STRENGTHS AND LIMITATIONS OF THIS STUDY

This was the first study to pool the results of numerous primary studies.

The included studies were of good quality.

The study did not use the Embase database to retrieve relevant studies.

There were high proportions of loss to follow-up across the included studies.

The study included only studies that had been conducted using English language.

Introduction

HIV is a leading global public health problem.1 2 As of 2021, it is estimated that more than 38.4 million people are living with HIV globally,3 25.8 million of whom reside in Africa.4 According to a global statistics report, 2019, there were approximately 1.7 million newly diagnosed HIV cases, and 700 000 HIV/AIDS-related deaths worldwide,5 which is far from the target set to end HIV/AIDS by 2023.6–8 The highest burden (60–90%) was reported for developing countries, including Ethiopia.9–11 According to the WHO global disease burden report, 2016, 768, 040 people were living with HIV, 39, 140 were newly diagnosed with HIV infections, and 28, 650 were diagnosed with HIV/AIDS-related deaths in Ethiopia.12

In the fight against the HIV/AIDS pandemic, early initiation and enhanced adherence to anti-retroviral therapy (ART) are essential to reduce the progression of the virus through viral load (VL) suppression and increasing the level of CD4+cell count.7 13–15 When the VL is adequately suppressed (<1000 copies/mL), damage to the immune system will be reduced, the occurrence of opportunistic infections (OIs) that lead to morbidity and mortality will decrease, and the risk of HIV transmission will also decrease.16–19 This approach improves patients’ survival and health-related quality of life and increases productivity.13–15

Virological failure (VF) after enhanced adherence to counselling has become a global public health problem among patients receiving ART.20 21 The WHO guidelines for the detection and management of VF 2017 reported that 58% of patients receiving ART who remain on the same regimen experienced non-suppressed VL at the next measurement.19 In Sub-Saharan Africa, the VL suppression rate ranged from 40.2% to 77.4%.22 The magnitude of VF was reported to be 11% in Uganda,23 11.9% in Rwanda,24 15.7% in Ghana,25 16% in Swaziland,26 19.3% in South Africa,27 24% in Mozambique,28 24.6% in Kenya,29 20.8% in China,30 30.6% in Cameroon,31 32% in Malawi,32 34% in Tanzania33 and 41.3% in Gabon.34

The risk factors of VF among patients receiving ART include younger age, being male gender, history of substance use, lack of awareness of the benefits of viral suppression, poor adherence to ART, type of ART regimen, ART-induced side effects, baseline CD4+cell count <200 cells/mm3, advanced WHO clinical stage (WHO clinical stages III and IV), baseline high VL >1000 copies/mL, not using cotrimoxazole prophylaxis, OIs, TB-HIV coinfections and financial insecurity.35–49 Failure to control VL increases the rate of transmission, which is a challenge in achieving an end to the HIV/AIDS epidemic by 2030.50 51 VF brings low productivity, increased economic dependence and social problems.51

Although several primary studies have been conducted to determine the prevalence of VF and its associated factors among patients receiving ART in Ethiopia, these studies have reported inconsistent findings. Therefore, this study aimed to determine the pooled prevalence of VF and identify its pooled associated factors among patients receiving ART.

MethodsDatabases and search strategy

The studies were subsequently retrieved from the Google Scholar, Scopus, PubMed and Web of Science databases using search terms and phrases. Moreover, a manual search of the reference lists of the included studies was performed (online supplemental file 1). The search was held from 18 January to 31 January 2024, and all the studies were conducted in Ethiopia and published in English between 2014 and 2023.

Eligibility criteria

All observational studies that were conducted among patients receiving ART in Ethiopia and reported that the prevalence of VF and/or at least one factor associated with VF and published in English were included in the study. However, articles without abstracts and/or full texts, systematic reviews, meta-analyses or qualitative studies were excluded from the study.

