Acute and early HIV infection screening among men who have sex with men, a systematic review and meta‐analysis

Abstract Introduction Screening for acute and early HIV infections (AEHI) among men who have sex with men (MSM) remains uncommon in sub‐Saharan Africa (SSA). Yet, undiagnosed AEHI among MSM and subsequent failure to link to care are important drivers of the HIV epidemic. We conducted a systematic review and meta‐analysis of AEHI yield among MSM mobilized for AEHI testing; and assessed which risk factors and/or symptoms could increase AEHI yield in MSM. Methods We systematically searched four databases from their inception through May 2020 for studies reporting strategies of mobilizing MSM for testing and their AEHI yield, or risk and/or symptom scores targeting AEHI screening. AEHI yield was defined as the proportion of AEHI cases among the total number of visits. Study estimates for AEHI yield were pooled using random effects models. Predictive ability of risk and/or symptom scores was expressed as the area under the receiver operator curve (AUC). Results Twenty‐two studies were identified and included a variety of mobilization strategies (eight studies) and risk and/or symptom scores (fourteen studies). The overall pooled AEHI yield was 6.3% (95% CI, 2.1 to 12.4; I2 = 94.9%; five studies); yield varied between studies using targeted strategies (11.1%; 95% CI, 5.9 to 17.6; I2 = 83.8%; three studies) versus universal testing (1.6%; 95% CI, 0.8 to 2.4; two studies). The AUC of risk and/or symptom scores ranged from 0.69 to 0.89 in development study samples, and from 0.51 to 0.88 in validation study samples. AUC was the highest for scores including symptoms, such as diarrhoea, fever and fatigue. Key risk score variables were age, number of sexual partners, condomless receptive anal intercourse, sexual intercourse with a person living with HIV, a sexually transmitted infection, and illicit drug use. No studies were identified that assessed AEHI yield among MSM in SSA and risk and/or symptom scores developed among MSM in SSA lacked validation. Conclusions Strategies mobilizing MSM for targeted AEHI testing resulted in substantially higher AEHI yields than universal AEHI testing. Targeted AEHI testing may be optimized using risk and/or symptom scores, especially if scores include symptoms. Studies assessing AEHI yield and validation of risk and/or symptom scores among MSM in SSA are urgently needed.

Acute HIV infection (AHI) is typically defined as the first weeks after HIV acquisition, during which HIV antibodies are undetectable [15]. AHI can be diagnosed with HIV-RNA testing using nucleic acid amplification testing (NAAT) and/or HIV p24-antigen testing [16,17]. Early HIV infection (EHI) is usually defined as the first months after HIV acquisition [18,19]. In this period, HIV antibody tests are often indeterminate. Therefore, diagnosis of EHI requires a combination of HIV antibody, HIV-RNA, and/or p24 assays [8,[18][19][20]. While AEHI testing, here defined as testing with a combination of HIV antibody, HIV-RNA and p24 assays, was not available in most of SSA until recently, the emergence of point-of-care HIV-RNA testing in SSA enables AEHI testing among a range of populations [21]. In some well-resourced countries, national guidelines recommend AEHI testing for people who report risk behaviour and symptoms associated with AEHI [22,23], and facility-based AEHI testing with HIV-RNA can successfully identify AEHI among MSM [16,[24][25][26][27][28][29]. Unfortunately, global policies do not recommend AEHI testing for MSM [30].
Modelling and phylogenetic transmission studies suggest that 10% to 50% of HIV transmission events occur during AEHI [8,[31][32][33][34][35]. In order to reduce HIV incidence among MSM, screening strategies should target MSM with the highest risk behaviour, as AEHI yield will be the highest [36]. Ideally, all people at risk of HIV acquisition should be tested for AEHI. However, this may not be feasible in less-resourced settings due to the high costs of AEHI testing. Focussing on yield would therefore limit the number of people that require AEHI testing, while increasing the number of people diagnosed with AEHI [36]. Behaviour risk scores can identify MSM with highrisk behaviour [37,38]. Thus, risk and/or symptom scores may assist in defining which subpopulations should be targeted for AEHI testing [39,40].
Recently, a systematic review assessed strategies to increase HIV testing among MSM [41]. Authors concluded that social network-based strategies, community-based testing, HIV self-testing and modifications to the traditional facilitybased model can effectively reach urban MSM. However, AEHI testing strategies were not reviewed. The aim of this study was to conduct a systematic review and meta-analysis of (1) AEHI yield among MSM mobilized for AEHI testing; and (2) assess which risk factors and/or symptoms could increase AEHI yield in MSM.

