Sexual behaviour change following HIV testing services: a systematic review and meta‐analysis

Abstract Introduction Learning one’s HIV status through HIV testing services (HTS) is an essential step toward accessing treatment and linking to preventive services for those at high HIV risk. HTS may impact subsequent sexual behaviour, but the degree to which this varies by population or is true in the setting of contemporary HIV prevention activities is largely unknown. As part of the 2019 World Health Organization Consolidated Guidelines on HTS, we undertook a systematic review and meta‐analysis to determine the effect of HTS on sexual behaviour. Methods We searched nine electronic databases for studies published between July 2010 and December 2019. We included studies that reported on at least one outcome (condom use [defined as the frequency of condom use or condom‐protected sex], number of sex partners, HIV incidence, STI incidence/prevalence). We included studies that prospectively assessed outcomes and that fit into one of three categories: (1) those evaluating more versus less‐intensive HTS, (2) those of populations receiving HTS versus not and (3) those evaluating outcomes after versus before HTS. We conducted meta‐analyses using random‐effects models. Results and discussion Of 29 980 studies screened, 76 studies were included. Thirty‐eight studies were randomized controlled trials, 36 were cohort studies, one was quasi‐experimental and one was a serial cross‐sectional study. There was no significant difference in condom use among individuals receiving more‐intensive HTS compared to less‐intensive HTS (relative risk [RR]=1.03; 95% CI: 0.99 to 1.07). Condom use was significantly higher after receiving HTS compared to before HTS for individuals newly diagnosed with HIV (RR = 1.65; 95% CI: 1.36 to 1.99) and marginally significantly higher for individuals receiving an HIV‐negative diagnosis (RR = 1.63; 95% CI: 1.01 to 2.62). Individuals receiving more‐intensive HTS reported fewer sex partners at follow‐up than those receiving less‐intensive HTS, but the finding was not statistically significant (mean difference = −0.28; 95% CI: −3.66, 3.10). Conclusions Our findings highlight the importance of using limited resources towards HTS strategies that focus on early HIV diagnosis, treatment and prevention services rather than resources dedicated to supplementing or enhancing HTS with additional counselling or other interventions.


| INTRODUCTION
HIV testing and knowledge of one's status is an essential first step towards linkage to HIV treatment, prevention and care [1]. Early linkage to antiretroviral therapy (ART) following an HIV-positive diagnosis reduces HIV-related mortality and morbidity and prevents HIV transmission from those who maintain viral suppression [2][3][4][5][6][7][8]. Linking HIV-negative individuals to relevant prevention interventions (e.g. pre-exposure prophylaxis [PrEP]) reduces the risk of HIV acquisition.
Counselling and messaging delivered through HIV testing services (HTS) may also offer an opportunity to influence subsequent sexual behaviour and thereby affect HIV acquisition and transmission [9]. These changes in behaviour may be attributable to counselling received as part of HTS, the act of testing, or knowledge of one's HIV serostatus. An early systematic review on behaviour change following HTS, which included studies conducted between 1985 and 1997, found HTS was associated with reductions in condomless sex among HIV-positive participants and serodiscordant couples, but not among HIV-negative individuals [9]. More recently, evidence from low-and middle-income countries has highlighted associations between HTS and reductions in the number of sex partners, increases in condom-protected sex among HIV-positive individuals [10] and risk reduction among HIV-negative serodiscordant partners [11]. However, there has been little assessment of the impact of HTS on sexual behaviour in the present era of new modes of HTS (e.g. self-testing), scale-up of ART and the introduction of prevention options, such as PrEPparticularly among key populations (e.g. men who have sex with men [MSM]) or by partner type (e.g. primary/non-primary) [9][10][11]. Thus, it remains unclear the extent to which contemporary HTS affects subsequent behaviour change in different populations.
In 2019, to update the World Health Organization (WHO) Consolidated Guidelines on HTS, WHO identified sexual behaviour change following HTS as an important area for review. This provided an opportunity to update the previous evidence presented by Fonner and colleagues [10]. The primary objective of this study was to synthesize the evidence on the effect of HTS on sexual behaviour.

| Inclusion criteria
We included studies published in a peer-reviewed journal or conference abstract between 1 July 2010 and 31 December 2019. The start date for inclusion represents the end date of the previous review by Fonner and colleagues [10]. We searched nine electronic databases and four conference abstract databases/books (Data S1). Studies were eligible if they prospectively compared outcomes of interest, fit into one of three a priori exposure/comparison categories, and reported one or more outcomes. These exposure/comparison categories and outcomes are listed in Table 1.