Study selection

All the retrieved studies were exported to the EndNote V.7 reference manager to remove duplicate studies. Initially, two independent investigators (TMA and SDK) screened the titles and abstracts, followed by full-text reviews to determine the eligibility of each study. Discrepancies between the investigators were resolved through discussion.

Data extraction

The two independent reviewers (TMA and WNA) extracted the data using a structured Microsoft Excel sheet. When inconsistencies were observed in the extracted data, the phase was repeated. When inconsistencies between the extracted data were continued, the third reviewer (SDK) participated. The name of the first author and year of publication, study area, study design, sample size, response rate and effect size of the included studies were extracted.

Outcome measure

The main outcome of interest was VF among patients receiving ART in Ethiopia.

Operational definition

VF is defined as a persistently detectable VL exceeding 1000 copies/mL of plasma in two consecutive VL measurements within a 3-month interval, with enhanced adherence support between measurements after at least 6 months of initiating the ART regimen following the first VL test.18

Data analysis

The extracted data were exported to STATA V.17 for statistical analysis. The pooled prevalence of VF was estimated using a random-effects DerSimonian-Laird model.53 A funnel plot was used to assess publication bias. The asymmetric distribution of the funnel plot is an indicator of publication bias, and Egger’s test with a p value of <0.05 was used to determine a significant publication bias.54 The heterogeneity across the included primary studies was assessed using I2 statistics.55 The I2 statistics ranged from 0 to 100%, and I2 values of 0%, 25%, 50% and 75% were considered to indicate no, low, moderate and high degrees of heterogeneity respectively.55 A p value of the χ2 test on Cochrane’s Q statistic <0.05 was used to indicate significant heterogeneity.56 57 A sensitivity analysis was also conducted to determine the effect of a single study on the overall estimate. A forest plot was used to display the effect of independent factors on the outcome, and a measure of association with a 95% CI was generated. The adjusted OR (AOR) was the most commonly reported measure of association in the included primary studies.

ResultsSearch results

The search strategy retrieved 2916 studies from PubMed (1512), Google Scholar (1298), Scopus (79) and Web of Science (27) databases. After the removal of irrelevant studies based on their titles and abstracts (2178) and duplicate studies (167), a total of 571 studies were selected for full-text review.

Consequently, full-text reviews were obtained, resulting in the removal of 457 studies because of a lack of complete texts. A total of 114 studies were assessed for full-text reviews, and 63 studies were excluded. Finally, 51 studies were found to be relevant for determining the pooled prevalence of VF and identifying the factors associated with VF among patients receiving ART. The PRISMA flow chart58 was constructed to show the selection process (figure 1).

Figure 1

PRISMA flow chart showing the studies selection process, 2024.

Characteristics of the included primary studies

Twenty-three studies,59–81 14 studies82–95 and 14 studies96–109 were conducted using cross-sectional, retrospective cohort and case-control study designs, respectively. Concerning the study areas, 26,59–62 64 66 69–71 73 79 83 85 87 90–92 97 99 100 103–105 107–109 11,67 75–77 81 82 86 89 96 98 106 five,72 84 94 95 102 four65 78 93 101 and three63 68 74 studies were conducted in the Amhara, Oromia, Addis Ababa, Southern Nations and Tigray regions, respectively. Additionally, a study88 was conducted in public health facilities in the Harari, Somali, East and West Hararghe zones of Oromia and Dire Dawa city. Another study80 was also conducted in Ethiopian national defence hospitals. The total sample size of the included studies was 38, 789, where the smallest (75) and the largest (19, 525) sample sizes were obtained in the Addis Ababa city and Tigray region, respectively. The prevalence of VF was obtained from 32 included studies.59–76 78 79 81 85–95 One study80 was excluded from the prevalence estimation after checking the forest plot. However, it was not excluded from the meta-analysis for determining the associated factors. The data regarding the associated factors were also obtained from 47 studies,59–66 68 69 71–75 77–91 93–109 with the response rate ranging from 73.8 to 100% (online supplemental table 2).

