|Home | About | Journals | Submit | Contact Us | Français|
Faith-based organizations have expanded access to antiretroviral therapy (ART) in community clinics across South Africa. Loss to follow-up (LTFU), however, limits both the potential individual and population treatment benefits and is an obstacle to optimal care.
To identify patient characteristics associated with LTFU six months after starting ART in patients in a large South African community clinic.
Patients initiating ART between April 2004 and October 2006 in one Catholic Relief Services HIV treatment clinic who had at least one follow-up visit were included in the analysis. Standardized instruments were used for data collection. Routine monitoring was performed every 6 months following ART initiation. Rates of LTFU over time were estimated by the Kaplan-Meier method. The log-rank test was used to examine the impact of age, baseline CD4 count, HIV RNA, gender and pregnancy status for women on LTFU. Cox proportional hazard regression was performed to analyze hazard ratios for LTFU.
Data from 925 patients (age > 14 years), median age 36 years, 70% female (16% pregnant) were included in the analysis. Fifty one patients (6%) were lost to follow-up six months after ART initiation. When stratified by baseline CD4 count, gender and pregnancy status, pregnant women with lower baseline CD4 count (≤200 /μl) had 6.06 times (95% CI: 2.20 – 16.71) the hazard of LTFU compared to men.
HIV-infected pregnant women initiating ART are significantly more likely to be lost to follow-up in a community clinic in South Africa. Interventions to successfully retain pregnant women in care are urgently needed.
South Africa has the highest number of people living with HIV in the world . The estimated number of HIV-infected individuals is 5.7 million, and 350,000 people died from AIDS-related complications in 2007 [1, 2]. About one-third of all pregnant women in South Africa are HIV-infected [3, 4]. The consistent growth of the HIV epidemic has led to a rapid expansion of HIV care and treatment .
Many non-governmental organizations operate mentoring and support programs for HIV-infected people in resource-limited countries [5, 6]. In South Africa, faith-based organizations play a significant and expanding role in providing HIV care and access to antiretroviral therapy (ART) [7, 8]. One of the largest ART treatment programs is jointly run by the Catholic Relief Services and the Southern African Catholic Bishops’ Conference, and is funded by the South African Department of Health, as well as the US President’s Emergency Plan for AIDS Relief (PEPFAR) . The ART treatment program operated by Catholic Relief Services and the Southern African Catholic Bishops’ Conference is spread widely across South Africa, and has been involved in locally-based HIV responses in South African communities .
Evaluating the outcomes of patients initiating ART is a critical task for ART treatment programs. However, outcome evaluation is generally based only on those patients who remain in care . High rates of loss to follow-up (LTFU) hinder optimal care, and substantially limit the effectiveness of ART treatment strategies [12, 13]. Furthermore the growing body of evidence suggests heterogeneity in rates of LTFU for men and women . Pregnancy may also contribute to higher rates of suboptimal retention in care . There is thus an urgent need to collect information on LTFU in patients who initiate ART. Our objective was to identify the impact of gender and pregnancy status and CD4 cell associated with LTFU six months after ART initiation in one of the large Catholic Relief Services treatment programs in South Africa.
This study was conducted in the Tapologo, one of the Catholic Relief Services HIV treatment clinics. The Tapologo HIV/AIDS program works with mine worker communities outside the Rustenburg region in the North West province of South Africa . Tapologo currently provides home-based care, local clinical consultation and support services to all mine workers and their families through a network of caregivers . The clinic population consists of urban adults (age >14 years) within the mining community who received ART from home-based caregivers. The caregivers are supported financially by the platinum mines, and the clinic employs doctors, nurses, counselors and adherence monitors . Treatment protocols closely followed World Health Organization (WHO) and South African National Department of Health guidelines [18, 19]. Adherence education was given both before and during ART treatment.
The cohort included HIV-infected patients who either developed AIDS or had a CD4 count which met criteria for ART initiation based on South African National Department of Health guidelines (CD4 count ≤200/μl) . Patients initiating ART between January 2004 and October 2006 who were eligible for at least one follow-up visit were included in the analysis. After ART initiation, patients were monitored every six months.
Standardized data reporting forms were used for data collection. Demographic characteristics at baseline included date of birth (or age) and sex. Clinical characteristics included: HIV status, WHO stage, height, weight, pregnancy status for women, functional status, and ART history. CD4 count and HIV RNA were measured at the initiation of ART and at each clinic visit. All records were maintained by the local site.
The outcome for the study was defined as the proportion of LTFU in patients initiating ART. We considered two definitions: the ‘clinic’-based definition and the ‘data’-based definition. Clinic based definition was determined by Tapologo-based provider as failure to return for a scheduled consultation or medication pickup within six months after ART initiation . We used clinic-based definition in main analyses.
