Reviewing the evidence on effectiveness and cost-effectiveness of HIV prevention strategies in Thailand

Interventions included in this review were those that showed evidence of reducing HIV incidence or risk behaviours likely to effect horizontal and vertical HIV transmission. The set of interventions was not restricted to those used in practice in Thailand or funded by the Thai government. It also covered interventions provided at all levels, i.e. individuals, groups, and communities which are likely to be beneficial in the reduction of the HIV/AIDS epidemic worldwide.

Given that a wide range of interventions were included in this study, it is imperative that they have clear definitions and detailed information to ensure a better understanding of, for example, their precise aims, what their delivery modes are, and at whom the interventions are targeted. A lack of clarity and descriptive detail of interventions makes it difficult to assess and compare either the effectiveness or cost-effectiveness in different settings. It is also impossible to make sensible recommendations with regards to policy decision making if there are no concise definitions for commonly implemented intervention approaches.

It was necessary, therefore, for this study to establish or adopt a standard structure to define and classify interventions for the prevention of HIV/AIDS. Based on the recommendations made by UNAIDS [10], HIV prevention interventions are grouped into five broad categories as follows:

However, as the fifth category was not related to HIV prevention, it was not included in this review. From the above recommendations, we provide a definition and classification of each HIV prevention intervention in additional file 1: Table S1.

We obtained literature published in Thai or English from 1997 to 2008. The studies conducted within a Thai setting were given a higher priority since they better recognise the resource and infrastructure limitations that are specific to the health care system in Thailand, as well as the effectiveness of interventions which are affected by context specific factors. The review of the Thai literature, therefore, included both published and unpublished (grey) literature such as research reports, Master's dissertations or Ph.D. theses, which are considered highly relevant to the Thai context. If local data regarding effectiveness or cost-effectiveness of an intervention were not available, then a systematic search of Pubmed and Cochrane databases was carried out. A broad search strategy with appropriate MeSH terms and keywords was used in combination with a filter to restrict type of studies to systematic reviews, meta-analysis, randomized controlled trial, case control studies or cohort studies. Search terms used included 'acquired immunodeficiency syndrome', 'HIV', 'prevention and control', 'primary prevention', 'intervention studies', or 'early intervention'. Exclusion criteria were: i) publications of the same study, ii) descriptive studies, iii) assessment of satisfaction, knowledge and attitude towards HIV/AIDS, risk behaviours and program activities (not outcomes), iv) reports of case studies, and v) cost analyses.

The effectiveness of interventions can be measured in a number of ways. Figure 1 (adapted from [11]) shows the concept of outcome hierarchies that emphasize the difference between 'immediate', 'intermediate' and 'final' outcomes of HIV interventions. Immediate measures of effectiveness of HIV interventions are characterised by change in knowledge, attitude, perception and skills of the individuals. In many HIV programs, the changes were reported in terms of trust, caution and received assurances. Further along the continuum, these immediate changes can subsequently affect health behaviours and health determinants, for example, condom use, abstinence, sharing injection equipments, or fewer sexual partners. Finally, changes in incidence or morbidity or mortality should be evaluated as the final or ultimate goal of the program. The primary criterion for selection of studies was that they report the effectiveness in terms of the changes in HIV risk behaviour and HIV incidence. We deliberately excluded descriptive or qualitative reports from the review as these most often do not provide comparable measures of outcome. Studies reporting outcomes in terms of improvement in attitude or knowledge were also excluded because changes in these are not always directly associated with behavior change [12].

The review included economic evaluation studies that presented their results in terms of cost per HIV infection averted, or cost per quality-adjusted life year (QALY) gained, or cost per disability-adjusted life year (DALY) averted. This review compared the value for money of different HIV/AIDS preventive interventions using a societal viewpoint since it is most relevant to priority setting in health care. However, if the evaluation took the more limited health care provider perspective, this was used instead.

Differences in monetary currencies and unit costs between locations and over time are among the most commonly cited obstacles to applying economic evaluation findings across different settings. This study adjusted all cost-effectiveness ratios in a common currency, the international dollar, at their 2008 value, using national consumer price index (CPI) and purchasing power parity (PPP) information from the International Monetary Fund (IMF) [13].

With regard to the thresholds for considering an intervention to be cost-effective, the Subcommittee for development of the National List of Essential Medicines in Thailand developed its own threshold as a criterion to include or exclude medicines from the reimbursement list [14]. Interventions with a cost per QALY gained below 100,000 THB (approximating the Thai GDP per capita) are considered favourably for inclusion. Given the fact that 11.23 QALYs could be saved from one HIV infection avoided [15], the thresholds for one HIV infection averted was approximately 90,000 PPP$.

Systematic reviews and meta-analyses of high quality RCTs were considered the most favourable data sources [2]. The advantages of using systematic reviews of clinical effects are twofold. First, a more precise estimate can be attained from combining the outcome data from a number of studies. Second, by using the results from studies carried out in a range of settings, assuming that these studies are sufficiently homogenous to be comparable, the estimate can then be applied to a more general patient population with different baseline risks, rather than specifically for a population group selected for an individual trial. In cases where a meta-analysis of RCT(s) was not available alternative evidence sources were used in accordance with the hierarchy shown in table 1.

