Secondary Logo

Journal Logo

Epidemiology

Health Impact and Cost-Effectiveness of HIV Testing, Linkage, and Early Antiretroviral Treatment in the Botswana Combination Prevention Project

Resch, Stephen C. PhDa; Foote, Julia H. A. BAb; Wirth, Kathleen E. ScDc,d; Lasry, Arielle PhDe; Scott, Justine A. MPHb; Moore, Janet PhDe; Shebl, Fatma M. PhDb,f; Gaolathe, Tendani MDj; Feser, Mary K. BAb; Lebelonyane, Refeletswe MDg; Hyle, Emily P. MDb,f,h,i; Mmalane, Mompati O. MDj; Bachanas, Pamela PhDe; Yu, Liyang MSb; Makhema, Joseph M. MDj; Holme, Molly Pretorius MSd; Essex, Max PhDd,j; Alwano, Mary Grace MPHk; Lockman, Shahin MDd,f,j,l; Freedberg, Kenneth A. MDa,b,f,h,i,m

Author Information
JAIDS Journal of Acquired Immune Deficiency Syndromes: August 1, 2022 - Volume 90 - Issue 4 - p 399-407
doi: 10.1097/QAI.0000000000002996

Abstract

INTRODUCTION

With an estimated 20% of the adult population living with HIV, Botswana has the third highest HIV prevalence globally.1 By 2015, Botswana had nearly achieved the UNAIDS 90-90-90 targets for HIV diagnoses, treatment, and viral suppression.2,3 Even so, annual HIV incidence among adults remained high at 1.3%.1,3

The Botswana Combination Prevention Project (BCPP, also known as the Ya Tsie trial) was a cluster-randomized trial designed to examine whether a combination of HIV prevention measures [combination prevention (CP)] could reduce HIV incidence compared with standard of care (SOC).4 Prevention measures in CP included activities approximating universal test and treat [intensive HIV testing campaigns, improved linkage to care, and expanded antiretroviral treatment (ART) eligibility] and increased referrals for voluntary medical male circumcision (VMMC).

In the trial, annual HIV incidence was 30% lower with CP than in the SOC arm over a 29-month follow-up period.4 The trial also reported increases in HIV status knowledge, treatment coverage, viral suppression, and male circumcision coverage in communities exposed to the CP package compared with the SOC.4,5 Around the same time, 2 other international trials, PopART and SEARCH, also showed that universal test and treat led to improved HIV outcomes and decreased transmission.6,7 Our objective was to estimate the cost-effectiveness of the CP package in the BCPP trial.

METHODS

BCPP Trial Overview

BCPP was a pair-matched, cluster-randomized clinical trial from October 2013 to June 2018 in 30 rural and periurban communities; it was designed to determine whether implementation of CP could reduce HIV incidence at the community level.4 Fifteen matched pairs of communities were randomized between the CP arm and the SOC arm. Each arm had a population (age 16–64 years) of ∼55,000 with an HIV prevalence of 26% and ART coverage of 72% at baseline. SOC communities received limited technical support at local HIV clinics. Both SOC and CP communities received HIV testing in a random sample of 20% of households (with clinic referral of persons with HIV) as part of a baseline household survey and annual follow-up surveys. In addition, CP communities received a combination of prevention activities including an intensive saturation campaign of door-to-door and mobile HIV testing and counseling, increased linkage-to-care efforts for those testing positive, and increased VMMC referrals for men testing negative. The CP arm offered expanded ART eligibility relative to SOC, although SOC changed over time with international guidelines (see Supplemental Digital Content, https://links.lww.com/QAI/B842).

All trial participants provided written informed consent. Participants aged 16–17 years provided written assent and written permission from their parents or guardians. The trial was approved by the institutional review boards at the Botswana Ministry of Health and Wellness and the US Centers for Disease Control and Prevention.

Indicators measured in a subset of 3 control and 3 intervention communities at baseline (random ∼20% sample) and study completion (remaining ∼80% of households) found an increase in ART coverage of 19 percentage points in intervention communities compared with a 10-percentage point increase in control communities and a corresponding prevalence ratio of 1.12 [95% confidence interval (CI): 1.07 to 1.17].5 VMMC coverage in men aged 15–49 years increased 10 percentage points in intervention communities (30%–40%) compared with 2 percentage points in control communities (33%–35%) (prevalence ratio 1.26; 95% CI: 1.17 to 1.35).5 A 31% HIV incidence reduction (unadjusted incidence ratio, 0.69; 95% CI: 0.46 to 0.90, P = 0.09) was demonstrated in the intervention arm. Additional trial details have been published previously.3,4

Modeling Analysis Overview

We used results from the BCPP trial to populate CEPAC, a computer microsimulation of HIV disease. Outside the model, we used trial data to estimate key outcomes that were used to scale the model output: the incremental number of people with HIV (PWH) started on ART and the number of infections averted during the trial due to CP. We then populated CEPAC with trial data to estimate lifetime HIV-related care costs and quality-adjusted life expectancy for people in the CP and SOC arms.