Data-based definition was determined by lack of recorded information in the data base on a patient returning for 6-month visit. Specifically, the patient was determined as LTFU according to data-based definition if no information was recorded with respect to the date of 6 month follow up visit and, laboratory testing (CD4 or/HIV RNA tests) within 6 months since ART initiation. The data-based definition was used in a sensitivity analysis.
The analysis was performed using data available on April 30, 2007 on all patients with at least six months of follow-up time. At six months after ART initiation, patients who had unknown drop-out status, had ART discontinued by their physician, voluntarily discontinued treatment, were transferred out of the program, or were deceased, were censored. Thus, study time for each patient was calculated from the date of ART initiation until the date of drop-out if known, or the date at six months after ART initiation if unknown. For patients who were lost follow-up at six months after ART initiation, study time was calculated from the date of ART initiation until the date of LTFU. Kaplan-Meier estimates were used to analyze the cumulative probability of LTFU. The log-rank test was used to examine differences in LTFU rates of subgroups of patient characteristics.
To enhance transparency of interpretation, all demographic and clinical variables were grouped as categorical: age (≤ 30, 31-40, > 40 years), gender and pregnancy status (pregnant women, non-pregnant women, men), baseline CD4 count (≤ 200, > 200/μl), and baseline HIV RNA (≤ 100,000, > 100,000 copies/ml). To determine the potentially joint effect of baseline CD4, gender and pregnancy status on LTFU, we stratified the pregnancy status of women by baseline CD4 counts which resulted in a variable with five categories: pregnant women with CD4 ≤ 200/μl, pregnant women with CD4 > 200/μl, nonpregnant women with CD4 ≤ 200/μl, nonpregnant women with CD4 > 200/μl, and men. For individual and stratified variables, hazard ratios with corresponding 95% confidence intervals (CI) for LTFU were analyzed through Cox proportional hazard regression. Two-sided p-values <0.05 were considered statistically significant. Analyses were performed using SAS software (version 9.1 or higher, SAS Institute Inc., Cary, North Carolina, USA).
Among adult patients (age >14 years) accessing Catholic Relief Services in Tapologo, 925 met criteria for analysis. The median age was 36 years (interquartile range (IQR), 29–44). Six hundred and forty-five of the 925 patients (70%) were female, of whom 16% were pregnant. At ART initiation, the median CD4 count was 111/μl (IQR, 41–214/μl) and HIV RNA was 4.9 log10 copies/ml (IQR, 4.4–5.4). Upon initiating ART, approximately 45% of patients had a baseline HIV RNA greater than 100,000copies/ml; 72% of patients and one-third of pregnant women had a baseline CD4 count ≤ 200/μl (Table 1).
Three hundred and twenty-two patients (35%) were confirmed to drop out of the study within six months of ART initiation. Among these patients, 37 (11%) were stopped by physician recommendation, 21 (7%) transferred out of the Tapologo program, 213 (66%) died, and 51 (16%) were lost to follow-up (using clinic-based definition). Patients with unknown drop-out status were considered as remaining in care and were subsequently censored six months after ART initiation.
Kaplan-Meier analysis indicated that during the period of observation, the cumulative probability of LTFU six months after ART initiation was 6% (95% CI: 4 – 7%) for the cohort. When stratified by age, baseline HIV RNA and CD4 count, and gender and pregnancy status, log-rank tests showed the cumulative probability of LTFU was significantly different across the sex and pregnancy subgroups (p=0.01). The cumulative probability of LTFU six months after ART initiation was 12% (95% CI: 5 – 18%) for pregnant women, 6% (95% CI: 4 – 8%) for non-pregnant women, and 3% (95% CI: 1 – 3%) for men (Table 2).
To quantify the effect of all characteristics on LTFU, Cox proportional hazard regression analyses were carried out (Table 2). Results showed that younger age (≤ 30 years) (HR: 2.14, 95% CI: 1.05 – 4.38), and pregnancy for women (HR: 3.75, 95% CI: 1.53 – 9.16) were significantly related to higher LTFU rates. However, in terms of the overall effect, gender and pregnancy status was the only factor showing a statistically significant association with LTFU (p=0.01), consistent with the result of the log-rank test.
Although we did not find an association between baseline CD4 itself and LTFU at 6 months, there was a great impact on LTFU related to pregnancy in lower CD4 stratum. If baseline CD4 ≤ 200/μl, pregnant women had 3.62 times of hazard of LTFU compared to nonpregnant women (95% CI: 1.52 – 8.62). To examine this joint effect, the gender and pregnancy status was further stratified by baseline CD4, and the proportional hazard regression analyses were conducted. The analysis results demonstrated that pregnant women with baseline CD4 ≤ 200/μl had the highest risk of LTFU six months after ART initiation (HR: 6.03; 95% CI: 2.20 – 16.71), followed by nonpregnant women with baseline CD4 > 200/μl (HR: 2.91; 95% CI: 1.19 – 7.11) and pregnant women with baseline CD4 > 200/μl (HR: 2.44; 95% CI: 0.80 – 7.46) (Table 3, Figure 1).
