Feasibility and cost-effectiveness of standardised second-line drug treatment for chronic tuberculosis patients: a national cohort study in Peru

Transcription

1 Feasibility and cost-effectiveness of standardised second-line drug treatment for chronic tuberculosis patients: a national cohort study in Peru Pedro G Suárez, Katherine Floyd, Jaime Portocarrero, Edith Alarcón, Elisabetta Rapiti, Gilbert Ramos, Cesar Bonilla, Ivan Sabogal, Isabel Aranda, Christopher Dye, Mario Raviglione, Marcos A Espinal Summary Background There are no data on the feasibility and costeffectiveness of using second-line drugs to treat patients with chronic tuberculosis, many of whom are infected with multidrug resistant (MDR) strains of Mycobacterium tuberculosis, in low or middle-income countries. Methods A national programme to treat chronic tuberculosis patients with a directly observed standardised 18-month daily regimen, consisting of kanamycin (3 months only), ciprofloxacin, ethionamide, pyrazinamide, and ethambutol, was established in Peru in Compliance and treatment outcomes were analysed for the cohort started on treatment between October, 1997, and March, Total and average costs were assessed. Cost-effectiveness was estimated as the cost per DALY gained. Findings 466 patients were enrolled; 344 were tested for drug susceptibility and 298 (87%) had MDR tuberculosis. 225 patients (48%) were cured, 57 (12%) died, 131 (28%) did not respond to treatment, and 53 (11%) defaulted. Of the 413 (89%) patients who complied with treatment, 225 (55%) were cured. Among MDR patients, resistance to five or more drugs was significantly associated with an unfavourable outcome (death, non-response to treatment, or default; odds ratio 3 37, 95% CI ; p=0 01). The programme cost US$0 6 million per year, 8% of the National Tuberculosis Programme budget, and US$2381 per patient for those who completed treatment. The mean cost per DALY gained was $211 ($165 at drug prices projected for 2002). Interpretation Treating chronic tuberculosis patients with high levels of MDR with second-line drugs can be feasible and cost-effective in middle-income countries, provided a strong tuberculosis control programme is in place. Published online May 14, National Tuberculosis Control Programme, Ministry of Health, Lima, Peru (J Portocarrero MD, G Ramos MD, C Bonilla MD, I Sabogal MD, I Aranda RN, P G Suárez MD, E Alarcón RN); and WHO, Geneva, Switzerland (K Floyd PhD, E Rapiti MPH, C Dye Dphil, M Raviglione MD, M A Espinal MD) Correspondence to: Dr Marcos Espinal, WHO, Stop TB Department, 1211 Geneva, Switzerland ( espinalm@who.int) Introduction Multidrug-resistant (MDR) tuberculosis ie, resistance to at least rifampicin and isoniazid is a threat to tuberculosis control efforts in some countries. 1,2 Patients with MDR tuberculosis often do not respond to shortcourse chemotherapy with first-line drugs. 3 Thus, treatment with second-line drugs may be needed to cure these individuals and to curtail transmission of such strains of Mycobacterium tuberculosis. However, data from low-income or middle-income countries on the management of MDR tuberculosis with second-line drugs under programme conditions are not available. As a result, international public-health institutions, including the WHO, do not have definitive policy recommendations in place; guidelines exist only for specialised centres where resources that permit treatment are available. 4 Treatment of MDR tuberculosis is, however, available in moredeveloped countries. 5,6 After results of a survey of drug-resistant tuberculosis showed an MDR prevalence of 16% among previously treated individuals, 1 Peru s National Tuberculosis Programme introduced a standardised treatment regimen, including second-line drugs, for patients who had failed to respond to standard treatment with first-line drugs. Our aim was to assess the feasibility, effectiveness, affordability, and cost-effectiveness of implementing such treatment services. Methods Setting Peru is a middle-income country that had a per capita Gross National Income of US$2390 in The DOTS strategy was introduced in 1991 and is now implemented in all health facilities. 8,9 WHO targets for tuberculosis control (70% case detection and 85% cure rate among new smear-positive pulmonary cases) have been achieved and sustained, as a result of which the incidence rate is declining at an estimated 6% per year. 9 Before 1997, the treatment policy in Peru was limited to short-course chemotherapy with first-line drugs. All newlydiagnosed tuberculosis cases are administered a treatment regimen consisting of an intensive phase of isoniazid, rifampicin, ethambutol, and pyrazinamide for 2 months, followed by a continuation phase with rifampicin and isoniazid for 4 months. Patients who do not respond to the treatment regimen or default with this regimen, or who are cured but later relapse, are given a re-treatment regimen, which involves five drugs in the 3-month intensive phase of treatment (isoniazid, rifampicin, ethambutol, and pyrazinamide for 3 months; streptomycin for first 2 months only), and a continuation phase with rifampicin, isoniazid, and ethambutol for 5 months. In both regimens, treatment is administered daily in the intensive phase and twice a week during the continuation phase. In each instance, treatment is directly observed by health workers, and is free of charge to all patients. THE LANCET Published online May 14,

