|Home | About | Journals | Submit | Contact Us | Français|
Voluntary Counseling and Testing (VCT) continues to play a critical role in HIV prevention, care and treatment. In recent years, different modalities of VCT have been implemented, including clinic-, mobile- and home-based testing and counseling. This review assesses the effects of all VCT types on HIV-related risk behaviors in low- and middle-income countries.
We updated a previous review from 1990–2005 by searching for eligible studies through July 6, 2010 on electronic databases, reference lists, and four key journals. To be included, studies had to evaluate pre-post or multi-arm biological, behavioral, or psychological results from an intervention in a developing country in which participants voluntarily received HIV testing and counseling and were administered pre-test counseling, HIV testing, and post-test counseling.
An initial search yielded 2808 citations. After excluding studies failing to meet the inclusion criteria, 19 were deemed eligible for inclusion. Of these studies, two presented duplicate data and were removed. The remaining 17 studies were included in the qualitative synthesis and 8 studies were meta-analyzed. Twelve studies offered clinic-based VCT, 3 were employment-based, 1 involved mobile VCT, and 1 provided home-based VCT. In meta-analysis, the odds of reporting increased number of sexual partners were reduced when comparing participants who received VCT to those who did not, OR= 0.69 (95% CI: 0.53–0.90, p=0.007). There was an insignificant increase in the odds of condom use/protected sex among participants who received VCT compared to those who did not, OR=1.39 (95% CI: 0.97–1.99, p=0.076). When stratified by HIV status, this effect became significant among HIV-positive participants, OR= 3.24 (95% CI: 2.29–4.58, p<0.001).
These findings add to the growing evidence that VCT can reduce HIV-related risky sexual behaviors, thus confirming its importance as an HIV prevention strategy. To maximize the effectiveness of VCT, more studies should be conducted to understand which modalities and counseling strategies produce significant reductions in risky behaviors and lead to the greatest uptake of VCT.
HIV voluntary counseling and testing (VCT) is an important component of HIV prevention programs and a critical entry point into HIV care and treatment. As a prevention strategy, it is believed that VCT can influence behavior change through a process involving individualized counseling, acquisition of HIV/AIDS knowledge and awareness, and learning one’s HIV status. Previous reviews have found that VCT can reduce risky sexual behaviors, thus reducing the likelihood of virus transmission, especially among serodiscordant couples (1, 2). In addition, results from a meta-analysis conducted by Weinhardt et al. (1999) substantiated evidence that VCT can be an effective behavior change strategy for people infected with HIV(3). The majority of studies included in these earlier reviews, however, were from developed countries. Given that implementing VCT programs in resource-limited settings can pose significantly different challenges from those presented in developed countries (4), Denison et al. (2008) conducted a systematic review and meta-analysis of VCT evaluations from low and middle-income countries that were published between 1990 and April 2005 (5). Results of this meta-analysis from seven studies provide evidence in support of VCT as a moderately effective strategy for reducing sexual risk behavior in developing country contexts.
Since the publication of the Denison et al. (2008) review, the field of VCT has continued to expand with the number of facilities globally offering VCT services increasing 35% from 2007 to 2008 alone. Despite this expansion, the majority of individuals worldwide remain unaware of their HIV status (6). In order to increase access and promote awareness of one’s HIV status, different models of VCT delivery have been developed to reach individuals, couples and families not only through different clinic settings and provider-initiated testing(7), but also by bringing VCT closer to people in their communities and homes through mobile and home-based testing approaches. More recently it has been debated whether HIV testing should become a routine part of medical care while still remaining voluntary and confidential. In 2007 the World Health Organization (WHO) and the Joint United Nations Programme on HIV/AIDS (UNAIDS) created guidelines for provider-initiated testing and counseling (PITC) which, as opposed to client-initiated testing and counseling, is performed in health facility as a standard service for all patients in a generalized epidemic and for high-risk patients in a concentrated epidemic setting (7).
This expansion and diversification reflects progress made towards achieving the Sixth Millennium Development Goal of providing HIV treatment to all who need it by 2015, as knowledge of one’s HIV infection status is necessary for accessing lifesaving antiretroviral therapy. The importance of VCT as an HIV prevention strategy, however, was reiterated in a resolution passed by the UN General Assembly in 2006(8), and the rapidly expanding field of HIV testing and counseling has provided new opportunities for evaluation of the service as an HIV prevention tool. This review seeks to examine the effectiveness of VCT, including these newer VCT delivery models such as mobile and home-based testing, on reducing sexual risk behaviors in low- and middle-income countries as a means of preventing future HIV infections.