Outcomes for each arm accounted for PWH started on ART and the people in whom infection was averted during the trial. We used CEPAC to estimate the additional infections that would be averted over a 10-year horizon after the trial because of the additional PWH started on ART. Therefore, the total health benefit comprises both quality-adjusted life-years (QALYs) gained by PWH who started ART earlier and QALYs gained by people in whom infection was averted because of CP. Lifetime care costs for these groups were combined with the programmatic cost of the CP intervention to estimate the total cost (2019 US$). We calculated the incremental cost-effectiveness ratio (ICER) of CP vs. SOC by dividing the additional cost of CP by the additional QALYs generated. An overview of the approach is shown in Figure 1 (additional detail, see Figure 1, Supplemental Digital Content, https://links.lww.com/QAI/B842).

F1
FIGURE 1.:
Overview of methods and model structure: cost-effectiveness of the BCPP. This figure outlines the CEPAC model runs used in the cost-effectiveness analysis of the Botswana CP Project. Prevalent cohort: The CP subcohort represents PWH who were detected and started on first-line ART during the BCPP trial. The counterfactual SOC represents PWH who were not detected or linked to care through the trial (they may be linked through SOC testing). Discounted QALYs and costs among prevalent PWH excluding the impact of transmissions (QALYCP/QALYSOC and CostCP/CostSOC) are shown in blue. Primary transmissions from PWH in the prevalent cohort (TxCP/TxSOC) are shown in green. Incidence cohort: the AHI subcohort represents participants who acquired HIV during the BCPP trial, and the counterfactual Averted Infection (HIV−) represents participants who did not acquire HIV during the trial. Model outcomes include discounted QALYs and costs (QALYAHI/QALYHIV− and CostAHI/CostHIV−) and are shown in gray. The lifetime difference in discounted quality-adjusted life-years and costs between the Acute HIV and Averted Infection cohorts represents the negative health impact (QALYTx) and additional costs (CostTx) per transmission. To determine prevalent model outcomes including transmission impact, we multiply the impact per transmission (QALYTx and CostTx) by the number of first-order transmissions in the CP and SOC subcohorts. We add the product to the discounted QALYs (QALYCP/QALYSOC) and costs (CostCP/CostSOC) for each subcohort. Prevalent cohort outcomes including transmission impact are found in Table 2. AHI, acute HIV infection; HIV−, no HIV infection; T0, time zero (model initiation); Tx, transmission.

CEPAC Model

CEPAC is a widely published and validated microsimulation of HIV infection, screening, disease progression, and treatment in resource-limited settings.8,9 Simulated patients are generated from user-defined distributions of sex, age, initial CD4 count, HIV RNA, and treatment adherence, then followed monthly from model entry until death. HIV diagnosis occurs via testing or presentation with an opportunistic infection (OI).

The model simulates probabilities of linkage to care, virologic suppression on ART, retention in care, and risk of OIs. ART is initiated according to the strategy-specific CD4 count and viral load thresholds and the occurrence of user-defined primary and secondary OIs. When virally suppressed on ART, patients experience increasing CD4 counts. In the absence of treatment, patients experience a monthly decline in the CD4 count and increased risk of OIs and mortality. Additional model details have been reported and are available at https://mpec.massgeneral.org/cepac-model/.9,10

Estimation of Incremental PWH Starting ART

In the trial's CP arm, 3065 PWH were started on ART, 455 via the baseline survey and 2610 via the testing campaign. Accounting for observed differences in linkage between the arms,4 we estimated that the baseline survey in the control arm resulted in 388 ART initiations. The trial did not collect data on HIV background testing or ART enrollment occurring before trial activities. However, background testing is important because the testing campaign could crowd out some background testing that would have otherwise occurred in the intervention arm. Case detection through background testing was estimated for each arm using trial data on ART coverage and case finding associated with trial activities (see Supplemental Digital Content, https://links.lww.com/QAI/B842).

