Rates and Reasons for Early Change of First HAART in HIV-1-Infected Patients in 7 Sites throughout the Caribbean and Latin America

Background HAART rollout in Latin America and the Caribbean has increased from approximately 210,000 in 2003 to 390,000 patients in 2007, covering 62% (51%–70%) of eligible patients, with considerable variation among countries. No multi-cohort study has examined rates of and reasons for change of initial HAART in this region. Methodology Antiretroviral-naïve patients > = 18 years who started HAART between 1996 and 2007 and had at least one follow-up visit from sites in Argentina, Brazil, Chile, Haiti, Honduras, Mexico and Peru were included. Time from HAART initiation to change (stopping or switching any antiretrovirals) was estimated using Kaplan-Meier techniques. Cox proportional hazards modeled the associations between change and demographics, initial regimen, baseline CD4 count, and clinical stage. Principal Findings Of 5026 HIV-infected patients, 35% were female, median age at HAART initiation was 37 years (interquartile range [IQR], 31–44), and median CD4 count was 105 cells/uL (IQR, 38–200). Estimated probabilities of changing within 3 months and one year of HAART initiation were 16% (95% confidence interval (CI) 15–17%) and 28% (95% CI 27–29%), respectively. Efavirenz-based regimens and no clinical AIDS at HAART initiation were associated with lower risk of change (hazard ratio (HR) = 1.7 (95% CI 1.1–2.6) and 2.1 (95% CI 1.7–2.5) comparing neverapine-based regimens and other regimens to efavirenz, respectively; HR = 1.3 (95% CI 1.1–1.5) for clinical AIDS at HAART initiation). The primary reason for change among HAART initiators were adverse events (14%), death (5.7%) and failure (1.3%) with specific toxicities varying among sites. After change, most patients remained in first line regimens. Conclusions Adverse events were the leading cause for changing initial HAART. Predictors for change due to any reason were AIDS at baseline and the use of a non-efavirenz containing regimen. Differences between participant sites were observed and require further investigation.


Introduction
An estimated 1.93 million people live with HIV in Latin America and the Caribbean, comprising 5.7% of all infected persons worldwide; the adult prevalence in this region is 0.5% [1]. Access to antiretroviral (ARV) therapy has improved and at the end of 2007 approximately 390,000 patients in this region were receiving antiretroviral therapy with an overall coverage of 62% (51%-70%), although considerable variation exists between countries [2,3]. Unfortunately, 75% of patients still initiate treatment at advanced stages of disease [4][5][6][7][8].
Treatment toxicities and adherence problems may lead to suboptimal therapy, discontinuation, and treatment failure. Early modification of initial highly active antiretroviral therapy (HAART) has been associated with poor clinical outcomes [9]. Therefore, knowing why patients modify therapy could improve our understanding of successful HAART, guide decisions regarding initiation and management of HAART in specific patient populations, and inform interventions to reduce HAART discontinuation.
The frequency and reasons for HAART change have been assessed by cohort studies from resource-rich and -limited settings, but Latin America and the Caribbean have been largely underrepresented in these studies [10][11][12][13][14][15][16]. Observational studies from sites in Argentina, Brazil, Haiti and Peru have described the occurrence of adverse events and durability of first regimen [17][18][19][20][21][22][23]. However, no multisite study has addressed frequency and reasons for change in this region.
The Caribbean, Central and South America Network for HIV Research (CCASAnet) collaboration includes sites from seven nations: Argentina, Brazil, Chile, Haiti, Honduras, Mexico, and Peru. In an earlier study of antiretroviral-naïve subjects starting HAART, mortality rates in the CCASAnet cohort were similar to those reported for resource-limited settings with a 1-year probability of death for the combined cohort of 8.3%, although this varied considerably across sites [8]. The purpose of the current study is to explore the frequency of, risk factors for, and reasons for changing/discontinuing HAART during the first year after initiation in the CCASAnet region.

Ethics Statement
This study was conducted according to the principles expressed in the Declaration of Helsinki. Institutional Review Board approval was obtained locally for each participating site and the coordinating centre: Comité  All data were de-identified prior to being transmitted to the Vanderbilt Data Coordinating Centre.
In each of the countries contributing data to this study, ethical regulations and policies permit retrospective analysis of deidentified clinical data without informed consent when the research is approved by an appropriately constituted ethics committee or Institutional Review Board. These approvals were obtained in all cases and the need to obtain informed consent was waived by all of the ethics committees of the participating sites.

