Dolores Albarracin's meta-analysis

The largest ever meta-analysis (study of studies) of HIV-prevention interventions¹ reviewed the effects of 354 HIV-prevention interventions and compared their effect with the results for 99 control groups within 33 countries over a 17-year span (though three-quarters of studies were US-based).

Altogether, 104,054 people took part in the HIV-prevention programmes, while another 34,751 were included in control groups. (The number in control groups was smaller because many studies compared the effect with matched cohorts or historical controls).

About 45% of participants were men (allowing for a few studies in which gender was not identified); their average age was 26 and 34% were white, 47% Afro-American or African and 13% Latino. Only 36% of participants had completed high school.

Many studies did not record which risk groups participants belonged to, but 11% were designed specifically for gay men, 15% for injecting drug users, 17% for multiple-partner heterosexuals, 14% for recreational drug users and 8.5% for sex workers (programmes could target more than one risk group). Fifty-five per cent of participants reported multiple partners.

Of note, very few studies recorded the HIV status of participants, possibly because of a presumption that participants were negative, although in studies where serostatus was recorded, it was 20%.

Before the interventions, 64% of participants ‘never or almost never’ used condoms, 34% ‘sometimes’ used them and only 2% ‘always or almost always’ used them. The total proportion of acts of intercourse in which condoms were used was 32%.

The size of the meta-analysis allowed Albarracin and colleagues to calculate the effectiveness of prevention interventions for particular groups of people, both in terms of demographic characteristics like gender, age and race, and in terms of risk category.

It also enabled them to calculate the effectiveness of specific kinds of intervention well enough to provide a set of ‘decision trees’ at the end of the study to help prevention workers decide on the best kind of intervention for a specific group in future.

There was one important limitation to Albarracin’s survey: she only used condom use as her primary endpoint. She did not, therefore, include studies which had other aims, such as abstinence, sexual delay or reduction in the number of partners, nor did she look at the ultimate effect, HIV incidence.

However, the size of the study did allow her to calculate also the effectiveness of programmes on intervening effects between the intervention and the condom use. In other words, she did not just measure the effect programmes had on condom use – she was able to categorise the programmes into the kind of changes of knowledge, skills and motivation they attempted to teach.

This is important because it is an aid to theoretical rigour of design. An intervention may be based on one of the theories outlined below and produce a positive result; but without measuring how participants’ psychological attitudes have changed, it leaves open the possibility that the change in condom use is due to other factors, such as the introduction of treatment. Or it could find out that the intervention did indeed produce the desired psychological effect – but that this change had a negative effect. This was what seemed to happen with threat-inducing arguments.

Albarracin analysed interventions according to the following categories (many studies would use more than one method):

• ‘attitudinal’ – containing arguments designed to induce a positive attitude about using condoms (48% of programmes)
• ‘normative’ – containing arguments designed to increase social responsibility or increase perceived peer-group or societal pressure to use condoms (15% of programmes)
• ‘behavioural’ – containing verbal training or arguments designed to improve participants’ condom-using behaviours (20% of programmes)
• ‘behavioural skills’ – containing training helping participants to practise behavioural skills (22% of programmes)
• ‘threat’ – containing “persuasive arguments designed to increase perceptions of threat [of HIV infection or poor sexual health] among recipients” (47% of programmes)
• most programmes (94%) provided information about HIV
• 22% of programmes distributed condoms to intervention groups and 7% to control groups
• 18% of programmes administered an HIV test
• 49% included ‘active’ interventions, such as HIV counselling and testing and behavioural-skills training
• Two-thirds of interventions (where it was recorded) were delivered to groups, 20% to individuals and 8% to both
• 30% were delivered in clinics, 31% in schools, 21% in community venues such as the street, community centres or gay bars, and just 3% consisted of a mass communication.