This review updates the results originally published by Denison et al. in 2008 using updated inclusion criteria that take into consideration the distinction PITC and VCT. Interventions originally included in the Denison et al. review that took place as a routine part of health services, such as receiving HIV testing and counseling at an antenatal care visit, are excluded from this review and analyzed separately in a review of PITC(9).
The primary objective of this review is to systematically examine the literature on the efficacy of VCT in reducing HIV-related risk behaviors in developing countries. This review compares the effectiveness of different approaches to VCT. The three specific aims are as follows:
We included studies conducted in low- and middle-income countries as defined by the World Bank (2010) (10). The period of inclusion for studies was January 1, 1990 through July 6, 2010. Articles published in any language were included if they met the inclusion criteria. Only peer reviewed articles were eligible; conference abstracts and unpublished reports were excluded from the review. We strove to include studies that maintained a high level of methodological rigor, such as randomized controlled trials, but also included interventions with weaker study designs. Specifically, the following study designs were eligible for inclusion: (1) Randomized trial (individual): Minimum two study arms; random assignment of individuals to study arm; (2) Randomized trial (group): Minimum two study arms; random assignment of groups (couples, classrooms, towns, etc.) to study arm; (3) Non-randomized “trial” (individual): Minimum two study arms; assignment of individuals to study arm, but not done randomly; (4) Non-randomized “trial” (group): Minimum two study arms; assignment of groups to study arm, but not done randomly; (5) Before-after study: Pre- and post-intervention assessment among the same individuals. One study arm and one follow-up assessment period; (6) Time series study: Pre-intervention and several post-intervention assessments among the same individuals. One study arm and multiple follow-up assessment periods; (7) Case-control study: Two groups defined by outcome measures, one consisting of cases and one consisting of controls. To be included, the study must compare outcomes between those who got the intervention and those who did not; (8)Prospective cohort: Two or more groups defined by exposure measures and followed over time; (9) Retrospective cohort: Two or more groups defined by exposure measures, but uses previously collected or historical data; (10)Cross-sectional: Exposure and outcome determined in the same population at the same time. To be included, the study must compare outcomes between those who got the intervention and those who did not; (11) Serial cross-sectional: When a cross-sectional survey is conducted in a population at multiple points in time with different people in that population.
Among the eligible study designs, we only included studies that presented the results of comparisons made between participants who received VCT to those who did not receive VCT or ones that made a pre-post test comparison before and after VCT was provided. We also included studies that compared more and less intensive versions of a VCT intervention. Studies that did not meet the inclusion criteria but presented relevant qualitative data, cost-effectiveness data, or other review articles/meta-analyses were collected and included in the review as background articles; a brief data abstraction was conducted for these studies.
This review included studies conducted with a variety of different populations; therefore, we had no restrictions on the types of participants. We only required that participants reside in the low- or middle-income country where the study took place and voluntarily consented to receive HIV testing and counseling, and participate in the study. We included both individual and couples- based VCT interventions. However, we did not include studies that compared individual to couples-based VCT as this comparison addresses a separate research question not posed in our current review.
Studies were included in the review if they evaluated a VCT intervention adhering to the international standards for VCT as initially presented by the US Centers for Disease Control and Prevention (CDC) and UNAIDS (11, 12). For the purposes of this review, we operationalized these guidelines into a concise definition of VCT, which consists of three criteria: (1) receiving pre-test counseling, (2) being tested for HIV, and (3) receiving post-test counseling and test results. In addition, this review included interventions where the client, as opposed to the provider, initiated the HIV test either directly or as a result of outreach services. Studies which are conducted in a healthcare setting where the provider initiates the HIV test for clients seeking other services were not included in this review as this is considered provider initiated testing and counseling (PITC) and was analyzed in a separate review (9).
This review focused on primary outcomes measured post-intervention where results are presented for either a pre/post test comparison or a comparison between at least two study arms. All other outcomes were considered secondary endpoints and were not emphasized in the analysis.
This review included studies reporting at least one outcome measure directly related to HIV prevention, including behavioral, biological, and psychosocial outcomes. Behavioral outcomes included indicators that measure participants’ risk reduction knowledge, attitudes, beliefs, and behaviors regarding HIV. Examples of behavioral outcomes include use of male/female condoms, frequency of having unprotected sex and number of sex partners. Biological outcomes included indictors, such as HIV/STI prevalence and pregnancy incidence, that relate to the biological manifestations of practicing or not practicing protective behaviors related to HIV transmission. Psychosocial outcomes included aspects of behavior or mental health influenced by social factors and includes examples such as self-efficacy, stigma, and support from family and friends. If additional behavioral, biological, or psychosocial outcomes were presented within the studies, we recorded these outcomes as well.
