health and Human Services (DHHS)/Kaiser Panel9 guidelines and is as follows: (1) 2 or more nucleoside reverse transcriptase inhibitors (NRTIs) in combination with at least 1 protease inhibitor (PI) or 1 nonnucleoside reverse transcriptase inhibitor (NNRTI), (2) 1 NRTI in combination with at least 1 PI and at least 1 NNRTI, (3) a regimen containing ritonavir and saquinavir in combination with 1 NRTI and no NNRTIs, and (4) an abacavir-containing regimen of 3 or more NRTIs (in the absence of PIs and NNRTIs). Combinations of zidovudine (AZT) and stavudine (d4T) with either a PI or NNRTI are counterindicated and were not considered HAART.
The study population for this analysis was restricted to WIHS participants who were HIV-seropositive at enrollment (n = 2059); had initiated HAART between October 1, 1995 and September 30, 2000 (n = 1167); had less than 1 year's difference between their last HAART-free and first reported HAART visits (n = 1089); and had more than 6 months of any ART use (n = 936) after HAART initiation. The date of HAART initiation was defined as the midpoint between the date of the last known HAART-free study visit and the date of the first visit in which HAART was reported. HAART discontinuation was determined from the first report of no ART use since the last WIHS visit. Downshifting from HAART to combination or monotherapy did not constitute HAART discontinuation. Discontinuation events were limited to the first report of complete long-term (ie, at least 6 months) discontinuation of all ART use.
The primary purpose of the current analysis was to identify temporal trends in discontinuation by comparing the hazard of HAART discontinuation in 3 consecutive periods in the era during which the benefits of HAART have been clearly established. Period 1 was defined as October 1, 1995 to March 31, 1998; period 2 was defined as April 1, 1998 to June 30, 1999; and period 3 was defined as July 1, 1999 to September 30, 2000. Periods were chosen to create an approximately equal number of “at-risk” person-years in each period. The analysis treated calendar period as an external time-varying exposure and incorporated staggered entries to compare the incidence of HAART discontinuation among individuals with equal durations of HAART use. A more detailed description of our analytic methods has previously been presented.5
Factors that were examined for association with HAART discontinuation included ethnicity, pre-HAART age, pre-HAART CD4 cell count, pre-HAART HIV RNA, pre-HAART antiretroviral experience, pre-HAART depressive symptom status, pre-HAART self-reported health status, pre-HAART report of injection drug use, pre-HAART AIDS, pre-HAART pregnancy status, complexity of the initial HAART regimen (eg, drug classes and number of drugs), and initial immunologic and virologic changes. Factors with univariate P values less than or equal to 0.100 in any of the 3 calendar periods were included in the multivariate models for each period. Depressive symptom scores were assessed using the Center for Epidemiologic Studies Depression (CESD) scale.10 Scores ranged from 0 to 60, and depressive symptom status was classified as high for scores ≥16. Self-reported health was scored on a scale of 0 to 100, and individuals with scores ≥50 were classified as having “high self-rated health.”
Finally, to investigate factors for which changes over time may be of greater relevance than initial values before HAART initiation (ie, CD4 cell count, HIV RNA level, depressive symptoms, self-reported health, and injection drug use), we analyzed these factors as time-varying covariates without adjusting for calendar period.
Of 936 women enrolled in the WIHS who initiated HAART between October 1, 1995 and September 30, 2000 and had at least 6 months of ART use, 220 (23.5%) persons discontinued all ART during 5 years of follow-up. The median age of the study population at the time of HAART initiation was approximately 38 years. Compared with those who remained on HAART, women who discontinued were more likely to be Latina (35% vs. 24%) and less likely to be white (10% vs. 21%) (P < 0.001). HAART discontinuers and non-discontinuers had similar median CD4 cell counts at the visit immediately before HAART initiation, and the slightly lower percentage of HAART discontinuers who had undetectable HIV RNA at the visit before HAART initiation was not statistically significant (P = 0.245). WIHS participants who initiated HAART were heavily pretreated, although there was no difference in previous ART experience between those who did and those who did not discontinue HAART. Discontinuers were more likely to have had AIDS (P = 0.032). Discontinuers also had a significantly (P = 0.010) lower median increase in CD4 cell count from pre-HAART initiation to the first on-HAART visit of 24 cells/μL, nearly 50% lower than the median increase for those who continued on HAART. Likewise, 28% of the discontinuers' HIV RNA levels remained or became undetectable compared with 38% of those who stayed on HAART (P = 0.011).
