Prevalence, trend and risk factors for antiretroviral therapy discontinuation among HIV-infected adults in Ethiopia in 2003-2015

Background It is well acknowledged that antiretroviral therapy (ART) discontinuation hampers the progress towards achieving the UNAIDS treatment targets that aim to treat 90% of HIV diagnosed patients and achieve viral suppression for 90% of those on treatment. Nevertheless, the magnitude, trend and risk factors for ART discontinuation have not been explored extensively. We carried out a retrospective data analysis to assess prevalence, trend and risk factors for ART discontinuation among adults in Southwest Ethiopia. Methods 12 years retrospective cohort analysis was performed with 4900 HIV-infected adult patients between 21 June 2003 and 15 March 2015 registered at the ART clinic at Jimma University Teaching Hospital. ART discontinuation could be loss to follow-up, defaulting and/or stopping medication while remaining in care. Because data for 2003 and 2015 were incomplete, the 10 years data were used to describe trends for ART discontinuation using a line graph. We used binary logistic regression to identify factors that were correlated with ART discontinuation. To handle missing data, we applied multiple imputations assuming missing at random pattern. Results In total, 4900 adult patients enrolled on ART, of whom 1090 (22.3%) had discontinued, 954 (19.5%) had transferred out, 300 (6.1%) had died, 2517 (51.4%) were alive and on ART, and the remaining 39 (0.8%) had unknown outcome status. The trend of ART discontinuation showed an upward direction in the recent times and reached a peak, accounting for a magnitude of 10%, in 2004 and 2005. Being a female (AOR = 2.1, 95%CI: 1.7–2.8), having an immunological failure (AOR = 2.3, 1.9–8.2), having tuberculosis/HIV co-infection (AOR = 1.5, 1.1–2.1) and no previous history of HIV testing (AOR = 1.8, 1.4–2.9) were the risk factors for ART discontinuation. Conclusions One out of five adults had discontinued from ART, and the trend of ART discontinuation increased recently. Discontinued adults were more likely to be females, tuberculosis/HIV co-infected, with immunological failure and no history of HIV testing. Therefore, it is vital to implement effective programs such as community ART distribution and linkage-case-management to enhance ART linkage and retention.


Methods
years retrospective cohort analysis was performed with 4900 HIV-infected adult patients between 21 June 2003 and 15 March 2015 registered at the ART clinic at Jimma University Teaching Hospital. ART discontinuation could be loss to follow-up, defaulting and/or stopping medication while remaining in care. Because data for 2003 and 2015 were incomplete, the 10 years data were used to describe trends for ART discontinuation using a line graph. We used binary logistic regression to identify factors that were correlated with ART discontinuation. To handle missing data, we applied multiple imputations assuming missing at random pattern.