Three different approaches were used to identify eligible articles for inclusion. First, five electronic databases – PubMed, Excerpta Medica Database (EMBASE), PsycINFO, Sociological Abstracts, and the Cumulative Index to Nursing and Allied Health Literature (CINAHL) – were searched using predetermined key words and phrases. These search terms were identical to those used for the Denison et al. (2008) review and were as follows: “HIV counseling and testing” OR “HIV VCT” OR “HIV voluntary counseling and testing” OR “HIV anonymous counseling and testing” OR “HIV ACT” OR “HIV testing and counseling and evaluation” OR “HIV testing and counseling and interventions” OR “HIV counseling and efficacy.” Second, hand-searching was conducted in four key journals including AIDS, AIDS and Behavior, AIDS Education and Prevention, and AIDS Care; the tables of contents of these four journals during the included time period were individually screened for relevant articles. Third, the reference lists of all articles included in the review were screened to identify any additional studies; this process was iterated until no additional articles were found.
Since the search methods were identical to those used in the Denison et al. 2008 review, a new search was conducted from April 2005 through July 2010. These search results were combined with the original search conducted by Denison et al. (2008) that covered dates from January 1990 through April 2005.
Search results from each database were merged and duplicate citations were removed. Trained staff (who received 15 hours of training on systematic review methods and training and practice with the specific methods for this review) screened the titles and abstracts of all search results to eliminate clearly unrelated citations. After preliminary screening, two senior study staff independently screened the remaining records for eligibility. These assessments were merged and discrepancies discussed until consensus was reached. Following this process, citations appearing to fit the inclusion criteria were retrieved and read thoroughly. Once a complete list of eligible articles was compiled, two trained staff independently abstracted the relevant data from the studies using a highly-detailed standardized form with fifteen content sections: (1) citation information, (2) inclusion criteria, (3) study methods, (4) study population characteristics, (5) setting, (6) sampling strategies, (7) study design, (8) unit of analysis, (9) attrition rates, (10) study group (arms or comparison populations), (12) intervention specific questions, (13) outcome measures, (14) outcome results, (15) additional information (e.g. costs and limitations).
All behavioral, biological, or psychosocial outcome variables reported in each study were recorded; however, only outcomes for which there is a pre/post test or multi-arm comparison were extracted in detail. Eligible outcome results were reported in a league table format and included the following: (1) the statistical analysis used, (2) the effect size and base rate, (3) the independent variables analyzed, (4) follow-up times (if applicable), (5) the confidence interval and/or P value, (6) the page number where the results are located, and (7) any additional information deemed of importance. After the two staff members independently coded a study, they compared their coding forms and discussed any discrepancies in data abstraction. If no resolution could be reached, they discussed the issue with a member of the senior project staff until all differences were resolved. We contacted authors if certain aspects of studies needed further clarification. Once resolution was completed, the two data abstractors created a final coding form, and all data were entered into a statistical database (SPSSTM, Chicago, IL) for analysis.
To standardize comparisons of rigor across studies, we used a rigor table previously developed for the synthesis project. This table recorded the following aspects related to study rigor: (1) Prospective Cohort analyses presented data from the same subjects followed over time; (2)Control or comparison groups compared those who received VCT to those who did not (including a more-versus-less intensive intervention); (3) Pre/post intervention data assessed participants before and after receiving VCT; (4) Random assignment of treatment groups of study subjects; (5) Random selection of subjects for assessment; (6) Attrition determined if the follow-up rate was 80% or more; (7) Comparison groups matching assesses if there were statistically significant differences in socio-demographic variables across study arms; and (8) Comparison group matching on outcome measures assesses whether studies had a statistically significant baseline difference in study outcomes. For articles reporting the results of randomized controlled trials, risk of bias was assessed using the Cochrane Collaboration’s tool for assessing risk of bias (13).
Standard meta-analytic methods (14) were used to standardize effect sizes of study outcomes that were comparable across studies using the software package Comprehensive Meta-Analysis (CMA) V.2.2. We assessed a study’s eligibility for inclusion in the meta-analysis based on a set of meta-analytic guidelines for social and behavioral interventions developed for the project Effect estimates from included studies were converted to the common metric of an odds ratio for studies that compared two groups (either intervention/control groups or before/after groups) using dichotomous outcome variables. If outcomes were presented as continuous variables, the effect size was entered into CMA and standardized as an odds ratio. If studies presented categorical data for an outcome of interest, the outcome was dichotomized when possible to utilize the common metric of the odds ratio. For example, if a study reported the number of sexual partners in the last year as 0–1, 2–3, or >3, the categories were collapsed into 0–1 or >1. When studies reported an outcome of interest in multiple ways, such as providing information on consistent condom use with primary partners and non-primary partners, the two outcomes were combined so that a singular effect size for consistent condom use reported among all partners could be utilized in meta-analysis. When studies presented results from more than one follow-up time period, data from one time point was selected for inclusion in meta-analysis based on its similarity to the follow-up times utilized in other included studies.
