Sexual partner concurrency: is it a useful concept for HIV prevention? A systematic review of the evidence for intervention effectiveness in low- and middle-income countries

Abstract

If sexual partner concurrency drives HIV transmission dynamics, shouldn’t HIV prevention efforts be addressing this behavior? We systematically reviewed studies evaluating interventions to reduce sexual partner concurrency in low- and middle-income countries using pre/post or multi-arm designs. Only two studies met our inclusion criteria; neither found significant differences by intervention exposure on self-reported concurrency. Overall, very few interventions have specifically targeted concurrency, and those that did have not been rigorously evaluated. In practice, concurrency may be difficult to separate from multiple partnerships more generally.

INTRODUCTION

One idea offered to understand geographic patterns in HIV epidemics is sexual partner concurrency (Morris & Kretzschmar, 1997). Concurrency is defined as sexual intercourse with one (new) partner occurring between acts of intercourse with an existing partner (UNAIDS Reference Group on Estimates, Modelling and Projections, 2010). Concurrency occurs when a new sexual partnership is formed before the previous one has concluded. Knopf et al. (2014) distinguish concurrency from serial monogamy. Concurrency, they contend, extends partnerships into “triplets”, connecting the practitioner more rapidly to new, possibly infected, partners (Morris & Epstein, 2010). This exposure places the original partner at risk of infection through no change in that person’s own behavior. If HIV transmission occurs in the newly formed partnership, the original partner will be exposed to an acute-phase HIV infection, with high viremia (Morris & Kretzschmar, 1997).

A lively academic debate has centered on the contribution of sexual partner concurrency to HIV epidemics in Africa (Padian & Manian, 2011; Morris, 2010; Lurie & Rosenthal, 2010). This debate has drawn attention from epidemiologists, mathematical modelers and social network analysts and persisted since the idea was originally proposed by Watts and May (1992). The debate has at times been contentious, with hardened positions and no sign of resolution.

Another way of considering the importance of sexual partner concurrency is by evaluating interventions specifically designed and implemented to decrease concurrency. If concurrency is indeed a useful concept in relation to disease transmission, then reports of efforts to address it should be found in the scientific literature, with evaluation results shedding light on how it might be applied to real-world disease prevention efforts. To understand the effectiveness of interventions to reduce sexual partner concurrency, we systematically reviewed evaluations of these interventions.

METHODS

We followed the UNAIDS consensus definition of sexual partner concurrency as “overlapping sexual partnerships in which sexual intercourse with a new partner or partners occurs between two acts of intercourse with another partner” (UNAIDS Reference Group on Estimates, Modelling and Projections, 2010). We defined interventions to reduce sexual partner concurrency as any planned activity that sought to deliberately reduce the number of concurrent sexual partnerships among the study population. Reducing sexual partner concurrency had to be a stated goal of the intervention, either explicitly stated, or demonstrated through clear identification of concurrency in the intervention logic model. We specifically focused on interventions to reduce concurrency, rather than interventions to reduce multiple partnerships more generally.

To be included, a study had to (1) evaluate an intervention explicitly designed to reduce sexual partner concurrency; (2) measure the effect on sexual partner concurrency; (3) take place in a low- or middle-income country setting as defined by the World Bank (2021); (4) use a study design that presented post-intervention data; (5) present a multi-arm, or pre-post-assessment of relevant outcomes; and (6) be published in a peer-reviewed journal, as a conference abstract, or as a program report between January 1990 and September 2021.

To identify articles, we searched five online databases (PubMed, CINAHL, PsycINFO, Sociological Abstracts, and EMBASE) on September 8, 2021 using a standard set of search terms adapted for the controlled vocabulary of each database (Appendix). We also used place-specific terms (e.g., “US”, “UK”) to restrict the search to low and middle-income countries. We searched scientific conference abstracts for available years from the International AIDS Conference, the International AIDS Society Conference on HIV Science, the Conference on Retroviruses and Opportunistic Infections, and the International Conference of AIDS and STIs in Africa. Finally, we searched websites of donor and implementing partner organizations known to fund or implement interventions to reduce sexual partner concurrency, (USAID and AED, 2012; USAID, AIDS Free, and PEPFAR, 2019) and reviewed reference lists of other related reviews (Astatke et al., 2012).

After removing duplicate citations, search results were screened by a single research assistant to remove those clearly not meeting the eligibility criteria. Next, two coauthors completed a second round of screening independently with discrepancies resolved by consensus. Final inclusion decisions were made after full-text review.

Two authors extracted data independently from each included study using a standardized form capturing: location, setting and target group, years; intervention description; study design; sample size; participant demographics; participant allocation method; follow-up length; outcome measures; comparison groups; effect sizes; confidence intervals; significance levels; and limitations. They also conducted a quality assessment of each study using the Evidence Project risk of bias assessment (Kennedy et al., 2019).

RESULTS

Of 990 unique citations identified, 86 were retained for duplicate screening and 61 for full-text review (Figure 1). We reviewed 24 program reports, websites, or available datasets from USAID-funded projects that targeted concurrency specifically, including C-Change – Lesotho, One Love – southern Africa, PSI’s campaigns – Botswana and Mozambique, Scrutinize – South Africa, and Synergy – Uganda. While these programs reported process data and qualitative findings, none included pre-post or multi-arm quantitative assessments of outcomes of interest. We also identified seven peer-reviewed studies and one conference abstract which measured concurrency as an outcome but did not clearly intend to target concurrency in the intervention. Ultimately, two publications met the inclusion criteria: one conference abstract (Halldorsdottir et al., 2012a) (presented in three PowerPoint slides (Halldorsdottir et al., 2012b)) and one journal article (Kajubi et al., 2020).