| Quality assessment
We assessed risk of bias for individual randomized controlled trials (RCT) and cluster-RCTs using the Cochrane Collaboration's tool [13]. We conducted a quality assessment for cohort studies and pre-post studies using the National Institutes of Health Study Quality Assessment Tools for observational cohort studies and before-after (pre-post) studies with no control group respectively [14].

| Data abstraction and analysis
A team of trained reviewers were involved in the review process. Search results from each database were merged and duplicate citations were removed. We used Covidence (Veritas Health Innovation Ltd, Melbourne, Australia) for screening and extraction. At least two reviewers independently screened titles and abstracts of all search results, reviewed full-text articles for those abstracts receiving two votes for inclusion and extracted relevant data. A third reviewer resolved conflicts. Non-English articles (n = 12) were reviewed for inclusion by WHO staff, but upon translation, these studies were deemed ineligible. The studies were reviewed from June 2018 to September 2020. The review was completed by September 2020.
All studies were categorized into one of three exposure/ comparison categories (Table 1). To the extent possible, outcomes were stratified based on the following a priori strata: HIV status of the participant, population (MSM, people who inject drugs [PWID], adolescent girls and young women [AGYW], pregnant women, female sex workers [FSW]), sex partner type (primary, non-primary), sex partner HIV status (HIV positive, HIV negative, unknown HIV status) and type of sex with sex partner (vaginal or anal). We only conducted stratum-specific meta-analyses when the studies were comparable. For all studies, we extracted data from the first follow-up time point unless otherwise noted. Due to heterogeneity in the  To standardize comparisons across studies, standard-of-care HTS was always considered to be "less" intensive, even in studies where the "more" intensive intervention (e.g. HIV self-testing) may have been operationalized as less intensive or abbreviated compared to standard-of-care HTS; b for studies that examined outcomes among individuals post-HTS compared to pre-HTS, we required that outcomes were ascertained in a manner that appropriately captured the period prior to HTS (i.e. before individuals knew their current HIV status) and after HTS (i.e. after individuals became aware of their HIV status). follow-up time points (e.g. six months, twelve months) and the recall period for behavioural outcomes (e.g. past 30 days, past three months), we did not stratify results by these factors.
We conducted meta-analyses using random-effects models for outcomes measured comparably for two outcomes: (1) condom use, defined as the frequency of condom use or condom-protected sex (henceforth referred to as "condom use") and (2) number of sex partners. For condom use, estimates from included studies were converted to a common metric of relative risk (RR) using dichotomous outcome variables. For comparability across studies, we did not consider the adjusted estimates that were reported, and instead utilized raw Ns from the studies to calculate summary estimates (e.g. the number of individuals who reported condom use out of the total number of individuals who reported engaging in sexual activity). If condom use data were reported as categorical, we dichotomized categories (e.g. always vs. not always). For all cluster-RCTs included in the meta-analysis, we used the intracluster correlation coefficient [ICC] of 0.026 obtained from a previous study to take into account the design effect using the formula: design effect = 1 + (cluster size-1) 9 ICC [15]. For continuous outcomes, we calculated the mean difference for studies in which the mean and standard errors were reported.
Meta-analysis and data summary were conducted using RevMan and R. Studies were excluded from meta-analyses if condom use could not be converted to an RR or could not be determined (e.g. studies that reported on sex acts), if complete outcome data were not available (e.g. no standard error/standard deviation for number of sex partners), or if there was heterogeneity in how outcomes were reported (e.g. number of sex partners reported categorically). These studies and all other outcomes were summarized descriptively.

| RESULTS
We screened 29 980 titles and abstracts and identified 441 full-text articles, of which 76 were included in the review (Figure 1). A summary of these 76 studies is provided in Table 2, and outcome categories for these studies are summarized in Table 3. 3.1 | Exposure/comparator categorization 1:

