Articles

Monitoring of HIV viral load, CD4 cell count, and clinical assessment versus clinical monitoring alone for antiretroviral therapy in low-resource settings (Stratall ANRS 12110/ESTHER): a cost-eff ectiveness analysis

Sylvie Boyer, Laura March, Charles Kouanfack, Gabrièle Laborde-Balen, Patricia Marino, Avelin Fobang Aghokeng, Eitel Mpoudi-Ngole, Sinata Koulla-Shiro, Eric Delaporte, Maria Patrizia Carrieri, Bruno Spire, Christian Laurent, Jean-Paul Moatti, on behalf of the Stratall ANRS 12110/ESTHER Study Group*

Summary Background In low-income countries, the use of laboratory monitoring of patients taking antiretroviral therapy Lancet Infect Dis 2013; (ART) remains controversial in view of persistent resource constraints. The Stratall trial did not show that clinical 13: 577–86 monitoring alone was non-inferior to laboratory and clinical monitoring in terms of immunological recovery. We Published Online aimed to evaluate the costs and cost-eff ectiveness of the ART monitoring approaches assessed in the Stratall trial. April 18, 2013 http://dx.doi.org/10.1016/ S1473-3099(13)70073-2 Methods The randomised, controlled, non-inferiority Stratall trial was done in a decentralised setting in Cameroon. See Comment page 560 Between May 23, 2006, and Jan 31, 2008, ART-naive adults were randomly assigned (1:1) to clinical monitoring *Members listed at end of paper (CLIN) or viral load and CD4 cell count plus clinical monitoring (LAB) and followed up for 24 months. We INSERM, UMR912 (SESSTIM), calculated costs, number of life-years saved (LYS), and incremental cost-eff ectiveness ratios (ICERs) with data Marseille, France (S Boyer PhD, from patients who had been followed up for at least 6 months. We considered two cost scenarios in which viral L March MSc, P Marino PhD, load plus CD4 cell count tests cost either US$95 (scenario 1; Abbott RealTime HIV-1 assay) or $63 (scenario 2; M P Carrieri PhD, B Spire MD, generic assay). We compared ICERs with a WHO-recommended threshold of three times the per-person gross Prof J-P Moatti PhD); Aix Marseille Université, domestic product (GDP) for Cameroon ($3670–3800) and an alternative lower threshold of $2385 to UMR-S912, IRD, Marseille, determine cost-eff ectiveness. We assessed uncertainty with one-way sensitivity analyses and cost-eff ectiveness France (S Boyer, L March, acceptability curves. P Marino, M P Carrieri, B Spire, J-P Moatti); Observatoire Régional de la Santé Provence– Findings 188 participants who underwent LAB and 197 who underwent CLIN were followed up for at least 6 months. Alpes–Côte d’Azur (ORS PACA), In scenario 1, LAB increased costs by a mean of $489 (SD 430) per patient and saved 0·103 life-years compared with Marseille, France (S Boyer, CLIN (ICER of $4768 [95% CI 3926–5613] per LYS). In scenario 2, the incremental mean cost of LAB was L March, M P Carrieri, B Spire, $343 (SD 425)—ie, an ICER of $3339 (2507–4173) per LYS. A combined strategy in which LAB would only be used in J-P Moatti); Central Hospital, UMI 233, Yaoundé, Cameroon patients starting ART with a CD4 count of 200 cells per μL or fewer suggests that 0·120 life-years would be saved at (C Kouanfack MD, an additional cost of $259 per patient in scenario 1 (ICER of $2167 [95% CI 1314–3020] per LYS) and $181 in scenario Prof S Koulla-Shiro MD); French 2 (ICER of $1510 [692–2329] per LYS) when compared with CLIN. Ministry of Foreign Aff airs, Yaoundé, Cameroon (G Laborde-Balen MSc); Institut Interpretation Laboratory monitoring was not cost eff ective in 2006–10 compared with clinical monitoring when the de Recherche pour le Abbott RealTime HIV-1 assay was used according to the $3670 cost-eff ectiveness threshold (three times per-person Développement (IRD), GDP in Cameroon), but it might be cost eff ective if a generic in-house assay is used. University Montpellier 1, UMI 233, Montpellier, France (A Fobang Aghokeng PhD, Funding French National Agency for Research on AIDS and Viral Hepatitis (ANRS) and Ensemble pour une Solidarité Prof E Delaporte MD, Thérapeutique Hospitalière En Réseau (ESTHER). C Laurent PhD); Virology Laboratory IRD/IMPM/

3 CREMER, Yaoundé, Cameroon Introduction second-line regimens, virological monitoring is still very (A Fobang Aghokeng, 4 Although access to antiretroviral therapy (ART) is still not rare in low-income and middle-income countries. E Mpoudi-Ngole MD); Faculté de universal in low-income and middle-income coun tries, In the coming years, a large proportion of patients Médecine et des Sciences large increases in donor funding in the past 10 years have receiving ART are expected to have therapeutic failure, Biomédicales, University of Yaoundé 1, Yaoundé, led to substantial progress, with about 8 million people in leading to a higher risk of not only HIV-related morbidity Cameroon (S Koulla-Shiro); and these countries receiving ART at the end of 2011.1 and mortality but also HIV transmission from drug- Department of Infectious and Widespread and rapid scale-up of ART has also been resistant strains. Evidence has shown that viral load Tropical Diseases, University supported by a public health approach that promotes the monitoring identifi es patients with ART failure more Hospital, Montpellier, France (E Delaporte) initiation and monitoring of ART based on clinical criteria accurately than does clinical or immuno logical moni- 2 5–7 Correspondence to: and, when available, on CD4 cell count. Although the toring, and does so earlier, which in turn leads to Dr Sylvie Boyer, INSERM, U912 2010 revised WHO guidelines now recommend the use of earlier use of second-line therapies (if available), a (SESSTIM), 13006 Marseille, HIV viral load, if available, to detect or confi rm treatment shorter duration spent with virological failure, and France failure and to inform decisions about switching to decreased drug resistance.6,8,9 [email protected] www.thelancet.com/infection Vol 13 July 2013 577 Articles

