A large meta-analysis of randomised controlled trials (RCTs) of behavioural interventions specifically aimed at reducing unprotected sex and sexually transmitted infection (STI) incidence in heterosexual African-Americans has found an overall significant effect in reducing unprotected sex and increasing condom use, and a near-significant reduction in STI incidence.
The overall efficacy across the 38 RCTs containing 14983 participants was a 25% reduction in ‘sexual risk behaviour’ as defined by the analysts, which was significant (95% confidence interval, 0.67 – 0.84), but became less so when participants were followed up more than six months after the end of the trial.
Trials using condom use as their outcome reported more efficacy (37% increase) than ones reporting unprotected sex (21% reduction).
The researchers commented that there was some evidence of ‘social desirability bias’ and publication bias in these reported outcomes; in other words negative findings were less common than a statistical analysis might have expected, due both to researchers being less likely to report negative results and participants being less likely to report negative behaviours.
The same bias was not evident in STI incidence considered as an outcome, however. The overall effect on reducing STI incidence, which was only reported by 10 trials containing 10,944 participants, was 12%, which was not quite statistically significant (95% confidence intervals 0.52-1.07). However when the trial identified as of ‘lowest methodological quality’, a small one containing 53 participants which had a negative outcome, was excluded, the reduction in STI incidence was 18%, which became statistically significant (95% confidence interval 0.69-0.98).
Furthermore the effect on STIs became more rather than less significant with longer follow-up times: in trials with follow-ups of more than 12 months there was a 23% reduction in STIs among participants.
The meta-analysis found that there was a clear ‘dose response’ in trials, with ones featuring more than one session and more than 160 total hours more effective than average. And it found trials with the following characteristics produced better-than-average results:
- Culturally tailored for African-Americans
- Delivered exclusively to African-Americans, rather than mixed
- Delivered by peers and containing an element of peer education
- Containing discussion of social norms
- Delivered in community rather than healthcare settings
- Containing the opportunity to learn and practise skills such as condom use and safer-sex negotiation.
Trials targeting women, young people and intravenous drug users were generally more effective than ones targeting the general adult heterosexual black population. The first trial studied was conducted in 1988 and the last in 2004. The largest contained 5758 participants, though it only had a three-month follow-up, and the smallest 38.
Sixteen trials measured condom use as their primary outcome, 20 unprotected sex and two STI re-infection. Very few trials split condom use or unprotected sex data into men and women or distinguished between casual and primary partners. Given that condom use is usually much higher with casual than regular partners, the authors comment, future trials should distinguish between these two groups.
Although a few trials were solely targeted at men, the authors comment that heterosexual men remain an under-targeted group, and none were aimed at prisoners. Of note also, trials directed specifically at people with HIV (three trials) were excluded from the analysis. Only trials where 80%-100% of the participants were African-American were included.
In terms of the 36 trials reporting sexual risk reduction, only two reported a negative effect (neither statistically significant). Of the 34 reporting positive effects, nine had statistically significant stand-alone results. One of the points of doing a meta-analysis is to increase the numbers analysed so that trials which in themselves do not report significant results end up contributing to a result with overall significance.
Given that only trials conducted in the USA were included in the analysis and with largely only black Americans as participants, some findings may not be generalisable to other countries and populations. For instance, some analyses of trials targeted at other groups have found that delivery in clinical rather than community settings works better, and by professionals rather than peers. It is therefore crucial to consider one’s target population when designing interventions. However the overall efficacies reported are broadly in line with previous meta-analyses of behavioural interventions in gay men, young people, white heterosexuals and intravenous drug users.
The authors comment that the challenge, as ever in behavioural interventions, is to translate this evidence-based research into robust packages that can be used in real-world settings and without the rigour of a PCT design to guide them. Several interventions for this target group, however, have been packaged or are in the process of being packaged for widespread dissemination within the USA – see the US CDC website for details.
The authors comment that future interventions for heterosexual African-Americans should include the characteristics listed above. “Future interventions…should take the unique needs of this community into account,” they conclude.
Darbes L et al. The efficacy of behavioral interventions in reducing HIV risk behaviors and incident sexually transmitted diseases in heterosexual African Americans. AIDS 22(10):1177-1194. 2008.