The promotion of safer sexual behaviour and the control of sexually transmitted infections (STI) are currently the most important methods of preventing HIV infection in developing countries. Three community randomized controlled trials have now assessed the effect of treatment for STI [1–3]. One of the trials, which was conducted in Masaka, Uganda, also assessed the effect of an information, education, and communication (IEC) behavioural intervention . Although the trial found some evidence of reported behaviour change in the intervention arms, it was not translated into a reduction in HIV incidence. This highlights the importance of using biological markers when assessing behavioural interventions .
The arguments for evaluating sexual behaviour interventions using randomized controlled trials have been well documented [5,6]. Few such trials have been published, however, and the Masaka trial is currently the only published community-based trial of a behavioural intervention for HIV control in Africa. The fact that it had no effect on HIV incidence is surprising, given that HIV is mainly transmitted through risky sexual behaviour. Possible explanations for not observing an effect include the suboptimal implementation of the intervention, perhaps as a result of insufficient coverage or contamination, or the intervention being inappropriate for the setting. Alternatively, there may have been shortcomings of the intervention package, which failed to encourage people to change their behaviour. It is possible that the intervention did result in behaviour change and a reduced risk of HIV incidence in some individuals, but this was not enough to observe an effect on HIV incidence in the communities as a whole. In this paper, we explore why the community-based behavioural intervention had no apparent effect on the incidence of HIV infection. The main objective was to assess the effect of attending IEC activities on an individual's risk of HIV acquisition.
Full details of the design of the trial have been published elsewhere [3,7]. The study was conducted in 18 rural parishes (hereafter referred to as communities) in Masaka district, south west Uganda. Each community comprised approximately 5000 adults (aged ≥ 13 years) residing in an average of 10 villages. Communities were randomly allocated to one of three arms to receive: (i) IEC activities alone (arm A); (ii) IEC activities with improved STI management (arm B); or (iii) routine government health facilities with general community development activities (arm C). Before randomization, the communities were matched to form six triplets by type of main road passing through, distance from the district capital and quality of the government health facility.
Information, education, and communication intervention
The IEC intervention was based on the ‘behavioural change for interventions model’ . This model focuses on knowledge acquisition, attitude development, motivational support and skills development. The aim of the intervention was to give knowledge and correct misconceptions about HIV/STI, and promote safer sexual behaviour and practices, through a community-oriented approach. Information was disseminated through large and small group meetings, village drama shows, village video shows, and group or one-to-one discussions with community educators . Each community had its own drama group, which performed specially written plays of approximately 30–60 min duration in each village in that community. These plays were performed and recorded on video, and one video was shown in each community every month. Information leaflets were distributed at each of the IEC activities. The social marketing of condoms and voluntary HIV counselling and testing services were implemented in all three arms.
A serological survey was conducted at three separate rounds (round 1 in 1994–1996, round 2 in 1997–1998, and round 3 in 1999–2000) on a subset of the target population residing in three to five villages closest to the health facility in each community. At each survey, a blood sample was taken and a questionnaire was administered to ascertain knowledge, attitudes, beliefs and practices (KABP). All adults in each household were randomly allocated to receive long (detailed) or short versions of the KABP questionnaire. Details of reported STI and condom use were included in both versions, whereas those for sexual behaviour and attendance at IEC activities were only in the long version. Data on attendance at IEC activities were collected only at round 2 and round 3. Data on STI and sexual behaviour were only collected from ‘sexually active’ adults (defined as all adults aged ≥ 25 years, or those aged < 25 years and either married or reporting sexual experience).
The sera of all adults were tested for HIV-1 infection using two independent enzyme immunoassays (EIA): Recombigen HIV-1/2 EIA (Trinity Biotech, Ireland) and Wellcozyme HIV-1 Recombinant EIA (Murex Biotech Ltd., UK) following set algorithms  and rigorous quality control procedures . The EIA results, when necessary, were confirmed by HIV-1 Western blot (Calypte Biomedical Corporation, USA). Samples with indeterminate results on Western blot were further tested for HIV-1 RNA (Cobas Amplicor version 1.5, Roche Diagnostics Corporation, USA).