2 Participants and treatment regimen Two categories of patient were eligible for the second-line drug treatment programme introduced in October, The first consisted of patients who remained sputumsmear positive after completing a fully supervised retreatment regimen. These patients are defined as chronic tuberculosis cases 10 and were included in two subcategories. The first subcategory comprised patients who had consecutively failed to respond to the standard treatment and re-treatment regimens with first-line drugs under direct observation by a health worker. The second subcategory included patients who had been declared cured in response to the standard treatment regimen with first-line drugs, but who later relapsed and failed to respond to the standard re-treatment regimen. The second category consisted of patients defined as exceptions for example, those whose previous treatment had not been administered under direct observation, or involved regimens not recommended by the National Tuberculosis Programme. We obtained written informed consent before giving eligible patients a standardised 18-month daily regimen, consisting of kanamycin (1 g injectable), ciprofloxacin (1 g orally), ethionamide (750 mg orally), pyrazinamide (1500 mg orally), and ethambutol (1200 mg orally). Kanamycin was administered only for the first 3 months. Treatment was provided on an ambulatory basis under direct observation at a local health clinic. Procedure Patients were enrolled between October, 1997, and March, 1999, in any of the 6435 health facilities that provided tuberculosis services in Peru. An MDR tuberculosis unit was established as part of the National Tuberculosis Programme in Lima, to manage the implementation of the treatment programme. This unit was staffed by a coordinator, two physicians, a nurse, and a data manager, and was in regular contact with the tuberculosis coordinators of the 34 health regions of Peru. A training programme, consisting of two national workshops and several regional workshops, was implemented for relevant staff. An expert committee of 12 lung specialists, publichealth specialists, and laboratory specialists was established in Lima to approve or reject requests from the general health facilities to enrol patients. Requests for enrolment followed a standardised approach, which included submission of the patients medical history, previous treatments received and outcomes, and the latest smear microscopy, culture, and drug susceptibility test results. Decisions of the committee to approve or reject access to drugs were by consensus. Patients were seen daily by nurses for administration of drugs and monitoring of any adverse effects associated with treatment, and monthly by doctors for a medical check-up. To discourage defaulting from treatment, the guidelines of Peru s National Tuberculosis Programme were followed: patients were provided, for example, with an appointment card and a weekly food parcel. Baseline and monthly follow-up cultures of sputum samples from which to isolate M tuberculosis were undertaken at any of the 75 laboratories that were doing cultures in Peru between 1997 and Cultures were done on a Löwenstein- Jensen medium. Drug susceptibility testing of first-line drugs was done with the proportion method at the National Institute of Health laboratory in Lima. Baseline and monthly follow-up sputum smear microscopy analyses were done by the Ziehl-Neelsen method in the network of 1168 laboratories available in Peru at the time of the study. Quality control for smear microscopy, culture, and drug susceptibility testing is done routinely in Peru. 11 In 1999, proficiency testing by the WHO Supranational Reference Laboratory in Argentina showed 100% agreement between the two laboratories for identification of susceptibility to rifampicin, isoniazid, ethambutol, and streptomycin. Cure was defined as two negative sputum smear microscopy results or two negative culture results, or both, at the end of treatment. Patients who did not respond to treatment were those who remained bacteriologically positive at the end of treatment. Defaulters were patients who did not attend to take their drugs for 1 month or more at any time after registration. Data management and statistical analysis At enrolment, we obtained data on demographic variables, clinical history, previous treatments for tuberculosis, history of diseases other than tuberculosis, previous imprisonment, and previous hospital admission. During treatment, staff recorded each dose of the drugs administered, any adverse effects and how they were managed, and radiological and bacteriological results. Data were collected in pretested standardised forms. Before analysis, electronic data were compared with 200 patient charts to verify their accuracy. Data were analysed with SPSS (version 7.5.2). Univariate analysis included the 2 test for the comparison of categorical variables. Crude and adjusted odds ratios and 95% CIs were calculated by logistic regression to test for the relation between an unfavourable outcome ie, death, non-response to treatment, or default and predictor variables. To adjust for possible confounding effects, all variables were entered into multiple logistic regression models. Variables were selected for removal according to their biological importance and by backward elimination with the likelihood-ratio test. Cost-effectiveness Cost-effectiveness was assessed from the perspective of the public sector in year 2000 prices (in US$) with standard methods. 12,13 The standardised treatment programme described above was compared with two other strategies using second-line drugs. The first was identical to the standardised treatment programme, except that patients who do not respond to treatment on this programme are given the additional option of treatment with an individualised regimen, tailored to their drug susceptibility pattern. The second was a new strategy proposed for national implementation in This approach also includes provision of individualised treatment to patients not responding to the standardised second-line drug treatment programme. However, in addition, patients who do not respond to the first-line treatment regimen are tested for MDR. Those identified with MDR tuberculosis are enrolled in the standardised second-line drug treatment programme, instead of the first-line re-treatment regimen. For each of the three alternative strategies, cost-effectiveness was calculated according to a strategy s total incremental costs and effects compared with a situation where second-line drugs are not available and treatment is confined to isoniazid monotherapy ie, the analysis focused on the increase in total costs and effects associated with the use of secondline drugs. This was justified on the basis that most lowincome and middle-income countries routinely provide treatment with first-line drugs through government 2 THE LANCET Published online May 14,

