Effects of PREPARE, a Multi-component, School-Based HIV and Intimate Partner Violence (IPV) Prevention Programme on Adolescent Sexual Risk Behaviour and IPV: Cluster Randomised Controlled Trial

Young South Africans, especially women, are at high risk of HIV. We evaluated the effects of PREPARE, a multi-component, school-based HIV prevention intervention to delay sexual debut, increase condom use and decrease intimate partner violence (IPV) among young adolescents. We conducted a cluster RCT among Grade eights in 42 high schools. The intervention comprised education sessions, a school health service and a school sexual violence prevention programme. Participants completed questionnaires at baseline, 6 and 12 months. Regression was undertaken to provide ORs or coefficients adjusted for clustering. Of 6244 sampled adolescents, 55.3 % participated. At 12 months there were no differences between intervention and control arms in sexual risk behaviours. Participants in the intervention arm were less likely to report IPV victimisation (35.1 vs. 40.9 %; OR 0.77, 95 % CI 0.61–0.99; t(40) = 2.14) suggesting the intervention shaped intimate partnerships into safer ones, potentially lowering the risk for HIV.


Introduction
Globally, HIV is ranked second among the leading causes of death among adolescents [1]. Among adolescents and youth in South Africa, there has been little progress in preventing new infections. Although declining somewhat, the HIV prevalence and incidence among young South Africans 15-24 years remains high, especially among women [2][3][4]. In the Western Cape, South Africa, the setting of this study, HIV is still the leading cause of premature mortality (http://www.mrc.ac.za/bod/WC2010Re port.pdf), and adolescents commonly report an early sexual debut and unprotected sex [5]. These behaviours increase the risk of sexually transmitted infections (STI) including HIV.
In South Africa, adolescents' intimate relationships are marked by a high incidence of violence [6]. Sexual violence and intimate partner violence (IPV) increase the risk of STIs including HIV among women [7]. This implies that to be effective, HIV prevention interventions should include a focus on preventing sexual violence and IPV. In the Western Cape and Limpopo provinces of South Africa, cluster RCTs of school-based HIV prevention interventions without a focus on IPV, failed to demonstrate an impact on the timing of sexual debut or condom use among the younger adolescent participants (average age 13 years) and also failed to impact on IPV [5]. However, in the Eastern Cape, a community-based HIV and IPV prevention programme for adolescents 16 years and older, which included a substantial focus on IPV, showed a beneficial impact on STI incidence and male IPV perpetration [8].
In our cluster randomised controlled trial, PREPARE, conducted among young adolescents (average age 13 years) in the Western Cape, we evaluated an HIV prevention programme that included a focus on IPV and sexual violence reduction. In the current trial we tested the hypothesis that the PREPARE programme would 1. delay sexual debut; 2. increase the use of condoms; 3. decrease the number of sexual partners among young adults. A secondary objective was to assess the effect of the intervention on IPV, and (not reported here) the three-year incidence of conceptions among female participants.

