Interventions to Improve Adherence to Antiretroviral Therapy (ART) in Sub-Saharan Africa: An Updated Systematic Review

Optimal adherence to antiretroviral therapy (ART) remains the bedrock of effective therapy and management of human immunodeficiency virus (HIV). This systematic review examines the effect of interventions in improving ART adherence in sub-Saharan Africa (SSA), which bears the largest global burden of HIV infection. In accordance with PRISMA guidelines, and based on our inclusion and exclusion criteria, PUBMED, MEDLINE, and Google Scholar databases were searched for published studies on ART adherence interventions from 2010 to 2019. Thirty-one eligible studies published between 2010 to 2019 were identified, the categories of interventions were structural, behavioral, biological, cognitive, and combination. Study characteristics varied across design, intervention type, intervention setting, country, and outcome measurements. Many of the studies were behavioral interventions conducted in hospitals with more studies being randomized controlled trial (RCT) interventions. Despite the study variations, twenty-four studies recorded improvements. Notwithstanding, more quality studies such as RCTs should be conducted, especially among key affected populations (KAPs) to control transmission of resistant strains of the virus. Reliable objective measures of adherence should replace the conventional subjective self-report. Furthermore, long-term interventions with longer duration should be considered when evaluating the effectiveness of interventions.


Introduction
Since highly active antiretroviral therapy (HAART) is the standard treatment for HIV-positive patients, the effectiveness of antiretroviral therapy (ART) varies majorly with patient's adherence observance to the daily medication regimen. One of the major concerns of public health for people living with HIV (PLWHIV) is the promotion of medication adherence [1]. Although structural, social, and personal factors could be reasons for failure to adhere to ART among patients, [2,3] other factors such as health-system-related barriers, food insecurity, supply-chain interruptions, and insufficient human health resources, are barriers to ART adherence in Africa [4]. Through the years, adherence has been found to be a fundamental predictor of ART treatment success acquiescent to intervention [2] however, several patients are found to lapse on the prescribed treatment regimen thereby increasing the risk of transmitting HIV, deteriorating health conditions [5], therapy failure, production of new resistant viral strains, progression to acquired immune deficiency syndrome (AIDS), more hospitalization and increased rates of mortality [6][7][8], and poor quality of life. Consequently, the resultant effect of not adhering to ART is increased cost

Eligibility Criteria
This review was updated from a previous review [18] that involved studies evaluating interventions' effectiveness to improve adherence to ART in adults in sub-Saharan Africa with adherence as the primary or secondary outcome. The definition of adherence in this review was operationally restricted to ART adherence, which implied the degree of medication (antiretroviral) intake by patients as recommended by their providers of healthcare. Studies relating to the distinct concepts of adherence such as clinic attendance or appointments and retention were equally reviewed. There was no restriction on the measures of assessment of ART adherence. The guide used for the inclusion criteria was the population, intervention, comparison, outcome, and time (PICOT) mnemonics [19,20].
Population: All adult HIV-positive patients on ART. Studies involving only children were excluded.
Intervention: Interventions to improve ART adherence and biological correlates of adherence.
Comparison: Studies with a comparison group or control group were included. However, studies with no direct comparison group, for instance, some quasi experimental studies were excluded from the study.
Outcome: Adherence to ART and correlates of adherence. Time: Studies published from 2010 to 2019 were included. There were no exclusion criteria for study designs. Unpublished trials were not included and only journal articles published in the English language were reviewed. The review contained only studies from SSA, and only studies involving SSA sites were included for multisite studies. According to the exclusion criteria, studies that were excluded include studies that were not journal articles, which involved only children, not reporting any adherence intervention, not involving a comparison group or control, and not reporting adherence-related outcomes. Studies that mentioned adherence in their titles but did not actually measure adherence were excluded.

Study Selection
Studies were reviewed based on strict adherence to ART appointments and medication as scheduled by their health care providers. By sequence, articles were screened according to title, abstract, and full text to ascertain their inclusion. Studies involving interventions relating to ART adherence in sub-Saharan Africa were included in the review, which reported adherence measurements conducted alongside interventions. From each article that passes the screening above, information on year of publication, type of intervention, country where study was conducted, health care setting, and outcomes were reported. Subjective and objective measures of adherence were recorded, including biological correlates of adherence, for instance, viral load and CD4 count. A total of thirty-one studies were included in the final analysis, and these studies were screened by two independent setting, and outcomes were reported. Subjective and objective measures of adherence were recorded, including biological correlates of adherence, for instance, viral load and CD4 count. A total of thirty-one studies were included in the final analysis, and these studies were screened by two independent reviewers (P.P.D and S.I) while R.A.M, S.M.S, and A.O reviewed the selection and resolved disagreements.