3 services, but they do not routinely provide treatment with second-line drugs. The important question for policymaking purposes is, therefore, what are the additional costs and effects associated with implementing second-line drug treatment programmes? Treatment costs in the absence of second-line drugs were assessed as zero. This was justified in view of the low price of isoniazid (US$0.01 per 100 mg dose in Peru) and the possibility that use of health services is infrequent. For the standardised second-line drug treatment programme implemented between 1997 and 2000, two groups of costs were considered: the average cost of individual components of treatment eg, drugs, a visit for direct observation of treatment (DOT), training and the average cost per patient treated. Wherever possible, the costs of individual components of treatment were calculated with an ingredients approach ie, the quantity of resources used was combined with unit price data. The exceptions were: nonpersonnel recurrent inputs, for which there was only expenditure data; and laboratory tests and radiographs, for which only prices were available. Overhead tuberculosis programme costs eg, supervision and management, monitoring, and surveillance were allocated equally across all notified tuberculosis patients in Peru. Since staff time was judged to broadly indicate overall resource use for different types of patient in outpatient health facilities, overhead clinic costs were allocated to tuberculosis patients in accordance with the fraction of clinic staff time spent on their care. Vehicle and equipment costs were annualised using current replacement prices, the assumption of a 5-year life expectancy, and a discount rate of 3% (the currently recommended international rate). 14 Start-up training costs were annualised over 3 years. Building costs per year were based on rental values per month. All local costs were converted into US$ according to the average exchange rate in 2000 (US$1=Peruvian Soles 3.5). The average cost per patient treated was calculated as the cost of each treatment component multiplied by the average number of times this cost was incurred. This calculation was done overall, and for three groups of patients: those who completed the full course of treatment ie, cures plus non-responders; those who died; and those who defaulted. Sources of data included expenditure records, standard salary scales, a database maintained by the MDR unit in Lima, interviews with the National Tuberculosis Programme staff, National Tuberculosis Programme reports, 11,15,16 patient case notes, and a study of the economic impact of tuberculosis in Peru. 17 In view of the uncertainty about several measures that affect effectiveness, and the cost of individualised treatment in middle-income countries, effectiveness and cost-effectiveness were estimated as part of a multivariate uncertainty analysis. This analysis involved three main steps. First, a spreadsheet model was developed in Microsoft Excel (version 7.0), in which treatment paths were defined for each alternative strategy. The earliest starting point for a treatment path was non-response to the first-line treatment regimen, since this is the earliest point at which patients become eligible for treatment with second-line drugs in the three strategies assessed. However, the treatment paths for exceptions and relapse patients who did not respond to the first-line re-treatment regimen started at a later point, to reflect when they become eligible for treatment with second-line drugs and because treatment before this time is the same in each strategy. Second, the number of patients following each treatment path, together with the associated costs and effects, was defined according to: the variables, variable values and distributions, and data sources 2 5,9,11,14,18 31 listed in table 1; the costs per patient reported in this article; and the treatment outcomes associated with the standardised second-line drug regimen. The high cost scenario for individualised treatment (US$ per patient for drugs plus all other costs) was based on drug prices reported for low-income and middle-income countries in the absence of concessional prices, 20 and allowance for other costs besides drugs such as those that applied to the standardised second-line drug treatment programme already implemented. The intermediate cost scenario for individualised treatment (US$5000 per patient for drugs plus all other costs) was chosen to reflect the cost per patient that might apply with both concessional drug prices 20 and with allowance for the costs of other inputs being higher than those of the standardised second-line drug treatment programme. The low-cost scenario for individualised treatment represented an optimistic situation in which drugs are available at highly concessional prices, and the costs of inputs besides drugs are similar to those of the standardised second-line drug treatment programme. Normal distributions were used when mean (SD) could be calculated. If only range information was available, a uniform distribution across this range was defined. The third step was to run a Monte Carlo simulation, involving 5000 iterations, (version 