In May 2015, the Thailand MOPH in collaboration with the CDC launched the Test, Treat, and Prevent HIV Programme to identify barriers to the immediate initiation of ART and the use of PrEP. Participants were enrolled and followed from 1 May 2015 to 30 April 2018 including 1880 Thai MSM and TG from the Khon Kaen, Lerdsin, Srinagarind University, Thammasat University, and Udon Thani Hospitals across four provinces (Bangkok, Khon Kaen, Pathum Thani and Udon Thani). PrEP was also offered at Lerdsin and Thammasat Hospitals 15.

We conducted a retrospective cost analysis of the Test, Treat and Prevent HIV Programme from 1 June 2015 to 31 May 2016 to derive the cost associated with HIV testing and PrEP services from a provider perspective. Costs of HIV testing and PrEP services were collected by programmatic activity (HIV testing and counselling, PrEP initiation and PrEP visit (adherence counselling, physical check, clinic visit, prescribing drugs and blood draw)), by input type (personnel, drugs and commodities, supplies, test kits, laboratory testing including creatinine, hepatitis B, syphilis screening, biochemistry and haematology), and by source of support (cohort study or MOPH). We used a time and motion analysis to measure the average time health and lab staff spent on each programmatic activity session. To calculate personnel costs, we collected information from facility and project staff on their salaries and allowances, and assumed staff worked eight hours per day for 20 days each month. We based the cost of supplies and equipment associated with laboratory testing on price per test using the quotes from the suppliers of lab reagents multiplied by the number of tests conducted during the study. The price paid per test covers the cost of the laboratory equipment, services, reagents and consumables for the term of the agreement excluding capital expenditures. The cost of drugs and commodities was determined by multiplying the MOPH unit price with the quantities used. The study excluded costs associated with HIV‐related morbidity and mortality, as well as morbidities and mortalities stemming from adverse events associated with care and treatment, costs borne by the health system, higher‐level overhead costs, and building maintenance and utility costs. Data on beneficiary volume were extracted from the outcomes of the cohort study to derive the unit cost per‐person‐per‐year (ppy) by dividing the total cost of the programme by the number of PrEP participants. Data collection was conducted from October to December 2016. All costs were collected in THB and converted to U.S. dollars (USD) at the market exchange rate (1 USD = 35.42174 THB) for the period the cost was incurred.

To estimate the epidemiological impact of providing PrEP to MSM, we used the Optima HIV tool (v2.6.6, available at http://hiv.optimamodel.com), which is a compartmental model of HIV transmission and disease progression linked to a programmatic response module for estimating the epidemiological and economic impact of interventions 16, 17. Of the approximately 571,000 MSM in Thailand 18, about one‐third are characterized as being high‐risk on the basis of having engaged in condomless sex with casual or known HIV‐positive partners 19, 20. We created a model of the MSM population in Thailand disaggregated into low‐risk and high‐risk groups and populated the model with available country‐specific data on population sizes, sexual behaviour, and testing rates, disaggregated by risk group where possible (Table 1). Non‐context specific parameters were also used to inform the model as documented in the Optima HIV user guide 21. We calibrated the model to historical HIV prevalence estimates to produce baseline estimates of the expected number of new HIV infections over the period from the beginning of 2017 to the end of 2022 in the absence of PrEP.

We assumed a five‐year implementation period (fiscal period 2017‐2018 to 2021‐2022) and measured HIV infections and disability‐adjusted life years (DALYs) averted under a range of different scenarios around PrEP provision. DALYs were calculated from the onset of infection and cost and effects were discounted at an annual rate ranging from 1.5% to 4.5%. Disability weights were obtained from the literature 21.

A knowledge of the baseline level of ART coverage among MSM is essential in assessing the impact and cost‐effectiveness of providing PrEP. In 2014, ART coverage among MSM was reported as 7% 18, but this is likely to be an underestimate as it only includes people who self‐reported as MSM. The ART coverage among the total population of Thailand was approximately 60% 24. To account for the uncertainty surrounding coverage of ART among MSM, we assumed baseline ART coverage of 30% and allowed coverage to vary between the estimates for MSM and for the total population. We further assumed that the scale‐up in ART coverage over the five‐year PrEP implementation will not exceed 20 percentage points. For example, if the baseline ART coverage is assumed to be 7%, we allowed ART coverage levels in 2021 to 2022 to range between 7% and 27%. With the baseline ART coverage of 60%, we allowed 2021 to 2022 coverage levels to range between 60% and 80%, with the upper range corresponding to the achievement of the UNAIDS 90‐90‐90 treatment targets.

We considered two core sets of scenarios based on the parameters listed in Table 2, one in which PrEP will be provided to all MSM and one in which PrEP will only be provided to high‐risk MSM. The uptake of PrEP in each scenario is linked to ART coverage levels, with PrEP uptake in each scenario ranging from 0% up to the assumed levels of ART coverage in 2021 to 2022. We consider the possible impact of PrEP on sexual disinhibition by modelling the impact of reductions in condom use among MSM. While we do not specifically model the effects of other changes in sexual behaviour (for example, an increase in the number of partners), we expect that the results would be similarly sensitive to other forms of sexual disinhibition 25.

We ran 60,000 model simulations, in which samples of each of the eight parameters listed in Table 2 were drawn from uniform distributions over their allowable ranges, and calculated the total number of new HIV infections and DALYs that were estimated over the implementation period for each simulation.