Quality Assessment
The Cochrane criteria were used for the systematic assessment of bias for the studies included in this review. The risk of bias was evaluated as either "low risk", "high risk", or "unclear risk" analyzed over seven domains [21][22][23]. Low risk indicates reported information with evidence of little or no possible bias while high risk implies evidence of possible bias. Unclear risk denotes a dearth of info or skepticism over possible bias. Among the domains were sequence generation, allocation sequence, concealment, blinding (participants and personnel), blinding of outcome assessment, incomplete outcome data, selective outcome reporting, and other potential sources of bias. The opinion of a third reviewer was sought to solve disagreements.

Results
A total of 4598 records (1560 in PubMed, 446 in Medline, and 2592 in Scopus) were identified, of which 4123 were excluded based on the content of their titles and abstracts, which was not in line with the inclusion criteria. Studies that did not measure adherence as an outcome were excluded from the systematic review. Thirty-one journal articles met the inclusion criteria  and were included in the systematic review. Figure 1 shows the flowchart of the systematic review process.  Table 1 presents the study characteristics, which include a summary of the authors, study design intervention category, intervention type, study country, setting, outcome, adherence measurement, and study findings. The study design refers to the methodology and statistical methods employed in a study to collect and analyze data. In this study, the site where the intervention was conducted is referred to as the intervention setting, which could be in a hospital (hospital-based) or in the community where the patients reside (community-based). The outcome measure was adherence or correlates of adherence. The adherence measurement states how adherence was measured in each study while the study findings present the results of adherence measured. Table 2 was adapted from a systematic review [18] and summarizes the intervention categories and classifies interventions into structural, biological, behavioral, cognitive, affective, and combination (mixture of some or all the categories).   Table 1 presents the study characteristics, which include a summary of the authors, study design intervention category, intervention type, study country, setting, outcome, adherence measurement, and study findings. The study design refers to the methodology and statistical methods employed in a study to collect and analyze data. In this study, the site where the intervention was conducted is referred to as the intervention setting, which could be in a hospital (hospital-based) or in the community where the patients reside (community-based). The outcome measure was adherence or correlates of adherence. The adherence measurement states how adherence was measured in each study while the study findings present the results of adherence measured. Table 2 was adapted from a systematic review [18] and summarizes the intervention categories and classifies interventions into structural, biological, behavioral, cognitive, affective, and combination (mixture of some or all the categories). (1) Self-report (2) Pharmacy refill 9 months

Risk of Bias Assessment
In the hierarchy of evidence according to the risk of bias, RCTs supersede observational studies, although this could be reversed in some instances where bias is present, as the strength of evidence is limited [55]. Most of the observational studies were high-risk, this is because unlike the RCTs, they are not characterized by random sequence generation, allocation concealment, and in some cases blinding. Unclear risk of bias with respect to blinding of participants and personnel and in some cases outcome assessment was observed in some studies. The Cochrane risk of bias assessment was used [21] and presented on Table 3. Table 3. Risk of bias ratings for each study included. Note. 1 Random sequence generation, 2 allocation concealment, 3 Blinding of participants and personnel, 4 Blinding of outcome assessment, 5 Incomplete outcome data, 6 Selective reporting, 7 Other bias. "H" = 0, "U" = 0, "L" = 1. Mean score = 36. Higher scores and percentages denote lower risk of bias.

Discussion
The goal of ART is lasting viral suppression to undetectable levels, and optimal adherence to ART is required to attain this. Several types of interventions have been used in sub-Saharan African countries to improve adherence to ART among HIV-positive patients. These interventions involving education and counseling, community-based adherence support, mobile devices, and food services resulted in long term or short term improvement in ART adherence. The thirty-one selected studies in this review support the drive to scale-up long-term ART success in SSA.