3.5), to estimate means and lower and upper bounds (5th and 95th centiles) for the main outputs of interest. These outputs were total costs, total and average (per patient treated) effects (cures, deaths averted, years of life gained, and DALYs gained), and cost-effectiveness ratios (cost per cure, cost per death averted, cost per year of life gained, and cost per DALY gained). To facilitate comparison with other health interventions, the cost-effectiveness results focus on cost per DALY gained figures. Sensitivity and threshold analysis A sensitivity analysis was done to assess the effect of reductions in the cost of the standardised second-line drug regimen on total costs and cost-effectiveness ratios. Projected drug prices were provided by WHO. A threshold analysis was done to identify the cost per patient at which use of individualised regimens for MDR patients who do not respond to the first-line treatment regimen would be as cost effective as the strategy proposed for national implementation in Role of the funding source The sponsors of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. Results Baseline characteristics and monitoring of participants During enrolment, the committee assessed 755 requests to treat tuberculosis cases with second-line drugs; 477 (63%) were accepted. 21 of 34 health departments in Peru enrolled patients. 11 (2%) of the enrolled patients were excluded from our analysis; reasons included use of a nonstandardised treatment regimen, and unfinished treatment. Table 2 shows demographic characteristics according to entry criteria. Susceptibility test results were available for 344 (74%) patients. 335 (97%) individuals were resistant to one or more drugs, 298 (87%) had MDR tuberculosis, and 273 (79%) were resistant to three or more drugs. THE LANCET Published online May 14,

4 Variable Variable value (distribution Reference where applicable)* Cost of individualised treatment Cost of drugs plus other costs per patient treated (US$) Three scenarios assessed: 4, 18, 19, 20 1) ) ) General variables related to effectiveness Prevalence of MDR among patients who do not respond to first-line treatment regimen (%) 58%, 68% (uniform) NTP survey Self-cure rate among chronic patients (%) 1% 21 Treatment outcomes on the re-treatment regimen for patients who do not respond to first treatment regimen Cure rate (%) 46%, 56% (uniform) 11 Death rate (%) 7% Default rate (%) 7% Non-response rate (%) 86% minus the cure rate Treatment outcomes on the re-treatment regimen for patients who do not respond to first treatment regimen and do not have MDR Cure rate (%) 65%, 75% (uniform) Assumptions Death rate (%) 5% Default rate (%) 7% Non-response rate (%) 88% minus the cure rate Treatment outcomes for patients who are treated with the standardised second-line drug regimen immediately after failing to respond to the first regimen and being found to have MDR (ie, the re-treatment regimen is not used) Cure rate (%) 55%, 65% (uniform) Assumptions Death rate (%) 5% Default rate (%) 7% Non-response rate (%) 88% minus the cure rate Treatment outcomes for patients treated with individualised treatment regimen after non-response to the standardised second-line drug regimen Cure rate (%) 50%, 62% (uniform) 5, 18, 22, 23, 24, 25 Death rate (%) 10% Default rate (%) 12% Non-response rate (%) 78% minus the cure rate Long-term death rate from tuberculosis, chronic patients not receiving treatment with 60%, 99% (uniform) 21, 26 second-line drugs (%) Long-term death rate from tuberculosis among patients who do not respond to treatment 32%, 7% (normal) 5 with second-line drugs (%) Long-term relapse rate among those initially cured with regimens that include second-line drugs (%) 14%, 4% (normal) 5 Death rate from tuberculosis among patients initially cured with regimens that include second-line 100% 5 drugs, but who later relapse (%) Death rate among patients who default (%) Set to equal death rate Assumption for chronic cases with no access to second-line drugs Average period of survival for patients who die after failing to respond to treatment with first-line drugs, in the absence of treatment with second-line drugs (years) 5 5 5, 21 Discount rate for valuing health gains arising in the future (%) 3% 14 Average health gain associated with one averted death (DALYs)** , 28, unpublished data Variables specifically related to effects associated with the impact of treatment on transmission Case reproduction number for chronic tuberculosis when no treatment with second-line drugs is 0 5, 1 (uniform) 2, 9, 29 available ie, average number of secondary cases produced by primary case in the next generation. Case reproduction number for chronic tuberculosis, in the presence of treatment with second-line drugs (Average duration of Assumption that infectiousness in presence amount of of treatment average transmission is duration of infectiousness directly related in absence of treatment) to duration case reproduction number of infectiousness for chronic tuberculosis when no treatment with second-line drugs is available Generation time (years) 2, 4 (uniform) 30 Total number of secondary cases arising g years in the future, where g=generation time in years Re 0 03 g 30 and R is the case reproduction number Total number of secondary cases arising in the future, where r is the total number of secondary r (1 r) Standard formula when cases arising g years in the future and g is the generation time in years, ie, r=re 0 03 g r<1 Average duration of infectiousness, chronic case not treated with second-line drugs (years) 2 31 Average duration of infectiousness, chronic case treated with second-line drugs whose outcome is 2, plus the period on the Assumption death or default (years) re-treatment regimen considered in the analysis Average duration of infectiousness, when final outcome is non-response to treatment (years) 2, plus the duration of Assumption treatment from point of entry into the analysis Average duration of infectiousness, non-responders to the treatment regimen who are cured on the 0 25 Assumption re-treatment regimen (years) Average duration of infectiousness, patients who are cured on the standardised second-line drug plus period on Unpublished data treatment regimen (years) re-treatment regimen considered in the analysis Average duration of infectiousness, patients who are cured on individualised treatment after not plus period on Assumption responding to the standardised second-line drug treatment regimen (years) re-treatment and standardised second-line drug regimen considered in the analysis 4 THE LANCET Published online May 14,

