Prevalence and predictors of virological failure in pediatric patients on HAART in sub-Saharan Africa: a systematic review and meta-analysis

Antiretroviral treatment failure has emerged as a challenge in the management of pediatric human immunodeficiency virus (HIV) patients, especially in resource-limited countries despite accessibility to Highly Active Antiretroviral Therapy (HAART). A systematic review and meta-analysis was conducted to synthesize virological failure (VF) prevalence and ascertain its predictors in children in sub-Saharan Africa. An electronic database search strategy was conducted from January to September 2021 on PubMed, EMBASE, SCOPUS, HINARI, and CINAHL. Further, manual searching was conducted on non-indexed journals. Utilizing the JASP© version 0.17.2 (2023) statistical software, a meta-analysis of pooled prevalence of VF was estimated using the standardized mean differences. Further, selection models were used to assess the risk of bias and heterogeneity. The pooled odds ratios were estimated for the respective studies reporting on predictors of VF. The overall pooled estimate of the prevalence of VF in sub-Saharan Africa among the sampled population was 29% (95% CI: 27.0-32.0; p<0.001). Predictors of VF were drug resistance (OR: 1.68; 95% CI: 0.88-2.49; p < 0.001), poor adherence (OR: 5.35; 95% CI: 5.26-5.45; p < 0.001), nevirapine (NVP)-based regimen (OR: 5.11; 95% CI: 4.66-5.56; p < 0.001), non-usage of cotrimoxazole prophylaxis (OR: 4.30; 95% CI: 4.13-4.47; p < 0.001), higher viral load at the initiation of antiretroviral therapy (ART) (OR: 244.32; 95% CI: 244.2-244.47; p <0.001), exposure to the prevention of mother to child transmission (PMTCT) (OR: 8.02; 95%CI: 7.58-8.46; p < 0.001), increased age/older age (OR: 3.37; 95% CI: 2.70-4.04; p < 0.001), advanced World Health Organization (WHO) stage (OR: 6.57; 95% CI: 6.17-6.98; p < 0.001), not having both parents as primary caregivers (OR: 3.01; 95% CI: 2.50-3.53; p < 0.001), and tuberclosis (TB) treatment (OR: 4.22; 95% CI: 3.68-4.76; p <0.001). The mean VF prevalence documented is at variance with studies in other developing countries outside the sub-Saharan region. The high prevalence of HIV cases contrasting with the limited expertise in the management of pediatric ART patients could explain this variance.


Introduction
Human immunodeficiency virus (HIV) remains a major global public health problem with people living with HIV/AIDS being estimated at 38.0 million as of December 2019 [1].However, the introduction of Highly Active Antiretroviral Therapy (HAART) has reduced HIV/AIDS-related morbidity and mortality, hence improving the quality of life of people living with HIV/AIDS [2].The newly proposed 95-95-95 strategy targets achieving viral suppression in 95.0% of HIVpositive patients on sustained HAART [1].In 2019, an estimated 81% of all people living with HIV knew their status with 67% receiving HAART of which 59% achieved viral suppression [1].In addition, 53% of children living with HIV globally received lifelong antiretroviral therapy (ART) [1].The number of children living with HIV accessing HAART has increased drastically [3].This has improved survival rates in pediatric HIV-positive patients including in sub-Saharan Africa [3].
However, ART treatment failure has emerged as a challenge in pediatric HIV patients on HAART [4].Pediatric ART treatment failure rates of 19.3% to over 32% in resource-limited countries including sub-Saharan Africa have been reported [5].Virological failure (VF) is a plasma viral load above 1000 copies/ml in an HIV patient after two consecutive viral load measurements 3 months apart, despite adherence to HAART for at least 6 months [6].Viral load (VL) monitoring for all patients on HAART is the most accurate available measure of the effectiveness of treatment response and a key in diagnosing treatment failure [1].Some associated factors with VF include poor adherence [7], inadequate dosing, and viral resistance [8].However, little is known about the magnitude of HIV virological treatment failure and its predictors in pediatric patients in sub-Saharan Africa.Therefore, this review discusses the prevalence of VF and associated factors in children aged above 6 months and adolescents on HAART in sub-Saharan Africa.Furthermore, the study provides proportionate effects of the associated factors of VF.