Incremental Lifetime Outcomes

We used the CEPAC model to project long-term differences in survival, onward transmission, and HIV-related care costs between CP and SOC among 2 cohorts. First, for CP, we simulated a cohort representing additional PWH started on ART during the trial, with characteristics matching those observed in the group started on ART during the trial in terms of age, sex, CD4 count at ART start, and HIV RNA. For SOC, we simulated the same cohort under a counterfactual in which they were not started on ART during the trial. Second, for CP, we simulated the cohort of people in whom infections were averted during the BCPP trial; characteristics matched those in both arms of the trial's incidence cohort diagnosed with HIV during the trial. For SOC, we modeled the counterfactual in which those people acquired HIV during the trial (see Figure 1, Supplemental Digital Content, https://links.lww.com/QAI/B842). Comparing the intervention to the counterfactual, we calculated the difference in lifetime QALY and cost outcomes for both cohorts. The simulation started at the end of the trial (T0 in Figure 1) and continued over a lifetime horizon.

Cost-Effectiveness

We calculated the ICER for CP vs. SOC by dividing the difference in cost between the 2 strategies by the difference in QALYs. As a benchmark for cost-effectiveness, we compared the ICER for CP vs. SOC to a threshold of $3,981, equal to 0.5x Botswana's 2019 annual per capita gross domestic product (GDP).11–13 We report undiscounted health outcomes but use discounted health and cost outcomes (3% per year) for the cost-effectiveness analysis, as recommended by the Second Panel on Cost-Effectiveness in Health and Medicine.14

Model Input Parameters

Cohort Characteristics

Characteristics of the cohort of PWH who started ART because of the intervention reflected trial participants with HIV previously undiagnosed or not linked to care at trial start, measured during the baseline survey. The mean (±SD) initial age was 37 years (±11 years), 68% were female, and the mean initial CD4 count was 449 cells/μL (±266 cells/μL) (Table 1).

TABLE 1. - Key Model Parameters
Input Parameter Base Case Value Reference
[Range Evaluated]
Cohort characteristics Prevalent Incident
 Sex, female/male, % 68/32 67/33 BCPP
 Age, mean (SD), yr 37 (11) 35 (11) BCPP
 Initial CD4, mean (SD), cells/μL 449 (266) 569 (226) BCPP
[182–715] [344–795]
HIV incidence, rate per 100 PY*
 Age, yr BCPP
 <18 0.54 [0.25–1.06]
 18–24 0.84 [0.38–1.65]
 25–29 0.77 [0.35–1.51]
 30–39 0.66 [0.30–1.30]
 ≥40 0.24 [0.11–0.47]
Treatment characteristics, 1st-line ART, TDF/FTC + DTG
HIV-1 RNA suppression at 48 wk, % 98 BCPP
CD4 increase, mean (SD), monthly cells/μL 32
 ≤2 mo 106.8 (29.9)
 >2 mo 5.3 (1.5)
Engagement in care
 Probability of loss to follow-up, yearly, % 3.5 [0.4–8.0] BCPP
 Probability of return to care after 12 mo, monthly, % 1.0 Assumption
 Probability of return to care after OI, one time, % 50.0 Assumption
Standard of care HIV screening
 Probability of HIV test offer and acceptance, monthly, % 3.0 [0.3–6.0] BCPP
HIV test characteristics, %
 Sensitivity 100 33
 Specificity 99 33
Probability of linkage to care if positive, one time, % 86 [50–95] BCPP
Quality of life (utility weights)
 HIV uninfected 0.909–0.860 BCPP
 HIV infected
  500 cells/μL 0.889
  350–500 cells/μL 0.888
  200–349 cells/μL 0.884
  <200 cells/μL 0.879
CP intervention outcomes
 Incremental number of PWH on ART 1418 [709–2836] BCPP
 Incremental cost, 2019 USD 1,693,921 [1.0–6.0 M] BCPP
*Applies to analysis of lifetime outcomes for those in whom an infection was averted.
This incremental number of PWH on ART translates to a 9-percentage point increase in ART coverage in the CP arm compared with SOC.
TDF/FTC + DTG, tenofovir disoproxil fumarate and emtricitabine with dolutegravir.