| METHODS
The Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA)-statement was followed, which provides items for reporting in systematic reviews and metaanalyses [42].

| Search strategy
On 25 May 2020, we searched PubMed, Embase.com, Clarivate Analytics/Web of Science Core Collection and Ebsco/ ERIC using search terms, including synonyms and related terms, and keywords such as "men who have sex with men, " "homosexuality, " "acute HIV infection, " "early HIV infection, " and "mobilization" from database inception to the search date mentioned earlier, without geographical or language restrictions. The keywords represented three domains: domains one and two identified studies pertaining to MSM and AEHI respectively. The third domain sought to capture studies that focused on mobilization strategies, which included methods of communication with MSM. The full search strategy is described in Table S1. Experts in the field and secondary reference searching on included studies identified additional studies.

| Inclusion criteria and screening
Studies were included when the following inclusion criteria were met: (1) the study described a strategy of mobilizing MSM for AEHI testing; or (2) the study described the development or validation of a risk and/or symptom score which could increase the yield of AEHI in MSM. Studies were excluded if they merely assessed knowledge of AEHI among MSM, assessed AEHI laboratory testing techniques, described AEHI testing among MSM who had already presented for HIV testing, did not include the number of AEHI cases, or described AEHI testing among MSM who had already presented for HIV testing (e.g. laboratory evaluations of pooled samples obtained from MSM who had tested for HIV). Peerreviewed articles and conference abstracts were included. For each conference abstract meeting the inclusion criteria, a specific search was set out to identify the subsequent peerreviewed article of the study, as such, no conference abstracts were included in the final review. Two independent reviewers (SP and MD) used rayyan.qcri.org to screen titles and abstracts of records identified through the search to remove non-relevant records. Full-text records were then assessed for eligibility. Discrepancies were resolved by discussion with a third and fourth reviewer (EJS and GJB). We assessed study quality using the Appraisal tool for Cross-sectional Studies (AXIS;

| Data extraction
Data were extracted by two independent reviewers (SP and MD) using a standardized form. If studies reported on both MSM and other populations, we extracted data for MSM only if disaggregated data were available, otherwise we included estimates of the whole sample. We contacted study authors when additional information was needed. A modified framework from Campbell et al. was applied [41]. Studies were categorized according to two principal testing categories: (1) mobilization for AEHI testing, and (2) risk and/or symptom score screening. Mobilization for AEHI testing included three subcategories: media campaigns, partner notification services (PNS) and community-based testing. The data extracted included the following: AEHI cases identified, the total number of visits during which AEHI was assessed, year of publication, year of conduct, country, study population and study design. For the papers concerning mobilization strategy, we extracted the mobilization strategy, eligibility criteria for AEHI testing, and AHI and EHI definitions. For risk and/or symptoms scores a list of risk factors and/or symptoms included in the score, the recall period, cut-off value of the score, the area under the receiver operator curve (AUC), sensitivity and specificity of the score.

| Mobilization for acute and early HIV infection testing
In literature, different definitions are being used for AEHI based on the interval between infection and evolution of HIV tests as well as dynamics in antibodies over time. We used AEHI definitions as proposed by authors of the included studies. These varying definitions may have biased the cumulative results of this systematic review, however, we were unable to standardize AEHI definitions across the included studies as study authors reported results based on the above-described definitions. We defined AEHI yield as the proportion of identified AEHI cases among the number of visits during which AEHI was assessed. Targeted AEHI testing was defined as testing among a selected subgroup of MSM based on high-risk behaviour and/or AEHI symptoms. This was opposed to universal AEHI testing, defined as testing all MSM. Outcomes included type of mobilization strategy, and AHI and AEHI yield.