Accordingly, the use of laboratory monitoring, in- in terms of im munological recovery was not shown. cluding viral load measurement, to optimise treatment Furthermore, 13 (6%) of 221 participants who were outcomes is becoming increasingly topical.10,11 Ran- assigned to both laboratory and clinical monitoring domised trials have shown that routine monitoring of switched to second-line regimens, whereas nobody in the CD4 cell count, with or without viral load measurement, clinical monitoring alone group did (p<0·0001). Other provides tangible clinical benefi ts in terms of mortality, predefi ned sec ondary outcomes, including viral disease progression, and immune recovery compared suppression, mortality, disease progression, HIV with clinical monitoring alone.12–14 In the Stratall trial12— resistance, and adherence, were very similar in both which assessed ART monitoring strategies at the district groups. level in Cameroon—non-inferiority of clinical monitoring Because available resources for HIV programme alone compared with laboratory and clinical monitoring funding are decreasing, one question that needs to be addressed is whether the benefi ts from laboratory monitoring are large enough to justify the associated LAB (n=188) CLIN (n=197) extra costs. With the exception of one study,15 investi- Sex gations assessing the cost-eff ectiveness of laboratory Women 135 (71·8%) 142 (72·1%) monitoring, including viral load, in low-income coun- Men 53 (28·2%) 55 (27·9%) tries have been based on mathematical models16–21 and Age (years) 37·5 (31·0–45·1) 35·9 (30·0–44·8) shown contradictory results.22,23 Additionally, until now, Marital status* no cost-eff ectiveness study had been done in the context Married or cohabiting 53 (29·0%) 68 (35·1%) of decentralised ART delivery with local non-specialist Divorced or separated 7 (3·8%) 4 (2·1%) health workers, a setting that represents the reality of Widowed 25 (13·7%) 25 (12·9%) health care in low-income settings. We therefore aimed Single 98 (53·6%) 97 (50·0%) to evaluate the cost and cost-eff ectiveness of the Level of education* laboratory and clinical monitoring approach versus the No formal education 4 (2·2%) 5 (2·6%) clinical approach alone assessed in the Stratall trial. Primary school 82 (44·8%) 93 (47·9%) Secondary school or higher 97 (53·0%) 96 (49·5%) Methods Head of the household† Stratall trial design, participants, and procedures No 82 (45·8%) 93 (49·2%) The Stratall trial was a randomised, controlled, open- Yes 97 (54·2%) 96 (50·8%) label, non-inferiority trial done in nine district hospitals in Cameroon. Its design is described in detail elsewhere.12 Active‡ Briefl y, adults infected with HIV, who were treatment- No 74 (44·3%) 73 (42·7%) naive and classifi ed as WHO clinical stages 3 or 4, or Yes 93 (55·7%) 98 (57·3%) WHO clinical stage 2 with a total lymphocyte count of Water supply§ fewer than 1200 cells per μL, were enrolled from May 23, Private tap or mineral water 22 (12·2%) 31 (16·2%) 2006, to Jan 31, 2008, and followed up for 24 months after Public tap 15 (8·3%) 9 (4·7%) ART initiation. At enrolment, participants were randomly Fitted well or bored well with pump 104 (57·8%) 106 (55·5%) allocated (1:1) to receive either laboratory and clinical Purchased or unfi tted well or spring 35 (19·4%) 38 (19·9%) monitoring (LAB; measurement of viral load and CD4 Other 4 (2·2%) 7 (3·7%) cell count plus clinical monitoring) or clinical monitoring WHO clinical stage alone (CLIN). Both groups underwent haematological Stage 2 0 1 (0·5%) and biochemical tests at baseline and during follow-up. Stage 3 140 (74·5%) 147 (74·6%) Clinical monitoring was done in both the LAB and Stage 4 48 (25·5%) 49 (24·9%) CLIN groups every 3 months. Treatment change was CD4 count (cells per μL) 194·5 (112·0–370·5) 177 (76·0–322·0) permitted for patients with drug-related adverse events HIV-1 viral load (log copies per mL)¶ 5·6 (5·0–6·0) 5·6 (5·3–6·1) 10 or treatment failure, defi ned in the LAB group as a viral Antiretroviral regimen load of more than 5000 copies per mL on two consecutive Stavudine, lamivudine, and nevirapine 130 (69·1%) 124 (62·9%) samples, 3 months apart, and in the CLIN group by Stavudine, lamivudine, and efavirenz 30 (16·0%) 36 (18·3%) persistent (new or recurrent) WHO stage 3 or 4 adverse Zidovudine, lamivudine, and nevirapine 12 (6·4%) 20 (10·2%) events after the fi rst 6 months. CD4 cell count Zidovudine, lamivudine, and efavirenz 16 (8·5%) 17 (8·6%) (FACSCount device, Becton Dickinson, Mountain View, Co-trimoxazole prophylaxis 179 (95·2%) 191 (97·0%) CA, USA) and plasma viral load (RealTime HIV-1 assay, Abbott Molecular, Des Plaines, IL, USA) were recorded at Data are number (%) or median (IQR). LAB=laboratory plus clinical monitoring. CLIN=clinical monitoring. *Eight missing values (n=377). †17 missing values (n=368). ‡47 missing values (n=338). §14 missing values (n=371). ¶One baseline and every 6 months thereafter. In patients with missing value (n=384). virological failure, genotypic mutations associated with antiretroviral drug resistance were assessed (Abbott Table 1: Baseline characteristics of participants followed up for ≥6 months Viroseq assay, Celera Diagnostics, Alameda, CA, USA).

578 www.thelancet.com/infection Vol 13 July 2013 Articles

We assessed CD4 cell counts, viral load, and resistance in opportunistic infections, including care delivered out side a reference HIV laboratory in Yaoundé, Cameroon. All the study sites. results for participants in the LAB group were returned We estimated costs of laboratory tests and ART, which immediately to doctors. Results for participants in the were partly subsidised by the Cameroon Ministry of CLIN group were retained at the reference laboratory Public Health and international donors, from prices until month 24. given by a national reference laboratory (the Centre The study protocol was approved by the National Pasteur du Cameroun) to value laboratory tests and the Ethics Committee (Cameroon) and the Institutional WHO Global Price Reporting Mechanism database,24 Ethics Committee of the Institut de Recherche pour le which provides information on prices and quantities of Développement (France). All participants provided antiretroviral medicines purchased by HIV/AIDS pro- written informed consent. grammes in low-income and middle-income countries (including Cameroon) to value ART costs. Outcomes measurement and study population We calculated all costs in 2008 CFA francs with the Outcomes in the cost-eff ectiveness analysis included Cameroonian Consumer Price Index25 and then costs, number of life-years saved (LYS), and incremental converted each value to US dollars with the European cost-eff ectiveness ratios (ICERs) computed over the Central Bank offi cial exchange rate for 2008 2 years of follow-up. For our base-case analysis, we used (US$1=446·00 CFA francs). For each category of data from participants in the Stratall trial who had been resources, we calculated the mean health-care use, the followed up for at least 6 months (ie, patients still alive mean related costs, and the mean and median cost per at month 6). We came to this decision mainly because patient-year for each study group. the monitoring strategy was unlikely to have aff ected We assessed survival benefi ts (ie, the number of LYS) by early mortality because baseline viral load and CD4 cell comparing the diff erence in the area under the LAB and count were not used in clinical decisions at enrolment CLIN Kaplan-Meier survival curves. We used a generalised (eg, treatment initiation time, which was based on log-rank test to compare survival times between groups. clinical criteria only, or choice of ART regimen), and the Data were censored at the 24 month visit or, if follow-up mortality rate was high and similar between the two was shorter, at the time of death or last visit.12 groups during the fi rst 6 months of ART (28 patients in We calculated ICERs by taking the diff erence in mean the CLIN group died [11·8%] vs 26 patients in the LAB costs between LAB and CLIN groups and dividing them group [11·8%]; p=0·25). Thus, because early mortality is by the diff erence in LYS. Costs and benefi ts occurring unlikely to explain diff erences between the LAB and after the fi rst year were discounted at 3·5% per year.26 We CLIN strategies, selection of patients with at least did not adjust costs to account for censoring because the 6 months of follow-up helps to place greater importance dropout rate over the study period was very low (18 [4·7%] on the specifi c eff ect of the monitoring strategy on mortality in the longer term. LAB CLIN p value* We measured costs for the complete study period—ie, Mortality from enrolment (date of ART initiation) to the censoring Deaths/participants 6/188 16/197 0·04 event. The costing perspective taken was that of the ≤200 cells per μL† 2/95 13/109 0·02 provider. We considered the following costs: inpatient >200 cells per μL† 4/93 3/88 0·76 costs, cost of antiretroviral drugs, cost of other drugs for Incidence per 100 patient-years 1·8 4·7 0·05 the treatment or prophylaxis of opportunistic infections, cost of laboratory tests, and cost of outpatient consult- ≤200 cells per μL† 1·2 7·0 0·02 ations. Thus, we calculated the total cost as the sum of all >200 cells per μL† 2·4 1·9 0·73 individual health-care resources used by participants Opportunistic infection multiplied by the unit cost of these resources. Stand- Events/participants 78/188 103/197 0·48 ardised case report forms were used to record individual ≤200 cells per μL† 51/95 68/109 0·87 clinical data and corresponding health-care resources >200 cells per μL† 27/93 35/88 0·16 consumed. Incidence per 100 patient-years‡ 21·4 27·1 0·02 We obtained unit costs of health-care resources—apart ≤200 cells per μL† 27·3 32·5 0·20 from subsidised resources (ie, laboratory tests and anti- >200 cells per μL† 15·2 20·4 0·06 retroviral drugs)—from trial accounting. To evaluate LAB=laboratory plus clinical monitoring. CLIN=clinical monitoring. *Numbers of routine outpatient consultation costs, we considered the mortality events were compared using the log-rank test, morbidity and mortality fees paid for a consultation with a nurse and a doctor in incidences were compared using a Cox regression model, and numbers of opportunistic infection events (WHO stage 1, 2, 3, or 4) were compared using a each of the study sites. We estimated the cost of inpatient Poisson regression model with time exposure included as an off set. †Baseline CD4 care for participants admitted to hospital from detailed cell count. ‡Data were censored at the time that the given event fi rst occurred. information about the prices and quantity of individual health-care resources used during hospital stay. We used Table 2: Mortality and incidence of opportunistic infections in participants followed up for ≥6 months the same approach to calculate the cost of treatment for www.thelancet.com/infection Vol 13 July 2013 579 Articles