Results
In total, 4900 adult patients enrolled on ART, of whom 1090 (22.3%) had discontinued, 954 (19.5%) had transferred out, 300 (6.1%) had died, 2517 (51.4%) were alive and on ART, and the remaining 39 (0.8%) had unknown outcome status. The trend of ART discontinuation showed an upward direction in the recent times and reached a peak, accounting for a magnitude of 10%, in 2004 [1]. Sub-Saharan Africa (SSA) contributed 76% (29 million) of the total HIVinfected people, 76% (1.9 million) of the total new HIV infections, and 75% (0.9 million) of the total HIV/AIDS deaths [1]. The advent of antiretroviral therapy (ART) in 1996 significantly reduced HIV-related deaths [2]. The global ART coverage in 2015 was low [3] (40.6%), with North Africa and Middle East having the lowest coverage (19%). High-income countries had the highest coverage (67%), and SSA had coverage of about 42% [1]. In Ethiopia, there were 768,040 HIV-infected people, 39,140 new HIV infections, and 28,650 HIV/AIDS deaths in 2015 [1]. The ART coverage in Ethiopia in 2015 was moderate [3] (52%) [1].
Optimum clinical and public health achievements of ART requires consistent long-term adherence [4]. Nevertheless, ART discontinuation-interruptions to ART due to loss to followup (LTFU), defaulting or total stoppage of the treatment-is a big challenge. Additionally, ART discontinuation (discontinuation) causes drug resistance [5], diminishes the immunological benefit of treatment [6,7], and increases AIDS-related morbidity and mortality [5,8]. Discontinuation has been recognized as an impediment for attainment of the second 90 of the Joint United Nations Program on HIV and AIDS (UNAIDS) 90-90-90 treatment targets (sustainable provision of treatment for 90% of patients diagnosed with HIV) as it affects the sustainable intake of the treatment. Furthermore, discontinuation affects the performance of the third 90 of the UNAIDS 90-90-90 that aimed at achieving 90% of the virological success of patients on ART. This is because ART interruption lowers the efficacy of the treatment and subsequently leads to diminishing the number of CD4 cells, increases the number of viral counts [5], and then to failing immunological or virological success. Uganda reported 84% of virological suppression due to strong retention [9].
The two studies [16,19] that were conducted in the west part of the nation were case control studies and assessed only factors contributing for defaulting, reflecting that the magnitude of the problem was not explored. LTFU, another major contributor for discontinuation was not assessed. Unlike the rest of Ethiopia, the Southwest region is composed of different population groups. A refugee camp located near Jimma city, which takes refugees from different east African countries, contributes a number of HIV-infected patients enrolled in the ART clinic in Jimma University Teaching Hospital (JUTH). Furthermore, Jimma is also near Gambella region (Southwest Ethiopia), a region known to have the highest prevalence rate (6.5%) of HIV in Ethiopia [25], and the JUTH caters for both Jimma and Gambella zones. Since the prevalence of HIV in Southwestern region is higher (6.5%) than in other parts of the nation (<2%), it is necessary to understand whether the high prevalence is associated with other than factors for discontinuation identified in similar studies in Ethiopia.
Given the above gaps, and the clinical and public health significance of discontinuation, we have assessed the prevalence, trend and risk factors for ART discontinuation among adults by using 12 years (2003-2015) ART clinic data from JUTH in southwest Ethiopia. It is imperative to comprehensively understand the magnitude of the discontinuation problem and its influential factors in order to contextualise interventions to retain HIV patients in in care and to contribute to the Nation's second and third UNAIDS treatment targets. . According to the current treatment guidelines, HIV infected adults are eligible to start ART if their CD4 cell count is 500 cells/mm3 irrespective of WHO clinical stage, their WHO clinical stage is 3 or 4 irrespective CD4 cell count, and they are pregnant, breast feeding women, sero-discordant couples or diagnosed with active tuberculosis irrespective CD4 cell count. For all HIV-infected children below 15 years of age, ART is recommended irrespective of WHO clinical stage and CD4 cell count. All HIV-infected adults aged !15 years enrolled in ART care in JUTH were the target population. If the recorded outcome were death, transferred out or unknown, participants would be excluded from this study.

Data source and procedures
The data were extracted from JUTH electronic medical records (EMR) system called comprehensive care center patient application database (C-PAD). This system was designed since 2007, and data registered before 2007 were copied retrospectively to the electronic record. Fig 1 presents the schematic presentation of data extraction of discontinuation among HIVinfected adults in JUTH. In total, 4900 HIV-infected adults on ART were eligible or the study in the period between 2003 and 2015, and 3607 of them were included in the analysis. Clinicians record the clinical and non-clinical characteristics of the patients on paper, and afterward, data clerks entered the data into the EMR.
Two data clerks perform the data entry process, and this ensures the accuracy and reliability of the data. The International Center for AIDS Care and Support (ICAP) at Colombia University has been providing expertise assistance on the data management process and has also been providing checkups on completeness of the data. Weekly patient summary generated from the EMR system helps to flag patients with conditions that seek follow-up. Records would be excluded from a complete case analysis if outcome status of the person was death, transferred out or not recorded.

Study variables and measurements
The dependent variable was discontinuation and was dichotomized as 1) alive and on ART, and 2) discontinued. Discontinuation refers to LTFU, defaulting and/or stopping medication while remaining in care. LTFU refers to patients who had been on ART treatment and had missed at least three clinical appointments but had not yet been classified as "dead" or "transferred out" (transferred). Defaulting refers to patients who had been on ART treatment and had missed less than three clinical appointments but had not yet been classified as "dead" or "transferred". Stopping medication refers to patients who had stopped treatment due to any reason while they have remained in care. Transferred is an official transferring of patient to another ART clinic within or outside a catchment area. The independent variables included age, sex, marital status, educational status, religion, ART adherence, cotrimoxazole adherence, clinical failure, immunological failure, treatment failure, late presentation for HIV care (LP), tuberculosis (Tb)/HIV co-infection, baseline functional status, history of HIV testing and ART shift. Educational status was categorized in to no education (could not read and write), primary (grade 1-8), and secondary and above (grades !9). Functional status was classified in to the following categories: i) work-able to perform usual work, ii) ambulatory-able to perform activity of daily living, and iii) bedridden-not able to perform activity of daily living. Table 1 reports the operational definitions of LP, adherence, clinical, immunological and treatment failures.