For the cost‐effectiveness analysis, we calculated the cost associated with each scenario by disaggregating the cost of PrEP and ART. To derive the cost of the PrEP intervention, we multiplied the number of HIV‐negative MSM by the percentage of MSM receiving PrEP and then by the unit cost ppy of the PrEP programme. We multiplied the number of HIV‐positive MSM by the percentage of MSM receiving ART, and then by the unit cost ppy of the ART programme to estimate the cost of ART provision. The median cost per person on ART per year including ARV medicines of US$369.57 was obtained from the Test, Treat and Prevent HIV Programme costing study from the same cohort 26. We used a cost‐effectiveness threshold of US$17,449, equal to three times the gross domestic product (GDP) per capita 27 to determine the cost‐effectiveness of PrEP 28.

From 1 June 2015 to 31 May 2016, 366 HIV‐negative MSM and TG participants were recruited at Lerdsin and Thammasat University Hospitals to participate in the PrEP sub‐study. Of the 366 participants, 163 (44.5%) accepted PrEP free‐of‐charge for 12 months, and this was used to determine the acceptance rate and to assess costs. Drug costs accounted for 82.5% of the total annual costs associated with PrEP provision (Table 3). HIV testing costs were US$824 per year. The unit cost ppy, including HIV testing for all clinic visits and the cost of project staff who provided PrEP was US$129. Project staff were hired to support PrEP services, but if the PrEP programme is implemented, project staff will be replaced by MOPH staff. The average unit cost of providing PrEP ppy without project staff is US$128. We assumed that the unit cost of providing PrEP to MSM will be the same as that derived for MSM/TG from the cohort study.

We projected the cost if each participant took PrEP continuously over the 12‐month study period, attended clinic and monitoring visits, and received HIV and lab tests following recommended PrEP programme guidelines 29. We applied a micro costing approach to estimate the unit cost ppy and quantities for personnel, HIV and lab testing, and drugs for each package (Table 4). We applied a 0.7% proportion of overhead costs from a similar existing study in Thailand 30 to the unit cost. In the cohort study, demand creation activities were included in each counselling session. Personal communications with a programme officer from a pilot PrEP project conducted previously in Thailand indicated that the spending for demand creation activities for both HIV testing and PrEP accounted for approximately 22% of the project budget. Although the amount included demand creation activities for PrEP (inclusive of activities during the counselling session) and for HIV testing, we applied this proportion to the unit cost to reflect additional demand creation activities that may be offered. The cost to implement the PrEP programme ranges from US$222.89 to US$310.99 ppy depending on the package options and demand creation activities chosen.

In Table 5 we present results on the number of HIV infections averted, lifetime treatment costs averted, and the incremental cost‐effectiveness ratio (ICER) for both HIV infections and DALYs averted over the five‐year implementation period. Results are presented for two core implementation strategies: PrEP for high‐risk MSM only and PrEP for all MSM, and are based on the scenarios generated using the parameter values shown in Table 2. Comparing these two strategies, we estimate that providing PrEP to all MSM would have a greater epidemiological and economic impact than providing PrEP to high‐risk MSM only, with almost 2.5 as many HIV infections averted (1368 vs. 555) and more than twice the discounted lifetime treatment costs averted (US$9.84 million vs. US$3.99 million). However, providing PrEP only to high‐risk MSM would be approximately 32% more cost‐effective than providing it to all MSM, with ICERs of US$4,836 and US$7,089 per DALY averted respectively. Applying the cost‐effectiveness threshold of three times GDP per capita of US$17,449, the ICERs per DALY averted for both targeting strategies demonstrate the cost‐effectiveness of PrEP services.

Figure 1 presents tornado diagrams summarizing the one‐way sensitivity analysis of the ICERs for both HIV infections and DALYs averted based on the model parameters defined in Table 2. The ICERs are particularly sensitive to both baseline ART coverage and the ART scale‐up rate. Higher ART baseline coverage and a more rapid scale‐up of ART would make the PrEP programme relatively less cost‐effective. These results are also sensitive to the PrEP programme unit cost. In scenarios where PrEP was only provided to high‐risk MSM with all other parameter values as indicated in Table 2, the ICER for HIV infections averted is estimated to lie between US$58,670 and US$83,587, while the ICER for DALYs averted is estimated to lie between US$3,767 and US$6,946. We estimate that these ranges would increase to US$78,855 to US$142,828 and US$5,266 to US$9,559 respectively if PrEP was provided to all MSM. We also find that the scenario for providing PrEP only to high‐risk MSM, if condom use while using PrEP were maintained, would be particularly cost‐effective.

Using a cost‐effectiveness threshold of three times GDP per capita, an estimated 80% of PrEP implementation scenarios (as constructed using the parameter values from Table 2) are calculated to be cost‐effective if the PrEP programme is provided solely to high‐risk MSM, compared to 66% for all MSM (Figure 2). The minimum and maximum ICERs for DALYs averted are estimated to be US$702 (assuming a PrEP unit cost of US$180, 95% PrEP programme effectiveness, no reduction in condom usage, ART coverage remaining constant at 7%, and a 1.5% discount rate) and US$218,802 ($310 PrEP unit cost, 55% PrEP programme effectiveness, 20% reduction in condom usage, 60% ART coverage increasing to 80% by 2021 to 2022, and a 4.5% discount rate) respectively.