Because this review built on the findings of the Denison et al. (2008) review, we employed similar meta-analytic methods and included the same outcome measures in meta-analysis: number of sex partners and condom use, which is the inverse of the unprotected sex variable used in the Denison review. However, since slightly different inclusion criteria and meta-analytic methods were used, not all studies meta-analyzed in the Denison review were included in this update. Additionally, many studies were not included in the meta-analysis because the effects were too heterogeneous to be synthesized. Sub-group analyses by modality of VCT were not feasible given that the majority of included studies took place in a clinic setting. Sub-group analyses based on gender and HIV status were possible and the results of these meta-analyses are presented below.
After searching the literature, 2808 records were identified, 2668 were eliminated during initial screening, and 85 additional records were eliminated after a second round of screening. The full texts of the remaining 55 articles were obtained to ascertain their eligibility. In total 19 articles met all inclusion criteria for the review. Among these 19 articles, two articles (14, 15) analyzed data from the same cohort study in eastern Zimbabwe. To avoid double-counting data from the same population, we only included the later article, Cremin et al., 2010, because this article provided data from a longer period of follow-up (from 3 time points instead of 2). Additionally, the later article recalculated the statistical analyses to take into account the degree of change between behavioral outcomes over time and the analyses were stratified by HIV infection status to provide a more accurate comparison group (15). Importantly, the earlier article found that VCT was associated with increased risk behavior among individuals who tested HIV-negative in terms of numbers of partnerships in the last month, the last year, and in concurrent partnerships (16). However, in the revised analyses, no increase in risk behavior was found for individuals who tested HIV-negative.
Two additional articles (17, 18) analyzed data from the same cohort of participants in Rakai, Uganda; however, the Matovu et al. 2007 article included a comparison of participants who had repeatedly gone through VCT as compared to first-time acceptors and those who refused (17). Since our research question does not address repeat testing and counseling, we choose to only include the Matovu et al., 2005 study as it was most comparable to the other studies included in the review (18).
The VCT Efficacy Study (2000) and the Grindstead et al. (2001) articles utilized data from the same randomized trial, but since the two articles assessed different behavioral outcomes, results from both articles were included in the review (19, 20). Each article presents outcomes separately for those who were randomized to individual VCT and those who were randomized to couples VCT; therefore, we considered individual VCT and couples VCT to represent two separate study populations and each was considered a separate study. Additionally, these studies took place in three locations: Trinidad, Tanzania, and Kenya. Although Trinidad is currently considered a high-income country (10), it was considered a middle-income country when the studies took place in 1995–1998; therefore, we included these two studies in the review.
After these decisions were made, 17 studies from 17 articles were included in the review (Figure 1). Nine of these studies were originally identified in the Denison et al. (2008) review; the remaining eight studies were found during the updated search.
Of the 17 articles included in the review, 10 took place in Sub-Saharan Africa, including four in Zimbabwe(15, 21–23), two in Uganda(18, 24), two in Zambia(25, 26), one in Mozambique (27), and one in Kenya (28). Two studies, one among individuals and one among couples, were multi-site, with locations in Trinidad, Kenya, and Tanzania (19, 20). The remaining five studies took place in the following locations: two in China (29, 30), two in Thailand (31, 32), and one in Guatemala (33). Most studies (N=12) took place in an urban location, while three took place in rural settings (15, 18, 30), one took place in a peri-urban setting (31), and one took place in a mixed setting involving both urban and rural locations (28). Combined, the 17 studies included a total baseline population of 40,309 participants. Study designs included one individual-randomized trial (19, 20), two group-randomized trials (19–21), three prospective cohort studies (15, 18, 27), eight before/after studies with no comparison group (23–25, 28–31, 33), one time-series study (22), and two cross-sectional studies (26, 32).
Nine included studies targeted sexually active adult men and women in the general population (15, 18–20, 27–28, 31–33). Five studies focused on couples (19, 20, 25, 26, 30), and three studies focused on employees (21–23). One workplace-based intervention in Zimbabwe included only male employees (22), whereas the other workplace-based interventions included both male and female employees (21, 23). Only one study focused on injection drug users (29). One study (32) focused solely on behavior among individuals who tested HIV-positive, whereas the remaining studies included both HIV-positive and HIV-negative participants. Table 1 presents study locations, target populations, participant demographics, and intervention descriptions for the included studies.
