Cost-Effectiveness of Preemptive Switching to Efavirenz-Based Antiretroviral Therapy for Children with Human Immunodeficiency Virus

The NEVEREST-3 (South Africa) and MONOD-ANRS-12206 (Côte d’Ivoire, Burkina Faso) randomized trials found that switching to efavirenz (EFV) in human immunodeficiency virus-infected children >3 years old who were virologically suppressed by ritonavir-boosted lopinavir (LPV/r) was noninferior to continuing o LPV/r. We evaluated the cost-effectiveness of this strategy using the Cost-Effectiveness of Preventing AIDS Complications-Pediatric model. Methods: We examined 3 strategies in South African children aged ≥3 years who were virologically suppressed by LPV/r: (1) continued LPV/r, even in case of virologic failure, without second-line regimens; continued on LPV/r with second-line option after observed virologic failure; and preemptive switch to EFV-based antiretroviral therapy (ART), with return to LPV/r after observed virologic failure. We derived data on 24-week suppression (<1000 copies/mL) after a switch to EFV (98.4%) and the subsequent risk of virologic failure (LPV/r, 0.23%/mo; EFV, 0.15%/mo) from NEVEREST-3 data; we obtained ART costs (LPV/r, $6-$20/mo; EFV, $3-$6/mo) from published sources. We projected discounted life expectancy (LE) and lifetime costs per person. A secondary analysis used data from MONOD-ANRS-12206 in Côte d'Ivoire. Results: Continued LPV/r led to the shortest LE (18.2 years) and the highest per-person lifetime cost ($19 470). LPV/r with second-line option increased LE (19.9 years) and decreased per-person lifetime costs($16 070). Switching led to the longest LE (20.4 years) and the lowest per-person lifetime cost ($15 240); this strategy was cost saving under plausible variations in key parameters. Using MONOD-ANRS-12206 data in Côte d'Ivoire, the Switch strategy remained cost saving only compared with continued LPV/r, but the LPV/r with second-line option strategy was cost-effective compared with switching. Conclusion: For children ≥3 years old and virologically suppressed by LPV/r-based ART, preemptive switching to EFV can improve long-term clinical outcomes and be cost saving. Clinical Trials Registration: NCT01127204. We used the Cost-Effectiveness of Preventing AIDS Complications (CEPAC)–Pediatric model to compare 3 modeled strategies for children ≥3 years of age virologically suppressed by initial LPV/r-based therapy (Figure 1). The first is continued LPV/r during suppression; with virologic failure (defined as confirmed viremia, 2 consecutive viral load measurements >1000 copies/µL), there are no subsequent options, as is often the case in resource-limited settings [16]. The second strategy is LPV/r with second-line option; that involves continuing LPV/r but also starting a nonnucleoside reverse-transcriptase inhibitor (NNRTI)–based regimen in children with observed virologic failure, per the 2013 recommendations from the World Health Organization (WHO) [3]. The third choice is switching, that is, using the preemptive EFV switch strategy evaluated in the NEVEREST-3 and MONOD-ANRS-12206 trials, which involves switching from suppressive LPV/r to EFV. Model outcomes included short- and long-term survival, per-person HIV-related healthcare costs, and life expectancy (LE). Diagram of the 3 modeled strategies. Continued LPV/r represents the current practice in most sub-Saharan African settings, where no alternative option is available after ritonavir-boosted lopinavir (LPV/r) failure. LPV/r with second-line option represents the current World Health Organization recommendations, where children failing first-line LPV/r should be switched to second-line antiretroviral therapy (ART). In the base case, we assumed this second-line option would be non-nucleoside reverse-transcriptase inhibitor based, with virologic outcomes shown here; we varied these virologic outcomes and costs widely to reflect other second-line ART options, including dolutegravir. The “switch” strategy is the strategy evaluated in the NEVEREST-3 and MONOD-ANRS-12206 trials. Abbreviations: EFV, efavirenz; VF, virologic failure. We calculated an incremental cost-effectiveness ratio (ICER) for each strategy, discounted at 3% per year, compared with the next least expensive alternative: the difference in lifetime costs divided by the difference in years of life saved (YLS). Following cost-effectiveness analysis convention, discounting was used to adjust future costs and LEs to their present value, reflecting common preferences for benefits that accrue in the present rather than in the future [17]. We considered interventions cost-effective if they had ICERs less than the country’s per-capita gross domestic product for 2016 ($5270 for South Africa and $1500 for Côte d’Ivoire). We recalculated virologic outcomes (initial suppression and