Table 1 presents descriptive statistics for individuals who contributed time at risk in at least 1 calendar period. Compared with the first period, there was an increased hazard of discontinuation in the later 2 periods (ie, period 2: relative hazard [RH] =1.61, P = 0.026; period 3: RH = 1.56, P = 0.074) after adjustment for factors univariately associated (P < 0.100) with HAART discontinuation in at least 1 of the 3 calendar periods.
In addition, using a multiple variable Cox proportional hazards model that allowed variables to act as time-varying exposures, high HIV RNA levels (RH = 1.36, P < 0.001) and depressive symptoms (RH = 1.53, P = 0.012) were significantly associated with an increased risk of HAART discontinuation.
Furthermore, in calendar-stratified analyses, we found that relative to white ethnicity, African-American (RH = 4.11) or Latina ethnicity (RH = 4.32) and injection drug use (RH = 4.15) were significantly associated with HAART discontinuation in the first calendar period and that ethnicity (RH = 2.86 for African-American ethnicity and RH = 5.00 for Latina ethnicity) and higher pre-HAART HIV RNA levels (RH = 1.35) were significantly associated with HAART discontinuation in the second calendar period. In the last calendar period, no factors were significantly associated with HAART discontinuation in the multivariate analysis.
Given the major public health gains associated with HAART for reducing the risk of AIDS and AIDS-associated mortality, increasing emphasis is being accorded to the long-term effects of these antiretroviral regimens. Given the recognition of the likelihood of discontinuation in some instances, we thought that an additional investigation into the factors and temporal trends associated with HAART discontinuation would be of interest. Using longitudinal data that were prospectively collected through September 2000 from the WIHS, we found that nearly 25% of individuals discontinued HAART for at least 6 months during study follow-up. After controlling for time since initiation of HAART and associated factors, we found that the risk of discontinuation was over 1.5 times greater after 1998 as compared with the prior 2.5 years, suggestive that discontinuation has become more acceptable as the burden of frequently associated side effects has become more recognized. Furthermore, Latina and African-American women were more likely than white women to discontinue HAART, an association that was significant in the initial 2 calendar periods. Given that the distribution of ethnicity did not change between the calendar periods, this observation is of concern and warrants additional investigation.
Consistent with the current findings, recent studies by Mocroft et al11 and Monforte et al12 found that individuals with higher recent HIV RNA levels were significantly more likely to discontinue HAART as compared with those with lower HIV RNA levels. This association between high HIV RNA levels and the discontinuation of HAART suggests that individuals discontinue regimens that are not successfully suppressing HIV RNA. It will be of interest for future studies to examine whether the most recently published guidelines9 result in a U- or J-shaped pattern of HAART discontinuation, whereby individuals with extremely low or extremely high HIV RNA levels discontinue HAART. Mocroft et al11 also found that women and older individuals were less likely to discontinue HAART, as were individuals who used nelfinavir-containing regimens. We did not find any significant association between HAART discontinuation and the number of drugs or the drug classes in the initial HAART regimen. A recent analysis by Dorrucci et al13 similarly found no difference in the time to discontinuation by specific type of regimen. The importance of psychosocial factors in modulating the risk of discontinuation is not unexpected, and our results suggest that factors other than response to treatment play an important role in predicting HAART discontinuation. Prior studies have shown that adherence is associated with depressive symptoms,14,15 and it is therefore possible that individuals became depressed because their lack of adherence led to therapeutic failure and elevated HIV RNA levels. Furthermore, given that depressive symptoms can be evaluated and depression can be treated, our findings emphasize that access to treatment of depression may have important implications for the management of HIV-infected individuals on ART.
There are several notable strengths of the current study. First, it involved a large number of individuals who initiated HAART and were followed for up to 5 years. Information on ART use was updated every 6 months based on self-report; the validity of such self-reported data has been previously demonstrated.14 Second, although several previous studies examined the risk of discontinuation,11-13 only 1 distinguished between discontinuation of all ART as compared with the modification of certain components of the HAART regimen.11 Consistent with our findings, this study found that 26% of 556 participants discontinued HAART. Third, the cohort study on which this analysis was based collects data on a prospective longitudinal basis as compared with previous studies that have relied on retrospective chart review.