5 Death rate for secondary cases on first-line treatment regimen (%) 9%, 2% (normal) 3 Death rate for secondary cases on first-line treatment regimen among patients who default from 7%, 2% (normal) 3 this regimen (%) proportion of untreated chronic cases that die Non-response rate for secondary cases on first-line treatment regimen (%) 21%, 3% (normal) 3 Average cost per patient treated on the first-line treatment regimen for drugs plus all other costs (US$) 322 Unpublished data Average cost per patient treated on the re-treatment regimen for drugs plus all other costs (US$) 482 Unpublished data *Uniform=analysis used values chosen at random between the lower and upper figures specified for the distribution; normal=first number specifies the mean and the second number the SD that define the distribution. Unpublished details available upon request. Values chosen on the assumption that the cure rate for patients without MDR would be higher than the overall cure rate among all patients who do not respond to the first-line treatment regimen. Values chosen on assumption that the cure rate would be slightly higher than for the cohort treated , due to earlier use of second-line drugs. Since most data were from industrialised country settings, the lower limit for the cure rate was deliberately set as lower than the minimum cure rate reported and approximately equivalent to the cure rate achieved with the standardised regimen used in Peru The higher rate was set as the proportion of those completing treatment multiplied by 0 8, to reflect the higher level of cure among those completing treatment reported in some studies, 24 and which seems plausible based on data reported from Lima, Peru. 22 Assumption justified on the basis that patients who default no longer receive treatment, and can therefore be expected to have treatment outcomes similar to those of chronic tuberculosis patients with no access to second-line drugs. **Formula for calculating DALYs provided in reference given. 28 Use of the formula requires data on the average age at onset of disease (30 years), and the average life expectancy at that age (40 years). Assumed to be between 0 5 and 1 to be conservative when assessing cost-effectiveness. A less conservative assumption, which would improve (ie, lower) cost per DALY gained figures, would have been to assume that the case reproduction number in the absence of treatment with second-line drugs would be equal to 1. Value is conservative with respect to effectiveness and costeffectiveness calculations (ie, value may lead to effectiveness being under-estimated) because the period on treatment is added to the duration of infectiousness expected in the absence of treatment. Value is conservative with respect to effectiveness and cost-effectiveness calculations (ie, value may lead to effectiveness being underestimated) because it is assumed that for those who cure, culture/smear conversion will take the same time as on the standardised second-line regimen, even though an individualised regimen is tailored to a patient s drug susceptibility pattern. Table 1: Variables and variable values used to estimate effectiveness and cost-effectiveness During treatment, 242 (52%) individuals had adverse effects (table 3), leading to permanent interruption of one or more drugs in 24 (5%) patients. An additional 61 (13%) patients underwent drug adjustments during treatment including the extension of kanamycin beyond 3 months and addition of drugs such as isoniazid, clavulanic acid, and sparfloxacin. Treatment outcome Table 4 shows treatment outcomes. Of the 466 patients analysed, 225 (48%) were cured. The proportion of cured patients increased to 55% (225 of 413) when patients who complied with treatment (cured plus died plus nonresponders) were evaluated, and to 63% (225 of 356) when patients who completed chemotherapy (cured plus non-responders) were evaluated. Of the 225 patients cured, 210 (93%) and 215 (96%) had converted to sputum smear negative status after 3 months and 6 months of treatment, respectively. Comparable figures for culture conversion were 153 (68%) and 193 (86%). Of the 298 proven MDR tuberculosis cases, 136 (46%) were cured, 96 (32%) did not respond to treatment, Entry criteria Patients with chronic disease* Exceptions (n=92) Overall (n=466) Patients who were declared cured, Patients who consecutively did not and who relapsed and did not respond respond to the standard treatment to the standard re-treatment and re-treatment regimens (n=189) regimen (n=185) Sex Male 105 (57%) 124 (66%) 50 (54%) 279 (60%) Female 80 (43%) 65 (34%) 42 (46%) 187 (41%) Age (years) <15 3 (2%) 2 (1%) 0 5 (1%) (34%) 78 (41%) 25 (27%) 166 (36%) (40%) 67 (35%) 47 (51%) 188 (40%) >34 45 (24%) 42 (22%) 20 (22%) 107 (23%) Health district Lima City 32 (17%) 34 (18%) 15 (16%) 81 (17%) Lima East 25 (14%) 17 (9%) 1 (1%) 43 (9%) Lima North 62 (34%) 51 (27%) 39 (42%) 152 (33%) Lima South 34 (18%) 51 (27%) 25 (27%) 110 (24%) Other 32 (17%) 36 (19%) 12 (13%) 80 (17%) Susceptibility test results MDR tuberculosis 113 (61%) 130 (69%) 55 (60%) 298 (64%) Tuberculosis no MDR 13 (7%) 16 (8%) 8 (9%) 37 (8%) Susceptible 4 (2%) 1 (1%) 4 (4%) 9 (2%) Not tested 55 (30%) 42 (22%) 25 (27%) 122 (26%) Resistance profile data 0 2 drugs 24 (13%) 32 (17%) 6 (7%) 62 (13%) 3 drugs 34 (18%) 43 (23%) 14 (15%) 91 (20%) 4 drugs 51 (28%) 56 (29%) 31 (34%) 138 (30%) 5 drugs 17 (9%) 15 (8%) 12 (13%) 44 (9%) Not tested 59 (32%) 43 (23%) 29 (31%) 131 (28%) Number of previous treatments 1 21 (23%) 21 (5%) (69%) 167 (88%) 36 (39%) 331 (71%) 3 57 (31%) 22 (12%) 35 (38%) 114 (24%) Data are n (%). *Patients who are sputum positive at the end of a fully supervised re-treatment regimen with first-line drugs. Table 2: Demographic characteristics and disease-related variables according to enrolment criteria THE LANCET Published online May 14,

