Inclusion and exclusion criteria

Studies were included if they met the following criteria: published or presented between January 2005 and April 2019 and conducted in SSA. We included randomized controlled trials and cohort studies that recruited adult males (≥15years of age). Studies were excluded if they had other study designs (qualitative, cross-sectional, case control). We also excluded previous systematic reviews, articles that measured other outcomes other than retention, conference abstracts and letters to the editor.

Review protocol

The review protocol was registered in PROSPERO, the International Prospective Register of Systematic Reviews on 29 April 2020. The registration number is CRD42020142923 and can be found online: https://www.crd.york.ac.uk/prospero/display_record.php?ID=CRD42020142923.

Data sources and search strategies

Studies were identified through a systematic search in bibliographic databases; using the OVID^® interface for (MEDLINE, EMBASE and Global Health). Search terms were developed using relevant key words and medical subject headings (MeSH). The search terms are summarized in S1 Table. Duplicate studies were removed from the results using the “de-duplicate” feature in OVID^® and exported to EndNote reference management software (X7).

Screening and selection

Two reviewers (SK and RN) reviewed the identified studies (titles, abstracts, keywords) independently for the eligibility criteria and articles that did not meet the eligibility criteria were excluded at this stage. Articles from stage one were subjected to full text review in which the methods were reviewed for relevance towards the eligibility criteria. More articles were excluded at this stage. Two copies of full text documents were printed for all articles that met eligibility criteria. SK and RN reviewed these independently and were blinded to the results from each other. Any inconsistencies between were discussed and where consensus was not reached, a discussion was held with a third member of the team. Any missing information was not collected from any corresponding authors.

Data extraction

A standardized data extraction form was used to document information from each included study on the following: Data items collected included (a) first author’s name, (b) publication year, (c) country, (d) sample size for males and total population, (e) study design and (f) interventions used to promote retention. Information on outcome measures was also collected.

Risk of bias assessment

Risk of bias within the included studies was assessed using the Cochrane risk of bias tool for non-randomized studies in interventions (ROBINS-I) [24].

Summary measures

We sought to establish retention proportion/rate, attrition rate/proportion, relative risk, hazard ratios, and odds ratios.

Data synthesis

Due to the heterogeneity of studies a meta-analysis was not considered and narrative synthesis of the included papers was conducted. Results were summarized using PRISMA flowchart [23].

Ethics statement

This work did not require an ethics statement.

Summary of search results

A total of 1620 records were identified through the literature search. Titles and abstracts were reviewed and 1570 articles were excluded because they did not meet the inclusion criteria. An updated search identified 557 records of which 334 were excluded because they did not meet the inclusion criteria.

Study selection

A total of 93 articles were subjected to full text review to determine eligibility (Fig 1), 18 met eligibility. During data extraction, five articles were excluded because they were protocols for planned studies. Therefore, 13 studies were included in this review. An updated search was conducted on 04 December 2020 and 4 additional records were subjected to full text review, and 3 articles were excluded because two did not have clear description of interventions and one did not disaggregate the outcomes by gender. Only one article was include bringing the total number of articles included in this review to 14.

Study characteristics

A summary description of the included studies is presented in Table 1 (adapted from the Cochrane library) [25]. The studies were conducted in eight African countries: Democratic Republic of Congo, Ethiopia, Malawi, Mozambique, Rwanda, South Africa, Tanzania and Uganda between 2005 and December 2020. The number of males in included studies varied and ranged from 122 to 21101, a proportion of 32% of the total population.

Table 1

Studies that met inclusion criteria.