| Data analysis
We pooled independent study estimates for AEHI yield using the Freeman-Tukey double arcsine transformation in random effects models based on the method of DerSimonian and Laird [44,45]. Exact binomial procedures were used to calculate 95% CIs [46]. Pooled estimates were back-transformed on their original scale. Heterogeneity across estimates was assessed using the I 2 statistic [47]. After observing large heterogeneity across the estimates, we performed sub-group analyses of studies assessing targeted AEHI and AHI testing and studies assessing universal AEHI and AHI testing. Analyses were performed using the Metaprop package [48] in Stata (version 15.1; StataCorp).

| Risk and/or symptom score screening
Outcomes included AUC, sensitivity and specificity for risk and/or symptom scores. We extracted (or calculated, if not provided by authors) sensitivity and specificity at the score cut-off as proposed by the authors of included studies. We defined internal validation as assessment of predictive ability (AUC, sensitivity and specificity) of a risk and/or symptom score in a different study sample from the same location as the study sample in which the score was developed (i.e. the dataset was randomly split in a development and validation dataset or split based on calendar year). We defined external validation as assessment of predictive ability of a risk and/or symptom score in a study sample from a different location as the study sample in which the score was developed.
Media campaigns aimed to target MSM to increase knowledge and awareness of AEHI, the increased transmission risk, AEHI symptoms, AEHI tests and early treatment. Furthermore, they aimed to increase motivation to test for AEHI and included referral for facility-based AEHI testing. The campaigns were developed and promoted in conjunction with MSM community-based organizations [51,56,57,61]. Resources included print advertisements, condom packs, billboards, posters, web-based advertisements (e.g. on dating websites and applications) and campaign websites. These were promoted at MSM community-based events and MSM venues such as bars and bathhouses, MSM-targeted magazines and HIV testing facilities.
One study offered PNS to people with AEHI (index clients) [58]. The target population included MSM sexual or injection drug equipment partners of index clients with AEHI. Referral was done by index clients, with or without assistance of a healthcare provider, or by a healthcare provider without disclosing the identity of the index client.
Three studies assessed community-based AEHI testing at MSM venues [55,59,60]. The target population consisted of MSM visiting the venues. Venues included bathhouses, saunas, spas, bars, clubs and local non-governmental organizations. Collection of samples, conduction of rapid antibody tests and delivery of rapid antibody test results took place on-site at the venues. AEHI testing was laboratory based.

| Definitions of acute and early HIV infection
AHI was defined as a positive HIV-RNA test and a negative antibody test in six included studies [55][56][57][59][60][61], as a positive HIV-RNA test and an indeterminate antibody test in one study [58], or as a positive HIV-RNA test and a positive antibody test and a documented negative antibody test in the previous 30 days in one study [51]. Five included studies defined and reported on EHI, varying from a negative or indeterminate Western blot test to a documented or self-reported negative antibody test in the previous six months [51,55,58,59]. HIV tests included (pooled) HIV plasma viral load, point-of-care HIV-RNA tests, fourth generation antigen/antibody tests, rapid antibody tests and Western blot.
3.9 | Variables included in risk and/or symptom scores The recall period for risk factors and symptoms included in the scores varied from two weeks to two years. The 13 scores comprised eight scores only including demographic or behavioural risk factors for HIV acquisition [17,39,49,50,52,54,62], four scores including risk factors and AEHI symptoms [38,53,65] and one score including only AEHI symptoms [40] (Table 3).