6 Side-effects* Number of patients (% of all patients) Nausea 80 (17%) Gastritis 54 (12%) Vomiting 46 (10%) Abdominal pain 33 (7%) Headache 20 (4%) General discomfort 9 (2%) Rash 9 (2%) Anorexia 6 (1%) Dizziness 5 (1%) Equilibrium disorder 4 (1%) Blurred vision 3 (1%) Insomnia 2 (0 4%) Arthralgia 2 (0 4%) Hipoacusia 2 (0 4%) Other 6 (1%) *Patients could have two or more side-effects simultaneously. Table 3: Adverse effects in 466 patients who underwent re-treatment with second-line drugs 32 (11%) died, and 34 (11%) defaulted. Of the 264 MDR tuberculosis patients who complied with treatment, 52% were cured. Patients without susceptibility test results were less likely to fail treatment than those with MDR (20 of 122, 16%, vs 96 of 298, 32%; p=0.001) and other patterns of drug resistance (20 of 122, 16%, vs 13 of 37, 35%; p=0 02). There was a significant increasing trend to fail the second-line drug regimen as the number of drugs to which patients were resistant increased (p=0 02). Of the 57 patients who died, 34 (60%) were bacteriologically positive at the time of death and 23 (40%) had 3 months or less of treatment. Among patients who defaulted from treatment, 27 (51%) were bacteriologically negative by the time they defaulted. The average duration of treatment before default was 6 months. Reasons for defaulting included feeling better, moving house, finding a new job, and mismanaged adverse effects. In multivariate analysis, the final model for explaining an unfavourable outcome included sex, the number of drugs to which patients were resistant, and entry criteria. An unfavourable outcome ie, death, non-response, default was significantly associated with two factors: resistance to five or more drugs (odds ratio 2 94, Cost item Average cost in US$ (% total cost) Second-line drugs (75 intensive phase doses at $3 1; 824 (35%) 375 continuation phase doses at $1 6) 450 DOT visits 507 (21%) Food parcels (one per week) 390 (16%) 18 physician consultations 213 (9%) MDR tuberculosis unit in Lima 133 (6%) Committee of treatment evaluation 56 (2%) 9 2 sputum cultures 68 (3%) 14 sputum smears 21 (1%) 3 2 radiographs 14 (1%) Start-up training 53 (2%) Routine training 5 (0 2%) Overall programme management, central level 34 (1%) Supervision at department and district levels 24 (1%) Tuberculosis register 21 (1%) Side-effects and changed treatment (eg, added 18 (1%) doses of kanamycin) Total cost per patient 2381 *Patients treated on the standardised second-line drug regimen who died on treatment received an average of 57 intensive phase doses and 64 continuation phase doses, whereas those who defaulted received an average of 59 intensive phase doses and 107 continuation phase doses. Average costs for patients who died and patients who defaulted were $1026 and $1209, respectively. Table 5: Average cost per patient treated for patients who completed the full course of treatment,* standardised secondline drug treatment programme implemented in Peru %CI ; p=0 01; reference group: drug resistance to two drugs or less), and not responding to two consecutive treatment regimens with first-line drugs (0 61, ; p=0 04; reference group: relapses of the treatment regimen who subsequently did not respond to the re-treatment regimen with first-line drugs). When MDR patients were assessed separately, the probability of an unfavourable outcome in the presence of resistance to five or more drugs increased (3 37, ; p=0 01). Cost and cost-effectiveness The average cost per patient treated on the standardised second-line drug treatment programme, for patients that completed the full course of treatment, was US$2381 (table 5). Drugs, at US$824 per patient, were the greatest Number (%) 3 months 6 months 12 months 18 months Treatment outcome (n=466) Defaulted 11 (2%) 21 (5%) 39 (8%) 53 (11%) Died 23 (5%) 38 (8%) 52 (11%) 57 (12%) Smear result* Negative 355 (76%) 317 (68%) 279 (60%) 249 (53%) Positive 70 (15%) 85 (18%) 94 (20%) 107 (23%) NA 7 (2%) 5 (1%) 2 (0 4%) Culture result* Negative 214 (46%) 255 (55%) 234 (50%) 224 (48%) Positive 102 (22%) 111 (24%) 134 (29%) 131 (28%) NA 116 (25%) 41 (9%) 7 (2%) 1 (0 2%) Non-response to treatment 131 (28%) Cured 225 (48%) Patients cured (n=225) Smear result Negative 210 (93%) 215 (96%) 219 (97%) 223 (99%) Positive 13 (6%) 8 (3%) 5 (2%) 2 (1%) NA 2 (1%) 2 (1%) 1 (0 4%) 0 Culture result Negative 153 (68%) 193 (86%) 207 (92%) 224 (99%) Positive 29 (13%) 17 (7%) 16 (7%) 0 NA 43 (19%) 15 (7%) 2 (1%) 1 (1%) NA=not available.*the number of smear and culture negative results declined over time due to the increase in the number of patients in the defaulted and died categories over time; in addition, not all patients maintained the same bacteriological status throughout treatment. Table 4: Follow-up and treatment outcome in patients who underwent re-treatment with second-line drugs 6 THE LANCET Published online May 14,