Author, year, Country	Journal	Sample size Males (Total)	Age (years)	Study design	Intervention	Outcome measure	RetentionMales (Female)
Decroo, 2017, Mozambique [15]	BMJ Open	884(2406)	≥15	Retrospective Cohort	Joining Community ART Groups (CAGs)	Retention at 12 and 24 months	88.2(92.4), 80.8 (88.9) Risk of attrition (aHR: 1.80, 95%CI 1.41–2.51)
Wringe, 2018, Malawi [16]	JIAS	7695(22633)	≥18	Retrospective Cohort analysis	Six-monthly clinical consultation schedule	Attrition rate	37.9/1000pys (30.5/1000pys) aHR = 1.3
Decroo, 2014, Mozambique [26]	Tropical Medicine and International Health	1746(5729)	30–43	Retrospective Cohort	Joining Community ART Groups	Attrition proportion, HRs	Attrition = 47.8 uHR = 2.07(95%CI: 1.59–2.70)
							aHR = 1.93(95%CI: 1.48–2.51)
Fatti, 2012, South Africa [27]	Implementation and Operational Research: Clinical Science	21101(66953)	29.4–42.3	Observational Cohort	Receiving community based adherence support	LTFU	uHR = 1.23 (95%CI: 1.16–1.30)
							aHR = 1.34 (95%CI: 1.24–1.44)
Nabaggala, 2018, Uganda [28]	BMC Research Notes	122(381)	23-35(IQR)	Retrospective Cohort	Tracking of PLHIV	Return to clinic proportion	56.6 (76.9)
Rich, 2012, Rwanda [29]	Implementation and Operational Research: Clinical Science	349(1041)	≥18	Retrospective Cohort	Enrolment in a community based ART program	Retention proportion	32.1(67.8)
						Attrition proportion	40 (40)
Tsondai, 2017, South Africa [30]	JIAS	948(3216)	≥16	Retrospective Observational Cohort	Enrolment in adherence clubs	LTFU HRs	No difference
Kipp, 2012, Uganda [31]	PLOS One	163(385)	≥18	Comparative Cohort	Community Vs Facility based ART delivery program	LTFU proportions	Community based = 50.0%(28/56)
							Facility based = 48.3% (28/58). p = 0.854
Megereso, 2016, Ethiopia [32]	BMC Health Services Research	834(1895)	≥18	Retrospective Cohort	Treatment in a primary health center Vs Hospital	Survival HR	aHR = 1.4 (95%CI: 1.1–1.7)
Akilimali, 2017, DRC [33]	PLOS One	238(717)	>18	Cohort	Disclosure of HIV status	LTFU rate per 1000pys	48.9 (25.5)
Siril, 2017, Tanzania [34]	AIDS Research and Therapy	208(824)	≥18	Prospective Cohort	NAMWEZA“Yes, together we can” Receiving psychosocial support	LTFU	Male gender was associated with higher risk of LTFU, p = 0.04†
Shearer, 2016, South Africa [35]	BMJ Open	4943(12219)	≥18	Observational Cohort	Reporting identification status	Attrition proportion	23.0 (15.8)
Bilinski, 2017, Malawi [36]	PLOS One	1422(3949)	33 (mean)	Retrospective Cohort	Travel distance to health center for care	Hazard ratio (HR)	uHR = 1.64 (95%CI: 1.46±1.84)
							aHR = 1.62 (95% CI: 1.44±1.82) p<0.0001
Bock, 2019, South Africa [37]	JIAS	166 (465)	≥18	Retrospective Cohort Analysis	Referral to adherence Clubs	LTFU (HR)	No difference by genderaHR 1.09 (95%CI:0.7–1.69) p = 0.704

aHR = adjusted hazard ratio, BMC = Biomedical central, BMJ = British Medical Journal, CI = Confidence interval, JIAS = Journal of International AIDS Society, IQR = Inter quartile range, LTFU = Loss to follow up, PLHIV = person living with HIV, PY = Person years, uHR- Unadjusted hazard ratio.

^† Measure of association not provided for males.

Risk of bias within studies

The quality of each study was assessed using the Cochrane risk of bias tool for non-randomized studies [24]. The risk of bias with regards to the different domains within included studies varied from low, moderate to serious. With regards to confounding, there was low to moderate risk of bias (Table 2). The bias in methods for selection of study participants was moderate in 10 studies [15,16,28,29,31–36], low in 3 [27,30,37], while there was a serious risk of bias due to absence of data on some variables that were considered confounders [26]. There was bias in some studies due to misclassification of interventions due to recall bias from study participants in one study and [33] and missing data in others [15,16,27–32,34–37].

Table 2

Risk of bias within studies.

		Study
		Decroo, 2017[15]	Wringe, 2018[16]	Dercoo, 2014[26]	Fatti, 2012[27]	Nabaggala, 2018[28]	Rich, 2012[29]	Tsondai, 2017[30]	Kipp, 2012[31]	Megereso, 2016[32]	Akilimali, 2017[33]	Siril, 2017[34]	Shearer, 2016[35]	Bilinski, 2017[36]	Bock, 2019 [37]
Risk of bias domains	Confounding	-	-	-	+	-	-	-	-	-	-	-	-	-	-
	Bias in selection of participants into the study	-	-	x	+	-	-	+	-	-	-	-	-	-	+
	Bias classification of intervention	+	-	-	-	-	-	-	-	-	-	-	-	x	+
	Bias due to deviations from intended interventions	-	-	-	-	-	-	-	-	-	+	-	-	-	+
	Bias due to missing data	-	-	-	-	-	-	-	x	-	x	-	+	-	-
	Bias in measurement of outcomes	-	-	-	-	-	-	-	-	-	-	-	+	-	-
	Bias in selection of the reported result	+	+	+	+	+	+	-	+	+	-	+	+	+	+

Key.