of 385 participants) and was unlikely to have aff ected the Economic and sensitivity analysis results.27 Furthermore, no adjustment was made to the In the base-case analysis, we estimated ICERs based on model to take into account variability in the cost- two diff erent scenarios for viral loads and CD4 cell counts eff ectiveness across the nine diff erent sites, because we that diff ered according to the technology used for viral noted low variability between hospitals, both in terms of load measurement. In both scenarios, CD4 cell count cost survival and cost. We estimated 95% CIs for ICERs with was $28, but in scenario 1, the unit cost of the viral load Fieller’s method,28 which is better at estimation than are measurement (RealTime HIV-1 assay, Abbott Molecular) bootstrap methods when the mean benefi ts diff erence of was $67, which when added to the cost of the CD4 cell See Online for appendix the ICER is close to zero.29 count, equates to $95. In scenario 2, viral load unit cost was $35 with a generic real-time assay (Biocentric, Base-case analysis One-way Bandol, France), giving a total cost of $63. The unit cost sensitivity for both technologies included consumables, staff , and analysis capital, mainly for the testing equipment. The generic Cost Sources Cost Sources real-time assay was developed by the French National ARV (unit cost per patient per year) Agency for Research on AIDS and Viral Hepatitis (ANRS) First-line therapy in response to the inability of standard assays to accurately Stavudine, lamivudine, and efavirenz $374 WHO database*24 $80† MSF38 measure viral load in non-B subtype and recombinant Stavudine, lamivudine, and nevirapine $149 WHO database*24 $56† MSF38 samples from west African patients. It was subsequently 30–32 Tenofovir, lamivudine, and efavirenz $294 WHO database*24 $111† MSF38 commercialised by Biocentric as a low-cost reagent kit. Tenofovir, lamivudine, and nevirapine $515 WHO database*24 $96† MSF38 Although the RealTime HIV-1 assay was used exclusively 12 Zidovudine, lamivudine, and efavirenz $424 WHO database*24 $139† MSF38 during the Stratall trial to measure viral load, we took the opportunity to assess the cost-eff ectiveness of the lower Zidovudine, lamivudine, and nevirapine $210 WHO database*24 $119† MSF38 cost Biocentric technology, which has been available in Tenofovir, emtricitabine, and efavirenz $502 WHO database*24 $137† MSF38 Yaoundé, Cameroon, since 2004. Second-line therapy The eff ect of cost variability on the cost-eff ectiveness Tenofovir, lamivudine, and ritonavir/ $767 WHO database*24 $435† MSF38 lopinavir results was assessed by deterministic one-way sensitivity Tenofovir, abacavir, and ritonavir/ $1023 WHO database*24 $594† MSF38 analyses that we did for both of the scenarios outlined. lopinavir First, we varied individual variable values such as the Zidovudine, lamivudine, and ritonavir/ $675 WHO database*24 $463† MSF38 costs of viral load plus CD4 cell count, inpatient day care, lopinavir antiretroviral drugs, and discount rate. Second, we Didanosine, abacavir, and ritonavir/ $1132 WHO database*24 $793† MSF38 examined the eff ect on the ICER estimations of the lopinavir choice of study population (restricted to patients still 24 38 Didanosine, tenofovir, and ritonavir/ $991 WHO database* $681† MSF alive at month 6) by doing an intention-to-treat analysis lopinavir including all participants with at least one follow-up visit Biological tests (unit cost) (n=459) and then considering only the period from HIV-1 viral load: Abbott RealTime HIV-1 $67 Central Hospital of Yaoundé $10–80 20,22 assay‡ month 6 of treatment until month 24, excluding all costs HIV-1 viral load: ANRS generic HIV-1 $35 Centre Pasteur du Cameroun ·· ·· and events that occurred in the fi rst 6 months (n=385). real-time RT-PCR test§ Finally, we considered only the subgroup of participants CD4 cell count: BD FACSCount device $28 Centre Pasteur du Cameroun $5† 15,19,39 with a CD4 cell count of fewer than 200 cells per μL at Haematology test: Hospitex Hema Screen 18 $13 Centre Pasteur du Cameroun ·· ·· ART initiation (n=204). 33 Pregnancy test $12 Centre Pasteur du Cameroun ·· ·· In accordance with guidelines suggested by WHO, we Biochemical tests: Hospitex Mini Screen p¶ $41 Centre Pasteur du Cameroun ·· ·· regarded an ICER of lower than three times the gross Resistance: Abbott Viroseq assay $220 CREMER ·· ·· domestic product (GDP) per person (ie, $3670–3800 in Follow-up and clinical consultation costs (unit cost) Cameroon, based on two sources for the yearly 2008 per- 25,34 Physician $2·2 Study sites ·· ·· person GDP at nominal values ) to suggest cost- Nurse $1·3 Study sites ·· ·· eff ectiveness. This cost-eff ectiveness threshold, which Hospital stay (cost per day) $14·1 Stratall trial $7–21 12 represents the Cameroonian Government’s ability to pay, would be applicable in a context in which health-care Costs are given in US dollars at 2008 prices. ARV=antiretroviral drugs. MSF=Médecins Sans Frontières. budgets could be extended so that treatment of all eligible CREMER=Centre de Recherche sur les Maladies Émergentes et Ré-émergentes. ANRS=French National Agency for Research on AIDS and Viral Hepatitis. *WHO Global Price Reporting Mechanism database,24 which provides patients under either monitoring strategy might be information on transaction prices and quantities of antiretroviral medicines, and tuberculosis and malaria drugs aff ordable. We also used an alternative cost-eff ectiveness purchased from April, 2005, by HIV/AIDS malaria and tuberculosis programmes in low-income and middle-income threshold of $2385, which represents the value for money countries; we calculated mean monthly prices of antiretroviral drug purchases for 2008 for each of the fi rst-line and of providing ART with clinical monitoring alone compared second-line regimens prescribed in the Stratall trial. †Upper bound (the lower bound is the value of the base-case 35,36 analysis). ‡Viral load test used in scenario 1. §Viral load test used in scenario 2. ¶Includes transaminase, amylase, with not providing ART in Cameroon (appendix). This creatinine, glycaemia, cholesterol, and triglycerides. alternative threshold represents the situation in Cameroon, in which strong budgetary constraints— Table 3: Unit costs for health-care resources in district hospitals in Cameroon resulting in fi xed health-care budgets—might mean that