7 Standardised second-line Standardised second-line Standardised plus individualised drug treatment drug treatment treatment strategy programme implemented programme plus proposed for implementation in individualised treatment 2001 for patients not responding to treatment with standardised secondline drug regimen Cost indicators Total annual cost in US$ millions 0 61 (0 58, 0 64) 1 34 (1 22, 1 46) 1 99 (1 84, 2 13) Affordability indicators Total annual cost as % National Tuberculosis 7 7 (7 3, 8 0) 24 5 (21 9, 27 1) 34 3 (31 5, 37 2) Programme budget Total annual cost as % total cost of tuberculosis 2 6 (2 4, 2 7) 5 6 (5 1, 6 1) 8 4 (7 7, 8 9) control in Peru Total annual cost as % total Ministry of Health budget 0 1 (0 1, 0 1) 0 3 (0 2, 0 3) 0 4 (0 3, 0 5) Effectiveness indicators Long-term cure rate, all patients included in the analysis (%) 57 (53, 60) 64 (62, 67) 64 (61, 67) Long-term cure rate, patients treated with second-line drugs (%) 41 (37, 46) 55 (50, 59) 59 (56, 63) Total deaths averted, excluding transmission benefits 106 (65, 150) 124 (86, 165) 133 (93, 182) Total deaths averted, including transmission benefits 145 (83, 215) 177 (112, 263) 206 (125, 315) Deaths averted per patient treated on second-line drugs, 0 36 (0 22, 0 51) 0 42 (0 29, 0 56) 0 33 (0 23, 0 45) excluding transmission benefits Deaths averted per patient treated on second-line drugs, 0 49 (0 28, 0 73) 0 60 (0 38, 0 89) 0 51 (0 31, 0 78) including transmission benefits Cost-effectiveness indicators Cost per DALY gained, excluding transmission benefits 270 (178, 415) 493 (350, 701) 672 (474, 947) Cost per DALY gained, including transmission benefits 211 (126, 339) 368 (227, 556) 484 (285, 737) *Year 2000 US$ prices for standardised second-line drug regimen and US$ per patient for individualised treatment; ie, as for strategy implemented , with two modifications: (a) individualised treatment for patients who do not respond to the standardised second-line drug treatment regimen and (b) standardised second-line drug regimen for patients who do not respond to the treatment regimen with first-line drugs and who are diagnosed with MDR. Affordability in relation to National Tuberculosis Programme (NTP) annual budget (approximately US$4 million) based on costs paid for from this budget; estimated annual cost of tuberculosis control US$23 8 million, 17 Ministry of Health annual budget US$497 million. n=500 (ie, treatment non-responders plus relapses and exceptions); total number treated on second line drugs estimated as 295 per year for the programme implemented , 295 for the strategy that adds individualised treatment for patients not responding to standardised second-line drug treatment regimen, and 404 for the new strategy proposed for national implementation in Average figure for proposed strategy is lower because although more deaths are averted in total, more patients are also treated on second-line drugs. Figures are not simply total annual costs divided by total effects, because they also incorporate the costs of treating the secondary cases that are associated with each strategy. These costs amounted to about US$ , dependent on the strategy; incremental cost-effectiveness ratios were calculated but are not shown; details available from the authors. Table 6: Mean cost, affordability, effectiveness, and cost-effectiveness indicators for alternative strategies, highest cost scenarios assessed* (5th and 95th centiles) single cost item. Food parcels, DOT visits, consultations with doctors, and the MDR tuberculosis unit in Lima also represented large costs. The total annual cost of the programme was about US$0 6 million, with costs funded by the National Tuberculosis Programme equivalent to 8% of their budget (table 6). An average of 0 49 deaths were averted per patient treated when benefits arising from prevented transmission were included; the figure was 0 36 when only deaths averted among treated patients were considered. The mean cost per DALY gained, including transmission benefits, was US$211 (range ). At drug prices projected for 2002, the mean cost per DALY gained fell to US$165 (table 7), and the cost of the standardised regimen would be reduced even if capreomycin was used instead of kanamycin, and ofloxacin was used instead of ciprofloxacin (drug cost US$753 vs US$824 for patients completing treatment). The total costs of the two alternative strategies with second-line drugs varied considerably. In the highest cost scenario, in which individualised treatment was assumed to cost US$ per patient, total annual costs were between US$1 3 million and US$2 0 million, and around one quarter to one third of the National Tuberculosis Programme budget (table 6). At lower estimates of the cost of individualised treatment, and with projected reductions in the cost of the standardised second-line drug regimen, total costs fell to around 10 20% of the National Tuberculosis Programme budget (table 7). Compared with the standardised programme, both strategies that included individualised treatment were estimated to be capable of raising cure rates and, depending on whether or not transmission was included in the analysis, of increasing the total number of deaths averted by 17 42% (table 6). In the highest cost scenario, the cost per DALY gained with these two strategies ranged from about US$200 to US$700, with means of US$368 and US$484 (table 6). However, if the price of the standardised second-line drug regimen fell to levels projected for 2002, and the cost per patient treated with individualised regimen was less than US$5000, the mean cost per DALY gained for both strategies would be US$ (table 7). The threshold cost at which use of individualised regimens for MDR patients who do not respond to the first-line treatment regimen would be as cost effective as the strategy proposed for national implementation in 2001 was US$ per patient (figure). Discussion These data from Peru provide evidence with respect to the feasibility and cost-effectiveness of providing standardised second-line drug treatment for chronic tuberculosis patients among whom MDR prevalence is high, under programme conditions and on a national basis, in a middle-income country. They show that Peru s National Tuberculosis Programme was able to achieve a compliance rate of nearly 90%, a cure rate of almost 50% among all patients enrolled, and a cure rate of 63% among those who completed the full course of treatment. The total programme cost was affordable in the context of the National Tuberculosis Programme s budget, and the mean cost per DALY gained was around US$ THE LANCET Published online May 14,