Methods
Registration of review: this review was registered in Prospero with registration: PROSPERO 2021 CRD42021230120.
Eligibility criteria: all cross-sectional, case-control, cohort, and randomized controlled studies conducted in sub-Saharan Africa documenting the prevalence and associated factors influencing VF in children above 6 months and adolescents on the treatment of HIV/AIDS were selected.The studies were reported in English and had a definable characterization of VF.Further, Studies outside the sub-Saharan African region or focused on non-HIV/AIDS-related treatment failure, studies that sampled adults, reviews, health demographic surveys, and editorials were not included.
Information sources: an electronic database search strategy was conducted from January 2021 to September 2021 on PubMed, EMBASE, SCOPUS, HINARI, and CINAHL.Further, manual searching was conducted on non-indexed journals: Web of Science, IBSS, BioMed Central, Directory of Open Access Journals (DOAJ), WHO electronic Library of Evidence for Nutrition Actions, and Google Scholar.The search for grey literature was conducted by contacting experts in retroviruses and opportunistic infections, following the reference list for potential articles and abstracts from HIV/AIDS scientific conferences.
Selection process: the articles retained after building up queries in the electronic databases were populated in RefWorks (2020) to remove duplicates.Two reviewers (NM and LL) independently retrieved the articles and screened them based on the title and abstract.The articles in doubt were further screened by a third review (PH) and any disagreements were resolved through discussions among the three reviewers.The primary outcome was plasma viral load above 1000 copies/ml based on two consecutive viral load measurements done 3 months apart in a patient who has been on HAART for at least 6 months.The secondary outcomes were the associative factors influencing VF.
Data extraction and management: data on study characteristics and statistical parameters deciphering correlation or causality were abstracted into Microsoft Excel Spreadsheet.Thus, a preconceived and standardized abstraction form was formulated based on the PRISMA [9] guidelines for conducting systematic reviews.Two independent reviewers (MN and LL) each populated the abstraction form with the name of authors, year of publication, the country where the study was conducted, the title of the research, study design, aim, sampling strategy, sample size, and prevalence, study setting, characterization of VF (i.e.measurement), and potential associative factors.
Methodological quality assessment: the Hoy [10] tool for assessing the risk of bias in prevalence studies was utilized to evaluate the methodological rigor of the potential studies.The two reviewers (NM and LL) independently evaluated the quality of the studies, and any discordancy was resolved through the third reviewer (PH).As assessment of the validity and reliability of the measurement tools, an indication of potential confounders, and the relevance of statistical analysis used were strictly followed in each of the studies which passed the inclusion criteria.Further, the studies were ranked based on meeting the quality checks in the Hoy [10] tool in order to ascertain the risk of bias.
Data synthesis and analysis: to statistically assess the pooled effects of VF among pediatric patients on HAART, a meta-analysis was conducted to provide an overall analysis for prevalence and predictors of VF, subgroup analysis according to study design and region, and heterogeneity and risk of bias.The JASP © version 0.17.2 (2023) statistical software was utilized for the metaanalysis.