580 www.thelancet.com/infection Vol 13 July 2013 Articles

the government will have to choose between initiating patient-year—ie, the LAB strategy cost 58% more than treatment with clinical monitoring alone or using clinical the CLIN approach. In the LAB strategy, the main cost and laboratory monitoring for patients receiving ART. We items were biological tests (52% of the total cost), calculated the probability of cost-eff ectiveness at diff erent followed by antiretroviral drugs, whereas in the CLIN threshold values—ie, the cost-eff ectiveness acceptability strategy, the main cost items were antiretroviral drugs curve—with a parametric technique based on the net- (47% of the total cost), followed by biological tests benefi t approach.26,37 (table 4). The mean number of inpatient days, the proportion of participants admitted to hospital, the mean Role of the funding source cost of a hospital stay, and the mean cost of opportunistic The sponsors of the study had no role in study design infection drugs were quite similar for both strategies. or implementation, collection, analysis, or interpretation In scenario 2 (generic in-house assay), the mean cost of the data, or preparation, reviewing, or approval of the of biological tests dropped to $296 (SD 76) per patient- report. The corresponding author had full access to all year in the LAB group, resulting in a total cost of the data in the study and fi nal responsibility for the $643·2 (253·6) per patient-year—ie, a 43% higher cost decision to submit for publication. than with CLIN. In the base-case analysis, the LAB strategy saved Results 0·103 life-years at an additional mean cost of $489 (SD 430) The Stratall trial randomly allocated 493 participants to either LAB or CLIN; 221 (93%) of 237 in the LAB group LAB (n=188) CLIN (n=197) p value* and 238 (93%) of 256 in the CLIN group were monitored Inpatient care by their assigned approach. The most common reasons Number of inpatient bed days 3·4 (12·0) 3·4 (9·5) 0·06 for participants not undergoing the allocated monitoring Patients admitted to hospital 32 (17·0%) 49 (24·9%) 0·06 intervention were death (six assigned to LAB and nine Patients with repeat hospital stays 9 (4·8%) 13 (6·6%) 0·44 assigned to CLIN), refusal (four in LAB and fi ve in CLIN), Cost per patient-year $37·4 (168·2) $37·9 (122·7) 0·07 and loss to follow-up (three in each group; appendix).12 Doctors, outpatient consultations During the fi rst 6 months, 26 deaths and seven losses to follow-up were reported in the LAB group compared Number of consultations 9·4 (1·7) 6·3 (1·5) <0·0001 with 28 deaths and 13 losses to follow-up in the Cost per patient-year $11·1 (2·1) $7·5 (1·7) <0·0001 CLIN group. Our base-case analysis therefore included Nurses, outpatient consultations 385 participants with at least 6 months of follow-up who Number of consultations 1·0 (1·2) 4·0 (1·4) <0·0001 underwent their assigned intervention (188 in the LAB Cost per patient-year $0·7 (0·8) $2·8 (1·0) <0·0001 group and 197 in the CLIN group). Baseline demo- Biological tests graphics and socioeconomic and clinical characteristics Viral load (scenario 1†) were similar between the two groups (table 1). Overall Number 4·7 (0·9) 0·2 (0·5) <0·0001 follow-up was 358·5 person-years in the LAB group and Cost per patient-year $165·4 (28·3) $2·8 (1·0) <0·0001 374·7 person-years in the CLIN group. CD4 cell count Six participants in the LAB group (3·2%) and 16 in the Number 4·7 (0·9) 0·2 (0·5) <0·0001 CLIN group (8·1%) died after the fi rst 6 months of Cost per patient-year $69·0 (12·2) $2·6 (6·5) <0·0001 treatment (p=0·04; table 2). Most deaths after the fi rst Resistance tests 6 months in the CLIN group were in participants with Number 0·1 (0·3) 0·0 <0·0001 baseline CD4 cell counts of fewer than 200 cells per μL Cost per patient-year $14·0 (47·3) $0·0 <0·0001 (13 [81%] of 16 deaths vs two [33%] of six in the LAB All laboratory tests group). Morbidity showed a similar pattern (ie, it was Cost per patient-year $375·6 (87·9) $117·6 (32·4) <0·0001 higher in the CLIN group than in the LAB group) Antiretroviral drugs although the diff erences between groups were not as Cost per patient-year $229·0 (106·7) $216·2 (104·5) 0·11 high as for mortality. Discounted mean survival time Drugs for opportunistic infections during follow-up was 797·0 days (95% CI 785·9–808·2) Cost per patient-year $68·7 (75·4) $75·5 (100·6) 0·39 in the LAB group and 759·5 days (748·8–770·3) in the Total cost for scenario 1† CLIN group. This 37 day diff erence corresponds to a Cost per patient-year $722·6 (258·7) $457·5 (218.9) <0·0001 signifi cant gain of 0·103 life-years in those assigned to Total cost for scenario 2‡ the LAB group at 2 year follow-up (p=0·04). Cost per patient-year $643·2 (253·6) $454·6 (218·8) <0·0001 Table 3 outlines unit costs of the main cost variables. Table 4 describes health-care use and corresponding Data are mean (SD) and number (%). Costs are given in US dollars at 2008 prices. LAB=laboratory plus clinical monitoring. CLIN=clinical monitoring. *χ² test for categorical variables and Wilcoxon rank-sum test for continuous mean costs per patient-year. In scenario 1 (Abbott variables. †Abbott RealTime HIV-1assay (viral load plus CD4 count tests cost of $95). ‡Generic in-house assay (viral RealTime HIV-1 assay), the mean cost of the LAB strategy load plus CD4 count tests cost of $63). was $722·6 (SD 258·7) per patient-year and the mean Table 4: Health-care use and cost of care per patient-year in participants followed up for ≥6 months cost of the CLIN strategy was $457·5 (218·9) per www.thelancet.com/infection Vol 13 July 2013 581 Articles