subsequent virologic failure risks) from primary trial data as needed for the structure of the CEPAC model (Supplementary Table A). Our primary analysis modeled NEVEREST-3 data in South Africa, and we conducted a secondary analysis using MONOD-ANRS-12206 data in Côte d’Ivoire (Table 1 and Supplementary Tables A and B). In sensitivity analyses, we varied key model input parameters individually and simultaneously. Selected Model Input Parameters Abbreviations: ART, antiretroviral therapy; EFV, efavirenz; LPV/r, ritonavir-boosted lopinavir; NRTIs, nucleos(t)ide reverse-transcriptase inhibitors; SD, standard deviation. aART efficacy was expressed as the probability of suppressing human immunodeficiency virus RNA levels to <400 copies/mL at the time specified (immediately in children continuing with suppressive LPV/r, otherwise 24 weeks in base case analyses after initiation of ART [19–21]. Owing to small numbers of children and similar suppression rates for second-line ART in the P1060 trial (second-line nonnucleoside reverse-transcriptase inhibitor [NNRTI], n = 9; 24-week suppression, 75%; second-line protease inhibitor, n = 48; 24-week suppression, 74%), we assigned a suppression rate of 75% to both second-line regimens. bThe monthly risk of virologic failure for those with initially suppression on ART was calculated from the difference in suppression risks at the earliest (24 weeks) and at 48 weeks in the NEVEREST-3 and MONOD-ANRS-12206 trials and the latest observed time point in the P1060 and PENPACT-1 trials. To fit the Cost-Effectiveness of Preventing AIDS Complications model structure, these values differ slightly from those reported in published reports of these trials. (See Supplementary Table B for full details of calculations of these parameters from the NEVEREST-3 and MONOD-ANRS-12206 data.) cThe second-line regimen is used after observed virologic failure with the previous ART regimen. In the LPV/r with second-line option strategy, in the base case, we assumed that this second-line regimen would be an NNRTI with 2 NRTIs. With the switch strategy, we assumed it would be LPV/r with 2 NRTIs. dCosts are given in 2016 US dollars. eMonthly ART drug doses were calculated for children aged 0–13 years old based on the World Health Organization weight-based dosing recommendations. Daily doses were then multiplied by unit drug costs from the May 2012 Clinton Health Access Initiative antiretroviral price list to determine monthly ART costs by age and weight. All children were assumed to receive liquid/syrup drug formulations until age 5 years for LPV/r. After this age, children were assumed to transition to pediatric or adult tablet formulations based on weight-based dosing recommendations. Fixed-dose combinations were assumed to be used where available [23].The initial NRTI backbone was Zidovudine/Lamivudine and children were switched to Abacavir/Lamivudine in case of toxicity. The CEPAC-Pediatric model is a validated patient-level, Monte Carlo, state-transition model of pediatric HIV disease in children (http://www.massgeneral.org/mpec) [15, 25, 26]. Children enter the model with CD4 cell count and age drawn from user-specified distributions and experience disease progression according to a range of literature-based parameters, including opportunistic infection (OI) and mortality risks, ART efficacy and toxicity and impact of viral load and laboratory monitoring (Supplementary Material). Effective ART decreases HIV viral load and increases CD4 cell count, leading to reduced risk of OIs and OI-related and HIV-related mortality rates. In each month, children can remain in care or be lost to follow-up, in which case they are assumed to stop ART, and return to care if a severe OI occurs. The model tracks clinical events, the amount of time spent in each health state, LE, and associated costs. In our base case, all children were suppressed on LPV/r-based ART at the start of the simulation (Table 1). In the continued LPV/r arm, they continued taking LPV/r regardless of virologic failure. For LPV/r with second-line option, they switched to NNRTI-based ART if virologic failure was later observed. With the switch strategy, children were switched immediately from LPV/r to EFV-based ART at the start of the simulation; if virologic failure observed with EFV, they were switched back to an LPV/r-based regimen (Figure 1). Monitoring and switching in all strategies followed WHO guidelines [4], including CD4 cell counts every 6 months and HIV RNA tests every 12 months (Table 1) [27]. The rate of loss to follow-up was 2%/mo. There were several key differences between NEVEREST-3 and MONOD-ANRS-12206 participants, including duration of suppressive LPV/r before the switch, age at switch, EFV dosing, use and dosing of previously received maternal and infant prevention of mother-to-child