+ = Low risk of bias: Low risk of bias -the study is comparable to a well-performed randomized trial with regard to this domain.

_ = Moderate risk of bias: Moderate risk of bias-the study is sound for a non-randomized study with regard to this domain but cannot be considered comparable to a well-performed randomized trial.

X = Serious risk of bias: Serious risk of bias-the study has some important problems in this domain.

Critical risk of bias-the study is too problematic in this domain to provide any useful evidence on the effects of intervention.

No information on which to base a judgement about risk of bias for this domain.

The risk of bias due to deviations from interventions was generally moderate with two studies having low risk of bias [33,37] and another providing no information on deviations [15,16,26–32,34–36]. There was moderate bias due to missing data but most studies adjusted for this in the analysis. However, one study had serious risk of bias in the same domain because they relied on patients to provide information on disclosure [33] while another indicated the lack of information as a limitation for the study [28].

The bias in measurement of outcomes was moderate in most studies as most authors adjusted for the factors that may affect the outcome in data analysis. The bias in selection of the reported result was moderate to low. In most studies, the results were reported as indicated while two studies had moderate risk of bias in this domain as there was no comparison of outcomes in the two groups being studied [30,33].

Retention strategies and outcome measures

In this review, seven studies (7/14) explored community based adherence support groups [15,26–30,37] while two (2/14) compared use of facility versus community-based delivery models [31,32]. Other studies measured the effect of disclosure of HIV status [33], giving psychosocial support [34], national identity cards [35], distance from the health center [36] and six-monthly clinic visits [16]. Retention of men was provided as proportions in two studies [26,29], an average of 56.5% at 24months. Five studies provided risk of attrition from care using hazard ratios ranging from 1.2–1.8 [15,26,27,32,36], four studies documented attrition proportions at an average of 40.0% [26,29,31,35] and two studies an average rate of attrition of 43.4/1000 person years [16,33].

The engagement of men in HIV care is important in ensuring epidemic control and achievement of UNAIDS targets [20]. Therefore, it is necessary to identify and set up strategies that will effectively ensure the retention of men in HIV care [38]. This review showed that few studies have investigated interventions that can be used to retain heterosexual males in HIV care. What is interesting is no study was found that focused on men as a population. Various studies have focused on involving men in PMTCT programs [39,40].

We also found that in this review the highest proportion of men retained in care was observed when community adherence groups were used as an intervention. Although retention in the population was high, the risk of attrition of men from care was at 80% [15]. Other studies also showed a high risk of attrition for men [16,26,27,32,36]. This is similar to what has been studied as predictors for poor retention in various studies in sub Saharan Africa where male gender has been highlighted as a risk factor for mortality and attrition from care [41–44]. Whereas retention is a bigger challenge for men relative to women, various studies did not provide a disaggregation of retention by gender; posing a challenge in identifying interventions that may work for men [45–50]. In contrast, most studies on engaging men who have sex with men have been undertaken outside SSA [51–55] with retention reported at about 64% [50]. A study done in Kenya showed that men who have sex with men (MSM) had lower retention at 12 months compared to heterosexual men and women when ART was received at the clinic compared to when it was not [56].

Although retention in care has been posed as a general challenge in SSA, most especially among men [44] similar sentiments have been seen elsewhere in the developed world like the United States of America where disparities in retention have been reported [57]. Gender and race account for differences in retention and also affect access to care with males and blacks being at higher risk of discontinuing care [58]. However, studies among those disproportionately affected by HIV have shown some improvement in retention in care with community based interventions facilitating engagement in care [59] while mobile phone used improved retention of these groups from 51% to 81% at 12 months [60].

Strengths and limitations

To the best of our knowledge, this is the first systematic review aimed at summarizing strategies to reduce attrition of men from HIV care. However, there are a few limitations to this review. Firstly, only a few studies on retention of heterosexual men in HIV care have been performed and only eight countries in sub Saharan Africa are represented. Eight of these were in southern Africa and six in East Africa. This may limit the generalizability of the findings to central and western Africa that may have different gender-related sociocultural practices. Additionally, the outcome measures used in the identified studies varied widely, making it impractical to perform a pooled analysis.

It is important to note that all the studies that included interventions to improve retention of men in HIV care were retrospective cohorts, making it challenging to account for the effects of confounding factors. While randomized controlled trials (RCT) are the gold standard in evaluating interventions including their effect sizes [61], none of the included studies in this review was an RCT. Therefore, the evidence from this review should be interpreted with caution.

Lastly, we didn’t explore whether the males lost in one program were identified or seen in another. Most of the studies used programmatic data that may not provide the true estimates for retention of men in care.