Quality assessment

Two independent investigators (TMA and WNA) appraised the quality of the included primary studies. The quality of each study was appraised using the Joanna Briggs Institute (JBI) quality appraisal criteria.110 Twenty-three,59–81 14,82–95 and 1496–109 studies were appraised using the JBI checklist for cross-sectional, cohort and case-control study designs respectively. Of the cross-sectional studies, 15 scored seven of the eight questions (87.5%, low risk), five studies scored six of the eight questions (75%, low risk), and three studies scored five of the eight questions (62.5%, low risk). Similarly, of the cohort studies, 11 scored eight of 10 questions (80%, low risk), and three studies scored seven of the 10 questions (70%, low risk). Likewise, among the case-control studies, scored eight of the 10 questions (80%, low risk), and three studies scored seven of the 10 questions (70%, low risk) (online supplemental table 3).

Studies were considered low risk or good quality when they scored 50% or higher on the quality appraisal indicators. The studies scored between 5 and 7 out of a total of 8 points. Therefore, all the included studies59–109 were considered low risk or good quality.

Risk of bias assessment

The standardised adoptive assessment tool111 was used to evaluate the risk of bias. Thus, of the total of 51 included studies, 46 scored eight of the 10 questions, and five studies scored seven of the 10 questions. Studies were classified as ‘low risk’ if eight or more of the questions received ‘Yes’, as ‘moderate risk’ if six to seven of the 10 questions received ‘Yes’ and as high risk’ if five or lower of the 10 questions received ‘Yes’. Accordingly, all the included studies59–109 had a low risk of bias (good quality) (online supplemental table 4).

Meta-analysisPooled prevalence of virological failure

A total of 51 eligible studies59–109 with 38, 789 participants were included in the final meta-analysis. In this study, the pooled prevalence of VF among patients receiving ART was 15.95% (95% CI: 12.63, 19.27); I2=97.99%; p<0.001) (figure 2).

Figure 2

Forest plot showing the pooled prevalence of virological failure with 95% CIs in Ethiopia, 2024.

Publication bias

The funnel plot indicated a symmetrical distribution (figure 3), and the p value of Egger’s test was 0.0934, which indicated the absence of publication bias in the study.

Figure 3

Funnel plot showing publication bias of virological failure among patients receiving anti-retroviral therapy in Ethiopia, 2024.

Investigation of heterogeneity

The percentage of I2 statistics in the forest plot presented significant heterogeneity among the included studies (I2=97.99%; p<0.001) (figure 2). Therefore, sensitivity and subgroup analyses were performed to investigate potential sources of heterogeneity.

Sensitivity analysis

A sensitivity analysis was also conducted to determine the effect of a single study on the overall pooled estimate. The forest plot showed that the estimate of a single study was closer to the pooled estimate, indicating the absence of a single study effect on the overall estimate (figure 4).

Figure 4

Sensitivity analysis of virological failure among patients receiving anti-retroviral therapy in Ethiopia, 2024.

Subgroup analysis

Subgroup analysis was performed based on the study area and study design. Therefore, the highest pooled prevalence of VF was detected among studies conducted in southern nations, nationalities and peoples (20.37, 95% CI: 12.92, 27.82; I2=88.46%; p<0.001), and the lowest pooled prevalence was also detected among studies conducted in Addis Ababa (9.79, 95% CI: 4.16, 15.42; I2=86.71%; p<0.001). Similarly, the higher pooled prevalence of VF was among studies conducted using a cross-sectional study design (16.39, 95%C: 11.99, 20.78; I2=98.50%; p<0.001]) followed by studies conducted using a retrospective cohort [(14.91, 95% CI: 11.60, 18.21; I2=89.14%; p<0.001]) (table 1). Therefore, the heterogeneity of the study could be because of variations in study areas and designs across the included studies.