Methods Intervention
The intervention was multi-component, comprising an educational programme, a school health service and a school safety programme, and described in detail in Table 1. The educational programme consisted of 21 sessions delivered once a week, immediately when school ended, in the school premises. The session duration ranged from 1 to 1.5 h. Sessions included up to 25 participants, and the education methods were interactive and skillsbased. The programme was built upon the Respect4U programme, an IPV and HIV prevention intervention which was developed based on the Jewkes conceptual framework [9], and piloted-tested among Grade 8 students in nine typical Western Cape high schools over 3 years [10]. It was also informed by social cognition models including the Reasoned Action Framework [11] and the I-Change theoretical model [12]. Staff employed by the PREPARE project, who had been screened for positive gender norms and comfort with sexuality education and condom demonstrations, facilitated groups of up to 25 participants. The facilitators received a two-week training course and subsequent weekly training, supervision and session preparation support.
The school health service (SHS) was implemented in collaboration with the Western Cape Department of Health, the City of Cape Town Health Department, and the Desmond Tutu HIV Foundation. A nurse from the public clinic nearest to the school delivered the service in the school premises, once a week immediately after school ended. The service was modeled on the new South African Integrated School Health Policy (http://www.health-e.org.za/2013/10/ 24/integrated-school-health-policy/), was free, and involved SRH education, identification of need for SRH services or commodities and referral for such services or commodities to the nearest community clinic, where they were provided free of charge. Some clinics were also able to send a health promoter to assist with health education. There are no known randomized controlled trials of the impact of SHS on adolescent SRH, however they are considered to be an accessible and acceptable strategy [13].
The school safety programme comprised two initiatives. School safety teams were invited to a two-day training at a central venue, conducted by the PREPARE team with the Centre for Justice and Crime Prevention (CJCP) (a nongovernment organisation). We implemented ''Photovoice'', a carefully piloted programme [14], for twenty, randomly selected students at each school, facilitated by two PRE-PARE researchers.

Control Condition
Participants in the control schools received school as usual, which excluded the after-school programme, the school health service and the safety programme.

Sampling
The study population comprised adolescents in Grade 8 (average age 13) in public high schools in the Western Cape. The sample size was calculated to have the power to show an intervention effect on the annual incidence of the primary outcome of sexual debut. We estimated 83 % of the participants would report they had not yet had their sexual debut at baseline [5]; 17 % of participants in the control arm would have their sexual debut during the PREPARE follow-up period of 1 year [5]; an intra-class correlation coefficient (ICC) of 0.06; and 20 % participant drop-out after 1 year. We used the Hayes and Bennett formula [15] to estimate that we would need 19 schools in each arm, with 62 of 75 participants pre-sexual debut at baseline in each school, to show a 50 % relative reduction in 1 year incidence of sexual debut (17 % in control schools and 8.5 % in intervention schools) with 80 % power for a 2-sided test with a significance level of 5 %.
We sampled 40 schools to allow for one school per arm to drop out of the study. They were randomly sampled using the database of 359 public high schools in the Western Cape Province. Of the 359 schools, before sampling we excluded: 1 school with Grade 12 pass rates below 40 % (indicating their inability to deliver on their core educational mandate); 33 schools with pass rates above 97 % (indicating well-resourced schools already able to offer students the types of interventions proposed by PREPARE); 67 schools in two of the eight districts

Randomisation and Masking
We stratified schools into two strata based on the Grade 12 pass rate which we assumed was an indication of how well the school functioned and its potential ability to benefit from the PREPARE programme. We found that pass rate was correlated with the amount of school fees charged, indicating it is also a reflection of socioeconomic status. Before allocating schools to conditions, we invited each of the 42 schools to participate, and all accepted.
To ensure allocation sequence concealment, a statistician at the South African Medical Research Council who did not have any knowledge of the schools, allocated them within each stratum to intervention and control arms of the study. Using a spreadsheet, he ordered the school names randomly within each stratum and then used a random number generator to give each school a number. Within each stratum, he allocated the schools with the lowest random numbers to the intervention arm, and the other half to the control arm. We selected classes randomly so that we would obtain a sample of at least 75 assenting students per school with signed parental consent.
After allocation assignment, the director of one of the Department of Health sub-districts requested that we exclude three of the participating schools because they had been identified for a Department of Health intervention. One had been allocated to the comparison arm and two to the intervention arm. We replaced these schools with randomly selected schools from the database, and the statistician randomly assigned one to the control arm and two to the intervention arm. When we were making plans to conduct the baseline survey, one school in the control arm withdrew from the study because they were unable to find time for the survey. We did not replace this school (Fig. 1). We were not able to mask intervention assignment.