Statistical analyses
Data cleaning and exploration were performed prior to analysis. Descriptive analysis was undertaken to explore the frequency tables and proportions for categorical data; and mean, median, range and line graph for continuous data. We described ten years trends for discontinuation (data for years 2003 and 2015 were excluded since the number of months was incomplete) using a line graph. The number of alive and on ART patients was calculated by subtracting the number of patients who dead, discontinued and Transferred from the total number of patients on the cohort. We checked for multicollinearity and potential interactions, and none was found. We used binary logistic regression to identify factors that were correlated with discontinuation. To choose the candidate variables (P<0.25 was considered) to multiple logistic regression, we applied bivariate logistic regression analysis. P-value 0.05 was considered as a cutoff value for statistical significance in the final model. To handle missing data, we performed multiple imputations (n = 5) assuming missing at random (MAR) pattern [

Characteristics of study participants
Of 8172 patients enrolled in HIV care program from 21 June 2003 to 15 March 2015, 4900 HIV-infected adults had documented commencement of ART (Fig 1). Table 2

Prevalence and trend of ART discontinuation
In total, 1090 (22.3%) had discontinued, 954 (19.6%) had transferred out, 300 (6.2%) had died, 2517 (51.4%) were alive and on ART, and the remaining 39 (0.8%) had unknown outcome status in the period between 2003 and 2015 (Table 3). Of the 1090 adults who had discontinued from ART, 906(83.1%) adults had defaulted, 145(13.3%) adults had LTFU and 39(3.6%) adults had stopped during the 12 years study. The prevalence of discontinuation was consistently high between 2004 and 2007. In the period between 2008 and 2011, the proportion of discontinuation showed a different trend with a sharp decline from 6% in 2008 to 2% in 2009, a sharp rise from 2% in 2009 to 6% in 2010, and then falling in 2011 to 3%. In the recent times, the proportion of discontinuation is increasing. Table 3 shows the trend in discontinuation among HIV-infected adults on ART.

Risk factors for ART discontinuation
The results from the multiple logistic regression analysis found from the analysis of a complete case and multiple imputations (MIs) are reported in Table 4. Being female (AOR = 2.1, 95%CI: 1.7-2.8), having immunological failure (AOR = 2.3, 1.9-8.2), having Tb/HIV co-infection (AOR = 1.5, 1.1-2.1) and no previous history of HIV testing (AOR = 1.8, 1.4-2.9) were the risk factors for discontinuation.

Multiple imputations (MI)
We have undertaken MI to address the missing data using five imputed data sets, and we reported a model with pooled imputed values (Table 4). In estimating factors associated with discontinuation among adults, results were similar in both MI and complete case analyses except for ART adherence and history of HIV testing. ART adherence was found to be statistically significant in the MI analysis but not in the complete case analyses, and to the contrary, history of HIV testing was not found to be statistically significant in the MI analysis unlike in the complete case analyses.