transmission regimens, and viral subtypes [12, 13]. Because the NEVEREST-3 trial had larger numbers and longer follow-up available, we chose to use NEVEREST-3 data for the base case analyses, including age at preemptive switch (3 years), initial suppression for each ART regimen, and subsequent monthly risk of virologic failure (Table 1, with full calculations in Supplementary Table A) [12]. The initial suppression probabilities for EFV in the switch arm were derived 24 weeks after the switch (proportion with RNA level <1000 copies/mL, 98.4%). The risk of virologic failure after this initial suppression (“late failure”) was 0.23%/mo with LPV/r and 0.15%/mo with EFV (Table 2 and Supplementary Table A). The rate of initial suppression (75%) and the late failure risk (0.91%/mo) for subsequent ART regimens initiated after failed LPV/r (for LPV/r with second-line option) or failed EFV (for the switch strategy) were derived from the P1060 and PENPACT-1 studies [19, 20]. Model-Projected Base Case Clinical And Economical Outcomes In South Africa Abbreviations: LE, life expectancy; LPV/r, ritonavir-boosted lopinavir. aLEs are mean values projected by the model for a cohort of children similar to those aged 3–5 years in the NEVEREST-3 trial at the time of switch. Discounted LEs, which value life-years in the future to be worth than those in the present, are not directly comparable to clinical experience. The NEVEREST-3 trial reported few neuropsychiatric adverse effects for EFV, so the base case rate of neurologic toxicity (leading to added costs and switching to LPV/r-based ART) was set to 0% and varied in sensitivity analyses [12]. Similarly, the NEVEREST-3 trial reported no toxicity that led to regimen change in children treated with LPV/r; as a result, the risk for toxicity with LPV/r was set to 0% [12]. In the base case, we included costs for the treatment of tuberculosis as follows. In the case of incident tuberculosis occurring with continued LPV/r or LPV/r with second-line option, we assumed a temporary switch to EFV for the duration of tuberculosis treatment; in the case of incident tuberculosis during second-line LPV/r in the switch strategy, we assumed the use of superboosted LPV/r for the 6-month duration of tuberculosis treatment. Costs were derived from the South African Health Review and the Clinton Health Access Initiative, in 2016 US dollars (Table 1) [19–21]. The monthly ART costs were calculated according to age and weight bands; for example, the cost of ART in 2–5-year-olds was $33/mo for LPV/r-based ART and $20/mo for EFV-based ART and second-line ART after LPV/r failure (Table 1). In a secondary analysis, we derived similar data from the MONOD-ANRS-12206 trial and applied these data to a modeled population of children in Côte d’Ivoire. Complete model input data are provided in Supplementary Table B. In univariate sensitivity analyses, we varied key model input parameters, including initial suppression (0%–100%), late failure rates (0.1%–2.0%/mo), major toxicity rates for EFV and LPV/r (1-time probability leading to regimen change, 1%–20%), and costs (0.5–2.0 times the base case) for each ART regimen; the frequency of viral load testing during first-line ART (every 3 months during EFV treatment and in all strategies); the costs of HIV care (0.5–2.0 times the base case) and HIV RNA monitoring (0.5–2.0 times the base case); age at preemptive switch (3–5 years); and loss to follow-up rate (0%–1%/mo) (Supplementary Table B). In 2-way sensitivity analyses, we varied the probabilities of late failure with EFV in the switch strategy and of LPV/r in with second-line option and with continued LPV/r. The specific pediatric ART regimen chosen to follow virologic failure with LPV/r varies by setting; DTG may soon become more widely available for this indication [7]. To reflect a range of possible antiretroviral formulations that would be used after failed LPV/r in children treated with second-line option, we varied the virologic outcomes (initial suppression and late failure) and costs of this regimen, including values likely to reflect DTG based on adult data (initial suppression rate, 94%; late failure risk, 0.21%/mo; cost, $4/mo) [28–30]. In a secondary analysis, we used data from the MONOD-ANRS-12206 trial [13]. Age at the start of the simulation was 26.8 months. The rate of initial suppression with EFV was 98.1%. The late failure risk was lower for LPV/r (0.34%) than for EFV (0.72%), which was directionally different from the findings in the NEVEREST-3 trial (Table 1). For this secondary analysis, we derived pediatric HIV care costs directly from the MONOD-ANRS-12206 trial for clinical care; clinical inputs were the same as the base case analysis for children <5 years old and were derived from adults in Côte d’Ivoire for older ages [22, 23, 31, 32].