In contrast with the main trial, all analyses in the present paper were performed at the individual level with adjustment for community (i.e. triplet and arm) in all models. We adjusted for community as a fixed rather than random effect because of the small number of communities (12) and the large number of individuals (> 100) within each community, although random effects models yielded similar results. Analysis is restricted to the arm that received the behavioural intervention alone, without improved STI management (arm A, hereafter referred to as ‘intervention') and the comparison arm (arm C). We excluded arm B because the effect of the behavioural intervention may be different in the presence of the STI intervention. Men and women were analysed separately. Sexual behaviour was measured using self-reported condom use, and the number of all sexual partners and casual sexual partners in the past year. Casual sexual partners were those with whom a person had ‘had sex with on only one or two occasions, for example at a party or disco'. Attendance at IEC activities was measured using self-reported attendance at meetings, videos, and dramas in the past year, or meeting with the community educator in the past year.
For the descriptive analysis, we estimated the proportion of individuals who reported attending the different types of IEC activities, and used logistic regression to identify factors associated with attendance. For the main objective, we used rate ratios (RR) to measure the effect of attending IEC activities in the past year on HIV incidence. HIV incidence was determined on all adults who participated in at least two serological surveys, and who were seronegative on at least the first occasion. Incident events and person-years at risk of seroconversion were combined over all rounds (rounds 1 to 2, rounds 2 to 3, and rounds 1 to 3 for those not contributing data at round 2) to give an overall incidence rate. RR were estimated using Poisson regression with adjustment for triplet, arm, age, and marital status as a priori confounders, and adjustment for religion if it was statistically significant (P < 0.10) after adjusting for all other variables in the model. Finally, in those individuals who reported being sexually active at two consecutive rounds, we used logistic regression to compare behaviour changes in those who reported attendance at IEC activities in the current round and those who did not. All data were double entered using Foxpro for Windows (Microsoft Corporation, USA) and EpiInfo version 6.04 and all analysis was performed using Stata version 7 (Stata Corporation, College Station, USA).
Demographic characteristics and reported attendance at information, education, and communication activities at rounds 2 and 3
Table 1 shows the demographic characteristics and reported attendance at IEC activities for individuals in intervention and comparison communities who answered a long KABP questionnaire. The analysis was restricted to individuals in the HIV-negative cohort who reported ever having had sex. Overall, 81% of individuals (2630/3244) in the intervention communities and 9% (288/3382) in the comparison communities reported attending at least one IEC activity in the past year. The mean number of activities attended in the past year was higher in men (2.59 at round 2 and 2.61 at round 3) than in women (2.03 at round 2 and 2.19 at round 3). The proportion of individuals in the comparison communities who reported attending any IEC activities ranged from 6 to 12%, depending on round and sex.
Factors associated with reported attendance at information, education, and communication activities at rounds 2 and 3
We examined the demographic and sexual behaviour characteristics associated with reported attendance at any IEC activity for individuals in the intervention communities (Table 2). Attendance was significantly lower in men and women aged 55 years or over compared with those aged 13–19 years, and in widowed men compared with never-married men. Associations between attendance and sexual behaviour were generally not statistically significant. In the comparison communities, the only factor significantly associated with IEC attendance was community (data not shown).
Association between attendance at information, education, and communication activities and HIV incidence
HIV incidence rates were lower in those who reported attending IEC activities in the past year compared with those who did not, and in women, most of these effects were statistically significant (Table 3). The robustness of the effect of attending any IEC activity in women [adjusted RR 0.41, 95% confidence interval (CI) 0.19–0.89, P = 0.024] was explored. First, we examined whether this effect was biased by non-attendance at IEC activities as a result of being away, which in turn may be associated with a higher HIV incidence. The effect of attending any IEC activity was almost identical when adjusted additionally for the number of times surveyed and for travel outside Masaka district in the past year. However, the effect of attending any IEC activity was diluted when those who reported not being sexually active were included (RR adjusted for triplet, arm and age group was 0.53, 95% CI 0.24–1.14, P = 0.103). There were too few incident events in those attending more than one activity to assess whether incidence reduced linearly with increasing frequency of attendance.