Recruitment of Participants
We sampled all Grade 8 classes in the selected schools, or, in large schools, we randomly selected Grade 8 classes to ensure at least 75 participants. We invited 6244 Grade 8 students in the 41 schools to participate in the PREPARE  1). In the intervention schools, participants were given an invitation to the afterschool programme after they had completed the baseline survey. To encourage attendance, we offered refreshments at each session and small stationary gifts at selected sessions. Each participant had a ''loyalty card'' which was stamped at each session or nurse consultation. We gave R50.00 (* US$5) supermarket gift voucher and a certificate for those who attended at least 15 sessions.

Measures
The instrument comprised a paper, self-administered questionnaire for each of the surveys (baseline and 6 and 12 months post-baseline) with a set of common questions across time points. Each question was provided in the three languages commonly spoken, and printed in full colour in an adolescent-friendly format resembling a ''teen magazine''. The questionnaire was informed by an instrument used in a previous study [5] and formative qualitative research we conducted with adolescents in six schools to identify the salient attitudes, beliefs, and social norms that served as barriers and facilitators to safe sexual behaviour. It was piloted (including conducting group cognitive interviews after students completed it in a classroom setting) and revised during 2011 to ensure maximum face validity and reliability of measures. The surveys were conducted during school hours and completion of the questionnaire took on average 45 min. At the 6-and 12-month surveys, we visited the school up to three times in an attempt to increase retention of absent participants. In the 12-month survey, we surveyed students who had left the participating school at their new school, at their home or at a convenient venue in the community in which they lived. (We had obtained consent to make such arrangements and had collected contact details of all participants). The behavioural outcome variables and the variables measuring theorized motivational determinants and their psychometric properties are described in Table 2. The items to measure IPV were adapted from the WHO multicountry study [16]. We measured participants' attendance at each of the 21 weekly after-school education sessions over 6 months between February and September 2013, using a register of participants, and roll calls at each session. School nurses kept registers of the participants who consulted them and     (1) to 'strongly agree' (5) e Scale from 'very difficult' (1) to 'very easy' (5) f Scale from 'never' (0), 'once' (1) to 'more than once' (2) g Scale from 'never' (1), 'sometimes' (2) to 'often' (3) h 'No' (0) and 'yes' (1) i Scale from 'not at all' (1), '1 time' (2), '2-5 times' (3), '6-10 times' (4), '11-20 times' (5) to 'more than 20 times' (6) j Scale from 'never' (1) to 'every time' (5) k Never (0), Once or more than once (1) individual patient records. Each participant's attendance records were linked with the data from their baseline survey.

Process Evaluation
Process data were collected in the 6-and 12-month followup questionnaires. To assess fidelity to the intervention, facilitators of the education sessions were observed at random, unannounced times by two independent observers, for a total of two sessions each in different schools. Their performance was scored from 1 to 5 on 13 aspects of session facilitation. A score of 1 indicated that aspect had not been achieved, and a score of 5 indicated that aspect had been excellent. The 13 aspects were grouped into following three dimensions: interaction with learners (for example, listened attentively); session fidelity (for example, implemented all activities); facilitation skills (for example, encouraged learners' involvement). Each of the 14 facilitator's scores was summed and averaged for the 4 observations.

Analysis
We computed the dimensionality and internal consistency of the scales measuring the motivational constructs and then calculated mean scores per construct. We performed a baseline descriptive analysis of the participants' demographics, self-reported sexual risk behaviour and scores on the motivational constructs in intervention and comparison arms. At the 6 and 12 months follow-ups, we report the estimates in the intervention and comparison groups for the three primary sexual behaviour outcomes and five other sexual behaviour outcomes; the two secondary IPV outcomes and four other variables measuring sexual coercion; and 17 motivational constructs related to the primary outcomes. Regression was undertaken to provide outcomes at 6 and 12 months with odds ratios for dichotomous variables and coefficients for continuous variables, adjusted for baseline demographics (age, gender, socio-economic status), the baseline measure in question, and clustering using the complex samples approach in SPSS Version 20. These analyses were carried out among the full sample, the subsample that attended at least 1 session, and the subsample that attended at least 10 sessions (50 %) in order to be able to detect the effect of attendance rate on outcomes. In addition to these complete case analyses, all analyses were repeated on a data file that had missing values on latent factors imputed by applying the expectation-maximisation algorithm (EM). The EM algorithm uses maximum likelihood estimators and is considered one of the most favoured ways to impute missing data [17].