When the cost of viral load plus CD4 count decreased to Mean survival Mean cost* per patient ICER (95% CI), cost* per life-year 15,19,38 years, discounted saved $15, the ICER dropped to $1527 per LYS, and when it increased to $108,20,22 the ICER rose to $6998 per LYS. Scenario 1† Scenario 2‡ Scenario 1† Scenario 2‡ Conversely, when taking the real reduction in anti- Base case analysis§ ·· ·· ·· $4768 $3339 retroviral drug costs into account by using 2012 prices (n=385) (3926–5613) (2507–4173) reported by Médecins Sans Frontières,39 we noticed only a LAB 2·182 (0·030) $1371 (404) $1219 (395) ·· ·· small reduction in ICERs ($4704 per LYS in scenario 1 CLIN 2·079 (0·046) $882 (453) $876 (453) ·· ·· and $3275 per LYS in scenario 2). This fi nding was a Intention-to-treat ·· ·· ·· $4721 $3368 result of the small diff erences in antiretroviral drug costs analysis¶ (n=459) (3742–5707) (2425–4314) between the two strategies, because only a few participants LAB 1·939 (0·055) $1216 (530) $1081 (496) ·· ·· switched to second-line regimens. Likewise, increasing CLIN 1·843 (0·060) $764 (489) $759 (487) ·· ·· (or decreasing) the cost of inpatient day care by 50% (to Data are mean (SD) or ICER (95% CI). ICER=incremental cost-eff ectiveness ratio. LAB=laboratory plus clinical $7–21) to take into account any possible underestimation monitoring. CLIN=clinical monitoring. *Costs are given in US dollars at 2008 prices. †Viral load plus CD4 count tests (or overestimation) had little eff ect on ICERs (range cost US$95. ‡ Viral load plus CD4 count tests cost $63. §Includes only participants followed up for at least 6 months. ¶Includes all randomly assigned participants who had attended at least one follow-up visit. $4709–4828 in scenario 1 and $3280–3399 in scenario 2) since the number of inpatient days was quite similar for Table 5: Survival, costs, and incremental cost-eff ectiveness ratios according to two scenarios using the two groups and inpatient costs were relatively low. diff erently priced technology to measure viral load When we analysed data from all participants with at least one follow-up visit (intention-to-treat analysis, n=459), we per patient in scenario 1, translating into an ICER of noted that the number of LYS (LAB group LYS minus $4768 (95% CI 3926–5613) per LYS (table 5). In scenario 2, CLIN group LYS=0·096) and ICERs ($4721 per LYS in the additional mean cost was $343 (SD 425) per patient— scenario 1 and $3368 per LYS in scenario 2) were quite ie, an ICER of $3339 (2507–4173). similar to those reported in the base-case analysis. As expected, the results of one-way sensitivity analyses However, when we analysed only the data obtained from (fi gure 1) for the two scenarios show that ICER results month 6 of treatment onwards (n=385), the eff ect on the were very sensitive to the cost of laboratory monitoring. ICERs was stronger, with values dropping to $3688 per

A Viral load plus CD4 tests=$95 ICER=$4768

Viral load plus CD4 test costs ($15–108)

2012 ART costs

Inpatient costs ($7–21)

Discount rate of cost and benefits (0–6%)

≤200 CD4 cells per μL at ART initiation (n=204)

Analysis starting on month 6 of ART (n=385)

Intention-to-treat analysis (n=459)

B Viral load plus CD4 tests=$63 ICER=$3339 Viral load plus CD4 test costs ($15–108)

2012 ART costs

Inpatient costs ($7–21)

Discount rate of cost and benefits (0–6%)

≤200 CD4 cells per μL at ART initiation (n=204)

Analysis starting on month 6 of ART (n=385)

Intention-to-treat analysis (n=459)

$1000 $2000 $3000 $4000 $5000 $6000 $7000 ICER (cost per life-year saved)

Figure 1: One-way sensitivity analysis of the incremental cost-eff ectiveness ratio Laboratory monitoring with Abbott RealTime HIV-1 assay plus CD4 count cost of $95 (A) versus generic in-house viral load assay plus CD4 count cost of $63 (B). The horizontal axis shows variations in the incremental cost-eff ectiveness ratio (ICER), in US dollars at 2008 prices per life-year saved, due to changes in variables listed on the vertical axis. Values in parentheses specify the upper and lower bounds assessed for each variable. Dashed vertical lines show the ICER of laboratory monitoring in the base-case analysis at a viral load plus CD4 count cost of either $95 (A) or $63 (B). ART=antiretroviral therapy.

582 www.thelancet.com/infection Vol 13 July 2013 Articles

LYS in scenario 1 and $2664 per LYS in scenario 2. This $2385 $3670 $3800 fi nding could be because most of the laboratory tests to 100 Scenario 1 monitor toxicity in the CLIN group were done during the Scenario 2 90 fi rst 6 months of ART, resulting in lower cost diff erences between LAB and CLIN from month 6 onwards, whereas 80 survival benefi ts during this period remained unchanged. 70 Finally, when we analysed data from the subgroup of participants with a CD4 cell count of fewer than 200 cells 60 per μL at ART initiation (n=204), ICERs reduced to 50 $3318 per LYS in scenario 1 and $2324 per LYS in scenario 2 because of a substantial increase in the number of LYS 40 (LAB group LYS minus CLIN group LYS=0·152). Figure 2 shows the probability of laboratory moni toring 30 Probability of cost-effectiveness (%) being cost eff ective as a function of the cost-eff ectiveness 20 thresholds per LYS in both scenarios. Scenario 1 is associated with a 0·5% probability of cost-eff ectiveness at 10 the $3670 threshold, which corresponds to three times the 0 GDP per person (lower limit), whereas in scenario 2, this $1000 $1500 $2000 $2500 $3000 $3500 $4000 $4500 $5000 $5500 $6000 probability is 78·2%. This country-specifi c cost-eff ective- Threshold of cost-effectiveness ness threshold ($3670) is reached at a CD4 cell count plus Figure 2: Cost-eff ectiveness acceptability curves of laboratory monitoring viral load total cost of $69. At the $2385 threshold, which Probability of laboratory monitoring being cost eff ective as a function of the cost-eff ectiveness threshold per corresponds to the ICER of clinical monitoring alone life-year saved in scenario 1 (CD4 cell count plus Abbott RealTime HIV-1 assay cost of $95) and scenario 2 (CD4 cell compared with no provision of ART, the probability of cost- count plus generic real-time assay cost of $63). Costs are given in US dollars at 2008 prices. Dashed vertical lines represent diff erent cost-eff ectiveness thresholds: the $2385 threshold corresponds to the incremental eff ectiveness is lower than 1·5% in both scenarios. This cost-eff ectiveness ratio (ICER) of providing antiretroviral therapy (ART) with clinical monitoring alone compared $2385 cost-eff ectiveness threshold is met at a CD4 cell with not providing ART; the $3670 threshold corresponds to three times the gross domestic product (GDP) per count plus viral load total cost of $20. person in Cameroon, lower end of range ($3670–$3800); and the $3800 threshold corresponds to three times the Because we noted greater survival benefi ts of LAB in GDP per person in Cameroon, upper end of range. those with a CD4 count of 200 cells per μL or fewer, we did a scenario analysis to examine the cost-eff ectiveness $2385 per LYS (which represents the value for money of of an alternative moni toring strategy not directly providing ART with clinical monitoring alone compared assessed in the Stratall trial. This hypothetical strategy with not providing ART), laboratory monitoring with a would combine laboratory and clinical monitoring for generic assay is not cost eff ective. In the present global patients beginning ART with a CD4 count of 200 cells economic context, in which Cameroon, like other sub- per μL or fewer with clinical monitoring alone for Saharan countries, is having to cope with budget con- patients with a CD4 count of more than 200 cells per μL straints that are threatening ART delivery, the use of this at ART initiation. When compared with the original alternative threshold to inform decisions about whether CLIN strategy, this hypothetical strategy would save to adopt laboratory monitoring might be more relevant 0·120 life-years at an additional mean cost of for programme managers than is the WHO-recom- $259 (SD 506) per patient in scenario 1 and $181 (485) mended cost-eff ectiveness threshold. per patient in scenario 2, translating into an ICER of Furthermore, our results show that the cost-eff ective- $2167 (95% CI 1314–3020) per LYS in scenario 1 and ness of laboratory monitoring strongly improves for $1510 (692–2329) per LYS in scenario 2. In scenario 1, participants beginning ART with fewer than 200 CD4 the probability of being cost eff ective would be 99·9% at cells per μL, because more life-years are saved. Our the $3670 threshold, but only 69·2% at the scenario analysis suggests that a combined monitoring $2385 threshold, whereas scenario 2 would have a 100% strategy in which only patients with fewer than 200 CD4 and 98·2% probability of cost-eff ectiveness, respectively cells per μL undergo laboratory and clinical monitoring (appendix). (and patients with >200 CD4 cells per μL undergo clinical monitoring alone), would maximise the LYS and Discussion reduce costs because laboratory costs for patients with a According to the WHO-recommended cost-eff ectiveness higher immunological status at baseline would be threshold of three times the GDP per person,33 our avoided. Our fi ndings suggest that this combined results suggest that laboratory monitoring plus clinical strategy, using either viral load assay technology, would monitoring is not cost eff ective in Cameroon compared be cost eff ective compared with clinical monitoring with clinical monitoring alone when using the Abbott according to the three times per-person GDP threshold. RealTime HIV-1 assay, but might be cost eff ective with a However, at the alternative threshold of $2385, this generic low-cost in-house test. However, when we used combined strategy would be cost eff ective only when the alternative lower cost-eff ectiveness threshold of using the generic assay. www.thelancet.com/infection Vol 13 July 2013 583 Articles