Association between attendance at information, education, and communication activities and behaviour change
The majority of men and women reported less than two sexual partners in the past year at two consecutive rounds, and this was not significantly associated with attendance at IEC activities (Table 4). Forty-five per cent of men reported no casual partners in the past year at consecutive rounds and 32% reported a reduction in casual partners, but this was irrespective of whether or not they attended IEC activities. The majority of women reported no casual partners at consecutive rounds, and there was no association between attending activities and a change in casual partners. Reported condom use remained low at consecutive rounds in men and women, and reported usage was not significantly associated with attendance at IEC activities.
Our findings suggest that the majority of individuals in the intervention communities reported attending at least one of the IEC activities. The fact that 9% of individuals in the comparison communities also reported attending some activities may have diluted the true intervention effect in the main trial. When our analysis was based on reported attendance at IEC activities rather than the intervention arm, we observed a lower HIV incidence in those who reported attending at least one intervention activity compared with those who attended none, and in women this effect was statistically significant. However, reported behaviour change was not associated with attending intervention activities.
Individuals in the intervention communities reported attending an average of at least two activities in the past year, which is consistent with process evaluation data for the trial . The lower attendance in women and in older age groups may reflect lower perceived risks of HIV infection in these groups . In addition, different channels of communication are likely to appeal to different sectors of the community. A multichannel approach may be necessary to reach all the different sectors of the community .
Reported attendance by individuals residing in the comparison communities supports anecdotal evidence from interviews with fieldworkers , although we cannot rule out misreporting. At round 2, attendance in the comparison communities was mostly by people living in a community situated relatively close to an intervention community. Future community randomized trials should employ measures to prevent contamination, for example, by means of geographically widely separated communities. In settings in which some contamination is inevitable, it should be monitored and a possible dilution effect should be allowed for in sample size calculations.
Possible explanations for observing a reduced HIV incidence in women who reported attending IEC activities include chance, confounding, bias and causation. The fact that the reduction in HIV incidence was strong and statistically significant for three out of four IEC activities makes chance an unlikely explanation. We have adjusted for the confounding effects of community, age group, marital status, religion, number of times surveyed and travel. When we also adjusted for sexual behaviour, the results were similar. A major strength of our study is that the primary outcome measure for assessing the behavioural intervention was biological, and misclassification bias for this will be minimal. The consistency of the data on reported attendance at IEC activities with our process data suggests that reporting bias was not a major problem for these data. Moreover, any reporting bias is unlikely to be differential with respect to HIV status, because most individuals were unaware of their HIV status.
A potential limitation of the study arises from measuring attendance at IEC activities and HIV incidence in concurrent rather than consecutive periods, so seroconversion could have occurred before, rather than after, attending IEC activities. When we addressed this by restricting the analysis to reported attendance at round 2 and seroconversion between rounds 2 and 3, we observed similar effects to those shown in Table 3, but wider confidence intervals because of the smaller sample size. Therefore, reverse causation is unlikely.
Despite observing a large reduction in HIV incidence in women who attended IEC activities, we observed no behavioural mechanism for this effect. We suggest three reasons why this may have occurred. First, our study, like other studies of reported sexual behaviour [15–17], is prone to reporting bias. This may arise through poor recall, misunderstanding the questions (such as the term ‘casual partners'), or not wanting to disclose such personal information to an interviewer. For example, individuals may have increased their use of condoms with casual sexual partners, but are reluctant to report having any casual partners or using condoms, because both are associated with promiscuity in this population. Reporting bias was demonstrated by the fact that 4% of men and 2% of women reported having used a condom at one round and never having used one at the following round. Low rates of reported condom use meant that data on condom use with particular types of sexual partners were difficult to interpret. For example, condom use with last casual partner refers only to those who reported casual partners and having ever used a condom.