| Predictive ability of the risk and/or symptom scores
The AUC ranged from 0.69 to 0.89 in development study samples, and from 0.51 to 0.88 in validation study samples (Table 4 and Figure 3). Sensitivity at the cut-off proposed by the authors ranged from 74% to 98% in development study samples, and from 25% to 94% in validation samples. Specificity was between 17% and 90% in development study samples, and between 15% and 96% in validation study samples. Internal and external validation resulted in lower predictive ability for most scores. For example the San Diego Early Test (SDET) score yielded an AUC of 0.74 (95% CI, 0.70 to 0.79) in the development study sample, and between 0.55 (95% CI, 0.44 to 0.66) to 0.70 (95% CI, 0.63 to 0.78) in external validation samples [37,39,63]. A study in Atlanta validated three scores (SDET, HIRI-MSM and the Menza score) in a cohort with a high proportion of HIV seroconversions among Black MSM, whereas the scores had been developed and previously validated in study samples consisting of predominantly white MSM [63]. The three scores performed poorly in this validation study sample among Black MSM and had markedly lower AUC values than in other validation study samples. This was also the case for a validation study in Chicago among young Black MSM [54]. Two scores showed high predictive ability in both the development and validation study samples: the Amsterdam score yielded AUC values of 0.78 (95% CI, 0.74 to 0.82) and 0.88 (95% CI, 0.84 to 0.91) in external validation study samples [64,65], the San Diego Symptom Score (SDSS) yielded an AUC of 0.85 (95% CI, 0.78 to 0.92) in internal validation [40]. Both scores included symptoms. Other scores, all from Kenya, with high AUC values in development study samples (0.76 to 0.89) have not been validated [17,38,53].

| DISCUSSION
In this systematic review and meta-analysis, we showed substantial AHI and AEHI yields when MSM were mobilized for AEHI testing in studies predominantly conducted in wellresourced settings. With the severe ongoing HIV epidemic among MSM in SSA [5][6][7], infrequent HIV testing and poor linkage to care and viral suppression outcomes [4], there is an urgent need to better identify AEHI in MSM. As such, targeted AEHI testing will likely result in high AEHI yields among MSM in SSA. Unfortunately, the World Health Organization (WHO) has no targeted AEHI testing recommendation for key populations, including MSM who have among the highest incidences [5][6][7]. Thus, AEHI testing should be offered to MSM, be supported by specific policy recommendations for MSM, and AEHI testing guidelines tailored to SSA need to be developed and endorsed by WHO.
Strategies mobilizing MSM for targeted AEHI testing resulted in higher AEHI yields than strategies mobilizing MSM for universal AEHI testing. Targeted AEHI testing may be optimized by screening with risk and/or symptom scores. The pooled AEHI yield was the highest when testing was targeted to MSM partners of people with AEHI, to partners of PLWH, or to MSM with AEHI symptoms who reported CRAI (11.1%). Although our review identified one study with a high AEHI yield resulting from PNS [58], two other studies did not assess and report on AEHI yield resulting from PNS for index clients with AEHI, and were therefore not included in this review [66,67]. When focussing only on AHI, the pooled AHI yield among studies assessing targeted testing was 3.3%.
Collaboration with MSM community-based organizations was key in successfully mobilizing MSM for AEHI testing, either through the design and promotion of AEHI media campaigns, or through the delivery of community-based testing [51,[55][56][57][59][60][61]. In the studies included in this review, onsite AEHI diagnosis was not possible in community-based testing settings, but required laboratory-based tests and skilled laboratory personnel. The emergence of point-of-care HIV-RNA tests may enable on-site community-based AEHI testing in SSA [21]. However, no study approached AEHI testing in a comprehensive, culturally sensitive and integrated fashion in SSA. As such, these strategies need to be urgently developed in close collaboration with local community-based organizations, including the need to include learning about point-ofcare HIV-RNA testing when locally available. While WHO recommends regular HIV testing for MSM, we suggest that MSM with unknown or HIV-negative status who experience AEHI symptoms or meet risk criteria be evaluated for AEHI, especially when PrEP initiation is considered [68].
Opportunities to diagnose AEHI are often missed, due to the non-specificity of symptoms and the costly diagnostic assays required for AEHI diagnosis [69][70][71][72]. The studies included in this review used several testing strategies to identify AEHI, including point-of-care HIV-RNA testing and (pooled) HIV viral load testing. A study in San Diego showed that AEHI testing with HIV-RNA testing was cost-effective in populations of MSM with an HIV prevalence above 0.4% [73]. Since HIV prevalence in MSM in SSA is estimated to be well above this threshold [2], AEHI testing among SSA MSM may also be cost-effective, although evidence hereof is lacking. Furthermore, targeting resources to specific subpopulations of  MSM (e.g. those reporting high-risk behaviour and/or symptoms) can substantially reduce costs compared with universal AEHI testing [36]. We identified 13 risk and/or symptom scores that may increase AEHI yield in MSM. Key risk factors included in these scores were age, number of sexual partners, CRAI, sexual intercourse with a PLWH, self-reported diagnosis of an STI and illicit drug use. Key symptoms were self-reported diarrhoea, fever and fatigue. As knowledge of symptoms of AEHI among MSM is low [74,75], these risk factors and symptoms may be used to educate MSM and help them self-recognize AEHI. Several risk and/or symptom scores have been included in MSM-targeted websites, facilitating self-assessment of HIV acquisition risk [39,52,61], although outcomes of these selfassessment tools need to be evaluated.
Predictive ability of the 13 risk and/or symptom scores varied greatly and was highest for scores that included symptoms [40,53,64,65]. Validation showed lower discriminate ability of most risk and/or symptom scores in the validation study sample than in the development study sample [52,54,63]. This was specifically the case for validation of risk and/or symptom scores among Black MSM in the USA, as the risk and/or symptom    Table 3. (Continued)