Results
Study selection: the systematic electronic database searching strategy employed retained the following number of articles in each database: PubMed (n=502), EMBASE (n=123), SCOPUS (n=2), HINARI (n=13) and CINAHL (n=14).A manual searching strategy conducted on non-indexed journals retained hundred and twenty (n=120) articles; of which forty (n=40) were editorials, and fifty-three (n=53) were systematic reviews, thus, were immediately excluded.This brought the total number of articles retained after manual searching to twenty-seven (n=27).Combining the articles obtained from manual searching with those from systematic electronic database searching brought the total number of articles retained to six hundred and eighty-one (n=681).However, thirtyeight (n=38) articles were duplicates and were, therefore removed.Two investigators (MN and LL) further conducted independently the title and abstract screening of six hundred and forty-three (n=643) articles.
One hundred sixty-four (n=164) articles passed the title and abstract screening while four hundred and seventy-nine (479) failed.Three hundred and forty-seven (n= 347) articles were conducted outside sub-Saharan Africa, one hundred and nine (n= 109) articles sampled study participants outside the pediatric age group and/or had a mixture of children and adults as study participants while twenty-three (n= 23) did not clearly state the prevalence or predictors of VF in children.The articles which passed the title and abstract screening (n=164) were further subjected to full-text screening against the inclusion criteria by two investigators (MN and LL).Forty-four (n=44) articles were within the inclusion criteria and included in the systematic review.The one hundred and twenty-one (n=121) articles that were excluded after a full-text screening did not highlight any prevalence of VF and/or its predictors in children.The reference list of the articles which passed the full-text screening was also checked for relevant articles.However, no article was obtained through this process.The electronic and manual search process is illustrated in the flow diagram in Figure 1.

Meta-analysis
Assessment of heterogeneity and publication bias: utilizing selection models in JASP© 0.17.2 (2023) statistical software, the test for heterogeneity gives a Q value of 457.7 (p <0.001) signifying evidence for heterogeneity.Further, using the recommended p value threshold of 0.1 for the selection model, the analysis indicates no evidence of publication bias (p = 0.697).The mean model estimate is illustrated in Figure 2. In addition, the Egger´s regression test also supports the absence of publication bias (p = 0.119) as illustrated in the funnel plot in Figure 3.
Influential analysis: an assessment of sources of heterogeneity indicated the absence of influential impact among the 42 studies considered for pooled estimates of prevalence.Thus, there were no significant changes in the fit models that were indicative of influencing heterogeneity.
Predictors of virological failure (VF): the predictors of VF were documented in 37 studies included in this review.These were drug resistance, poor adherence, NVP-based regimen, non-usage of cotrimoxazole prophylaxis, higher viral load at the initiation of ART, exposure to PMTCT, increased age/older age as a risk factor for VF, advanced WHO stage, not having both parents as primary caregivers, and TB treatment.
Poor adherence: of the 44 studies considered in this review, eleven studies [14,21,23,29,30,36,[38][39][40]44,53] documented poor adherence to HAART as a predictor of VF from a sample population of 3928.Study participants with poor adherence were more likely to have VF when compared to those who had good adherence.This was statistically indicated by a pooled odds ratio of 5.35 (95% CI: 5.26-5.45;p < 0.001) (Figure 6).
Non-usage of cotrimoxazole prophylaxis: it has been established that non-usage of cotrimoxazole prophylaxis likely increases the occurrence of VF.In two studies [27,49] with a sample population of 880, a pooled odds ratio of 4.30 (95% CI: 4.13-4.47;p < 0.001) was indicative of the significant influence of the non-usage of cotrimoxazole prophylaxis has on VF.
Higher viral load at the initiation of ART: a viral load of >1000 copies/ml at ART initiation was an important predictor of VF.It has been demonstrated that children with an initial viral load > 1000 copies are more likely to experience VF.In five studies [12,23,24,30,48] with a sample population of 7266, a pooled odds ratio of 244.32 (95% CI: 244.2-244.47;p <0.001) indicated the significant influence of higher viral load at ART initiation on VF.
Exposure to PMTCT: the review also documented exposure to prevention of mother to child transmission especially when NVP regime is used likely influenced VF.This was documented in four studies [12,23,30,41] with a sample population of 7050 and a pooled odds ratio of 8.02 (95%CI: 7.58-8.46;p < 0.001).
Increased age/older age as a risk factor for VF: adolescent age was associated with an increased risk of VF.This was reported in three studies [16,22,28] with a sample population of 1439 and a pooled odds ratio of 3.37 (95% CI: 2.70-4.04;p < 0.001).
Advanced WHO stage: the advanced WHO HIV clinical stage was associated with an increased risk of VF.This was reported in six studies [4,24,37,38,42,54] with a sample population of 2290 and a pooled odds ratio of 6.57 (95% CI: 6.17-6.98;p < 0.001).