Participants and Settings
The CCASAnet cohort (www.ccasanet.vanderbilt.edu) has been described elsewhere [24] Data audits were performed at each site by a team from the VDCC. The present analysis used data for the first year of followup after starting HAART collected through June 2008. Included were antiretroviral-naïve HIV-infected patients prescribed HAART at age 18 years or older with at least one follow-up visit.
Initiation of HAART at each site followed either national or World Health Organization guidelines [25][26][27][28][29][30]. Guidelines from Argentina, Brazil, Chile, Honduras and Mexico recommend drug substitutions after toxicity and switching regimens after virologic failure [25,[27][28][29]. In contrast, in Haiti and Peru failure was defined according to WHO clinical and immunologic criteria [30]. Table 1 lists site-specific practices related to initiation criteria, laboratory monitoring, and regimen availability.

Outcomes
The primary outcome was first change of regimen during the first year of HAART. Regimen change was defined as any alteration-switch or discontinuation-of $1 antiretroviral. Discontinuation was defined as simultaneous stopping of all antiretrovirals without initiation of a subsequent regimen for more than 30 days. Dosage adjustments and interruptions of therapy shorter than 30 days were ignored because of inconsistent recording of short interruptions across sites. Reasons for change were collected by each site and classified at the coordinating centre. Specific definitions of reasons for regimen change, including definition of treatment failure, were not standardized across sites and only included if they prompted a regimen change. Secondary analyses classified patients who died or were lost to follow-up (LTFU) while on their first HAART as having discontinued treatment. Patients without a visit for 6 months were classified as LTFU. The 6-month interval was chosen to include the longest interval between regular visits in participant sites, although most sites scheduled visits every 3 months.