Effectiveness of Interventions
Majority of the studies in this systematic review that reported effective interventions were RCTs [27,31,[39][40][41][42][43][44]46,49,50,54]. In determining the effectiveness of interventions, the RCT has proven to be the most reliable in providing evidence and has been considered the gold standard for evaluating the effectiveness of interventions over the past decade [57]. While the concept of gold standard relates to research design, a broader perspective to fully appraising the evidence of interventions as the gold standard is demonstrated in its ability to function, be implemented and serve its purpose [58]. Thus, when searching for answers to the clinical research question concerning the evaluation of diverse treatments, the RCT is primarily recommended because of its propensity to minimize bias [59]. Although a significant improvement in the intervention group denotes the intervention's success, it is more beneficial to consider the effect size, which explains the magnitude of the effect and not just the statistical significance. Additionally, effect size is independent of sample size whereas p value depends on both the effect size and sample size [60]. Unfortunately, information regarding the effect size for most studies in this systematic review was not clearly stated. Subsequent interventions should base judgment of their primary findings on effect size and not solely on statistical significance.
Furthermore, in concluding on the effectiveness of an intervention, a considerable length of time may be examined in order to determine a substantial impact. In this review, one study [39] reported a reverted improvement in adherence following cross-over after three months, implying the inauthenticity of the intervention. This suggests that the short-term effect of interventions may not be generalizable as its sustenance is not guaranteed. Additionally, since ART is a life-long behavior, and optimal adherence is required for achieving maximum viral suppression, interventions with ephemeral effectiveness may just offer diminutive impact to treatment success. Thus, in order to validate the effectiveness of an intervention, prospective studies may need to observe the effectiveness of these interventions for longer duration so as to ensure lasting impact. Less than half of the included studies in this systematic review were observed for a minimum of one year [19,21,22,[24][25][26]28,31,35,36,39,42], the remaining were mostly six months and below. It is suggested that more studies in sub-Saharan Africa adopt interventions with longer duration, as this may further authenticate study findings. Additionally, further studies could focus on developing a clear standard for evaluating successfulness of adherence interventions and duration of observation.

Adherence in Key Affected Populations
Additionally, evident in this review was the paucity of studies on interventions relating to ART adherence among HIV key affected populations such as men who have sex with men (MSM), injection drug users (IDU), sex workers, people in prisons and other closed settings, and transgender people. In 2018, it was reported that these groups together with their sexual associates accounted for over half of the global incidence [61]. As such, it is important to conduct studies among these populations, to investigate issues peculiar to them such as linkage to care and commitment to treatment regimen, most importantly their adherence to ART, which is fundamental in managing HIV infection. Additionally, noncompliance to medications has been reported to be a characteristic behavior of these affected populations, evidenced by low adherence rates [62,63], although some studies reported over 90% adherence among MSM [64]. Many factors responsible for poor adherence in these populations include HIV stigmatization, fear of healthcare-seeking and denial of care, social isolation, poor access to health services, and psychological issues such as depression.
In respect of this, further intervention studies on adherence should be considered in order to eschew the implications of nonadherence, which include transmission of resistant strains, thereby limiting the therapeutic options of newly infected patients.
There is no "gold standard" for measuring adherence as each of these assessments have strengths and weaknesses; however, the choice of the assessment method will greatly depend on the economic setting of the study. This is because some assessment methods are capital-intensive, and some study locations are resource-rich while others are resourcelimited. These are some of the challenges associated with the choice of adherence measurement for instance, pharmacy refill and self-report are mostly employed in HIV/AIDS hospital settings while Medication Event Monitoring System (MEMS) are commonly used in clinical studies [65]. Due to its ease of use and affordability, self-report has been the most commonly used in resource-limited settings (RLS) [65]. This is consistent with the findings of the present systematic review, which revealed fifteen [25][26][27]29,30,32,36,39,41,[44][45][46]49,50,54] out of thirty-one studies employed a self-report; with twelve studies reporting significant results [25][26][27][28]36,39,41,44,46,49,50,54]. It is also note-worthy to state that these twelve studies constituted half of the twenty-four studies with significant findings. Selfreport is associated with many advantages, which makes it the most commonly used measure of adherence [66].
Besides its ease of use and validity that propels it to be the most widely used adherence measure, the self-report is consistent with objective methods of measuring adherence such as plasma viral load monitoring and MEMS [67]. Other advantages of a self-report in RLS include affordability and low staff requirements, it is also considered to be robust and an apt indicator of adherence [65]. The major demerit of self-report is the overestimation of adherence due to recall bias and social desirability. This mostly stems from the patient's fear of being judged by the healthcare providers or the consequences of providing negative feedback, which compels them to give inaccurate adherence reports [68]. Despite its demerits, majority of studies in sub-Saharan Africa, especially in RLS, employ its use. We recommend that concrete justification for further use is researched.
Appointment scheduling, which is also an early warning indicator (EWI), is also considered to be subjective, although the results can be fetched from the clinic's attendance records [65]. It is similar to the subjective self-report assessment but more objective. However, it is prone to manipulations by clinic staff [69]. Pill count and pharmacy refill on the other hand are the commonly used objective measures due to their relatively inexpensive nature and ease of use. Pharmacy refill is a validated measure of ART adherence that relates to viral load [70]. The draw-backs of pharmacy refill include pill dumping or sharing [65], the need of a closed pharmacy system, its dependence on accurate and reliable records [69], and its inability to predict or detect viral rebound in patients [71]. The disadvantage of pill count, which is also an EWI, include pill dumping and limited availability. Another drawback is that it is difficult to keep record of pharmacy visits and refills when patients obtain their medications from different pharmacies. Even though most patients in RLS return to their primary healthcare providers for free treatment and refill; this makes pharmacy refill a more feasible adherence measurement [65].
Other adherence measures seldom used as reported in this review include EAMD, viral load monitoring, hair concentration, medication possession ratio, and appointment diary. The EAMD entails recording every medication bottle opening thus providing a more reliable proof of medication-taking behavior, nonetheless this is not without demerits. In the event of a single opening, misclassification bias might occur; a situation where multiple doses could be taken out for future dosing (pocket doses) or no doses taken out at all despite opening (curiosity openings) thereby altering with its accuracy [72]. In both cases, evaluation is achieved mostly during a clinic visit or at the time of a study, which is probably long after the occurrence of the adherence gap [73]. To avoid that, real-time adherence monitoring (RTAM) devices were introduced, which are EAMD designed to deliver instant information on dosing events, this has proven to be more beneficial in monitoring adherence actively and promptly between clinic visits or in study visits [74]. RTAM devices that have proven to be feasible and reliable lately are automated medication bottles that possess lasting battery half-lives capable of containing medication supplies for a period of 30 days. It functions by transmitting a time-stamped cellular signal to a central web-based server at each opening of the device; this denotes a dosing event and is recorded [75]. Though information on adherence could be examined in real-time thus enabling prompt adherence intervention, internet connection is required for this task thereby rendering this measure less feasible especially in RLS. Other cons of this measure would be its inability to confirm medication ingestion [76], and a lack of privacy as patients may have to travel around with the device [77]. Other accurate measures although expensive, include direct methods such as measuring drug levels or its metabolite in urine or blood, detecting an added biomarker to the drug formulation, and direct observed therapy [78]. Deliberations on measuring stool and urine samples daily could be considered for further studies.
Validating the measurement of adherence against viral load is beneficial [65], and attaining undetectable viral load is also considered to be one of the most common measures of ART adherence. A high adherence level of 95% was previously associated with undetectable viral load [6] thus equating viral suppression with adherence. However, in recent times, adherence levels between 80 and 85% is sufficient for viral suppression, thereby making undetectable viral load an unsatisfactory proxy for maximum adherence. It should also be noted that viremia is evident long after the occurrence of an adherence gap [69].
In a bid to manage the inevitable limitations of the various measures of adherence, newer pharmacological measures [79] have been introduced that possess the ability to quantify medication adherence and exposure over time. The advantage of these new measures is its ability to expose both medication adherence and pharmacokinetics, which involves absorption, distribution, metabolism, and excretion in one evaluation. Dried blood spots (DBSs) and hair are the obtainable mediums that aggregate the measurement of ART adherence exposure [69].