vs LYS), both studies led to similar conclusions regarding Panel: Research in context the cost-eff ectiveness of laboratory monitoring. Systematic review Additionally, our cost-eff ectiveness study was derived We searched Medline from Aug 24, 1996, to Nov 15, 2012, from the Stratall trial, which was done in a decentralised without language restrictions, for cost-eff ectiveness studies setting that represents real-world health-care delivery in derived from randomised controlled trials that compared low-income countries; this setting diff ers from those 15,36 clinical monitoring of antiretroviral therapy with laboratory used in existing studies and is contextually relevant. monitoring in low-income countries. We used the search The Stratall trial’s model of HIV care is facility-based at a terms “cost-eff ectiveness”, “antiretroviral”, “monitoring” OR decentralised district level in accordance with the WHO- “laboratory” OR “viral load” OR “CD4”, “biological” OR recommended strategy for decentralised and integrated 2 “biologic”, “resource-limited” OR “resource-poor” OR delivery of care. ART was started on the basis of WHO “resource-constrained” OR “Africa” OR “developing”, and clinical criteria alone (ie, CD4 cell count was not used) “trial”. We identifi ed two cost-eff ectiveness studies derived and viral load and CD4 cell count were measured every 3 from randomised trials: the HBAC trial in Uganda11 and the 6 months, as recommended by WHO. Additionally, DART trial in Uganda and Zimbabwe.21 We discuss the fi ndings patients were followed up by local non-specialist health- of both trials here. We also searched abstracts from major care workers and sometimes consulted by nurses instead conferences but did not identify any other relevant study. of doctors, in accordance with WHO recommendations on task shifting.2,40 Interpretation Some limitations to the study should be acknowledged. Laboratory monitoring is not cost eff ective in Cameroon First, the short study period of 2 years probably does not compared with clinical monitoring alone in the current account for the full benefi ts of laboratory monitoring, context of fi xed health-care budgets. The cost-eff ectiveness especially because participants were ART naive. The of laboratory monitoring could, however, be improved by use Stratall clinical study clearly showed that laboratory of generic tests to measure viral load. Additionally, the monitoring resulted in earlier detection of treatment cost-eff ectiveness of the laboratory strategy strongly failure.12 If the longer-term eff ects of earlier treatment improved in patients starting ART with a CD4 cell count of failure detection on patients’ survival were taken into fewer than 200 cells per μL, because more life-years are saved. account, the cost-eff ectiveness of the laboratory strategy These fi ndings suggest that targeted laboratory monitoring would improve. A second limitation is intrinsic to the with generic assays in patients beginning ART with a low study design, which did not include diff erent confi g- immunological status (ie, <200 CD4 cells per μL) could be a urations of laboratory monitoring (eg, immunological valuable strategy even in a context in which budgets are and clinical monitoring or clinical and virological fi xed. Our fi ndings also underline the need for the monitoring). The HBAC trial, which followed up development of low-cost generic tests that are easy to use in patients for 3 years after ART initiation, reported no low-resource settings with infrastructure constraints. additional benefi t of viral load monitoring when it was added to CD4 count and clinical monitoring.13 However, in our study, it was not possible to disentangle the Two other cost-eff ectiveness studies derived from eff ects of viral load monitoring from those of CD4 randomised trials and comparing diff erent strategies to monitoring and to assess the potential specifi c benefi ts monitor adults taking ART in low-income countries have of one compared with the other. Third, our study been recently published: the Development of population was restricted to participants who had been AntiRetroviral Therapy (DART) trial in Uganda and followed up for at least 6 months. We excluded early Zimbabwe36 and the Home-Based AIDS Care (HBAC) mortality to take greater account of the specifi c eff ect of trial in Uganda (panel).13,15 Compared with clinical the monitoring strategy on mortality. However, the fact monitoring alone, the DART trial showed that CD4 that we did not fi nd any signifi cant diff erences in the monitoring was not cost eff ective, with a mean survival baseline characteristics of the two groups suggests that benefi t of 0·112 life-years (41 days) at an additional mean this choice did not introduce any selection bias. cost of $765—ie, an ICER of $7386 per LYS. Viral load Furthermore, as suggested by the one-way sensitivity testing, on which most of the debate regarding ART analysis in which we considered all partici pants with at monitoring is centred, was investigated only in the HBAC least one follow-up visit (intention-to-treat analysis, trial. The investigators concluded that clinical and CD4 n=459), excluding the fi rst 6 months of follow-up had cell monitoring compared with clinical moni toring alone little eff ect on the ICER results and did not modify our (ICER of $174 per disability-adjusted life-year [DALY]) was cost-eff ectiveness con clusion. Fourth, we cannot exclude more cost eff ective than when combined with viral load the fact that attitudes and behaviours of health workers monitoring (ICER of $5181 per DALY). Although the and patients might have been aff ected by the open magnitude of the ICERs reported in the HBAC and nature of the trial. However, our personal observations Stratall studies are not directly comparable, especially in clinics suggest that this potential limitation probably because of the use of diff erent eff ectiveness criteria (DALY did not occur. Finally, we did not consider the