Risk of bias was assessed for the three articles (3 studies in total) that utilized a randomized study design (19–21). In all studies, participants and personnel were not blinded as the intervention arms included activities, such as VCT or health information, that were impossible to conceal. Due to the nature of the interventions, we do not believe the lack of blinding greatly influenced the study outcomes. In all five studies, it is unclear whether interviewers collecting behavioral information from participants were blinded; however, Corbett et al. states that lab staff members were blinded during specimen analysis. The VCT Efficacy studies (19, 20) showed low risk of attrition bias and selective reporting bias, whereas Corbett et al., 2007 demonstrated a risk of attrition bias as attrition rates differed between intervention and comparison groups. Additionally, Corbett et al., 2007 was not powered to detect differences in HIV incidence, which was the main outcome presented in the study. Risk of bias assessments are presented in Table 2.
For the remaining 15 studies that utilized non-random study designs, the quality of the evidence varied. All but three studies (26, 31, 32) studied a cohort of participants over time. Seven studies included some type of a control or comparison group (15, 18, 22, 26, 27, 31, 32) and 12 studies included data from pre- and post-intervention (15, 18, 21, 22–25, 27–30, 33). Only two studies randomly selected individuals for assessment (18, 26). Among studies that followed one group of people through time, three studies (22, 29, 30) reported an overall follow-up rate of 80% or greater. Quality assessment results are presented in Table 3.
There were not enough studies to assess publication bias in this review.
Despite the various implementation modalities for VCT, including clinic-based, mobile, and home-based approaches, the vast majority of included studies were clinic-based (19, 20, 25, 26, 27, 28, 30, 31, 32, 33). One study in China was specifically geared toward couples obtaining a marriage license (30). Although HIV testing and counseling is currently not required in China, this program sought to test the feasibility of integrating VCT into the existing required premarital counseling and health exams (30). One study, Matovu et al., 2005, offered VCT in the participants’ homes or in a venue of their choosing; however, the article did not report how many participants received VCT in their home versus at a clinic or other setting (18). Cremin et al., 2010 provided mobile VCT services; however, participants who reported having had VCT through other means (e.g. at government clinics or non-governmental organizations) were categorized as having “received VCT” (15). The article does not specify the percentage of participants who received VCT through mobile services and the percentage of those that received VCT through other modalities (15). Three studies (21–23) offered VCT in worksites. The remaining 12 studies offered clinic-based VCT.
The length of time between being tested for HIV and receiving test results varied across studies. Only seven studies reported that participants received same-day HIV test results (21, 23, 25–26, 28–30). However, many studies took place prior to the availability of rapid HIV testing.
Three studies measured HIV incidence as a primary outcome (18, 21, 22). Corbett et al., 2007 utilized data from a cluster-randomized trial among 22 businesses in Harare, Zimbabwe to assess HIV incidence among employees in businesses randomized to receive rapid on-site VCT as compared to employees in businesses randomized to receive paid vouchers for VCT with an external provider. The results showed no statistical difference between unadjusted HIV incidence in the intervention group (1.37/100 person-years) as compared to the control group (0.95/100 person-years) after the completion of follow-up (average 1.7 years between HIV tests). The adjusted incidence rate ratio was 1.49 (95% CI 0.79–2.80). However, uptake of VCT was quite different across sites; in the rapid, on-site group, VCT uptake was 70.7% compared to only 5.2% in the group receiving off-site vouchers. Machekano et al., 1998 (N=1683) examined HIV incidence among a cohort of male factory workers in Harare, Zimbabwe and observed that participants who completed VCT and decided to receive their test results had a higher HIV incidence, IR= 3.29 (2.31–4.56) per 100 person-years, as compared to those who chose not to receive their test results, IR= 1.76 (1.01–2.86) per 100 person-years, IRR= 1.87 (1.01–3.61, p=0.030). However, this study also found that among those who received their results, HIV incidence was lower following the period after obtaining the results, IR=3.04 (1.86–4.70) per 100 person-years as compared to the period prior to obtaining results, IR=4.82 (2.75–7.81) per 100 person years, IRR= 0.63 (0.31–1.30, p=0.180). Matovu et al., 2005 followed a cohort of participants (n=10,694) to assess their willingness to participate in VCT and compared HIV incidence among participants who accepted VCT to those who refused. This study observed no significant difference between those who accepted VCT (IR= 1.6/100 person-years) and those who refused (IR= 1.4/100 person-years), p=0.60.