Characteristics of the cohort of persons in whom HIV was averted reflected study participants who tested negative at baseline and became infected during the trial. Mean age was 35 years (±11 years), and 67% were female. For those who eventually acquired HIV (projected after the trial period), the mean initial CD4 count was 569 cells/μL (±226 cells/μL) at diagnosis.4

Transmissions

To project first-order transmissions averted after the trial period due to incremental ART initiation in prevalent PWH, we used model-based estimates of community viral load in conjunction with viral load-specific monthly transmission rates ranging from 0.00 to 9.03 transmissions/100 person-years (PY), with an acute transmission rate of 62.56 transmissions/100 PY.15

HIV Incidence

In modeling the lifetime impact of averting an infection, the monthly probability of acquiring HIV after the trial period was from age-specific incidence rates observed in the trial's CP intervention arm, ranging from 0.24 to 0.84/100 PY.4,5

Treatment Efficacy and Engagement in Care

Data on adherence, ART efficacy, and engagement in care were from the BCPP trial. Average virologic suppression at 48 weeks across both adherence groups was 98%. For individuals in care, we modeled adherence-dependent loss to follow-up ranging from 0.23% monthly probability for those most adherent to 0.64% for those less adherent, resulting in about 7% of individuals experiencing loss to follow-up over 2 years—similar to rates observed in the trial population.16 Once lost to follow-up, individuals do not receive ART and experience the disease progression typical of untreated HIV. We assumed that individuals return to care at a monthly probability of 1.0% (after 12 months out of care) or with an OI.

Natural History

We derived monthly probabilities of HIV-related mortality stratified by CD4 count, history of OI, and treatment status from mortality in the BCPP trial (see Table 1, Supplemental Digital Content, https://links.lww.com/QAI/B842). Using World Population Prospects and WHO data, we derived monthly non–AIDS-related mortality in Botswana by age and sex.17,18

Intervention Cost

For cost outcomes, we first estimated the cost of delivering the CP intervention from an analysis of the BCPP testing campaign costs19 and program data (see Table 2, Supplemental Digital Content, https://links.lww.com/QAI/B842). In the CP arm, we applied the scaled $32.76/person cost to each person assessed through CP campaigns or the baseline survey. We added the cost of linkage counselors, background testing, and VMMC procedures. Each of the 15 intervention communities was assigned a cost of $13,110 for all linkage counseling activities throughout the trial period (Personal communication, Arielle Lasry). We used costs of $11.15/person for background testing and $116/VMMC procedure (see Table 2, Supplemental Digital Content, https://links.lww.com/QAI/B842).20,21 In the SOC arm, we applied the $32.76/person cost to each person tested through the baseline survey because the door-to-door household testing was similar to the intervention arm testing campaign. Unit costs for background testing and VMMC were assumed the same as in the CP arm.

HIV-Related Care Costs

We calculated HIV-related care costs by multiplying model-estimated resource utilization (eg, outpatient visits and inpatient days) by unit costs. Annual first- and second-line ARV drug costs from the Clinton Health Access Initiative were $77.76 and $288.36.22 We used the Botswana consumer price index and the average 2019 exchange rate to convert all costs to 2019 US dollars (see Table 2, Supplemental Digital Content, https://links.lww.com/QAI/B842).23,24

Sensitivity Analysis

In sensitivity analysis, we considered uncertainty in main trial outcomes including infections averted during the trial, increase in ART coverage, and CP intervention costs. We also considered uncertainty in parameters used in the CEPAC model to simulate long-term outcomes after the trial including HIV care costs, monthly probability of receiving an HIV test, probability of linkage to ART if tested positive, and viral load-based monthly probabilities of transmitting HIV. We varied each parameter from 10% to as much as 300% of its base case value. For multiway sensitivity analysis, we varied the incremental increase in ART coverage from 0.5 to 9.0 percentage points (base case: 9 percentage points) and simultaneously adjusted the number of infections averted from 5% to 100% of the base case level, as ART coverage and number of infections averted are expected to be correlated. This range covers the lower bound of uncertainty intervals for main trial outcomes regarding infections averted and ART coverage.4 We simultaneously varied the incremental cost of the CP intervention from 100% to 250% the base case (base case: $1.7 million).

Role of the Funding Source

The study sponsor had no role in study design, data collection, analysis, or interpretation, presentation of the findings, or the decision to submit the manuscript.

RESULTS

Base Case

Calibrating to observed differences in the increase in ART coverage between arms, we estimate that CP resulted in 1418 more PWH started on ART than SOC (Table 1 and see Supplemental Digital Content, https://links.lww.com/QAI/B842). During the trial, 262 infections were averted over 29 months in the intervention communities because of improved case finding and linkage, earlier ART initiation, and VMMC coverage increases. Based on simulation of the lifetime of persons started on ART during the trial because of CP, we estimate an additional 42 infections averted in the 10 years after trial, for a total of 304 infections averted (Table 2).