Response Rate and Retention
We sampled 6244 adolescents in 41 schools, and 3451 (55.3 %) obtained signed parental consent and assented to participate (Fig. 1). The non-responders included 69 students and 281 parents who declined, and the remaining were students who did not bring back signed parental consent forms. Retention at 6 months was 94.1 % and at 12 months 87.9%.

Exposure at the School Level
We implemented the after-school educational programme in all 20 intervention schools, but in two schools we were unable to complete it. In one of the two, the programme was interrupted by a religious fast. In the other, the school could no longer find a free afternoon for the sessions. We trained nurses and health promoters for 17 of the 20 intervention schools, and the school health service operated in 17 schools. The public health services did not have capacity to provide school health nurses for two of the remaining schools and in one, the school was not able to allocate an afternoon session for the school health service. The school safety teams of 18 of the 20 intervention schools participated in the school safety training, and we trained 53 school safety team delegates in total. None of the schools implemented participatory safety audits or develop safety plans during the 6 months following the safety training. We implemented the Photovoice programme in 10 of the 20 intervention schools. In two of these schools we did not complete the implementation of Photovoice because students did not feel safe taking photographs in the school premises (one school) and because attendance had dwindled (one school). An overview of exposure at the school level shows that seven schools were exposed to all four components of the intervention (educational sessions, school health service, school safety training and Photovoice); seven schools had all components except Photovoice; two schools had all components except the health service; two schools had all components except the safety training; and two schools were exposed to only the educational sessions and the school safety training.

Fidelity to Programme
The average facilitator performance scores ranged from 32.3 to 58.7 out of a maximum score of 65. Eleven of the fifteen facilitators scored above 50/65 indicating that the sessions were conducted with a moderately high degree of fidelity by the majority of facilitators. Three facilitators scored below 50. On closer inspection 2 of these 3 facilitators scored low on interaction with learners, yet had acceptable scores on session fidelity and facilitation skills. The other facilitator had low scores on all 3 dimensions, but none of the scores was below 25.

Effect Evaluation
At baseline there were no significant differences between participants in the intervention and comparison arms of the study ( Table 3). The adjusted outcomes at first follow-up ( The adjusted outcomes at the 12-month follow-up (Table 5) showed no differences between intervention and control arms in 1-year incidence of self-reported sexual debut (12. Baseline rates of past 6-month IPV victimization were high, but these dropped in absolute terms, by over 10 % in the intervention arm and 4.6 % in the control arm. Participants in the intervention arm were more likely to report using contraception (other than condoms) than those in the control arm (28.1 vs. 22.1 %) but the difference was not statistically significant.

Effect of Rate of Session-Attendance on Outcomes
Since attendance was voluntary and attendance rates ranged substantially among those in the intervention arm [18], additional subgroup analyses were performed firstly among participants who attended at least 1 intervention session, and secondly among participants who attended at least 10 intervention sessions (Table 6). After 6 months, participants who attended at least 1 session reported better knowledge of condoms and HIV and more positive attitudes towards delaying sex in comparison to the control arm. Participants, who attended more than 10 sessions, reported better knowledge of condoms and HIV, more positive attitudes towards condom use and delaying sex, less IPV victimization, but also more sexual debut in comparison to the control arm. After 12 months, participants who attended at least 1 session reported better knowledge of condoms and HIV, and less IPV victimisation. Participants who attended more than 10 sessions reported better knowledge of condoms and HIV, and more positive attitudes towards condoms and delaying sex ( Table 6).
When all previously mentioned analyses were repeated using the expectation-maximisation algorithm to impute missing data, the results were similar: there were no additional significant effects or effects that ceased to exist in comparison to the complete case analyses (results not shown).