Five studies measured STI prevalence or STI incidence (19–22, 28). Arthur et al., 2007 (N=401) found that the proportion of participants reporting any STI symptom decreased 6 months after receiving VCT (15.3%) as compared to baseline (39.7%), p<0.02. Matambo et al., 2006 (N=366) observed no significant pre-to post-intervention difference in participants reporting a genital ulcer or discharge (1.3% and 1.6%, respectively, p=0.73). Machekano et al., 1998 (N= 2060) detected no pre/post difference in STI incidence among participants who completed VCT and received their test results (IRR=1.30, 95% CI 0.95–1.80, p =0.100); however, the authors did find that subjects who went through VCT and received their test results had a significantly higher incidence of STIs than those who went through VCT but chose not to receive their test results (IRR=4.37, 95% CI=3.47, 2.51–4.89, p=0.001).
Only three studies (19, 28) presented outcomes related to positive and negative life events associated with VCT. Grinstead et al., 2001 (n= 3120 enrolled as individuals, 1173 enrolled as couples) found that among participants receiving individual and couples VCT, positive life events were relatively common, such as strengthening a sexual relationship (42% overall) and negative life events, such as physical abuse (4.5% overall), were relatively rare. The most common negative life event was the break-up of a sexual relationship (27% overall). Arthur et al., 2007 (n=401) found no significant changes in the rates of life events following VCT, regardless of HIV status. Both studies (19, 28) found that rates of HIV serostatus disclosure were significantly higher among those who tested HIV-negative as compared to those who tested HIV-positive.
Meta-analysis was conducted on two primary outcomes: number of sexual partners and condom use. The VCT Efficacy Study (2000) measured unprotected sex with primary and non-primary partners among individuals and unprotected sex with enrollment partner and with non-enrollment partners among couples. To include this outcome in meta-analysis, the two effects were combined using CMA to obtain one overall effect. Muller et al., 1995 and Matovu et al., 2005 differentiated between consistent, inconsistent, and no condom use. To dichotomize this outcome for meta-analysis, we compared consistent condom use to inconsistent or no condom use. Mola et al., 2006 contained several different condom use outcomes including: condom use at last sex, condom use at last sex among those reporting having had sex with friends/prostitutes in the last month and always/sometimes condom use. We chose to include the effect size for always/sometimes condom use as we felt this was the most comparable outcome to the other included studies. Additionally, the VCT Efficacy Study (2000) measured unprotected sex at two time-points—the first follow-up visit occurring on average 7.3 months after baseline and the second follow-up visit occurring on average 13.9 months after baseline. Mola et al., 2006 also included two follow-up times: 4 and 6 months after receiving the intervention. To have the most accurate comparison, we chose to use the 1st follow-up time for the VCT Efficacy study (2000) (avg. 7.3 months after baseline) and the 2nd follow-up time for Mola et al., 2006 (6 months after baseline).
Twelve studies reported condom use or unprotected sex as a primary outcome (15, 18, 20, 23–24, 27–29, 31–33). For studies that reported unprotected sexual activity (20, 28), we took the inverse of this outcome to make it comparable to condom use outcomes presented in other studies. Of studies reporting either condom use or protected sex, seven were considered eligible for meta-analysis: VCT Efficacy 2000 (2 studies), Muller et al., 1995, Matovu et al., 2005, Mola et al., 2006, Arthur et al., 2007, and Cremin et al., 2010. After disaggregation by sex and HIV status within the results of these seven studies, a total of 14 discrete effects were used to calculate an overall effect size estimate (N= 23923). The random-effects pooled odds ratio across all studies showed a positive but not statistically significant effect of VCT on condom use (OR: 1.39, 95% CI: 0.97–1.99, p=0.076). There was significant heterogeneity, Q=94.10, p<0.001, which indicates the findings across studies were inconsistent. When stratified by gender, the random-effects pooled odds ratio for condom use was not statistically significant for either men (OR: 1.31, 95% CI: 0.64–2.67, p=0.46) or for women (OR: 1.17, 95% CI: 0.76–1.80, p=0.49). However, when stratified by HIV status, the random-effects pooled odds ratio for HIV-positive participants, combining males and females, was statistically significant (OR=3.24, 95% CI: 2.29–4.58, p<0.001). These results suggest that participants who received VCT and were given positive results had greater odds of using condoms than HIV-positive participants who did not receive VCT. However, these results should be interpreted with caution as this analysis contained only two studies (N=8873) (15, 32). Since there was only one study (15) that provided results separately for HIV-negative participants, this outcome could not be meta-analyzed.