Discussion
The UNAIDS target has declared three new and ambitious goals by 2020 to end AIDS epidemic by 2030: diagnosing 90% of people living with HIV, providing ART for 90% of those diagnosed with HIV, and achieving viral suppression for 90% of patients receiving treatment [37]. Nevertheless, discontinuation from treatment challenges the success of the last two goals [38]. One in five people on ART (22.3%) had discontinued from the treatment in the current study. When compared to findings elsewhere in Ethiopia, discontinuation rate in the current study was higher than findings from the Tigray's study [14] and lower than findings from the Amhara's study [15]. In addition, this rate was lower than findings from studies conducted in Guinea-Bissau (51.1%)[39], Nigeria (28%) [40], and a multi-clinic study from Republic of Congo, Cameron and Burundi (83%) [41]. The dissimilarity in measurement [42], access to HIV care services [43], and innovation, adoption and implementation of cost-effective retention strategies could be the possible reasons of differences [44,45]. For example, HIV care services such as ART accessibility affects the discontinuation rate. The estimated ART coverage [1] in Ethiopia (52%) is higher than from Guinea-Bissau (25%), Nigeria (29%), Republic of Congo (25%), Cameron (22%) and Burundi (38%). This could be one of the reasons why discontinuation rate in our finding is lower than those countries. Furthermore, the introduction of innovative programs such as Health Extension Workers  (HEWs) and Health Development Army (HDA) to HIV care services [9]-programs that are not implemented in Guinea-Bissau, Nigeria, Republic of Congo, Cameron and Burundicould affect the discontinuation rates. The trend for discontinuation was different across years. The proportion of discontinuation in the beginning year (2004) was very high (10%) because ART was not freely available. Discontinuation had also increased from 6% in 2006 to 9% in 2007. This could be because: i) ART was universally scaled up without preparations [7]; and ii) there was inequity and limitations in access to HIV care services, a justification corroborated by 24% of the health facilities in Ethiopia provide ART care and 59% of the health facilities in the country provide prevention of mother to child transmission services [46]. The trend of discontinuation had dramatically decreased from 9% in 2007 to 2% in 2009. The overall HIV care services have been improved [6] and had a profound contribution to the reduction of discontinuation. Also, HIV was included as one of the 16 packages in a new program called health extension program (HEP)-an innovative community based health service delivery system aiming at provision of essential primary health care services-, and this program could reduce the proportion of discontinuation [47]. Nevertheless, as of 2011, the discontinuation has been growing, because non-governmental organizations has been phasing out. To the contrary, local governments have been allocating little budget for health developments, particularly for HIV/AIDS [48]. We should seriously heed the issue since the developmental assistance for health budgeted by local governments is expected to grow only slightly [49].
Hence, the magnitude of discontinuation in the current study reflects that it is a considerable number and its trend is increasing in the recent times. Programs such as community based ART distribution [9,50], improving adherence via medication diary for care givers [51], home based nursing interventions [52] and adherence clubs [53] could increase linkage to and retention in care and further reduce discontinuation. People living with HIV should also meaningfully be involved in the continuum of care to improve retention programs [54]. The universal test, treat and keep strategy would dramatically improve ART retention and is fundamental in cost-effective HIV care [55].
Compared to males, females were more likely to discontinue from ART, a similar finding to previous studies [56,57]. Part of the reasons for this could be due to: (i) HIV related stigma which is known to be higher in females than in males, and prohibits them out from HIV care seeking [58,59]; (ii) Lower literacy status in females than males, which is a big challenge that prevent women from optimising the benefit of HIV care [60], (iii) higher usage in females than males of traditional healers which hinders females from taking ART consistently [61], and (v) lack integration between modern and traditional medicines which has been identified as one of the key factors in for a consistent uptake of ART programs in Ethiopia [62]. Similar to findings by Meloni and colleagues [63], discontinuation was higher among patients with immunological failure than their comparator. HIV-infected patients with immunological failure are susceptible to multiple opportunistic infections and do progress to the advanced stage of HIV/AIDS rapidly leading to quick deterioration of their health status [64,65]. Poor health status can also be among significant factors that deter them from uptaking HIV care services consistently [66]. To address ART discontinuation problem, Programs such as linkage-casemanagement (LCM) [67] that focus on multiple points of HIV care to enhance retention have been suggested as a good practice.
Although it has been stated that Tb/HIV co-infection mortality and morbidity was reduced dramatically in the era of ART [68,69], in the current study, the odds of discontinuation was higher in Tb/HIV con-infected patients than in patients with HIV infections alone. These results are consistent with findings from earlier studies [70][71][72]. The intimate linkage between HIV and Tb enables the progression of HIV disease to advanced stage rapidly and thereby disallowing patients from regular treatment intake [26,73]. In addition, the double stigma related to Tb/HIV coinfection as well as the double burden of having to take multiple pills for both conditions (pill effect) could be a compounding factor for discontinuation within this cohort. Because the burden of Tb/HIV confection is a serious problem in Ethiopia and in similar settings, it is vital to implement intervention strategies that strengthen access to universal Tb/ HIV co-infection care throughout the country. It is necessary for the special attention to be given to Tb/HIV co-infection because it is well known that Tb causes the highest mortality among HIV-infected patients [38]. The lack of history of HIV testing in association with discontinuation could be linked with poor awareness of the care [7,74], high HIV-related stigma [59], and feeling of wellbeing [75]. Although the outcomes status of current study was not known, previous studies in Ethiopia and elsewhere have shown that 9% of transferred out and 40-86% of LTFU cases failed to re-engage to the care [76], and half of LTFU patients after tracing were found deceased [7,23].
It is necessary to acknowledge limitations of the current study including: i) some variables such as stigma and mental illness were not assessed due to the nature of the study design-previous studies reported that stigma [77] and mental illness[24] affect ART discontinuation; ii) the outcome status of discontinued patients was not known, iii) findings of this study might not infer the other level of health facilities such as health centers or private hospitals; and iv) the comparison for discontinuation may not be appropriate due the dissimilarity in definition between studies.

Conclusions
The magnitude of discontinuation was recorded among one in five patients, and the trend has increased recently. Discontinued patients were more likely to be females, Tb/HIV co-infected, with immunological failure and no history of HIV testing. Additionally, in order to ensure the best outcomes and enhance the course of HIV care it is necessary that patients should be tracked to establish whether they have defaulted, transferred out, or died. We recommended implementing community ART distributions, strengthening adherence clubs, and adopting benchmarking programs such as linkage-case-management to enhance ART linkage and retention which have demonstrated to be effective in similar settings. Ethiopia has to subscribe into the East African International epidemiologic Databases to Evaluate AIDS (EA-IeDEA) Consortium [78], not only to share uniform measurements and gain research networks, but learn good practices in implementing HIV/AIDS treatment and overall care and management.