TABLE 2. - Clinical and Economic Outcomes of the CP Intervention Compared With SOC
Incident HIV Infections* Total Life-Years (Discounted) Total Quality-Adjusted Life-Years (Discounted) Total Cost, $ (Discounted) ICER $/QALY
Long-term impact of BCPP Trial
 SOC 972 23,586 20,803 24,124,000
 CP 668 25,847 22,804 24,282,000
 Difference −304 2261 2001 157,000 $79
*Infections include those during the trial period, measured in the trial, plus additional first-order transmissions over 10 years estimated with the CEPAC model for the cohort that started ART earlier because of the intervention and its counterfactual.
Discounted at 3% per year.
Total costs are rounded to the nearest $100.
LY, life year.

Earlier HIV detection and ART initiation in the CP arm increased quality-adjusted life expectancy by 0.90 QALYs per person and lifetime HIV-related care costs by $869 per person started earlier on ART (see Table 3, Supplemental Digital Content, https://links.lww.com/QAI/B842). For people in whom an infection was averted, quality-adjusted life expectancy increased by 2.43 QALYs (from 17.03 to 19.46 QALYs), and $9200 in HIV-related medical care costs were saved per person (from $9970 to $730, see Table 4, Supplemental Digital Content, https://links.lww.com/QAI/B842). Using CEPAC model results, we estimate that 2086 discounted QALYs were generated by CP, split about evenly between the 1418 PWH started earlier on ART and the 304 persons in whom an infection was averted.

Testing and linkage activities in the intervention arm, including those related to the baseline survey, cost $2.22 million. We estimated that another $111,000 was spent on background testing and $207,000 on 1777 VMMC procedures in the intervention arm. In the SOC arm, we estimate a cost of $208,000 for testing and $41,000 for VMMC related to the baseline survey and $592,000 in background testing costs. Thus, the incremental total direct cost of CP activities was $1.69 million (Figure 2).

F2
FIGURE 2.:
Cost breakdown of the incremental cost of the CP intervention. This waterfall chart reports the positive and negative incremental costs (y-axis) of each component (x-axis) of the CP intervention arm compared with the standard of care arm. Positive incremental costs are represented in orange, negative incremental costs in blue, and total or net incremental costs in gray. Exact cost values appear above or below the corresponding bar.

In addition to the incremental cost of delivering the intervention, we estimated the incremental cost of HIV care for the 1418 PWH started on ART because of CP at $1.23 million. The $2.8 million in HIV care costs saved due to the 304 infections prevented offset most of the cost of the CP activities and consequent HIV care for PWH detected and started on ART, resulting in a net cost of $157,000 (Table 2 and Figure 2) and an ICER of $79 per QALY gained.

Sensitivity Analysis

In univariate sensitivity analysis, the ICER for CP vs. SOC was consistently <0.5x Botswana annual per capita GDP. Other parameters used in simulating long-term impacts did not substantially affect the ICER when varied across wide ranges (Figure 3).

F3
FIGURE 3.:
One-way sensitivity analyses on the cost-effectiveness ($/QALY) of CP compared with SOC in Botswana, including the impact of first-order HIV transmissions over 10 years. This tornado diagram represents the ICERs (x-axis) for CP compared with SOC after input parameters (y-axis) were varied. The base case value for each input parameter is listed in parentheses before the semicolon. The range across which we varied each parameter is listed after the semicolon, with the value resulting in the lowest ICER before the hyphen and the value resulting in the highest ICER after the hyphen. The range of ICERs for each varied parameter is indicated by the horizontal bars. Longer horizontal bars indicate parameters to which the model results are most sensitive. The solid black line indicates the ICER for CP vs. SOC in the base case ($79/QALY). The dotted black line indicates 0.25x Botswana per capita GDP in 2019.

When we simultaneously varied the incremental cost of the CP intervention and the impact of the program on ART coverage and infections averted, CP remained cost-effective over a wide range of parameters (Figure 4). When the impact of the intervention was two-thirds of the base case, with an increase in ART coverage of 6 percentage points and 203 infections averted, CP remained cost-effective even when intervention costs were 2.5x the base case. When the impact of the intervention was reduced to one-third of the base case (a 3-percentage point increase in ART coverage and 102 infections averted), CP remained cost-effective if the intervention cost did not exceed 1.5x the base case. If the impact of the intervention was only 10% of the base case (a 0.9-percentage point increase in ART coverage and 31 infections averted), the CP intervention would not be cost-effective at base case cost. In this case, costs would need to be 42% lower to meet the threshold of 0.5x GDP per capita per QALY gained. Even with a 3-percentage point increase in ART coverage and ∼10% reduction in infections, corresponding to the lower bound of the 95% CI for the incidence reduction observed in the trial, the ICER remained below 25% of annual per capita GDP in Botswana per QALY gained.