584 www.thelancet.com/infection Vol 13 July 2013 Articles

transportation costs needed for laboratory tests to be S Ngo Yebga, C Paul Bile (IRSA, Catholic University of Central Africa, done in the reference laboratory located in Yaoundé, Yaoundé, Cameroon); S Abada, T Abanda, J Baga, P Bilobi Fouda, Cameroon, because they were relatively marginal com- P Etong Mve, G Fetse Tama, H Kemo, A Ongodo, V Tadewa, H D Voundi (District Hospital, Ayos, Cameroon); A Ambani, M Atangana, J-C Biaback, pared with the total cost of CD4 cell count and viral load M Kennedy, H Kibedou, F Kounga, M Maguip Abanda, E Mamang, testing (about $1 per test; <1·6% of the total cost). A Mikone, S Tang, E Tchuangue, S Tchuenko, D Yakan (District Hospital, Decentralising viral load testing throughout the whole Bafi a, Cameroon); J Assandje, S Ebana, D Ebo’o, D Etoundi, G Ngama, territory (eg, in regional hospitals) would, however, P Mbarga Ango, J Mbezele, G Mbong, C Moung, N Ekotto, G Nguemba Balla, G Ottou, M Tigougmo (District Hospital, Mbalmayo, involve additional costs related to staffi ng and training Cameroon); R Beyala, B Ebene, C Eff emba, F Eyebe, M-M Hadjaratou, that would in turn reduce the cost-eff ectiveness of T Mbarga, M Metou, M Ndam, B Ngoa, E B Ngock, N Obam (District laboratory monitoring. Hospital, Mfou, Cameroon); A M Abomo, G Angoula, E Ekassi, Essama, J J Lentchou, I Mvilongo, J Ngapou, F Ntokombo, V Ondoua, R Palawo, To improve the cost-eff ectiveness of laboratory moni- S Sebe, E Sinou, D Wankam, I Zobo (District hospital, Monatélé, toring, market barriers need to be overcome to bring Cameroon); B Akono, A L Ambani, L Bilock, R Bilo’o, J Boombhi, about a large reduction in the cost of viral load testing. F X Fouda, M Guitonga, R Mad’aa, D R Metou’ou, S Mgbih, A Noah, This market is dominated by one company, Roche, with M Tadena, Ntcham (District Hospital, Nanga Eboko, Cameroon); G Ambassa Elime, A A Bonongnaba, E Foaleng, R M Heles, R Messina, limited competition from Siemens, bioMérieux, and O Nana Ndankou, S A Ngono, D Ngono Menounga, S S Sil, L Tchouamou, 4 Abbott. Products off ered by these companies rely on B Zambou (District Hospital, Ndikinimeki, Cameroon); R Abomo, closed systems that also act to impede competition J Ambomo, C Beyomo, P Eloundou, C Ewole, J Fokom, M Mvoto, because only the manufacturer’s own branded reagents M Ngadena, R Nyolo, C Onana, A Oyie (District Hospital, Obala, Cameroon); P Antyimi, S Bella Mbatonga, M Bikomo, Y Molo Bodo, or cartridges can be used after the instrument is S Ndi Ntang, P Ndoudoumou, L Ndzomo, S O Ngolo, M Nkengue, Nkoa, purchased. High-entry barriers and low competition in Y Tchinda (District Hospital, Sa’a, Cameroon). this market have resulted in marginal price decreases Confl icts of interest throughout the past decade.4,22 Several strategies have We declare that we have no confl icts of interest. been identifi ed to reduce prices, including price Acknowledgments transparency and the development of open platforms We thank our partners at the Ministry of Public Health in Cameroon, that stimulate competition.10 So far, the only available all the patients and staff of the district hospitals who participated in the generic testing open platform is that assessed in our study, and Jude Sweeney (Rome, Italy) for the English revision and editing of the manuscript. The study was supported by a grant from the 10,41,42 study, commercialised as a low-cost reagent kit. French National Agency for Research on AIDS and Viral Hepatitis Open systems enable diagnosis of various infectious (ANRS 12110) and Ensemble pour une Solidarité Thérapeutique diseases with the same molecular technology but Hospitalière En Réseau (ESTHER). diff erent reagent kits or cartridges. This technology References might also be suitable for low-income settings outside 1 UNAIDS. Global report: UNAIDS report on the global AIDS epidemic 2012. Geneva: Joint United Nations Programme on HIV/AIDS, 2012. of national laboratories, such as regional hospitals, 2 Gilks CF, Crowley S, Ekpini R, et al. The WHO public-health because it has a small laboratory footprint and can be approach to antiretroviral treatment against HIV in done by adequately trained technicians. In rural resource-limited settings. Lancet 2006; 368: 505–10. 3 WHO. Antiretroviral therapy for HIV infection in adults and settings, however, the implementation of viral load adolescents: recommendations for a public health approach testing will only be possible if even simpler technology (2010 revision). Geneva: World Health Organization, 2010. http:// becomes available at a low cost.10,41 Our fi ndings suggest whqlibdoc.who.int/publications/2010/9789241599764_eng.pdf (accessed Jan 4, 2012). that eff orts should focus on overcoming market barriers 4 Médecins Sans Frontières. Undetectable—how viral load and providing tests that are low cost and easy to use to monitoring can improve HIV treatment in developing countries. improve the cost-eff ectiveness of laboratory monitoring Geneva: Médecins Sans Frontières, 2012. http://www.msfaccess. org/content/undetectable-how-viral-load-monitoring-can-improve- in the near future. hiv-treatment-developing-countries (accessed Oct 31, 2012). Contributors 5 Reynolds SJ, Nakigozi G, Newell K, et al. Failure of immunologic CK, ED, and CL designed and coordinated the Stratall clinical trial. SB and criteria to appropriately identify antiretroviral treatment failure in J-PM designed and coordinated the cost-eff ectiveness analysis. SB Uganda. AIDS 2009; 23: 697–700. contributed to the data analysis and wrote the fi rst draft of the manuscript. 6 Hosseinipour MC, Van Oosterhout JJG, Weigel R, et al. The public LM did the statistical analysis. CK and GL-B were responsible for the health approach to identify antiretroviral therapy failure: high-level implementation of the Stratall clinical trial in the fi eld. SB, GL-B, and SK-S nucleoside reverse transcriptase inhibitor resistance among contributed to collection of data used in the cost-eff ectiveness study. AFA Malawians failing fi rst-line antiretroviral therapy. AIDS 2009; 23: 1127–34. did the main laboratory analysis in the Stratall clinical trial. All authors contributed to the interpretation of data and reviewed the manuscript. 7 Rawizza HE, Chaplin B, Meloni ST, et al. Immunologic criteria are poor predictors of virologic outcome: implications for HIV Stratall ANRS 12110/ESTHER Study Group treatment monitoring in resource-limited settings. Clin Infect Dis M Biwolé-Sida, C Kouanfack, S Koulla-Shiro (Central Hospital, Yaoundé, 2011; 53: 1283–90. Cameroon); A Bourgeois, E Delaporte, C Laurent, M Peeters (IRD, 8 Sigaloff KCE, Hamers RL, Wallis CL, et al. Unnecessary University Montpellier 1, UMI 233, Montpellier, France); G Laborde-Balen antiretroviral treatment switches and accumulation of HIV (French Ministry of Foreign Aff airs, Yaoundé, Cameroon); M Dontsop, resistance mutations; two arguments for viral load monitoring in S Kazé, J-M Mben (IRD, Yaoundé, Cameroon); A Aghokeng, M G Edoul, Africa. J Acquir Immune Defi c Syndr 2011; 58: 23–31. E Mpoudi-Ngolé, M Tongo (Virology Laboratory, IMPM/CREMER/IRD- 9 Gupta RK, Hill A, Sawyer AW, et al. Virological monitoring and UMI 233, Yaoundé, Cameroon); S Boyer, M P Carrieri, F Marcellin, resistance to fi rst-line highly active antiretroviral therapy in adults infected with HIV-1 treated under WHO guidelines: a systematic J-P Moatti, B Spire (INSERM, IRD, University Marseille, UMR 912, review and meta-analysis. Lancet Infect Dis 2009; 9: 409–17. Marseille, France); C Abé, S-C Abega, C-R Bonono, H Mimcheu,