Discussion
This evaluation of the PREPARE after-school behavioural HIV prevention programme, which included a focus on IPV prevention, provided no evidence that it reduced sexual risk behaviour. Despite beneficial effects on knowledge about HIV prevention, our evaluation gave no indication that participants in the intervention arm of the PREPARE trial were less likely to have their sexual debut, were more likely to use condoms or had fewer sexual partners than those in the comparison arm. There was also no indication that students who attended a greater number of education sessions reported less sexual risk behaviour than those in the control condition. Nor was there any suggestion that the ''quality'' of sexual debut was superior among the intervention participants: there were no differences between arms in an indicator of a ''safe and good'' sexual debut, defined by use of condoms and contraception at first sex, absence of coercion at first sex, and absence of regret about first sex (data not shown).
These findings are consistent with a previous trial of an in-school HIV prevention intervention in the Western Cape, the SATZ trial, which failed to impact sexual risk  behaviour [5]. They are also consistent with the SATZ trial in Limpopo province [5]. However, they are not consistent with the findings of a meta-analysis of South African sexual risk reduction interventions among adolescents and youth, which found the interventions delayed sexual intercourse, increased condom use and reduced the number of sex partners [19]. The meta analysis suggested that features of interventions successful at delaying sexual intercourse and increasing condom use relative to the control condition were a focus on social norms, condom skills, gender inequalities and alcohol [19]. (All these factors were addressed in PREPARE).
One explanation for the failure of PREPARE in reducing sexual risk behavior is that, in the Western Cape setting, the contextual constraints on safe sexual behaviour might mitigate against any positive impact of the Adjusted effects were adjusted for clustering, age, gender, SES, stratification and the baseline measure; * p \ 0.05; ** p \ 0.01; *** p \ 0.001 a Only students who indicated to have had oral, anal or vaginal sex were included b Only students who had their sexual debut during the study period were included AIDS Behav (2016) 20:1821-1840 1833 PREPARE intervention. Apart from IPV and sexual violence, there are numerous social and environmental factors which undermine adolescent sexual and reproductive health, which were not addressed in the PREPARE programme, and which might be particularly relevant in the study setting. For example, the PREPARE programme did not include interventions to ensure adolescents had safe, supportive homes, secure livelihoods, and social protection, factors important for adolescent sexual and reproductive health [1,20]. However, these structural barriers to safe sexual behaviours apply to the other South African settings in which effective sexual risk reduction interventions have been demonstrated [19]. Another explanation for the absence of a beneficial effect on sexual behaviour is that adolescents might need to have a much greater intensity of exposure to the components of the PREPARE intervention, to achieve reductions in sexual risk behaviour. We have shown that exposure to the PREPARE after-school education sessions and school health service was sub-optimal. The findings of an HIV prevention trial in the Eastern Cape, South Africa suggest that greater intervention exposure, with a smaller   facilitator-to-adolescent ratio are important factors in reducing sexual risk. Young adolescents who received intense exposure an in-school HIV/STI risk reduction intervention comprising twelve 1-h modules in smallgroups, with very low participant-to-facilitator ratios (two facilitators to between nine and 16 participants), reported less unprotected sex over several subsequent years [21]. Further evidence in support of this explanation comes from a meta-analysis of the effects of South African sexual risk reduction interventions. It found that interventions were successful in delaying sexual intercourse and increasing condom use among youth when they used more facilitators to deliver the intervention and participants received a higher ''dose'' [19]. The most efficient way to ensure young adolescents have adequate exposure to HIV risk reduction interventions is to embed such interventions in the school curriculum. After-school programmes often do not achieve high levels of exposure because it is difficult to obtain high attendance rates and such programmes tend to selectively attract participants who are less vulnerable to adverse outcomes [18].
Our study suggests that behavioural HIV prevention programmes which include a focus on IPV prevention do indeed reduce self-reported intimate partner violence. In our evaluation, we found high baseline rates of IPV victimisation, a reduction in rates in both intervention and control arms over the course of the study but a significantly greater reduction in the intervention group. We observed an even greater impact on IPV victimisation among those with higher rates of education session attendance, compared with the control arm. The PREPARE trial is one of several adolescent IPV prevention programmes with evidence of beneficial effects on reported IPV [22,23]. However, to our knowledge PREPARE is one of only two that have been trialled among adolescents in sub-Saharan Africa. The other trial conducted in this region of the world was Stepping Stones, conducted out of school, among older adolescents than those in the PREPARE trial [8]. The PREPARE study is one of several trials with a combined IPV/ HIV prevention focus [8,[24][25][26], but only PREPARE and Stepping Stones [8] were conducted predominantly among adolescents. PREPARE is the only one of these combination IPV/HIV trials that failed to impact both IPV and HIV risk (measured either by self-reported sexual risk behaviour or STI/HIV infection).
Given that the participants of the PREPARE trial had sub-optimal exposure to the PREPARE intervention components, how did the programme achieve a beneficial effect on IPV? We speculate that the explanation is that participants were more likely to be exposed to the PREPARE IPV intervention components than components focussing on sexual risk behaviour. The after-school education sessions were ordered so that the sessions addressing gender inequities and gender power came before the sessions addressing condom use, delaying sex and number of partners. Attendance was higher at the earlier sessions [18], and consequently more participants received exposure to the intervention content related to IPV, than content related to sexual risk behaviour. Furthermore, the school-level intervention (the school safety programme), focussed on reducing IPV and sexual violence and did not have any direct focus on sexual risk reduction.