F4
FIGURE 4.:
Two-way sensitivity analysis: Cost-effectiveness as a function of the incremental increase in ART coverage and the cost of the CP intervention. This heat map reports the ranges of incremental cost-effectiveness ratios of CP vs. SOC as a function of the 2 most influential parameters in Figure 3: incremental cost of the CP intervention (vertical axis) and incremental increase in ART coverage and infections averted (horizontal axis). Colors indicate the incremental cost-effectiveness ratio achieved by each combination of these parameters, ranging from very cost-effective in green (<0.25x Botswana annual per capita GDP of $8000) to cost-effective in yellow (0.25–0.5x GDP) and orange (0.5–1x GDP) and not cost-effective in red (>1x GDP). The base case combination (9-percentage point incremental increase in ART coverage for an incremental CP cost of $1.7 million) is indicated by ** in the upper left cell.

DISCUSSION

We projected the long-term clinical impact, cost, and cost-effectiveness of a combination HIV prevention intervention implemented in the BCPP. Based on trial outcomes, we estimated that 1418 additional PWH were started on ART and 304 infections were averted because of the CP intervention compared with the SOC. The 2086 QALYs of long-term total health gain produced by CP were split equally between prevalent cases starting earlier on ART and those in whom an infection was averted. Although the CP intervention required a large initial investment—about $1.7 million more than SOC prevention to cover a target population of 55,000—and will generate additional HIV care costs in those already infected who start ART earlier, these costs will likely be largely offset over time because of the prevention of new infections. Indeed, our analysis projects that HIV care costs saved by averting infections will offset nearly all CP program costs and treatment costs for the additional PWH started on ART, resulting in a long-run net cost of $157,000 and an ICER of $79 per QALY.

These results were most dependent on the cost of the CP intervention, the cost of HIV care, and the impact of the intervention on infections averted and ART coverage gained. Within plausible ranges, the CP intervention would remain cost-effective.

This study can also be interpreted in the context of 2 other major international trials assessing the test and treat paradigm. The PopART trial (HPTN 071) in South Africa and Zambia tested 2 versions of CP and showed that CP improved ART coverage and viral suppression. A decrease in HIV incidence was observed with one version of CP, but not the other. A report of cost-effectiveness suggests that for a comparable scenario, the intervention as implemented in the trial had an ICER of $326/disability-adjusted life year in South Africa and $258/disability-adjusted life year in Zambia.25 The PopART results, though slightly less favorable when comparing base cases, are broadly consistent with our findings regarding the value of CP considering differences in epidemiological context, case finding, and HIV treatment costs, adherence and viral suppression in people on ART, modeled time horizon, and parameter and model uncertainty in both analyses. The SEARCH study of over 350,000 people in Kenya and Uganda showed an increase in viral suppression in PWH from 68% to 80% but no change in HIV incidence. Cost-effectiveness results have not yet been reported.26,27

Other case-finding approaches such as index contact testing, social network testing, and self-testing have also proven cost-effective.30,31 These approaches are not intended as alternatives; they complement one another as part of a strategic mix, which depends on the context in which they are implemented. Although the unit costs of an intensive saturation campaign are high relative to facility-based approaches, the testing campaigns led to the identification of nearly all remaining unaware and out-of-care PWH in the intervention communities, in turn leading to observed incidence reduction within the study period.

This analysis of the BCPP trial has several limitations. We could not evaluate the contribution of each aspect of the CP intervention to the benefits observed in the trial. We did not have HIV care cost data from BCPP and used cost estimates from other studies.22,28,29 Furthermore, we included only first-order transmissions over 10 years, so we may be underestimating the total number of infections averted over time from the intervention. The incremental cost of the testing campaign and the number of PWH started on ART attributable to the intervention are uncertain. The amount and cost of testing and the number of PWH started on ART in the control arm were not measured in the trial, so our estimates of these rely on extrapolations from changes in ART coverage measured in a subset (3 of 15 pairs) of the communities in both trial arms. In addition, our analysis compares the CP intervention to an SOC counterfactual that included a baseline survey with HIV testing in 20% of households. If the SOC had not included the baseline survey, both the incremental cost of CP and the incremental health effect would be higher and the net impact on cost-effectiveness would likely be small. Finally, this analysis, based on the BCPP trial, compared CP with an evolving SOC that was short of universal test and treat. Despite these limitations, sensitivity analysis showed that our cost-effectiveness results were robust to variations in efficacy and cost: we found that at base case effectiveness, the intervention would remain cost-effective at 2.5x base case cost. Even if effectiveness was substantially reduced, with only a 3-percentage point increase in people starting on ART and a 10% reduction in incident infections (102 infections averted), the CP strategy would remain cost-effective compared with a threshold of 50% of Botswana annual per capita GDP.