Figure 1.

PRISMA flow chart showing the disposition of articles through the systematic review process

The conference abstract evaluated Botswana’s Concurrent Partnerships Campaign (Halldorsdottir et al, 2012a, 2012b). This campaign used mass media, social mobilization, and interpersonal communication in three phases, to (1) stimulate discussion, (2) address risk perceptions, knowledge, and awareness of concurrent partnerships and (3) assess consequences and personalize risk. To evaluate the effect of campaign exposure, data from a national two-stage cluster sampling survey were used, with coarsened exact matching to select a sample and compare those exposed or not to the intervention (N=1138). On the risk of bias tool, the study scored well for having a comparison group, but limited information did not allow for assessment on many criteria.

Using multiple logistic regression, the authors reported that people who were exposed versus unexposed to the campaign showed no statistically significant differences in reported concurrency, measured through either point prevalence (adjusted odds ratio [aOR] = 1.3, 95% confidence interval [CI]: 0.8, 2.1) or cumulative prevalence (aOR: 1.1, 95% CI: 0.8, 1.6). However, those exposed to the campaign did report less risky attitudes and behaviors on several other HIV-related outcomes compared to those who were unexposed. They were significantly more likely to report peer pressure not to engage in concurrent partnerships (aOR: 1.7, 95% CI: 1.1, 2.7), and holding negative attitudes to having variety of partners (aOR: 1.6, 95% CI: 1.1, 2.3). They also reported more consistent condom use (aOR: 1.3, 95% CI: 1.01, 1.7), having been tested for HIV (aOR: 1.6, 95% CI: 1.1, 2.4), higher condom use self-efficacy (aOR: 1.4, 1.02, 1.8), and higher perceived HIV risk (aOR: 1.5, 95% CI: 1.1, 2.0).

The second study used a mixed-methods approach to conduct and assess an intervention with couples in Uganda (Kajubi et al., 2020). The methods consisted of focus group discussions which informed a baseline survey, couples-focused multi-session workshops conducted by respected religious leaders and an endline survey. Though many intervention participants expressed satisfaction with the intervention and perceived improvements in their relationships, no substantive change in key sexual risk behaviors, including sexual partner concurrency, between the first and second surveys was seen. With no substantive change in behaviors, the authors chose not to present findings of difference-in-difference analysis in detail.

DISCUSSION

Despite much academic discussion, our review found that sexual partner concurrency has been the focus of relatively few interventions, most of which have not been rigorously evaluated using pre/post or multi-arm designs. The two studies included in this review, one only reported as a conference abstract, both presented limited information from which to judge the rigor of the analyses. Results suggested that neither program had an impact on concurrency but did have moderately positive effects on either HIV-related behaviors (Halldorsdottir et al, 2012a, 2012b) or on relationship satisfaction (Kajubi et al., 2020).

Although concurrency represents a specific pattern of sexual partnerships, it has often been conflated with multiple partnerships more generally, as evidenced by the phrase “multiple and concurrent partnerships” (MCPs) (Astatke et al., 2012). Based on the limited number of interventions specifically addressing concurrency, concurrency may not be sufficiently different from serial monogamy – or multiple partnerships more generally – to allow for specific messaging in HIV prevention interventions. Timing is an important element in concurrency, as the sexual activity of one dyad must be interrupted by sexual acts with another (Masuda, Miller & Holme, 2021). This may be too difficult to address in an intervention, with the result that, on an actionable behavioral level, efforts to reduce sexual partner concurrency are few. The reasons for this are manifold: individuals may not always know whether they will have sex again with their previous sexual partner, concurrency might not always map on to meaningful local conceptualizations of sexual relationships, and program planners may struggle to craft and evaluate meaningful messages about sexual partner concurrency that are distinct from multi-partner prevention messages. For example, both a polygynous marriage and a casual sexual liaison would technically be considered concurrent sexual partnerships, though their cultural meaning, acceptability and HIV risk may be very different (Reniers & Watkins, 2010).

Developing practical behavioral recommendations addressing concurrency specifically may be too challenging. Instead, messages about reducing partner numbers or avoiding multiple partnerships are more common and may be more easily understood. If successful, they should reduce concurrent partnerships as well. Given these complications, it is not surprising that interventions to decrease concurrency are so rarely reported in the scientific literature.

Table 1.

Evidence Project risk of bias tool

Study	Cohort	Control or comparison group	Pre/post intervention data	Random assignment of participants to the intervention	Random selection of participants for assessment	Follow-up rate of 80% or more	Comparison groups equivalent on socio-demographics	Comparison groups equivalent at baseline on outcome measure
Halldorsdottir et al., 2012	No	Yes	No	No	NR	NA	NR	NA
Kajubi et al., 2020	Yes	No	Yes	No	No	Yes	NR	NA

NA: Not applicable; NR: Not reported

Appendix: Search strategy for PubMed

“sexual partner concurrency” OR “sexual concurrency” OR “partner concurrency” OR (multiple OR concurrent AND (partner OR partner’ OR partner” OR partner’s OR partner’s’ OR partnered OR partnered’ OR partnership OR partnerships) OR “concurrent partner” OR “sexual network” OR “sexual networks” OR “Sexual Behavior/statistics & numerical data”[MeSH terms] OR “Safe Sex/statistics & numerical data” [MeSH terms].

DATA AVAILABILITY STATEMENT

All data are publicly available from the peer-reviewed articles cited in this manuscript.