More-intensive versus less-intensive HTS
Thirty-six studies examined more-intensive HTS versus less-intensive HTS. All were either individual-RCT (n = 22) or cluster-RCT (n = 14) ( Table 2). The specific population, intervention and outcomes for these studies are summarized in Table 4.
Nineteen studies were included in the meta-analysis (Figure 2). After disaggregation by participant HIV status, sex and partner type, there were a total of 36 estimates from 19 studies. There was no significant difference in condom use after receipt of HTS among individuals receiving more-intensive HTS compared to those receiving less-intensive HTS (RR = 1.03; 95% Confidence Interval [CI]=0.99 to 1.07) ( Table 5 and Figure 2).
These findings were largely consistent for outcomes reported among MSM (RR = 1.27; 95% CI = 0.92 to 1.76) and FSW (RR = 1.02; 95% CI = 0.95 to 1.10). The strongest association between receipt of more-intensive HTS and increased condom use was in a study of Chinese MSM [91]. In that study, Hao and colleagues found that HIV-negative MSM randomized to receive HIV testing plus enhanced posttest counselling, plus a video narrated by HIV-positive Chinese MSM, plus a bracelet to serve as a reminder to engage in safer sex (more-intensive HTS) was significantly more likely to report no condomless anal sex with non-primary partners (RR = 3.5; 95% CI = 2.4 to 5.2) and with primary partners (RR = 1.5; 95% CI = 1.0 to 2.3) compared to those receiving standard HTS.
Thirteen studies were not included in the meta-analysis because the outcomes could not be pooled or because there were not sufficient data to include in the summary estimate. Eight of these studies observed no significant difference in the overall number of condomless sex acts at follow-up among those receiving more-versus less-intensive HTS [16][17][18]39,42,45,46,50], though Metsch and colleagues found that participants who received rapid HIV testing plus individual risk-reduction counselling (i.e. more-intensive HTS) reported lower rates of condomless sex with non-primary partners compared to individuals who received rapid HIV testing with information only (IRR = 0.66; 95% CI = 0.55 to 0.79) [39]. In addition, an RCT by Homsy and colleagues among pregnant women [30] found that HIV-uninfected pregnant or lactating women in Uganda who received individual or couple-enhanced counselling quarterly for up to two years post-partum did not report any difference in the frequency of condom use relative to those receiving standard HTS. The remaining studies reported increases in condom use/decreases in condomless sex for those randomized to more-intensive versus less-intensive HTS. A study among African-American women aged 18 to 65 in the United States found those randomized to more-intensive HTS reported a similar number of condom-protected vaginal sex acts at follow-up compared to less-intensive HTS, but were significantly more likely to report increasing the frequency of condom use for anal sex [29]. In India, Mimiaga and colleagues observed fewer condomless sex acts at follow-up among MSM engaged in exchange sex after receiving more-intensive HTS compared to those receiving less-intensive HTS [40]. In a US study from Mimiaga and colleagues [41], MSM with crystal methamphetamine dependence who received more-intensive HTS (enhanced frequency of sexual risk-reduction counselling) reported significantly fewer condomless anal sex acts with partners who were living with HIV or whose HIV status they did not know at 3-month follow-up compared to MSM who received less-intensive HTS. Wechsberg and colleagues [49] found that South African women who received a gender-focused HIV prevention intervention in addition to HTS reported more condom use with a main partner compared to women who received only HTS. In an US RCT that enrolled individuals in community supervision programmes and their female sex partners, El-Bassel and colleagues [24] noted a significantly lower number of condomless sex acts with study partners and non-study partners for individuals randomized to multi-session risk-reduction counselling (more-intensive HTS) versus one-time counselling.
Three studies noted fewer sex partners [42], fewer new anal sex partners [50], or a lower percentage of reporting multiple male sex partners [47] among participants randomized to more-intensive HTS compared to less-intensive HTS. Five studies did not observe significant differences in the mean number of condomless sex partners at follow-up [18,32,39], the number of sex partners [23] or percent reporting at least five clients per day [22] for those receiving moreversus less-intensive HTS. However, in a subgroup analysis, Metsch et al. [39] found MSM receiving more-intensive HTS reported fewer condomless sex partners at follow-up compared to those receiving less-intensive HTS (IRR = 0.71; 95% CI = 0.61 to 0.83).

| HIV incidence
Eleven studies reported on HIV incidence [17,22,25,26,28,30,33,35,39,45,91]; ten did not identify statistically significant differences in HIV incidence at follow-up among individuals receiving more-versus less-intensive HTS. The only study to observe a statistically significant difference was a cluster-RCT among FSW in Tanzania [33]. In that study, FSW in communities randomized to more-intensive HTS had a significantly lower HIV incidence compared to FSW in communities with standard HTS (5.0% vs. 10.4% respectively).