Table 1

Subgroup analyses of virological failure among patients on anti-retroviral therapy in Ethiopia, 2024

Factors associated with virological failure

Three studies77 88 101 reported a significant association between substance use and VF. The pooled AOR of VF for patients receiving ART who were substance users was 2.37 (95% CI: 1.28, 4.39; I2=5.48%; p<0.16).

Seven studies71 84 96 105 107–109 showed that non-disclosure of HIV status was significantly associated with VF. The pooled AOR of VF for patients who did not disclose their HIV status was 4.63 (95% CI: 3.51, 6.11; I2=20.85%; p<0.27).

Three studies59 66 71 indicated a significant association between interruption of ART and VF. The pooled AOR of VF for patients who interrupted their ART was 2.61 (95% CI: 1.55, 4.40; I2=0.00%; p<0.78).

Thirty-two studies59 61–66 68 71 77 79 81–84 86 87 89–91 93–97 99–101 103 104 107 109 also reported a significant association between non-adherence to ART and VF. The pooled AOR of VF for patients who were not adherent to ART was 4.31 (95% CI: 2.73, 6.79; I2=92.54%; p<0.001).

Twenty studies59 62 63 65 72 73 79 81–83 87 95–99 103–105 107 showed that a baseline CD4 cell count <200 cells/mm3 was significantly associated with VF. The pooled AOR of VF for patients with a CD4 count <200 cells/mm3 was 2.65 (95% CI: 1.62, 4.33; I2=92.59%; p<0.001).

Seven studies59 63 64 66 75 89 94 revealed a significant association between baseline advanced WHO clinical stage and VF. The pooled AOR of VF for patients with a baseline advanced WHO clinical stage was 3.02 (95% CI: 1.67, 5.45; I2=75.41%; p<0.001).

Six studies69 85 88 100 108 109 indicated that baseline VL >1000 copies/mL was significantly associated with VF. The pooled AOR of VF for patients with a baseline VL >1000 copies/mL was 4.63 (95% CI: 2.60, 8.26; I2=66.83%; p<0.01).

Eight studies66 71 78 84 98 100 101 107 reported a significant association between OIs and VF. The pooled AOR of VF for patients who developed OIs was 3.58 (95% CI: 2.58, 4.96; I2=0.00%; p<0.66).

Six studies59 69 74 78 95 98 showed that TB-HIV coinfection was significantly associated with VF. The pooled AOR of VF for patients who developed TB-HIV coinfection was 3.97 (95% CI: 2.41, 6.54; I2=47.81%; p<0.09).

Discussion

This study aimed to determine the pooled prevalence of VF and its associated factors among patients receiving ART in Ethiopia. This study reported that the pooled prevalence of VF among patients receiving ART in Ethiopia was 15.95% (95% CI: 12.63, 19.27; I2=97.99%; p<0.001), which was congruent with the findings of studies conducted in Ghana (15.7%)25 and Swaziland (16%).26 However, this study finding was higher than the study findings conducted in Uganda (11%)23 and Rwanda (11.9%).24 Conversely, this study finding was lower than the study findings conducted in South Africa (19.3%),27 China (20.8%),30 Mozambique (24%),28 Kenya (24.6%),29 Cameroon (30.6%),31 Malawi (32%),32 Tanzania (34%)33 and Gabon (41.3%).34 These discrepancies could be due to variations in study designs, sample sizes, eligibility criteria, inconsistency of treatment approaches and variability in VF measurements across the studies.66 87

Moreover, this study finding reported that patients who were substance users were 2.37 times more likely to develop VF than their counterparts. This study finding was supported by a study finding conducted in Vietnam.112 This might be because substances are prone to immune dysfunction and behavioural changes, resulting in poor adherence to ART, ultimately leading to VF.113 114

In this study, patients who did not disclose their HIV status were 4.63-fold more likely to experience VF than those who disclosed their HIV status. This study finding was in line with the findings of previous studies conducted in Uganda115 and Zimbabwe.116 ART is a lifelong treatment that demands the support of others. Therefore, individuals who disclose their HIV status could receive the necessary moral, psychological and material support when they are in need. Additionally, patients who declare their status to their loved ones and friends can take their medication freely in front of family members and even be reminded by others to take drugs on time in contrast to their counterparts.84

