Reducing violence and increasing condom use in the intimate partnerships of female sex workers: study protocol for Samvedana Plus, a cluster randomised controlled trial in Karnataka state, south India

Female sex workers (FSWs) have a disproportionately high HIV burden compared to women in the general population [1]. In 110 countries with available data, the prevalence of HIV infection is almost 12 times higher among sex workers than for the population as a whole, with prevalence at least 50-fold higher in four countries [2]. This heightened risk of HIV acquisition and transmission among FSWs varies by sex work typology [3‐6] and the nature of the underlying HIV epidemic [1]. Increased risk is due to a variety of biological (e.g. concurrent STI infection), behavioural (e.g. partner volume, partner concurrency, injecting drug use), biomedical (e.g. hormonal contraception) and structural factors (e.g. structural organisation of sex work, poverty, discrimination, housing and financial insecurity, legal and criminalisation policies, and violence) [7‐10].

A growing body of evidence suggests that violence exposure is strongly associated with HIV infection, both among women in the general population [11‐19], as well as among female sex workers [20‐24]. However, the mechanisms through which violence operates to enhance vulnerability are multiple and complex, with HIV transmission risk occurring directly during coerced sex, as well as indirectly: FSWs who experience violence are less likely to visit STI/HIV clinics or use condoms, and more likely to experience condom breakage, report anal sex and have a concurrent STI infection, compared to FSWs who don’t experience violence [25‐35]. They are also more likely to report mental health morbidity, including alcohol and drug use, which in turn can reduce their ability or will to negotiate condoms with their sexual partners [36‐41]. In many settings, sex work is either illegal or the legal status is ambiguous, meaning violence by police, clients, intimate partners (IPs) and others is likely to go unreported, and the perpetrators can continue unchecked [5, 8, 42‐45]. Men who are violent are more likely to report a clustering of ‘risk’ behaviours, including multiple concurrent partnerships, low/no condom use, anal sex, and substance use. These behaviours place the men and their sexual partners at increased risk of STI and HIV [14, 46‐52]. In the context of female sex work, women can experience violence from a variety of perpetrators, both in their working lives (from the police, clients, pimps and street criminals) [3, 25, 45, 53‐55], as well as in their ‘domestic’ lives from their non-paying, IPs (husband/lover) [35, 39, 56].

In some settings, domestic violence can be as important as ‘work-place’ violence in enhancing FSWs vulnerability to HIV and diminishing the quality of their lives [36, 57]. Domestic violence can inhibit women from negotiating condom use [15, 39, 40]. Within FSWs’ intimate partnerships, women’s ability to negotiate condom use is often severely compromised due to norms around female obedience and fidelity, men’s ignorance of her sex work status, men’s expectations that their partner will abandon sex work and are therefore no longer a source of risk, and women’s emotional and sometimes financial dependence on their partners [57‐59]. All of these challenges are exacerbated if a male partner is violent.

In terms of addressing violence, among the general population, there have been several examples of successful interventions in low and middle income countries, which have worked at the community level to successfully reduce violence within intimate partnerships [60‐64]. Among FSW populations, evidence suggests that comprehensive HIV prevention programming, which includes community mobilisation, empowerment and violence prevention components, can successfully reduce violence by ‘non-partners’ such as clients and the police [24, 29, 65]. However, few programmes have attempted to address violence perpetrated against FSWs by their IPs. In this paper we present the protocol for a cluster randomised control trial in northern Karnataka state in South India, to evaluate a multi-level intervention called Samvedana Plus which works with FSWs, IPs, sex worker community-based organisations (CBOs), and the wider community (general population) where FSWs and their IPs live, to reduce violence and increase consistent condom use within the intimate relationships of FSWs.

The cluster randomised design will reduce problems associated with confounding factors, and there is sufficient geographic dispersion of the villages to minimize spill-over effects. While randomization is conducted to ensure that the arms of cluster RCTs are not systematically imbalanced, it is possible that imbalances will arise due to chance. We expect that the arms of this trial will be suitably balanced for the following reasons. First, this trial has a relatively large number of clusters and individual respondents. Second, the villages included in the design have similar characteristics in terms of sex work and relationships between sex workers and their IPs, as well as the norms that govern such relationships. Third, we stratified the randomization by the number of FSWs with IPs in the village, and then by village size. The similarity of the intervention and control samples will be assessed at baseline, and any differences adjusted for in the final analysis.

We aim to keep selection bias to a minimum through the randomization of the study villages and the selection of all FSWs with a current or recent IP living in the study villages. The effects that are observed, and any variability, will be explored using data on the fidelity and intensity of delivery in different sites, and the data on the context of intervention delivery. Selection bias will be largely addressed as a result of the careful process of randomization and selection of subjects. Selective loss to follow-up could also lead to bias, for example, if women who are experiencing IPV are more likely to drop out of the study. We plan to minimize this effect in the intervention villages, through the longstanding relationship that the implementation teams have established with sex workers and the trust created through the informed consent process. In addition, we expect some loss to follow-up of participants, for example due to migration for sex work or other reasons. We have allowed for 10 % loss to follow-up in our sample size calculations. We plan to reach women who have migrated away by networking with the CBO in the area to which they have migrated, if it is known. It will not be possible to mask the interviewers to the study arms; interviewer bias will be addressed by rigorous training of the field investigators. The study may also be subject to reporting bias, especially with an issue as sensitive as IPV and condom use within intimate partnerships. We will be assessing the main study outcomes through face-to-face interviews, primarily as most FSWs in this region are illiterate, and could not self-complete a questionnaire. We plan to minimise reporting bias by drawing upon the long-standing experience the research team has in conducting research on violence. The violence exposure questions used in the questionnaire have been validated [16] and will be preceded by an introduction, framing domestic violence as something that many people experience in their lives, reassuring participants on confidentiality, and explaining the importance to the study of answering questions honestly. Previous experience suggests that recruiting IPs for the study may be a challenge, and we recognize that it may not be feasible to reach a sample size large enough to measure statistically significant changes among IPs. To increase the response rate among IPs, men will be recruited from a line-list of IPs, and initially contacted by peer educators from the CBO, rather than directly by the research team. In any case, changes in condom use and violence as reported by IPs are not considered as primary outcomes, but will still provide useful information. Finally, due to the implementation of the Avahan Intervention programme in the district since 2004, which included a violence prevention component aimed at addressing violence by clients, police and others [29, 67], the control villages are not pure controls; this may mean the level of effect is smaller than might otherwise have been the case.

In conclusion, despite increasing efforts to reduce IPV among women in the general population, and to address ‘workplace’ violence by clients, police, pimps and others among FSWs, this is the first RCT of which we are aware to evaluate the impact of an intervention which aims to reduce domestic violence and increase condom use in the context of intimate partnerships of FSWs. Addressing the violence experience of FSWs will be key in ensuring that HIV prevention efforts are successful. This trial will provide important findings that can be used to inform programme design and policy, both in the HIV and violence prevention fields. The trial will also provide evidence on the feasibility of the approach of addressing IPV among FSWs in this setting.