To detect the impact of different definitions on LTFU, a sensitivity analysis was performed, in which a revised data-based definition of LTFU was used which included patients missing visit information for at least six months.
Using data-based definition of LTFU, 95 patients (10%) were lost to follow-up six months after initiating ART. The proportional hazard regression analysis showed pregnant women with baseline CD4 ≤ 200/μl had the highest risk of LTFU six months after ART initiation (HR: 2.67; 95% CI: 1.12 – 6.35), followed by pregnant women with baseline CD4 > 200/μl (HR: 2.40; 95% CI: 1.14 – 5.05) and nonpregnant women with baseline CD4 > 200/μl (HR: 2.29; 95% CI: 1.24 – 4.24) (Table 3). The overall joint effects of baseline CD4, gender and pregnancy status on LTFU were similar for both definitions of LTFU.
Several studies have described the reasons for LTFU in ART treatment programs in South Africa [20, 21]. These studies used chart reviews to investigate patient characteristics predicting LTFU in large ART treatment programs in public hospitals in South Africa. Causes of LTFU during ART in the studies included: financial difficulty, lack of patient knowledge that ART needs to be lifelong, hospitalization or illness, medication toxicity, CD4 count increase and death. Little work has focused on the relationship between LTFU and gender and pregnancy . The HIV prevalence among pregnant women in South Africa is 20-30% [8, 15]. Further understanding of LTFU among pregnant women on ART could help improve retention of these patients in care and may also decrease risk of mother to child HIV transmission . This study identified patient characteristics including gender and pregnancy status associated with LTFU in patients who began ART in the Tapologo program, one of the Catholic Relief Services HIV care and treatment clinics in North West province, South Africa.
In the community-based cohort of 925 adult patients who were eligible for ART, we conducted analyses to identify patient characteristics associated with LTFU six months after ART initiation. It was found that the gender and pregnancy status was significantly related to LTFU and pregnant women had the highest risk of LTFU. Although the data provided insufficient statistical evidence to support an association between baseline clinical characteristics (HIV RNA or CD4) and LTFU six months after ART initiation, the distributions of these clinical characteristics in pregnant women were different from nonpregnant women and men. In addition, the stratified analysis in baseline CD4 ≤ 200/μl showed pregnant women were more likely to be lost to care than nonpregnant women. All these evidence indicated the possible joint effects of clinical characteristics, gender and pregnancy status. To detect the joint effects of baseline CD4, gender and pregnancy status, we stratified the pregnancy status by baseline CD4 count for women. A Cox proportional hazard model was analyzed to quantify the joint effects on LTFU. To examine the impact of different definitions on LTFU, we carried out sensitivity analyses using a second standardized (data-based) definition of LTFU. The results showed that, although HIV-infected nonpregnant women who initiated ART and had higher CD4 were significantly more likely to be lost to care, pregnant women at all stages of HIV disease initiating ART were at substantial risk of LTFU with its attendant risk of developing an AIDS-defining illness or of death.
This study had several limitations. First, the reasons why pregnant women stopped treatment were not available. Although ART programs in this study were local clinics that provided community-based care at no cost, pregnancy itself represented a financial burden for women and their families. The higher proportion of LTFU among pregnant women may arise from unemployment as well as increased prenatal care costs. Another explanation for LFTU among pregnant women may be that some of these patients may transfer out of the Catholic Relief Services ART treatment programs to maternal care facilities or hospitals without informing local clinic staff. Second, patient records were maintained by the local site, however, no data existed for approximately 10% of the Tapologo patients. In order to perform this analysis, we assumed that patients with missing current treatment status were in care. This likely underestimated the rates of LTFU .
This was the first systematic study of LTFU in the programs of the Catholic Relief Services and the Southern African Catholic Bishops’ Conference in South Africa. Pregnant women with HIV disease initiating ART were at the greatest risk of being lost to follow-up. To ensure the best outcomes for HIV-infected women, as well as to prevent HIV transmission from mother to child, interventions to successfully retain pregnant women in care are urgently needed.
The authors wish to thank Lauren Uhler, Jennifer Chu and Ji-Eun Park for their invaluable support and expert technical assistance.
Presented in part at the 15th Conference on Retroviruses and Opportunistic Infections, February 3-6, Boston, MA, USA 2008 (abstract 839).
CONFLICT OF INTEREST STATEMENT
The authors declare no conflict of interest.