Data Sources and Measurements
Baseline CD4 count was defined as the measurement closest to HAART initiation but not more than 6 months prior to, or 7 days after, the date of HAART start. Baseline HIV-1 plasma viral load (PVL) was defined as the pre-HAART measurement closest to, but not more than 6 months prior to, HAART initiation. Baseline weight and hemoglobin were defined as the measurements closest to HAART initiation within +/2 30 days. HAART was defined as protease inhibitor (PI)-based (1 ritonavir-boosted or unboosted PI plus $2 nucleoside reverse-transcriptase inhibitors [NRTI]), nonnucleoside reverse transcriptase (NNRTI)-based (1 NNRTI plus $2 NRTIs), or other combinations (including triple NRTI regimens and any other regimen containing a minimum of three drugs). Clinical stage of disease was defined as AIDS (WHO stage 4, CDC stage C, or 1986 CDC stage 4), non-AIDS, or unknown.
The closing date was defined separately for each site as the date of the most recent visit recorded in the database, and ranged from March 2007 to June 2008. The relationship between time to change and baseline variables was assessed using Cox proportional hazards models applied separately for each site. The primary multivariable analyses only included baseline predictors whose hazard ratio could be computed for all sites. Secondary, sitespecific multivariable analyses included other routinely collected predictors with .50% non-missing data. In multivariable analyses, missing values of baseline predictors were accounted for using multiple imputation techniques applied separately within each site [31]. CD4 count and date of HAART initiation were included in models as continuous variables and expanded using restricted cubic splines to avoid linearity assumptions [32]. The combined hazard ratios and 95% confidence intervals (CI) were computed based on the results of site-specific hazard ratios using the metaanalysis approach of DerSimonian and Laird [33], a random effects method which makes no assumption regarding proportional hazards across sites [34]. All analyses were performed using R statistical software, version 2.8.1 (http://www.r-project.org). Analysis scripts are available at http://ccasanet.vanderbilt.edu/ files/public/switch.nw
Most regimen changes were simple drug substitutions followed by other regimen changes and discontinuations.
Of 1288 living patients whose initial HAART was changed, 1147 (89%) switched to a different regimen, 104 (8%) did not start a second regimen during the observed follow-up, and 37 (3%) re-  Clinical AIDS at HAART initiation and non-efavirenz based regimens were associated with a higher hazard of change in unadjusted analyses for most sites and in the combined cohort (data not shown). Multivariable analyses for each site and pooled across sites are given in Table 5. After adjusting for sex, age, baseline CD4 count, year of HAART initiation, and type of regimen, the hazard of change was 1.3 times higher for a person with clinical AIDS prior to HAART initiation than a person without (95% CI: 1.1 to 1.5) ( Table 3). Using EFV-based regimens as the reference category, the hazard ratios for change for NVPbased regimens and non-NNRTI-based regimens were 1.7 (95% CI: 1.1 to 2.6) and 2.1 (95% CI: 1.7 to 2.5) respectively. The increased hazard for change for NVP was not observed in FA-Chile and GHESKIO-Haiti whereas it was especially pronounced in IMTAvH-Peru. Except at GHESKIO-Haiti, patients who started 3TC,ZDV,EFV generally had lower rates of change than those starting other regimens (Table S1, online supplemental material). Overall and by site there were no consistent associations between gender, age, CD4, year of HAART initiation, or ZDVcontaining regimens and change. Results were similar when those who died and those who died or were lost to follow-up were assumed to have discontinued regimens (Tables S2-S3, online  supplemental material).
Multivariable analyses including HIV-1 RNA, hemoglobin, weight, and more refined regimen categories were performed for sites with sufficient data and are shown in Table S4 in the online supplemental material. Higher baseline weight was associated with a lower risk of changing regimens at IMTAvH-Peru. Higher baseline hemoglobin was predictive of a decreased risk of changing regimens at GHESKIO-Haiti. Patients treated with d4T had lower hemoglobin at baseline (medians of 9.7 vs. 11.0 mg/dl, p,0.0001). Higher baseline HIV-1 RNA was predictive of changing regimens at FA-Chile, but was not an independent predictor at FH-Argentina, INNSZ-Mexico, or IMTAvH-Peru.
The reported reasons for change during the first year are given in Table 6. Adverse events (AE) prompted change in 14.4% of HAART initiators, and were the most common reason for six of the seven sites. Other reasons for change were failure (1.3%), the availability of a better regimen or simplification (1.5%), drug supply problems (1.8%), and abandonment/adherence failures (1.1%). Of the patients who initiated HAART, 2.9% changed regimens for an undocumented reason.
The most common AE were hematological toxicity (6.7%), skin rash (3%) and gastrointestinal intolerance (1.9%), with substantial heterogeneity between sites. Of HAART initiators in IHSS/HE-Honduras, 3.7% changed regimens during the first year due to hematological adverse events, compared to 15.8% in IMTAvH-Peru. Among those with hematological adverse events, 73% were anemia, 4%. The distribution of HAART initiators changing due  to skin rash also varied with FH-Argentina, FA-Chile and IMTAvH-Peru reporting 4.0%, 7.3%, and 5.5%, respectively, and other sites reporting ,2%. Within the first 3 months, AE were also the most common reported reason for changing regimens. Ten percent of patients changed regimens due to adverse events: 4.7% due to hematological toxicity and 2.8% due to skin rash. Forty-two patients from GHESKIO-Haiti who were on 3TC,ABC,ZDV switched to 3TC,ZDV,EFV in April/May of 2003 because this regimen became available and was deemed superior; each of these 42 patients had been on their initial regimen for less than 3 months.