Score name
Menza [49] Facente [62] HIRI-MSM [50] CDRSS [38] UMRSS [38] S D E T [ 39] Sanders [53] Beymer [52] Amsterdam score [65] CDC [54] Gilead [54] Wahome [17] SDSS [40] Symptoms Body pains/ myalgia assessment of predictive ability of the score in a different study sample from the same location as the study sample in which the risk and/or symptom score was developed; cc assessment of predictive ability of the risk and/or symptom score in a study sample from a location different to the study sample in which the score was developed.
scores poorly predicted HIV acquisition [54,63]. This underlines the importance of external validation of risk and/or symptom scores [76]. Importantly, none of the MSM risk and/or symptom scores developed in SSA were validated [17,38,53]. Furthermore, no risk and/or symptom scores developed in well-resourced settings have been validated in less-resourced settings. Scores including symptoms may be particularly useful in SSA, where stigma and discrimination towards MSM behaviour is high, and social desirability bias may prevent MSM from disclosing high-risk behaviour to healthcare providers [77][78][79]. However, symptoms may vary by HIV-1 subtype [80], limiting the generalizability of symptom-based scores across SSA.
Risk-based scores may assist targeted AEHI screening, but may also be of use in identifying and prioritizing candidates for pre-exposure prophylaxis (PrEP) [81]. Recent studies using machine learning of routine health care data from electronic patient records to identify potential PrEP candidates among the general population showed high predictive ability of generated risk-based scores, but included more than 20 variables [82][83][84], which may limit practical use. Simpler risk and/or symptom scores, consisting of a smaller number of variables, which requires simple summation to calculate an individual's score, could be implemented in resource-limited settings.
However, risk and/or symptom scores are imperfect, and using a risk and/or symptom score to define who will be tested for AEHI will inevitably exclude people with AEHI [85,86]. Thus far, no AEHI yield has been reported resulting from screening MSM with published AEHI risk and/or symptom scores.
This study has some limitations. First, the database search strategy did not identify seven out of 22 included studies. Some of the included studies not identified by the search strategy focused on PrEP screening scores rather than AEHIscreening scores. Because these scores may also assist AEHI screening, we included these studies in this review. Second, heterogeneity across study estimates was large. This was partly explained by different testing strategies; heterogeneity was smaller when we stratified for testing strategy. Another possible explanation is the variable definitions for AEHI as proposed by study authors. This has possibly overestimated the AHI yield in studies that included indeterminate or positive antibody tests in their AHI definition [51,58]. Additionally, the variable study designs may have increased heterogeneity. For risk and/or symptom scores, the high variability in recall periods (two weeks to two years) will have likely resulted in variable outcomes. Likewise, the risk and/or symptoms recorded varied considerably between studies depending on the local context and how their data collection was set up, thus impacting the comparability of different scores. Furthermore, studies originated from various locations with different HIV epidemics, which has likely increased heterogeneity. Third, we did not standardize the cutoff at which sensitivity and specificity were assessed for the risk and/or symptom scores, and as a result, these values varied across studies. This has limited the comparison of sensitivities and specificities for the risk and/or symptom scores.