Second, if our data on reported sexual behaviour are accurate, then only a relatively small number of individuals are engaged in high-risk behaviour (Table 4). Under this assumption, the study would not have enough power to detect changes in behaviour in such small numbers. Whereas we believe that many high-risk men and women in our study will be misclassified as low risk, it is plausible that a substantial proportion will be truly low risk. Qualitative studies in this area found that a substantial proportion of individuals did not reduce their number of sexual partners since becoming aware of AIDS because they already had few sexual partners (R. Pool, A. Kamali, J.A.G. Whitworth, 2004, in preparation). Furthermore, in a mature HIV epidemic, reporting only one sexual partner may not be a marker of low risk if this partner is HIV infected and has a high viral load . Third, it is possible that attending IEC activities was associated with behaviour change, but our KABP questionnaire did not capture the variables that changed. Increased alcohol intake  and coercive sex  have been associated with HIV infection in Uganda, but such data were not collected in our trial. Our measures of sexual behaviour are crude, with inevitable limitations and may not accurately reflect HIV risk.
Why was the reduction in HIV incidence in women who attended IEC activities not translated into an effect in our main trial ? The present results are based on reported attendance (`per protocol') rather than allocated intervention arm (`intention to treat'), and thus take account of non-attendance in arm A (up to 26%) and attendance in arm C (up to 12%). The present results are restricted to those individuals who were allocated the long KABP questionnaire and reported being sexually active, whereas the main trial included all individuals aged 13 years or more, even the 28% who did not report being sexually active. Individuals were randomly allocated to long or short KABP questionnaires, so this is unlikely to bias our results, but restricting to sexually active individuals did result in stronger and more significant RR. The community level analysis of the main trial measured effects on both HIV transmission and acquisition in the community, whereas the present analysis measures HIV acquisition at the individual level. In summary, we believe that the intervention resulted in a reduced risk of HIV acquisition in the women, and to a lesser extent the men, who attended it, but that this effect was not large enough to impact on HIV transmission and acquisition in the community as whole.
What are the lessons learned from this study? First, data on reported coverage are useful for identifying subgroups not reached by behavioural interventions. Second, measures to prevent contamination in future intervention studies should be implemented, and contamination should be monitored and reported. Sexual behaviour data are prone to reporting bias, and therefore behavioural interventions need to be evaluated using biological markers. Randomized controlled trials are important for evaluating behavioural interventions, but may not be sufficient to explain the mechanisms underlying why the interventions do or do not succeed. Implementing and evaluating behavioural interventions remain priorities for controlling the HIV pandemic. Future research on behavioural interventions needs to identify what characteristics of sexual behaviour matter and how valid measures of these characteristics may be collected.
This study was supported by the Medical Research Council and the UK Department for International Development with partial funding from initially the Global Programme on AIDS and later United Nations Programme on AIDS. Maria Quigley is supported by the Medical Research Council. We are grateful to Linda Morison for her comments on this paper. We would like to thank the study population, and the survey, intervention and analytical teams.
Sponsorship: This study was supported by the Medical Research Council and the UK Department for International Development with partial funding from initially the Global Programme on AIDS and later United Nations Programme on AIDS.