| STI incidence/prevalence
Eight studies reported on STI incidence, either as a composite STI outcome [32,36], STI-specific outcome [22,23,45,91]    3.2 | Exposure/comparator categorization 2: received HTS versus did not receive HTS Six studies examined outcomes among individuals who received HTS compared to those who did not receive HTS: four were cohort studies [53][54][55][56], one was a cluster-RCT [52] and one was an individual-RCT [27] (Table 2). Due to the small number of studies and heterogeneity in how outcomes were reported, we did not conduct meta-analyses for any outcomes. No study in this category reported on STI incidence.

| Condom use
None of the three studies that reported on the condom use observed statistically significant differences between individuals who did and did not receive HTS [27,52,54]. Baird and colleagues found that 66% of AGYW who received home-based HTS reported always using condoms or not having sex at follow-up compared to 67% among those who did not receive any HTS [52]. In a series of serosurveys, Cawley and colleagues found no differences in the proportion of men and women in Tanzania who reported using condoms with their spouse, non-primary partner or regular co-habiting partner among those who did or did not receive HTS [54]. In a cohort study of cross-border truck drivers in China, Lau et al. found no significant differences in condom use with FSWs or nonprimary partners at follow-up among individuals who did and did not receive HTS [27].

| Number of sex partners
Two studies reported on the number of sex partners [52,54]. In Malawi, a cluster-RCT of immediate versus delayed homebased HTS among AGYW (aged 13 to 22) found those receiving immediate home-based HTS reported a higher mean number of sex partners at follow-up, though the result was only statistically significant for HIV-negative participants [52]. In contrast, in Tanzania, HIV-negative individuals who received HTS were significantly more likely to decrease their number of sex partners in the last year between serosurveys, compared to those who did not receive HTS [54].

| HIV incidence
Four cohort studies reported on HIV incidence with mixed results [53][54][55][56]. The two studies [53,55] [55]. Two other cohort studies did not observe a significant difference in HIV incidence by receipt of HTS [54,56]. However, in a secondary analysis weighted for risk factors, youth who received HTS in South Africa had a significantly lower risk of HIV compared to those never tested (aHR = 0.59; 95% CI = 0.45 to 0.78) [56].

| Exposure/comparator categorization 3: post-HTS versus pre-HTS
Of 34 studies that examined outcomes among individuals in a period after receiving HTS compared to a period prior to HTS, 32 were cohort studies, one was a study of serial crosssectional surveys and one was a two group pre/post-test quasi-experimental study ( Table 2). None of these studies reported on HIV incidence.
Five studies were not included in the meta-analysis [64,67,78,79,90] because the outcomes were reported in a manner inconsistent with other studies. In Zimbabwe, Cremin and colleagues reported an increase in consistent condom use among individuals testing newly HIV positive compared to before, but the difference was not significant [64]. In Malawi, 28% of individuals newly diagnosed HIV positive reported increasing their condom use since learning their HIV status, but the authors did not report on the statistical significance of this finding [67]. Lin et al., noted a substantial increase in condom use among individuals newly diagnosed HIV positive from  the time period pre-HTS (9%) to post-HTS (91%) [78]. Among individuals newly diagnosed HIV negative, M€ oller and colleagues found that the odds of condomless anal sex were higher pre-HTS compared to post-HTS (aOR = 2.1; 1.2 to 3.6) [79]. Cremin and colleagues did not observe differences in consistent condom use pre-HTS compared to post-HTS among HIV-negative individuals [64]. Finally, Wall and colleagues found that serodiscordant couples in Zambia undergoing couples HTS had significantly fewer condomless sex acts prior to HTS compared to three and six months after HTS [90].