Searching the Thai databases initially identified a total of 932 abstracts (see figure 2). Of these, 890 abstracts were excluded based on our exclusion criteria. From the full text review of the remaining 42 papers, only fourteen papers were found to be relevant and included in the analysis. Of the 28 papers excluded, 25 papers reported only immediate outcomes of the HIV prevention programs. For example, two papers that reported the effectiveness of the distribution of condom vending machines in the communities, used only numbers of condoms sold per machine and/or customer's satisfaction as their outcome measures [16, 17]. Three other papers that evaluated drug regimens for the prevention of vertical HIV transmission, were excluded because the regimen under investigation(AZT regimens), is no longer used in clinical practice in Thailand [18‐20]. The remaining 14 studies [21‐34] were included in the analysis.

We identified 1,394 abstracts from the international literature (see figure 3). After reading the abstracts, 1,203 studies were eliminated because they were editorials, descriptive, or qualitative reports. In addition, we also excluded a number of studies that assessed the effectiveness and cost-effectiveness of programs for the prevention of mother-to-child HIV transmission because relevant Thai studies had already been identified. The full text of the remaining 191 studies was reviewed and 66 studies [35‐100] were considered relevant and included in the analysis in the final stage.

Additional file 2: Table S2 summarizes the effectiveness and cost-effectiveness of each HIV prevention intervention based on the reviews of domestic and international studies. It was not surprising that a much larger proportion of effectiveness and cost-effectiveness studies were conducted in international settings, mainly the US and Sub-Saharan Africa. In both the Thai and the international settings there were more effectiveness studies than cost-effectiveness ones (11 and 48 effectiveness studies in the Thai and international settings, respectively, versus 3 and 18 cost-effectiveness studies in each of the respective settings).

Most of the assessments focused on interventions affecting knowledge, attitudes and beliefs (36/63 or 57%), followed by biological/biomedical interventions (16/63 or 25%), harm reduction interventions (9/63 or 15%) and, lastly, mitigation of barriers to prevention and negative social outcomes of HIV infection (2/63 or 3%).

The findings demonstrated that male condoms use, street outreach programs, programs for the prevention of mother-to-child HIV transmission, circumcision, and needle and syringe programs were the only interventions to show strong evidence of reducing HIV infection among target populations.

Figure 4 compares the cost per HIV infection averted of each HIV prevention intervention. It can be seen that the cost-effectiveness ratios vary largely, ranging from 70 PPP$ per HIV infection averted for screening blood product to 2,000,000 PPP$ per HIV infection averted for community-based education for low income women. It is likely that biological/biomedical interventions (highlighted in blue) are more cost-effective than those interventions affecting knowledge, attitudes and beliefs (highlighted in pink).

Figure 5 presents the modification of the figure 4 according to levels of HIV prevalence in settings where economic evaluations were conducted. It indicates that interventions performing in high HIV prevalence settings (e.g. screening blood products and donated organ for HIV or school-based sex education program in settings with high HIV epidemic) are likely to be cost-effective whereas interventions performing in low HIV prevalence settings (e.g. school-based sex education program in settings with low HIV prevalence or community-based education for low income women) are less likely to be cost-effective.

Figure 6 summarises the findings from the review and prioritises HIV prevention interventions based on effectiveness and cost-effectiveness evidence. The table presents results by targeted population including female sex workers, IDUs, men who have sex with men and serodiscordant couples, who are currently the major sources of HIV infection in Thailand.

Those interventions proven to be both effective and cost-effective for female sex workers were: voluntary HIV counselling and testing, peer education, and male and female condom use. Community-based education and workplace-based education proved to be effective, but no evidence regarding the value for money among female sex workers was found. Although the use of microbicides is still in the trial phase, the evidence so far suggests that microbicides and sexually transmitted infections (STI) control were not effective in preventing HIV transmission amongst female sex workers.

Condom use was proven to be the only effective and cost-effective intervention for men who have sex with men while voluntary HIV counselling and testing demonstrated effectiveness but lacked cost-effectiveness information. Community-based education was clinically effective but cost-ineffective. Peer education and STI control were shown to be ineffective amongst this population.

For injecting drug users, voluntary HIV counselling and testing, condom use, needle and syringe programs, and street outreach were amongst the programs shown to be both effective and cost-effective. Needle social marketing, peer education, and substitution treatment demonstrated clinical effectiveness but were unsupported by economic evidence. Community-based education, HIV vaccines, STI control and post-exposure prophylaxis were shown to be ineffective in the prevention of HIV transmission amongst IDUs.

Condom use was the only interventions proven to be both effective and cost-effective for serodiscordant couples. Voluntary HIV counselling and testing was proven to be clinically effective but no cost-effectiveness information was available. STI control was not effective in preventing HIV transmission amongst this population.

With respect to information availability, voluntary HIV counselling and testing and condom use were the only interventions with extensive evaluations concerning effectiveness and cost-effectiveness across the different population groups. The main identified information gaps were for the following interventions: routine (provider-initiated) voluntary HIV screening at healthcare settings, introduction of female condoms, HIV vaccine, male circumcision, microbicides, and post-exposure prophylaxis. The evidence was particularly scarce for some of the targeted populations, namely serodiscordant couples, prison inmates and health care workers.