www.thelancet.com/infection Vol 13 July 2013 585 Articles

10 Roberts T, Bygrave H, Fajardo E, Ford N. Challenges and 27 Glick HA. Economic evaluation in clinical trials. Oxford: Oxford opportunities for the implementation of virological testing in University Press, 2007. resource-limited settings. J Int AIDS Soc 2012; 15: 17324. 28 Willan AR, O’Brien BJ. Confi dence intervals for cost-eff ectiveness 11 Hamers RL, Kityo C, Lange JMA, De Wit TFR, Mugyenyi P. Global ratios: an application of Fieller’s theorem. Health Econ 1996; threat from drug resistant HIV in sub-Saharan Africa. BMJ 2012; 5: 297–305. 344: e4159. 29 Siani C, Moatti J-P. The handling of uncertainty in economic 12 Laurent C, Kouanfack C, Laborde-Balen G, et al, for the Stratall evaluations of health care strategies. Rev Epidemiol Sante Publique ANRS 12110/ESTHER study group. Monitoring of HIV viral loads, 2003; 51: 255–76 (in French). CD4 cell counts, and clinical assessments versus clinical monitoring 30 Rouet F, Chaix M-L, Nerrienet E, et al. Impact of HIV-1 genetic alone for antiretroviral therapy in rural district hospitals in diversity on plasma HIV-1 RNA Quantifi cation: usefulness of the Cameroon (Stratall ANRS 12110/ESTHER): a randomised Agence Nationale de Recherches sur le SIDA second-generation non-inferiority trial. Lancet Infect Dis 2011; 11: 825–33. long terminal repeat-based real-time reverse transcriptase 13 Mermin J, Ekwaru JP, Were W, et al. Utility of routine viral load, polymerase chain reaction test. J Acquir Immune Defi c Syndr 2007; CD4 cell count, and clinical monitoring among adults with HIV 45: 380–8. receiving antiretroviral therapy in Uganda: randomised trial. 31 Rouet F, Foulongne V, Viljoen J, et al. Comparison of the Generic BMJ 2011; 343: d6792. HIV Viral Load assay with the Amplicor HIV-1 monitor v1.5 and 14 DART Trial Team. Routine versus clinically driven laboratory Nuclisens HIV-1 EasyQ v1.2 techniques for plasma HIV-1 RNA monitoring of HIV antiretroviral therapy in Africa (DART): quantitation of non-B subtypes: the Kesho Bora preparatory study. a randomised non-inferiority trial. Lancet 2010; 375: 123–31. J Virol Methods 2010; 163: 253–7. 15 Kahn JG, Marseille E, Moore D, et al. CD4 cell count and viral load 32 Rouet F, Ekouevi DK, Chaix M-L, et al. Transfer and evaluation of monitoring in patients undergoing antiretroviral therapy in an automated, low-cost real-time reverse transcription-PCR test for Uganda: cost eff ectiveness study. BMJ 2011; 343: d6884. diagnosis and monitoring of human immunodefi ciency virus type 1 16 Bendavid E, Young SD, Katzenstein DA, Bayoumi AM, Sanders GD, infection in a West African resource-limited setting. J Clin Microbiol Owens DK. Cost-eff ectiveness of HIV monitoring strategies in 2005; 43: 2709–17. resource-limited settings: a southern African analysis. 33 Sachs JD. Macroeconomics and health: investing in health for Arch Intern Med 2008; 168: 1910–18. economic development. Report of the Commission on 17 Kimmel AD, Weinstein MC, Anglaret X, et al. Laboratory Macroeconomics and Health (CMH) of the World Health monitoring to guide switching antiretroviral therapy in Organization. Geneva: World Health Organization, 2001. resource-limited settings: clinical benefi ts and cost-eff ectiveness. 34 International Moneraty Fund. World economic outlook database, J Acquir Immune Defi c Syndr 2010; 54: 258–68. September 2011. http://www.imf.org/external/pubs/ft/ 18 Phillips AN, Pillay D, Miners AH, Bennett DE, Gilks CF, weo/2011/02/weodata/index.aspx (accessed Feb 6, 2012). Lundgren JD. Outcomes from monitoring of patients on 35 Ramsey S, Willke R, Briggs A, et al. Good research practices for antiretroviral therapy in resource-limited settings with viral load, cost-eff ectiveness analysis alongside clinical trials: the ISPOR CD4 cell count, or clinical observation alone: a computer simulation RCT-CEA Task Force report. Value Health 2005; 8: 521–33. model. Lancet 2008; 371: 1443–51. 36 Medina Lara A, Kigozi J, Amurwon J, et al. Cost eff ectiveness 19 Bishai D, Colchero A, Durack DT. The cost eff ectiveness of analysis of clinically driven versus routine laboratory monitoring of antiretroviral treatment strategies in resource-limited settings. antiretroviral therapy in Uganda and Zimbabwe. PLoS One 2012; AIDS 2007; 21: 1333. 7: e33672. 20 Vijayaraghavan A, Efrusy MB, Mazonson PD, Ebrahim O, 37 Löthgren M, Zethraeus N. Defi nition, interpretation and calculation Sanne IM, Santas CC. Cost-eff ectiveness of alternative strategies for of cost-eff ectiveness acceptability curves. Health Econ 2000; initiating and monitoring highly active antiretroviral therapy in the 9: 623–30. developing world. J Acquir Immune Defi c Syndr 2007; 46: 91–100. 38 Offi ce of William Jeff erson Clinton. Clinton Foundation announces 21 Hamers RL, Sawyer AW, Tuohy M, Stevens WS, De Wit TFR, landmark agreement on major reduction in price of HIV/AIDS lab Hill AM. Cost-eff ectiveness of laboratory monitoring for tests, Jan 14, 2004. http://www.bd.com/globalhealth/pdfs/Clinton_ management of HIV treatment in sub-Saharan Africa: Foundation_Press_Release.pdf (accessed Mar 9, 2012). a model-based analysis. AIDS 2012; 26: 1663–72. 39 Médecins Sans Frontrières. Untangling the web of antiretroviral 22 Walensky RP, Ciaranello AL, Park J-E, Freedberg KA, Mayer KH. price reductions 15th edn, July 2012. Geneva, Médecins Sans Cost-eff ectiveness of laboratory monitoring in sub-Saharan Africa: Frontières, 2012. utw.msfaccess.org (accessed Nov 15, 2012). a review of the current literature. Clin Infect Dis 2010; 51: 85–92. 40 WHO. The World Health Report 2006: working together for health. 23 Loubiere S, Meiners C, Sloan C, Freedberg KA, Yazdanpanah Y. Geneva: World Health Organization, 2006. http://www.who.int/ Economic evaluation of ART in resource-limited countries. whr/2006/en/ (accessed Jan 5, 2012). Curr Opin HIV AIDS 2010; 5: 225–31. 41 Wang S, Xu F, Demirci U. Advances in developing HIV-1 viral load 24 WHO. Global price reporting mechanism for HIV, tuberculosis assays for resource-limited settings. Biotechnol Adv 2010; 28: 770–81. and malaria. http://www.who.int/hiv/amds/gprm/en/ (accessed 42 Rouet F, Rouzioux C. The measurement of HIV-1 viral load in Oct 25, 2011). resource-limited settings: how and where? Clin Lab 2007; 25 The World Bank. Indicators. http://data.worldbank.org/indicator/ 53: 135–48. (accessed Dec 6, 2011). 26 Petrou S, Gray A. Economic evaluation alongside randomised controlled trials: design, conduct, analysis, and reporting. BMJ 2011; 342: d1548.

586 www.thelancet.com/infection Vol 13 July 2013