| Number of sex partners
The results from the 14 studies that examined this outcome were not summarized in meta-analyses because of heterogeneity in how outcomes were reported. Nine studies included individuals testing newly HIV positive [58,60,64,70,72,77,86,88,89]), seven studies included individuals testing HIV negative [64,65,70,72,76,79,86] and two studies included individuals testing newly HIV positive and negative but did not disaggregate the outcomes by testing status [57,80].
All nine studies among individuals testing newly HIV positive found that individuals reported fewer sex partners after an HIV-positive diagnosis versus before, although not all studies identified significant differences. Four studies that included general populations of men and women [58,64,86,89] observed significant declines in the mean number of sex partners after an HIV-positive diagnosis versus before. However, Venkatesh and colleagues found no difference in the proportion of women reporting >1 sex partner in the past three months before HTS versus after an HIV-positive diagnosis [89]. Among the four studies that included MSMtwo from the USA [70,88], one from Hong Kong (China) [77] and one from Thailand [72] all found that the number of sex partners was higher in the period prior to an HIV-positive diagnosis compared to after. The only study including FSW [60] noted a higher median number of clients before an HIV-positive diagnosis (median = 10; interquartile range [IQR] = 5 to 18) than after (median = 3; IQR = 2 to 5).
Of the seven studies that compared the number of sex partners reported by individuals before and after testing HIV negative, four included general populations of men and women [64,65,76,86] and three included MSM [70,72,79],all but   The two studies that aggregated outcomes of individuals testing newly HIV positive and newly HIV negative observed mixed results. In a Ugandan study, Mulogo and colleagues noted a reduction in the number of sex partners post-HTS compared to pre-HTS [80]; this reduction was larger for those receiving facility-based versus home-based HTS. In Nigeria, a study of cantonment residents did not observe a difference in the number of casual sex partners reported at 3-month post-HTS compared to the period of time before HTS [57].

| STI incidence/prevalence
Two studies reported on STI incidence [62,72]. Calvo and colleagues noted a higher proportion of MSM and transgender women in Peru was diagnosed with an STI after an HIV-positive diagnosis (68%) compared to three months before the HIV-positive diagnosis (47%), though this difference was not statistically significant [62]. In another study of MSM and transgender women in Thailand, Hiransuthikul and colleagues found that the prevalence of any STI declined significantly at 12-month post-HTS compared to the time period prior to HTS but only for those individuals who were not diagnosed with HIV. There was no significant change in STI prevalence post-HTS compared to pre-HTS among individuals diagnosed with HIV [72].

| Risk of bias and quality assessment
Risk of bias assessment for 23 individual-RCTs revealed that there was low risk of biased allocation to interventions due to inadequate generation of a randomized sequence in all but four studies [27,30,42,50] (Table 6). One of these four studies was a conference abstract [30]. Considering the nature of the intervention that involved interaction between HTS counsellors and individuals who tested for HIV, we assessed a high risk of bias due to blinding of participants and personnel and outcome assessors for the majority of the included individual RCTs. There was a high risk of bias due to incomplete outcome data for eight studies [16,19,23,25,30,37,48,91], mainly due to a high proportion of missing outcome data that were not balanced across HIV testing and comparator groups.
Risk of bias assessment of 15 cluster-RCTs revealed low risk of bias for most of the studies except Hawk et. al [29], for which there was evidence of high or unclear risk of bias related to four out of five domains (Table 7). Overall, all the pre-post studies and cohort studies were judged to be of "good" or "fair" quality, and none were deemed to be of "poor" quality (Tables 8 and 9).