Limitations
Self-report measures are prone to bias and and biological measures provide the most convincing evidence of effects. The biological measure included in our trial (3 years incidence of conceptions among female participants) will only be available in 2016. Students who had signed parental consent and therefore participated in the study might not have been those most in need of interventions such as PREPARE. Therefore we might not have reached those who would benefit most from the programme. Our analyses of participants who attended more than one, or more than ten PREPARE education sessions compared with the control arm, are not based on an intention-to-treat analysis and are therefore potentially undermined by selection bias. We do not have the statistical power to perform genderstratified analyses for the primary outcomes.

Conclusions
Recent evidence shows adolescents are the only age group in which AIDS-related mortality is increasing, while in all other age groups it is decreasing [27]. Preventing new infections among adolescents and young people is a key goal in reaching the target of ending the AIDS epidemic by 2030. The PREPARE study aimed to prevent HIV among adolescents by reducing sexual risk behaviours and IPV. Despite adolescents' low participation rate in the PRE-PARE intervention, we observed a reduction in IPV, which suggests the intervention shaped intimate partnerships into more safe and appropriate ones. Thus, we have demonstrated the potential of interventions such as PREPARE to have a beneficial effect on one of the factors which strongly affect adolescents' risk of STIs and HIV. Reducing IPV is regarded as a critical goal for HIV prevention [28]. Women exposed to IPV have limited ability to assert their choices about sexuality [29] and are at a higher risk of incident HIV infection [28,30]. Reducing IPV is thus likely to lower risk for HIV/STIs in the longer term.
The PREPARE intervention did not lead to any reductions in sexual risk behaviours related to HIV prevention, despite that the intervention focussed on, and had a beneficial effect on one of the key structural barriers to HIV prevention, gender violence. Young adolescents probably need more intense, sustained exposure to interventions such as PREPARE to have an impact on sexual risk behaviour. With after-school programmes, it is difficult to achieve such exposure. However, we believe our findings probably also imply that reducing HIV risk among adolescents requires interventions which address a greater range of structural, social and environmental barriers to behaviours that prevent HIV infection.