Only one study, Matovu et al., 2005, found that participants who received VCT had reduced odds of condom use as compared to participants who did not receive VCT (18). Self-selection bias was an issue in this study as participants either agreed to receive VCT or refused; there was no randomization. The authors suggest that condom users may exhibit more risky behaviors, leading to increased fear they are HIV infected and discouraging them from accepting VCT (18). Cremin et al., 2010 also showed that the odds of condom use were reduced among HIV-negative female participants who received VCT compared to those who did not, especially with non-regular partners; however, the confidence interval for this outcome is extremely large, suggesting a high degree of imprecision and a non-significant effect (15). Results from all analyses related to the condom use/protected sex outcome are detailed in Figure 2 and Table 4.
The VCT Efficacy Study (2000) was the only randomized controlled trial that reported results on protected sex. For both individual and couples-based VCT, the results from this trial show that receiving VCT did not have a significant effect on protected sex comparing those who received the intervention to those who did not (20).
Six studies reported on the number of sexual partners, either comparing the same individuals before- and after- receiving VCT (28, 31, 33) or comparing intervention groups who received VCT to comparison groups who did not receive VCT (15, 18, 32). Of these, five studies were included in the meta-analysis: Arthur et al., 2007, Matovu et al., 2005, Muller et al., 1995, Samayoa et al., 2010, and Cremin et al., 2010 (N=18573). The overall pooled random effect odds ratio was 0.69 (95% CI: 0.53–0.90, p=0.007), demonstrating that participants who did not receive VCT had significantly higher odds of reporting greater numbers of sexual partners as compared to participants who did receive VCT. The Q statistic, 61.84, was significant (p<0.001), which suggests substantial amounts of heterogeneity between study results. When stratified by gender, the effect of VCT on the number of sexual partnerships became statistically insignificant; the random-effects pooled odds ratio was 0.86 (95% CI 0.57–1.29, p=0.46) for males and 0.70 (95% CI: 0.47–1.04, p=0.08) for females. When stratified by HIV status, the random-effects pooled odds ratio was 0.61 (0.37–0.997, p=0.048) for HIV-positive individuals and 0.90 (95% CI: 0.77–1.10, p=0.195) for HV-negative individuals. All individual studies, including all sub-populations, showed a positive trend toward reduced number of partners for those participants receiving VCT compared to those not receiving VCT; no results suggested that receiving VCT increased the number of sexual partners of participants. Results for number of partners are presented in Figure 3 and Table 4 below.
This systematic review and meta-analysis synthesized results across 17 studies and 5 HIV-related biological, behavioral, and psychological outcomes, including HIV incidence, STI prevalence/incidence, positive and negative life events, condom use/protected sex, and number of sex partners. Overall, there were no significant differences in HIV incidence and STI incidence/prevalence comparing participants who received VCT and those who did not. The two studies reporting on positive and negative life events suggest that individuals who receive VCT do not experience a significant increase in negative life events as compared to those not receiving VCT. However, these studies also found that reported disclosure of HIV status was more common among participants who received VCT and tested negative as compared to those who tested positive. As disclosure of one’s HIV status remains an important strategy for HIV prevention (34), VCT counselors should work with clients, especially those who test HIV positive, to discuss ways of disclosing their HIV status to sexual partners, and should refer clients to further counseling if necessary.
For outcomes relating to HIV-related risk behavior, results from our meta-analysis demonstrate that people who received VCT were more likely to report reducing their number of sexual partners than those who did not receive VCT. This result stands in contrast to the random-effects pooled odds ratio in the Denison et al., 2008 review, which found no significant difference in odds of reported number of sex partners comparing those who received VCT to those who did not. However, the current meta-analysis incorporated only two studies originally included in the Denison et al., 2008 review for this outcome (18, 32) and added three new studies (15, 28, 33). Therefore, it is possible that the inclusion of additional studies enhanced the statistical power necessary to detect a significant difference in the reported number of sex partners comparing participants and non-participants of VCT interventions. This result is encouraging in that it shows VCT can have a significant effect on reducing HIV related risk behaviors for all participants, regardless of serostatus.