Not having both parents as primary caregivers:
interestingly, not having both parents as primary caregivers was a major predictor of VF.It was documented that HIV-positive children whose both parents died were likely to experience VF.A motherless orphan had a higher risk of experiencing VF compared with those whose mothers were alive.This was reported in five studies [31,34,37,51,54] with a sample population of 9388 and a pooled odds ratio of 3.01 (95% CI: 2.50-3.53;p < 0.001).
A child being on TB treatment: concomitant treatment of tuberculosis in an HIV-infected child on HAART can lead to VF.In addition, children who had tuberculosis at baseline were more likely to have VF.This was reported in four studies [14,15,31,38] with a sample population of 8388 and pooled odds ratio of 4.22 (95% CI: 3.68-4.76;p <0.001).

Discussion
The findings of this review indicate a mean prevalence of 29% of VF among children aged below 15 years in sub-Saharan Africa.The regional analysis ranked West Africa (21%) with the lowest VF pooled estimate rate while Central Africa (60%) had the highest.However, it is important to highlight that only one study was considered from Central Africa.Therefore, the reported mean value might not reflect the picture across the Central African region.The predictors of VF identified in this systematic review and meta-analysis were drug resistance, poor adherence, NVP-based regimen, non-usage of cotrimoxazole prophylaxis, higher viral load at the initiation of ART, exposure to PMTCT, increased age/older age as a risk factor for VF, advanced WHO stage, not having both parents as primary caregivers, and TB treatment.
The findings of this review are at variance with what has been documented in studies in other developing countries outside the sub-Saharan region.In Thailand, Cambodia, and India, the mean VF ranged from 3.3 to 27% in children treated for HIV 1 using the WHO standard first-line ART for a period not less than six months [55][56][57][58][59].This highlights a higher success rate in the management of HIV compared to rates obtained in sub-Saharan Africa.Differences in settings could explain this variance since sub-Saharan Africa accounts for most HIV cases coupled with several challenges in the management of pediatric ART patients such as limitations in resources availability and lack of expertise in the management of pediatric ART patients.

Concerning
the predictors of VF, immunosuppression and high VL at the initiation of HAART were documented.In agreement with this review, several studies have shown the association of better HAART response with less HIV-1-associated immunodeficiency and lower HIV RNA viral load at diagnosis and before starting HAART [55,56,[58][59][60].All these findings reinforce the importance of initiating an early HAART in HIV pediatric patient.Furthermore, this review showed that poor adherence and increased age for patients on HAART were risk factors for VF.These findings are similar to other studies conducted in the United Kingdom and Asia [57,61].Poor adherence has been long known to be associated with VF as missing doses of HAART are likely associated with reduced levels of drugs below the therapeutic levels in blood consequently leading to increased risk of VF.In addition, poor adherence in the adolescent age group remains common especially with older age likely associated with rebellious behavior leading to poor adherence-subsequently to resistance and VF.Comorbidity of TB in an HIVpositive patient on HAART was equally a predictor of VF.This finding is congruent with what was documented in pediatric populations in six countries in Asia [57], a setting that is similar to sub-Saharan Africa in terms of disease burden.
Caregivers have a critical role to play in the management of pediatric ART patients and this has a bearing on the ART outcomes.In this review, the non-availability of biological parents as primary caregivers has been associated with an increased risk of VF [34,37,40].This finding is in consonant with a multicenter retrospective cohort study in the Asian-Pacific [59].Having family members other than biological parents/grandparents as primary caregivers increased the risk of subsequent VF among Asian children living with HIV.Thus, it emphasizes that dedicated caregivers are critical in assuring care and adherence in children who are not able to fend for themselves.Another factor predictive of VF based on this review was advanced WHO stage [4,24,37,38,40,42], a finding that is consistent and associated with worsening immunodeficiency which in turn leads to advanced WHO stage.The advanced WHO stage is the tail end of the consequence of VF indirectly i.e. clinical failure.
Various studies in this review documented perinatal exposure to ARVs as a predictor of VF for patients who were put on the first-line ART drugs [12,23,30,41].This was a common case for patients who received single doses of NVP without a tail-end cover which in turn raised the risk of resistance to the NNRTI-based regimen.These findings are in congruence with studies conducted in the United Kingdom (UK) which highlighted maternal exposure to ART perinatal increased the risk of VF were NVP based regimen [61].
In addition, this review has demonstrated drug resistance as a major contributor to VF with several studies documenting drug resistance mutations for both NNRTI and PI as significant risks for VF [28,40,42,44,47,53].These findings are similar to a cohort study [62] in France in which the prevalence of resistance to any drug was 82.4%.Further, resistance ranged from 76.5% for NRTI, to 48.7% for NNRTI and 42.9% for PIs.Resistance to at least one drug of two classes and three classes (triple resistance) was 31.9% and 26.9%, respectively.These findings highlight drug resistance is one of the key factors contributing to VF in children and should be a priority concern in patients failing on ART.