Other Results
In this systematic review, interventions in the affective category and behavioral category were the most common intervention categories, while counseling and treatment supporter were the most common intervention types. The use of counselors though cumbersome has been found to be effective in improving adherence as reported in a systematic review [80]; treatment supporter intervention also yielded similar success [81,82].
Additionally, in this systematic review, more studies came from South Africa [27,33,37,40,41,46,[48][49][50], Uganda [26,28,42,43,47], and Nigeria [30,32,36,44]. This finding is not surprising because the highest global disease burden of HIV lies in South Africa and Nigeria [83]. Furthermore, these countries account for about half of all new infections in sub-Saharan Africa annually [84]. This explains why more studies emanate from these countries and it is also a reason for the substantial funding of HIV research in these countries.

Limitations
The limitations of this systematic review include the unavailability of studies targeted at a key affected population such as MSM, female sex workers (FSWs) and orphans and vulnerable children, and the elderly. Some interventions as reported by some studies were at risk of bias, as study protocols were not duly followed. Additionally, because adherence is a life-long behavior, and there is no clear set standard period for observing interventions, the authors utilized the information from the studies to evaluate the successfulness of the interventions.

Conclusions
In conclusion, a wide range of studies on ART adherence interventions was done among HIV positive adults in sub-Saharan Africa. Many quality studies such as RCTs and cohorts were present; despite the high-cost and ethical limitations of RCTs. Additionally, various types of interventions were used in both hospital and community settings in different countries to improve adherence; although the majority proved effective in both settings, some failed to show any effect. In addition, among the various methods of assessing adherence, subjective self-report though unreliable, proved to be the commonly used measure of adherence. It is recommended that objective methods of assessment that are more reliable be used in future studies. Lastly, further studies should focus on closing significant evidence gaps on interventions for improving adherence. These gaps include effectiveness in key affected populations, long-term effectiveness, and quality studies.