| DISCUSSION
In this review of sexual behaviour change following HTS, we found that receipt of more-intensive HTS was not significantly associated with subsequent increases in condom use relative to less-intensive HTS. Likewise, we did not observe differences in subsequent condom use among individuals who did and did not receive HTS. However, we found that receipt of HTS was significantly associated with increases in condom use after receipt of HTS among individuals newly diagnosed HIV positive, but only marginally significant among individuals diagnosed HIV negative. This finding was consistent when we stratified studies to include only MSM or only serodiscordant couples. Taken together, these results suggest that enhanced counselling or other components included in more-intensive HTS may not have a large impact on subsequent sexual behaviour, but that receipt of HTS may affect subsequent sexual behaviour among individuals diagnosed HIV positive.
The finding that more-intensive HTS was not associated with changes in behaviour or HIV/STI incidence is noteworthy. For the majority of these studies, the "more" intensive intervention included additional or enhanced counselling sessions or support groups above and beyond that which is included within pre-test information in standard HTS, and thus our findings indicate that these additional counselling components may not have a strong influence on subsequent sexual behaviour or HIV/STI incidence. Furthermore, some of the moreintensive HTS may be difficult or expensive to implement at a population-level (e.g. HTS with multi-session behavioural counselling); thus, in a setting of limited resources, these more-intensive HTS interventions may not provide large gains in HIV prevention for the cost. That said, there were several studies that observed lower rates of condomless sex, increases in condom use, or fewer sex partners among key populations [29,39,40,43,91], indicating that more-intensive HTS may be beneficial for some populations or within some settings. We included HIV and STI incidence as an outcome in this review as a proxy for sexual behaviour change. Given that we did not observe significant differences in condom use comparing more-intensive HTS versus less-intensive HTS, it is not surprising that most studies did not observe significant differences in HIV or STI incidence among those two groups. However, three community-based HTS studies demonstrated a slight reduction in HIV incidence [17,21] or a significant reduction in HIV incidence [33] among communities randomized to more-intensive HTS. But these studies involved multi-component interventions; thus, it is difficult to disentangle the impact of an individual component. Of the studies that compared the receipt of any HTS to no HTS, two [53,55] examined the frequency of HTS and found those testing more frequently had a higher subsequent risk of HIV acquisition. However, it remains unclear if this is due to more frequent testing (i.e. "the more you test, the more you find"), due to the fact that those individuals at high risk for HIV tested more frequently, or that there is minimal or no impact of HTS on subsequent behaviour change and corresponding risk of HIV acquisition.
A key finding of this review is that individuals substantially change their condom use behaviour after being diagnosed with HIV, a finding that confirms the previous meta-analysis by Fonner and colleagues [10]. Our results were consistent when we calculated stratified estimates for MSM and couples. In contrast, we only observed a marginally statistically significant change in condom use after an HIV-negative diagnosis. Given that our review is the third to demonstrate that    receiving an HIV-negative diagnosis does not lead to substantial increase in condom-protected sex, our results suggest that it is the HIV-positive diagnosis, not the testing itself, that likely has an impact on subsequent behaviour, and somewhat calls into question the notion that HIV testing itself (in the absence of an HIV diagnosis) directly leads to modifications in sexual behaviour. This finding also highlights the importance of integrating PrEP referrals or PrEP provision within HTS as part of an HIV prevention package for individuals diagnosed HIV negative who may be at ongoing HIV risk. From a public health standpoint, the modification in behaviour immediately following a new HIV diagnosis is critically important, as it has the potential to reduce HIV transmission to HIV-uninfected partners at a time when individuals newly diagnosed with HIV may not yet be linked to ongoing HIV medical care or may not yet be virally suppressed. However, in our meta-analysis we only used outcome data from studies' first follow-up time point, which was typically between three and twelve months after the HIV-diagnosis. So, although we observed an immediate change in behaviour following an HIVpositive diagnosis, it remains unclear if the change is durable beyond a few months. Several studies included in this review have observed subsequent decreases in condom use after the initial increases observed immediately following HIV diagnosis [69,74,77,82]. In the era of undetectable-untransmissible, future studies will also likely start to observe longer term decreases in condom use among these populations. Our findings therefore confirm the importance of immediate linkage to ART and sustained engagement in HIV care to achieve continuous viral suppression.
The largest effect of sexual behaviour change following HIV diagnoses that we observed was among couples, where there was a nearly sixfold increase in condom use after HTS. These studies varied in the type of HTS offered (i.e. couples HTS [59,84] vs. standard HTS [83,87]), and in the HIV status of enrolled couples. Despite these differences, the increase in condom use was large and relatively consistent across the studies, suggesting that the effect of HTS on subsequent sexual behaviour is particularly robust among couples, and underscores the importance of couples HTS programmes.
There are several notable limitations. First, there was marked variation in study outcome measures, particularly in terms of the follow-up time-points and recall periods, which prevented an assessment of behavioural outcomes for different time intervals. Second, the outcome of condom-protected sex was not clearly labelled in some studies and was obtained by inverting the outcome of "condomless sex. " Doing so might not have always precluded "abstinence" from condom-protected sex and hence, the outcome might not have always been consistently extracted across studies. Third, there was a high degree of heterogeneity, supported by the heterogeneity test in our meta-analyses (Table 5). While we acknowledge this challenge, we were unable to explore the source of additional heterogeneity beyond the a priori subgroup and stratified analyses. Fourth, the majority of pre-post studies did not always include the same number of people in the post-and the pre-HTS group. Fifth, we did not consider the adjusted estimates that were reported, and instead utilized raw Ns from the studies to calculate summary estimates. Sixth, the decision to meta-analyse estimates from observational studies may be prone to bias including confounding; however, we carefully considered the quality and methodological homogeneity of these studies prior to metaanalysis. Sixth, we cannot exclude the possibility of recall bias or social-desirability bias for self-reported outcomes. Seventh, we did not extract data such as individuals' or partners' PrEP use, ART use or viral load, or changes in seroadaptive behaviours, all of which could affect an individual's decision about whether or not to use condoms. Eight, this review only included testing services related to HIV, and not STI, though some of the "more-intensive" HTS may have included STI testing. Ninth, there is heterogeneity in the goals of HTS, and it is possible that not all HTS explicitly have a goal of modifying subsequent condom use and number of sex partners. Finally, we did not assess publication bias as part of this review but note that this review only included studies published in peer-reviewed literature or accepted as conference abstracts.