1. Grosskurth H, Mosha F, Todd J, Mwijarubi E, Klokke A, Senkoro K, et al
. Impact of improved treatment of sexually transmitted diseases on HIV-1 infection in rural Tanzania: randomised controlled trial.Lancet
2. Wawer MJ, Sewankambo N, Serwadda D, Quinn TC, Paxton LA, Kiwanuka N, et al
. Control of sexually transmitted diseases for AIDS prevention in Uganda: a randomised community trial.Lancet
3. Kamali A, Quigley M, Nakiyingi J, Kinsman J, Kengeya-Kayondo J, Gopal R, et al
. Syndromic management of sexually transmitted infections and behaviour change interventions on transmission of HIV-1 in rural Uganda: a community randomised trial.Lancet
4. Cowan F, Langhaug LF, Mashungupa GP, Nyamurera T, Hargrove J, Jaffar S, et al
., for the Regai Dzive Shiri Project. School based HIV prevention in Zimbabwe: feasibility and acceptability of evaluation trials using biological outcomes.AIDS
5. Oakley A, Fullerton D, Holland J, Arnold S, France-Dawson M, Kelley P, McGrellis S. Sexual health education interventions for young people: a methodological review.BMJ
6. Stephenson J, Imrie J. Why do we need randomised controlled trials to assess behavioural interventions?BMJ
7. Kamali A, Kinsman J, Nalweyiso N, Mitchell K, Kanyesigye E, Kengeya-Kayondo JF, et al
. A community randomised controlled trial to investigate impact of improved STD management and behavioural interventions in rural Masaka, Uganda: trial design, methods and baseline findings.Trop Med Int Health
8. King AJC, Wright NP. AIDS and youth. An analysis of factors inhibiting and facilitating the design of interventions
. Review commissioned by the Global Programme on AIDS, World Health Organisation, January 1993.
9. Kinsman J, Kamali A, Kanyesigye E, Kamulegeya I, Basajja V, Nakiyingi J, et al
. Quantitative process evaluation of a community-based HIV/AIDS behavioural intervention in rural Uganda.Health Educ Res
10. Nunn AJ, Biryahwaho B, Downing RG, Van Der Groen G, Ojwiya A, Mulder DW. Algorithm for detecting antibodies to HIV-1: results from a rural Ugandan cohort.AIDS
11. Nunn AJ, Biryahwaho B, Downing RG, Ojwiya A, Mulder DW. Computer-assisted quality assurance in an HIV serology laboratory.Meth Inform Med
12. Kengeya-Kayondo JF, Carpenter LM, Kintu PM, Nabaitu J, Pool R, Whitworth JAG. Risk perception and HIV-1 prevalence in 15 000 adults in rural south-west Uganda.AIDS
13. Mitchell K, Nakamanya S, Kamali A, Whitworth JAG. Community-based HIV/AIDS education in rural Uganda: which channel is most effective?Health Educ Res
14. Mitchell K, Nakamanya S, Kamali A, Whitworth JAG. Exploring the community response to a randomised controlled HIV/AIDS intervention trial in rural Uganda.AIDS Educ Prevent
15. Dare OO, Cleland JG. Reliability and validity of survey data on sexual behaviour.Health Trans Rev
16. Lagarde E, Auvert B, Chege J, Sukwa T, Glynn JR, Weiss HA, et al
. Condom use and its association with HIV/sexually transmitted diseases in four urban communities of sub-Saharan Africa.AIDS
2001; 15 (Suppl. 4)
17. Buve A, Lagarde E, Carael M, Rutenberg N, Ferry B, Glynn JR, et al
. Interpreting sexual behaviour data: validity issues in the multicentre study on factors determining the differential spread of HIV in four African cities.AIDS
2001; 15 (Suppl. 4)
18. Quinn TC, Wawer MJ, Sewankambo N, Serwadda D, Li C, Wabwire-Mangen F, et al
. Viral load and sexual transmission of human immunodeficiency virus type 1. Rakai Project Study Group.N Engl J Med
19. Mbulaiteye SM, Ruberantwari A, Nakiyingi J, Carpenter LM, Kamali A, Whitworth JAG. Alcohol and HIV: a study among sexually active adults in rural southwest Uganda.Int J Epidemiol
20. Quigley MA, Morgan D, Malamba SS, Mayanja B, Okongo MJ, Carpenter LM, Whitworth JAG. A case–control study of risk factors for incident HIV infection in a rural, Ugandan population.J Acquir Immune Defic Syndr Hum Retrovirol