Discussion
This is the first multi-cohort study in Latin America and the Caribbean to describe rates of and reasons for changing initial HAART regimen. We found high rates of change early after treatment initiation with substantial variation across sites, ranging from 8-28% in the first 3 months and 18-41% in the first year. These rates are similar to those reported in other cohorts [10,[12][13][14][15]35].
Also in agreement with other studies [10][11][12][13][14]35,36], adverse events were the main reason for change early after HAART initiation, with significant heterogeneity in the distribution of adverse events across sites. Hematological adverse events, .70% of which were anemia, were most common. This was most frequent in IMTAvH-Peru at 67%. Previous studies from Peru also reported anemia as a main reason for discontinuation (68%), and associated this finding with the use of standard 600 mg ZDV in low weight patients [21]. ZDV use was associated with an increase risk of discontinuation in the first 120 days of therapy and this early toxicity was associated with low baseline body weight. The high rates of HAART change due to anemia in IMTAvH-Peru may also be a reflection of baseline anemia and the fact that this site closely monitors anemia and changes HAART soon after its occurrence. The distribution of change in regimen due to skin rash also varied, with high rates in FH-Argentina, FA-Chile and IMTAvH-Peru and low rates elsewhere. This could be in part related to ethnicity, although this remains controversial [37][38][39].
GHESKIO-Haiti, with more advanced disease at baseline, had unexpectedly lower rates of change/discontinuation due to adverse events than other sites. However, the rates were similar when deaths were included as discontinuations. The availability of alternative drugs more than the occurrence of adverse events may explain this low rate.
As expected and previously reported [10,12], patients were more likely to change therapy shortly after HAART initiation because of adverse events rather than treatment failure. Failure was given as the reason for change in 5% of changes, corresponding to 1.3% of HAART initiators. This low rate primarily may be explained by the short duration of follow-up.
Interruption in drug supply prompted changes in 2% of HAART initiators per site: its importance cannot be minimized since interruptions ,30 days were ignored. Continuous provision of therapy is a key component of any successful HIV program, as treatment interruptions affect program effectiveness [5,21]. Consistent with previous studies, individuals who died while on their first regimen were censored at the time of death in our primary analyses [10,13,14]. This analysis implicitly assumes that the frequency with which these individuals would have changed regimens had they continued to live is similar to the frequency of changing for those patients who remained in care. To examine the sensitivity of results to this assumption we performed additional analyses which categorized individuals who died while on their first regimen as having discontinued regimens. We also performed analyses, assuming those lost to follow-up stopped therapy. This latter assumption seems reasonable in sites where there were few other options for HIV care, but less reasonable for sites located in areas with several other points of care. Characteristics at HAART initiation of those subsequently lost to follow-up for the CCASAnet cohort have been described elsewhere [8]. Risk factors for changing/discontinuing first regimen were similar regardless of how those who died or were lost were classified.
For the combined cohort, clinical AIDS prior to HAART initiation was identified as an independent predictor for treatment change. For FH-Argentina, in contrast to other sites, there was an increased risk of change at higher CD4 counts. A previous report suggested that patients with higher CD4 counts were at higher risk of GI intolerance whereas misclassification of gastrointestinal intolerance can occur in patients with low CD4 and associated opportunistic diseases [40]. Comorbidities in patients with advanced disease and concurrent treatments for opportunistic diseases may affect antiretroviral tolerance and thereby increase risk of toxicities. Late HAART initiation was associated with higher rates of treatment change. Approximately 50% of the patients in the combined cohort started therapy at less than 100 CD4/mL, highlighting the urgent need for timely diagnosis and treatment of HIV-positive patients.
Efavirenz-based regimens had the lowest hazard for change. The increased hazard for change of NVP-based regimens was especially pronounced in IMTAvH-Peru. This may be explained by the use of fixed dose combinations containing ZDV and anemia frequency. In spite of the high proportions of hematological toxicity, the hazards of change for ZDV-and non-ZDVcontaining regimens were similar. We believe that since patients with anemia at baseline were typically assigned to non-ZDV containing regimens (primarily d4T), ZDV-treated patients were ''protected'' from subsequent change due to anemia. However, baseline anemia status was not consistently collected for all sites. We failed to identify consistent associations between gender or age and risk of antiretroviral change, although other studies have found that younger age and female gender predict change [13].
After the first change, most patients remained on regimens within the same class. The outcome of second regimens was not assessed as it was beyond the scope of this study.
Our study had several limitations. Adverse events or regimen failures were computed only if they prompted regimen change. Therefore, their frequency cannot be used to estimate their occurrence, but rather the frequency of events deemed significant enough to prompt a regimen change. Strategies for changing regimens varied throughout the region and have evolved over time. Therefore, differences between sites regarding rates of toxicities, for example, may reflect site differences in guidelines for changing regimens or availability of alternative regimens. In addition, these data were collected retrospectively, and differences between sites in reasons for change and adverse events may reflect varying levels of data capture. Results of pooled analyses should be treated with caution given the heterogeneity seen in our cohort.
We did not consider treatment modifications or interruptions shorter than 30 days because registration of such events varied across sites and shorter discontinuations were less likely to have been recorded. We recognize the potential impact of minor interruptions on treatment outcomes, particularly using NNRTIbased regimens; such interruptions occur frequently in real life but generally have not been considered in other cohort studies [14,41].
Although we controlled for key variables such as CD4 count and clinical stage, patients were not randomly assigned to their initial regimens, so rates of change may be higher for certain regimens due to baseline characteristics rather than the regimen itself. However, most of the associations observed in this study are similar to those reported elsewhere [10,[12][13][14][15][16]35,39].
In conclusion, the high rate of change due to adverse events is consistent with studies from other cohorts. Heterogeneity between sites may be explained by differences in baseline characteristics at HAART initiation, programmatic differences, demographics and population genetics. Unfortunately, with only 7-sites, we are unable to perform analyses to investigate the impact of site-specific factors on regimen change. Efavirenz-based regimens were widely used and showed a lower rate of discontinuation compared to nevirapine or PI-based treatments.