| CONCLUSIONS
Our review indicates that enhanced counselling or other components often included in more-intensive HTS may not have a . Was the research question or objective in this paper clearly stated? C2. Was the study population clearly specified and defined? C3. Was the participation rate of eligible persons at least 50%? C4. Were all the subjects selected or recruited from the same or similar populations (including the same time period)? Were inclusion and exclusion criteria for being in the study prespecified and applied uniformly to all participants? C5. Was a sample size justification, power description, or variance and effect estimates provided? C6. For the analyses in this paper, where the exposure(s) of interest measured prior to the outcome(s) being measured? C7. Was the timeframe sufficient so that one could reasonably expect to see an association between exposure and outcome if it existed? C8. For exposures that can vary in amount or level, did the study examine different levels of the exposure as related to the outcome (e.g. categories of exposure, or exposure measured as a continuous variable)? C9. Were the exposure measures (independent variables) clearly defined, valid, reliable and implemented consistently across all study participants? C10. Was the exposure (s) assessed more than once over time? C11. Were the outcome measures (dependent variables) clearly defined, valid, reliable and implemented consistently across all study participants? C12. Were the outcome assessors blinded to the exposure status of participants? C13. Was loss to follow-up after baseline 20% or less? C14. Were key potential confounding variables measured and adjusted statistically for their impact on the relationship between exposure(s) and outcome(s). CD, cannot determine; NA, not applicable; NR, not reported.
large impact on condom use, but that receipt of an HIV-positive diagnosis likely affects condom use, at least in the time period immediately following an HIV diagnosis. These results provide reassurance that most populations increase condom use during a period in which they may not yet be fully engaged in HIV care or virally suppressed. Among most populations, we did not observe changes in sexual behaviour after being diagnosed HIV negative. This underscores the need to incorporate PrEP provision or PrEP referral services into HTS. The findings from this review suggest that limited HTS resources should be focused on expanding efforts that promote early HIV diagnosis and linkage to treatment and prevention services instead of more-intensive approaches that incorporate enhanced behavioural counselling.

A U T H O R S ' C O N T R I B U T I O N S
CMK and BB developed the initial search strategy with input from CJ, MSJ, RB, MBD, KS. CMK, RT, JW, HH, BB, DK, NK, ATT, KAC, SF, BG and MA were involved in title and abstract screening. RT, NK and LS extracted the data. CMK verified eligibility of the included studies, verified extracted data and resolved discrepancies at each step of the review. RT conducted the analysis with input from CMK. RT and CMK wrote the initial draft of the manuscript. All authors critically reviewed and commented on the drafts and approved the final version of the manuscript.

A C K N O W L E D G E M E N T S
We thank the systematic review teams led by Dr. Anjuli D. Wagner, and Dr. Brandon Guthrie and Dr. Alison L. Drake for their valuable comments and suggestions throughout the review process. We thank the University of Washington Health Sciences Library for their support in developing the search strategy. We thank the Global HIV, Hepatitis and STI programme at WHO and the 2019 Consolidated HTS Guidelines Development Group and Steering Committee for their input on the search strategy, protocol and interpretation of the findings. We are very grateful to Nandi Siegfried for her guidance on the methods and analysis, and for her review of the manuscript.

F U N D I N G
This research was supported by the WHO #018/CDS/HIV/004, WHO #2018/ 865307-0 USAID GHA-G-00-09-00003, the Bill and Melinda Gates Foundation OPP1177903, the University of Washington Center for AIDS Research (NIH/ NIAID P30-AI027757). KAC has received the STD/AIDS Research Training Predoctoral Fellowship, which is funded by NIH T32 AI07140.

D I S C L A I M E R
The manuscript represents the views of the authors and does not necessarily represent the views or decisions of the World Health Organization.