As with any study involving long-time horizons, results are subject to uncertainty related to advances in testing, treatment, prevention, or other factors that may impact future transmission dynamics or the outcomes of HIV treatment. Despite these uncertainties, the results of sensitivity analyses increase our confidence that the CP intervention, as implemented in Botswana during the Ya Tsie trial, will be cost-effective. Even with substantially higher testing campaign costs, much smaller increases in ART coverage, and fewer infections averted, the ICER for CP would remain under 50% of Botswana annual per capita GDP, a reasonable cost-effectiveness threshold for a middle-income sub-Saharan African country.12,13 At the time of the study Botswana had, and still has, among the highest pre-existing rates of HIV diagnosis, linkage to ART, and viral load suppression of any country, particularly high HIV-burden countries—and despite this, the CP intervention was cost-effective. It is likely that similar interventions, if they can improve case finding and provide robust linkage to ART services, would be cost-effective in other settings as well.

CONCLUSIONS

Despite high upfront costs, large-scale CP interventions featuring intensive saturation testing campaigns that reach PWH not in care and link them to ART are likely to be cost-effective in settings that are comparable to Botswana during the period of the Ya Tsie trial in terms of HIV prevalence, undiagnosed HIV, and care costs. Even in settings with high baseline testing and high ART coverage like Botswana, CP may further improve health among PWH and improving population health and lowering HIV care costs by reducing HIV incidence.

ACKNOWLEDGMENTS

The authors gratefully acknowledge BCPP trial investigators, personnel, and participants. They also thank Yiqi Qian, Kara Bennett, and Jean Leidner for statistical assistance and Taige Hou and Christopher Panella for programming assistance.