Beyond these strengths, our analysis contributes to the literature on HAART discontinuation by applying methods that allow for the incorporation of the duration of HAART use. Although the use of calendar period as a time-dependent external variable has been established for the purpose of estimating population effectiveness,3,5 we have applied the methodology to evaluate temporal trends in HAART discontinuation among women with equal durations of HAART use.
In summary, our understanding of the long-term clinical implications of ART is greatly facilitated by the existence of large cohort studies in which the date of HAART initiation is well defined. Studies with ongoing follow-up of HIV-infected individuals allow for the investigation of landmark events in the treated history of HIV infection, including HAART initiation, switching between antiretroviral regimens, treatment response and failure, and HAART discontinuation. Furthermore, such studies are essential to track changes in the anticipated increase in rates of those toxicities associated with the prolonged use of HAART.
Data in this manuscript were collected by the WIHS Collaborative Study Group with centers (Principal Investigators) at the New York City/Bronx Consortium (Kathryn Anastos); Brooklyn, NY (Howard Minkoff); the Washington DC Metropolitan Consortium (Mary Young); The Connie Wofsy Study Consortium of Northern California (Ruth Greenblatt, Phyllis Tien); the Los Angeles County/Southern California Consortium (Alexandra Levine); the Chicago Consortium (Mardge Cohen); and the Data Coordinating Center (Alvaro Muñoz, Stephen Gange).
1. Schapiro JM, Winters MA, Efron SF, et al. The effect of high-dose saquinavir on viral load and CD4+ T-cell counts in HIV infected patients. Ann Intern Med
2. Hammer SM, Squires KE, Hughes MD, et al. A controlled trial of two nucleoside analogues plus indinavir in persons with human immunodefi-ciency virus infection and CD4 cell counts of 200 per cubic millimeter or less. N Engl J Med
3. Detels R, Muñoz A, McFarlane G, et al. Effectiveness of potent antiret-roviral therapy on times to AIDS and death in men with known HIV infection duration. JAMA
4. Gange SJ, Barrón Y, Greenblatt R, et al. Effectiveness of highly active antiretroviral therapy among HIV-1 infected women. J Epidemiol Community Health
5. Tarwater PM, Mellors J, Gore ME, et al. Methods to assess population effectiveness of therapies in HIV incident and prevalent cohorts. Am J Epidemiol
6. Grant LA, Silverberg MJ, Palacio H, et al. Discontinuation of potent an-tiretroviral therapy: predictive value of and impact on CD4 cell counts and HIV RNA levels. AIDS
7. Barkan SE, Melnick SL, Preston-Martin S, et al. The Women's Inter-agency HIV Study. Epidemiology
8. Carpenter CCJ, Cooper DA, Fischl MA, et al. Antiretroviral therapy in adults. Updated recommendations of the International AIDS Society-USA Panel. JAMA
9. DHHS/Henry J. Kaiser Family Foundation Panel on Clinical Practices for the Treatment of HIV Infection. Guidelines for the use of antiretroviral agents in HIV-infected adults and adolescents. 2/2001 version. Available at: http://hivatis.org
10. Bozzette SA, Hays RD, Berry SH, et al. Derivation and properties of a brief health status assessment instrument for use in HIV disease. J Acquir Immune Defic Syndr Hum Retrovirol
11. Mocroft A, Youle M, Moore A, et al. Reasons for modification and discontinuation of antiretrovirals: results from a single treatment center. AIDS
12. Monforte A, Lepri AC, Rezza G, et al. Insights into the reasons for discontinuation of the first highly active antiretroviral therapy (HAART) regimen in a cohort of antiretroviral naive patients. AIDS
13. Dorrucci M, Pezzotti P, Grisorio B, et al. Time to discontinuation of the first highly active antiretroviral therapy regimen: a comparison between protease inhibitor- and non-nucleoside reverse transcriptase inhibitor-containing regimens. AIDS
14. Kleeberger CA, Phair JP, Strathdee SA, et al. Determinants of heterogeneous adherence to HIV-antiretroviral therapies in the Multicenter AIDS Cohort Study. J Acquir Immune Defic Syndr
15. Wagner JH, Justice AC, Chesney M, et al. Patient- and provider-reported adherence: toward a clinically useful approach to measuring antiretroviral adherence. J Clin Epidemiol
. 2001;54(Suppl 1):S91-S98.