| CONCLUSIONS
In conclusion, strategies mobilizing MSM for targeted AEHI testing resulted in higher AEHI yields than universal AEHI testing. Targeted AEHI testing may be optimized using risk and/or symptom scores, in particular scores including symptoms.

C O M P E T I N G I N T E R E S T S
GJB has received grants through her institution from Bristol-Meyer Squibbs and Mac Aids Fund; honoraria to her Institution for scientific advisory board participations for Gilead Sciences and speaker fees from Gilead Sciences and Takeda. The remaining authors declared no competing interests.

A U T H O R S ' C O N T R I B U T I O N S
SP, MD, EW, JW, EG, EME and EJS designed the study. JK designed the search strategy. SP and MD independently assessed records for eligibility, and conducted data extraction, supported by EW. GJB and EJS had oversight in study selection and data extraction. MD conducted the statistical analysis and drafted the manuscript. MFSVL had oversight in the statistical analysis. All authors critically reviewed and revised the manuscript and approved the final version for publication.  . Area under receiver operator curves of published risk and/or symptom scores to assist screening for acute and early HIV infection among men who have sex with men. The black dots represent point estimates, the coloured lines 95% confidence intervals. If no coloured lines are displayed, the study did not report 95% confidence intervals. For each risk and/or symptom score, the first point estimate represents the area under receiver operator curve of the development study sample, the latter point estimate(s) represents the area under receiver operator curve of the validation study sample(s). The development outcomes of scores Facente, UMRSS, CDC and Gilead have not been included in this review, therefore, only validation outcomes are represented. CDC, Center for Disease Control and Prevention; CDRSS, Cohort Derived Risk Screening Score; D, Development study sample; HIRI-MSM, HIV Incidence Risk Index for MSM; MSM, men who have sex with men; NL, the Netherlands; NS, not specified; SDET, San Diego Early Test; SDSS, San Diego Symptom Score; UMRSS, University of North Carolina Malawi Risk Screening Score; USA, United States of America; V, Validation study sample. Development (USAID). This work was also supported in part through the sub-Saharan African Network for TB/HIV Research Excellence (SANTHE), a DELTAS Africa Initiative [DEL- . The DELTAS Africa Initiative is an independent funding scheme of the African Academy of Sciences (AAS) Alliance for Accelerating Excellence in Science in Africa (AESA) and is supported by the New Partnership for Africa's Development Planning and Coordinating Agency (NEPAD Agency) with funding from the Wellcome Trust [107752] and the UK government. EJS receives research funding from IAVI, NIH (grant R01AI124968) and the Wellcome Trust. MD receives funding through a PhD Scholarship from the Graduate School of Amsterdam UMC. The contents are the responsibility of the study authors and do not necessarily reflect the views of USAID, NIH, the US or UK government, AAS, NEPAD Agency, or the Wellcome Trust. This report was published with permission from the director of KEMRI. Table S1. Database search strategies Table S2. The appraisal tool for cross-sectional studies Table S3. Critical appraisal of included studies using the Appraisal Tool For Cross-Sectional Studies