REFERENCES

1. UNAIDS. Botswana country factsheet [Internet]; 2020. Available at: https://www.unaids.org/en/regionscountries/countries/botswana. Accessed February 11, 2022.
2. Essex M, Makhema J, Lockman S. Reaching 90-90-90 in Botswana. Curr Opin HIV AIDS. 2019;14:442–448.
3. Gaolathe T, Wirth KE, Holme MP, et al. Botswana's progress toward achieving the 2020 UNAIDS 90-90-90 antiretroviral therapy and virological suppression goals: a population-based survey. Lancet HIV. 2016;3:e221–30.
4. Makhema J, Wirth KE, Pretorius Holme M, et al. Universal testing, expanded treatment, and incidence of HIV infection in Botswana. N Engl J Med. 2019;381:230–242.
5. Wirth KE, Gaolathe T, Pretorius Holme M, et al. Population uptake of HIV testing, treatment, viral suppression, and male circumcision following a community-based intervention in Botswana (Ya Tsie/BCPP): a cluster-randomised trial. Lancet HIV. 2020;7:e422–e433.
6. Hayes RJ, Donnell D, Floyd S, et al. Effect of universal testing and treatment on HIV incidence—HPTN 071 (PopART). N Engl J Med. 2019;381:207–218.
7. Havlir DV, Balzer LB, Charlebois ED, et al. HIV testing and treatment with the use of a community health approach in rural Africa. N Engl J Med. 2019;381:219–229.
8. Goldie SJ, Yazdanpanah Y, Losina E, et al. Cost-effectiveness of HIV treatment in resource-poor settings—the case of Côte d'Ivoire. N Engl J Med. 2006;355:1141–1153.
9. Walensky RP, Borre ED, Bekker LG, et al. The anticipated clinical and economic effects of 90-90-90 in South Africa. Ann Intern Med. 2016;165:325–333.
10. Walensky RP, Ross EL, Kumarasamy N, et al. Cost-effectiveness of HIV treatment as prevention in serodiscordant couples. N Engl J Med. 2013;369:1715–1725.
11. The World Bank. GDP Per Capita (Current US$)–Botswana [Internet]; 2020. Available at: https://data.worldbank.org/indicator/NY.GDP.PCAP.CD?locations=BW. Accesses December 4, 2020.
12. Edoka IP, Stacey NK. Estimating a cost-effectiveness threshold for health care decision-making in South Africa. Health Policy Plan. 2020;35:546–555.
13. Woods B, Revill P, Sculpher M, et al. Country-level cost-effectiveness thresholds: initial estimates and the need for further research. Value Health. 2016;19:929–935.
14. Sanders GD, Neumann PJ, Basu A, et al. Recommendations for conduct, methodological practices, and reporting of cost-effectiveness analyses: second panel on cost-effectiveness in health and medicine. JAMA. 2016;316:1093–1103.
15. Attia S, Egger M, Müller M, et al. Sexual transmission of HIV according to viral load and antiretroviral therapy: systematic review and meta-analysis. AIDS. 2009;23:1397–1404.
16. Lebelonyane R, Bachanas P, Alwano MG, et al. Interventions Successfully Lead to Increases in Knowledge of HIV Positive Status, Antiretroviral Treatment, and Viral Suppression in the Ya Tsie Botswana Prevention Project. Amsterdam, the Netherlands: International AIDS Society; 2018.
17. United Nations. World Population Prospects: Botswana [Internet]; 2019. Available at: https://population.un.org/wpp/Download/Standard/Population/. Accessed January 29, 2020.
18. World Health Organization. Global Health Estimates 2016: Disease Burden by Cause, Age, Sex, by Country and by Region, 2000-2016. Geneva, Switzerland; 2018.
19. Lasry A, Bachanas P, Suraratdecha C, et al. Cost of community-based HIV testing activities to reach saturation in Botswana. AIDS Behav. 2019;23:875–882.
20. Sharma M, Ying R, Tarr G, et al. Systematic review and meta-analysis of community and facility-based HIV testing to address linkage to care gaps in sub-Saharan Africa. Nature. 2015;528:S77–S85.
21. Pineda-Antunez C, Martinez-Silva G, Cerecero-Garcia D, et al. Meta-analysis of average costs of HIV testing and counselling and voluntary medical male circumcision across thirteen countries. Afr J AIDS Res. 2019;18:341–349.
22. The Clinton Health Access Initiative (CHAI). 2017 Antiretroviral (ARV) CHAI Reference Price List [Internet]; 2017. Available at: https://www.clintonhealthaccess.org/2017-chai-arv-reference-price-list/. Accessed January 15, 2021.
23. OANDA. Botswana Pula Historical Exchange Rates [Internet]. Available at: https://www1.oanda.com/fx-for-business/historical-rates. Accessed January 29, 2020.
24. The World Bank. Botswana Consumer Price Index [Internet]. Available at: https://data.worldbank.org/indicator/FP.CPI.TOTL?end=2018&locations=BW&start=1996. Accessed January 29, 2020.
25. Thomas R, Probert WJM, Sauter R, et al. Cost and cost-effectiveness of a universal HIV testing and treatment intervention in Zambia and South Africa: evidence and projections from the HPTN 071 (PopART) trial. Lancet Glob Health. 2021;9:e668–e680.
26. Kwarisiima D, Kamya MR, Owaraganise A, et al. High rates of viral suppression in adults and children with high CD4+ counts using a streamlined ART delivery model in the SEARCH trial in rural Uganda and Kenya. J Int AIDS Soc. 2017;20:21673.
27. Shade SB, Osmand T, Luo A, et al. Costs of streamlined HIV care delivery in rural Ugandan and Kenyan clinics in the SEARCH Study. AIDS. 2018;32:2179–2188.
28. Cleary S, Boulle A, McIntyre D, et al. Cost-effectiveness of Anti-retroviral Treatment for HIV-Positive Adults in a South African Township. Médecins Sans Frontières and the Health Systems Trust; 2004.
29. Anglaret X, Chêne G, Attia A, et al. Early chemoprophylaxis with trimethoprim-sulphamethoxazole for HIV-1-infected adults in Abidjan, Côte d'Ivoire: a randomised trial. Cotrimo-CI Study Group. Lancet. 1999;353:1463–1468.
30. Johnson LF, van Rensburg C, Govathson C, et al. Optimal HIV testing strategies for South Africa: a model-based evaluation of population-level impact and cost-effectiveness. Sci Rep. 2019;9:12621.
31. Cambiano V, Johnson CC, Hatzold K, et al. The impact and cost-effectiveness of community-based HIV self-testing in sub-Saharan Africa: a health economic and modelling analysis. J Int AIDS Soc. 2019;22(suppl 1):e25243.
32. Walmsley SL, Antela A, Clumeck N, et al. Dolutegravir plus abacavir-lamivudine for the treatment of HIV-1 infection. N Engl J Med. 2013;369:1807–1818.
33. Mine M, Chishala S, Makhaola K, et al. Performance of rapid HIV testing by lay counselors in the field during the behavioral and biological surveillance survey among female sex workers and men who have sex with men in Botswana. J Acquir Immune Defic Syndr. 2015;68:365–368.
Keywords:

cost-effectiveness; combination prevention; HIV testing; modeling; economic analysis; Botswana

Supplemental Digital Content

Copyright © 2022 The Author(s). Published by Wolters Kluwer Health, Inc.