This study finding revealed that patients who interrupted their ART were 2.61 times more likely to experience VF than patients who did not interrupt. ART suppresses viral replication, resulting in boosting patients’ immunity, decreasing the incidence of OIs and reducing VL.117 Thus, drug interruption could be associated with decreased immune system efficacy and increased risk of drug resistance. As a result, patients may lose the ability to suppress their VL, ultimately leading to VF.34 118

Similarly, patients who were not adherent to ART were 4.31 times more likely to be exposed to VF than patients who were adherent to the ART regimen. This study finding was consistent with the study findings conducted in Kenya,119 Cameroon31 and Uganda.23 Poor adherence to ART reduces viral suppression due to suboptimal drug concentrations and subsequently increases viral replication, which leads to VF.18 120

In addition, this study finding showed that patients with a CD4 cell count of <200 cells/mm3 were 2.65 times more likely to be at risk of developing VF compared with patients with a CD4 cell count of ≥200 cells/mm3. This study finding was congruent with the findings of studies conducted in Cameroon31 and Tanzania.121 This could be explained that patients with a lower CD4 cell count (compromised immunity) are more susceptible to different OIs that ultimately lead to increased VF.122

Similarly, this study’s findings indicated that patients with a baseline advanced WHO clinical stage were 3.02 times more likely to encounter VF compared with their counterparts. This study finding was similar to those studies conducted in Swaziland26 and Uganda.123 This might be because patients who present with advanced WHO clinical stage are at greater risk of developing OIs. The presence of OIs during ART initiation increases the pill burden, which results in drug-drug interactions, and ultimately leads to VF.66 Moreover, VF occurs first, followed by immunological failure and clinical failure (WHO clinical stages III and IV).124

Moreover, this study finding revealed that patients with a baseline VL >1000 copies/mL were 4.63-fold more likely to develop VF than those patients with a baseline VL ≤1000 copies/mL. This study finding was supported by the findings of studies conducted in Swaziland26 and Uganda.125 A baseline of high VL indicates high virus replication, and the virus is resistant to antiretroviral drugs, which can easily result in VF.126

Furthermore, this study’s findings showed that patients who developed OIs were 3.58 times more likely to encounter VF than their counterparts. This study finding was consistent with a study conducted in Nigeria.127 OIs are common among patients living with HIV. Some OIs are recurrent and cause the patients to take multiple drugs; this in turn forces patients to focus on their current illness and give less attention to chronic conditions, possibly leading to pill fatigue. When patients become overwhelmed by the pill burden, they fail to take their ART regimen properly, leading to VF.128

Additionally, this study’s findings indicated that patients who developed TB-HIV coinfection were 3.97 times more likely to develop VF than those patients who did not develop TB-HIV coinfection. This study finding was supported by the findings of studies conducted in Cameroon31 and Uganda.23 The intimate linkage between HIV and TB could compress the immunity of the patient and enable the progression of HIV to the advanced stage rapidly, prohibiting patients from regular treatment intake and hence leading to VF.129

Conclusions

The pooled prevalence of VF among patients receiving ART was high in Ethiopia. This study’s findings showed that substance use, non-disclosure of HIV status, interruption of ART, poor adherence to ART, baseline CD4 cell count <200 cells/mm3, baseline advanced WHO clinical stage, baseline VL >1000 copies/mL, OIs and TB-HIV coinfection were the pooled independent predictors of VF. Therefore, due attention should be given to patients with these identified factors. Moreover, we recommend that researchers to conduct a triangulated study that can address additional factors associated with VF.

Data availability statement

All data relevant to the study are included in the article or uploaded as supplementary information.

Ethics statementsPatient consent for publication

Not applicable.

Ethics approval

Not applicable.

Acknowledgments

We would like to thank Henok Andualem for his unlimited support throughout the study.