Tables and figures
Table 1: characteristics of the 44 included studies reporting either the prevalence or the predictor(s) of virological failure among pediatric patients on HAART in sub-Saharan Africa Table 1 (suite): characteristics of the 44 included studies reporting either the prevalence or the predictor(s) of virological failure among pediatric patients on HAART in sub-Saharan Africa Figure 1: flow chart illustrating the study selection process for the systematic review and metaanalysis of the prevalence and predictors of virological failure among pediatric patients on HAART in sub-Saharan Africa Figure 2: mean model estimates assessing heterogeneity through the standardized estimated effects from included studies (42 studies on prevalence) Figure 3: funnel plot for publication bias, Egger's regression test (p = 0.119) supporting the absence of publication bias using effect sizes on the x-axis and standard error on the y-axis

Figure 4 : 6 : 7 :
Figure 4: Forest plot of the prevalence (effect size) of virological failure with respective 95% confidence intervals from the 42 studies which estimated the prevalence (presented as ratios) Figure 5: the pooled effects of drug resistance mutation (DRM) on virological failure from 6 studies with a total sample population (N) of 1,479 pediatric patients in sub-Saharan Africa Figure 6: the pooled effects of poor adherence on virological failure from 11 studies with a total sample population (N) of 3,928 pediatric patients in sub-Saharan Africa Figure 7: the pooled effects of nevirapine-based regimen (NVP) on virological failure from 6 studies with a total sample population (N) of 7,408 pediatric patients in sub-Saharan Africa

Figure 1 :
Figure 1: flow chart illustrating the study selection process for the systematic review and meta-analysis of the prevalence and predictors of virological failure among pediatric patients on HAART in sub-Saharan Africa

Figure 2 :Figure 3 :
Figure 2: mean model estimates assessing heterogeneity through the standardized estimated effects from included studies (42 studies on prevalence)

Figure 4 :Figure 5 :Figure 6 :
Figure 4: forest plot of the prevalence (effect size) of virological failure with respective 95% confidence intervals from the 42 studies which estimated the prevalence (presented as ratios)

Figure 7 :
Figure 7: the pooled effects of nevirapine-based regimen (NVP) on virological failure from 6 studies with a total sample population (N) of 7,408 pediatric patients in sub-Saharan Africa

Table 1 :
characteristics of the 44 included studies reporting either the prevalence or the predictor(s) of virological failure among pediatric patients on HAART in sub-Saharan Africa

Table 1 (
suite): characteristics of the 44 included studies reporting either the prevalence or the predictor(s) of virological failure among pediatric patients on HAART in sub-Saharan Africa