For the condom use/protected sex outcome, results from the meta-analysis show that VCT had no significant overall effect. Although VCT was not statistically effective in increasing condom use across all VCT participants, those who received a positive test result were significantly more likely to report condom use than HIV positive participants who did not receive VCT. The significant effect of VCT on condom use for HIV-positive individuals has been demonstrated in previous findings (3) and also among serodiscordant couples (1, 2). The Denison et al., 2008 review also found a significant effect of VCT on increased condom use, including both HIV positive and negative participants (5). In contrast, this meta-analysis found an insignificant effect of VCT on condom use when combining both HIV-positive and HIV-negative participants. As a potential explanation for this discrepancy, the Denison et al., 2008 review included studies that took place in antenatal settings. In the current review, these studies were moved into a separate category comprising provider-initiated testing and counseling (PITC), which is being analyzed in another review (9). Additionally, the current review comprised 4 studies originally analyzed in the Denison et al., 2008 review: VCT Efficacy 2000—2 studies, Matovu et al. 2005, and Muller et al. 1995, in addition to including three new studies: Cremin et al. 2010, Arthur et al. 2007, and Mola et al. 2006. Estimated effects from the current review and the previous review (Denison 2008) show similar results, as the odds of condom use/protected sex were increased for VCT participants as compared to non-participants; however, significant heterogeneity was found in both analyses, suggesting inconsistent results across studies.
Originally this study aimed to assess the behavioral effects of different VCT modalities, including clinic-, mobile-, and home-based testing and counseling. However, since the vast majority of studies took place in clinic-based settings, differences in effects by modality could not be meta-analyzed. Although our review contained mostly clinic-based VCT, some studies demonstrated a much higher uptake of VCT when different modalities were offered. For example, one included study demonstrated that VCT uptake was 70.1% for onsite employment-based testing as compared to only 5.2% for offering vouchers for an off-site provider (21). Given the increasing importance of testing and counseling in accessing HIV care and treatment, more research should be conducted to assess which modalities are most successful at both increasing uptake of VCT and influencing behavior.
There are several limitations to this systematic review and meta-analysis that should be considered. Firstly, it is possible that our extensive search methods failed to identify eligible studies. We sought to minimize this limitation by searching not only electronic databases but also the reference lists of included articles as well as hand-searching the table of contents of four relevant journals. Additionally, since this review only included published studies, it is possible that unpublished studies exist that show negative findings, thus resulting in publication bias. This review spans two decades of HIV interventions related to testing and counseling. During this time, there have been numerous advances in the field of HIV, including increased availability of anti-retroviral therapy (ART) and increased availability of rapid, same-day VCT services. Therefore, it is possible that participants who volunteered for VCT prior to the wide availability of ART and rapid-testing are different in some aspects from those that volunteered in more recent years. Additionally, VCT interventions must be understood in the context of the larger HIV prevention efforts that have been underway since the early 1990s. The majority of studies in this review have relatively short follow-up time periods of one year or less. Therefore, we were not able to assess any larger time trends surrounding HIV prevention or any long-term effects of VCT. Finally, outcomes presented in the included studies often varied. For example, some studies reported consistent condom use in the past two weeks while others reported the amount of unprotected sex in the past month. In the future, developing standardized HIV-related behavioral outcomes would make comparisons between studies more valid.
In conclusion, results from this review and meta-analysis bolster the growing evidence (1–3, 5) that VCT reduces risky behaviors related to HIV by significantly reducing the number of sex partners of participants. Additionally, people living with HIV who received VCT exhibited increased odds of using condoms and engaging in protected sex than people living with HIV who did not receive VCT.
In the past year, innovative, effective interventions for HIV prevention have been discovered, including treatment as prevention (35) and pre-exposure prophylaxis (PrEP) (36). However, these interventions will only succeed if more people become aware of their HIV status, highlighting the continued importance of VCT in preventing HIV and accessing care and treatment. Although this review adds to the evidence that VCT reduces risky behaviors, relatively few details are known about which modalities of VCT are most effective. Additionally, the included studies revealed little information about how intervention components were implemented, such as which counseling strategies were employed during pre- and post-test counseling. These details are precisely what must be understood in order to maximize the effectiveness of VCT in regards to changing sexual risk behavior, thus maximizing this intervention’s ability to prevent HIV, and for those who test positive, increasing the uptake of HIV care and treatment.
This research was supported by the US National Institute of Mental Health, grant number 1R01MH090173. We wish to thank Krissy Brinsely, Elizabeth McCarthy, Amolo Okero, Sarah Finocchario, Jennifer Gonyea, Andrea Ippel, Amanda Rosecrans, Devaki Nambiar, Stephanie Mullins-Sweat, Marta Mulawa, Samantha Dovey, Jewel Gausman, Alexandria Smith, Eugenia Pyntikova, Hieu Pham, Tina Dickenson, Erica Layer, Jeremy Lapedis, and Aisha Yansaneh for their coding work on this review. We would also like to thank Amy Medley and Elena Tuerk for their work as coordinators for The Evidence Project.