8 Mean total annual cost in US$ millions (costs funded by NTP Cost per DALY gained (5th and 95th centiles)* as % NTP budget) Standardised regimen Standardised regimen Standardised regimen Standardised regimen Standardised regimen Standardised regimen at projected prices for at projected prices for at projected prices for at projected prices for at projected prices for at projected prices for 2002, individualised 2002, individualised 2002, individualised 2002, individualised 2002, individualised 2002, individualised treatment US$ treatment US$5000 treatment US$2500 treatment US$ treatment US$5000 treatment US$2500 per patient per patient per patient per patient per patient per patient Strategy Standardised second (4%) 0 49 (4%) 0 49 (4%) 165 (96, 267) 165 (96, 267) 165 (96, 267) line drug treatment programme implemented Standardised second (21%) 0 85 (12%) 0 67 (8%) 331 (202, 504) 226 (134, 346) 173 (100, 267) line drug treatment programme plus individualised treatment for patients not responding to treatment with standardised secondline drug regimen Standardised plus 1 79 (30%) 1 20 (18%) 0 90 (12%) 434 (254, 664) 282 (160, 435) 206 (112, 321) individualised treatment strategy proposed for implementation in 2001 NTP=National Tuberculosis Programme. *Includes costs and effects associated with transmission. ie, as for strategy implemented , with two modifications: (a) individualised treatment for patients who do not respond to the standardised second-line drug treatment regimen and (b) standardised secondline drug regimen for patients who do not respond to the treatment regimen with first-line drugs and who are diagnosed with MDR. Table 7: Cost and cost-effectiveness indicators for alternative strategies under different cost assumptions We believe that the feasibility of using second-line drugs nationwide relied on three key factors. First, the new service was built on a tuberculosis control programme that had already achieved WHO s global tuberculosis control targets with first-line drugs. 8,9 Second, there was a special committee to filter requests for access to second-line drugs, which allowed the National Tuberculosis Programme to benefit from the best expertise available in the country, and to screen out candidates for whom treatment with first-line drugs was a more appropriate option. Third, there was an MDR tuberculosis unit operating within the overall framework of the National Tuberculosis Programme, facilitating coordination and management. Our results are consistent with previous observations that cure rates for patients with chronic disease who complete treatment range from around 60% to 80% when second-line drugs are used, 24 and with a study of treatment outcomes for patients with MDR-tuberculosis in the USA. 5 They are poorer than figures reported for patients with MDR tuberculosis in the Netherlands and Canada, but in these studies a large proportion of patients 7000 Threshold cost per patient treated (US$) individualised treatment US$ per patient, including transmission individualised treatment US$ per patient, excluding transmission individualised treatment US$5 000 per patient, including transmission individualised treatment US$5 000 per patient, excluding transmission individualised treatment US$2 500 per patient, including transmission individualised treatment US$2 500 per patient, excluding transmission Cure rate (%) 85 Threshold costs for individualised treatment of patients with MDR who do not respond to the first-line treatment regimen and are diagnosed with MDR when compared with strategy proposed for national implementation in THE LANCET Published online May 14,

9 had no history of previous treatment. 18,23 Cure rates are never likely to be as high as those for new drug susceptible cases patients have already failed treatment with firstline drugs, and second-line drugs are less effective and more toxic than first-line drugs. 32 The data suggest that cure rates with second-line drugs can be improved in Peru. Half the patients who defaulted were bacteriologically negative at the time of default. Individuals who abandoned therapy could be traced by health workers for adjustment of the time and location of treatment. Improved training for health workers with respect to management of adverse effects, and strengthening the standardised treatment regimen, might also help. Substitution of capreomycin for kanamycin, and ofloxacin for ciprofloxacin could soon be possible with no increase in cost to the National Tuberculosis Programme, and use of cycloserine could be considered in the presence of MDR and ethambutol resistance. Further options for increasing cure rates include use of individualised treatment for some patients, and earlier use of second-line drugs. A promising experience with individualised treatment in Northern Lima is already underway; 33 what remains to be proven is feasibility and effectiveness on a wider scale. Choice among strategies should also be informed by economic considerations. Cost was a major factor when the standardised programme was first implemented in 1997, with the regimen chosen being the most affordable in the context of the National Tuberculosis Programme s budget. Other options remain more expensive, but are becoming increasingly realistic with substantial reductions in drug prices. 20 Treatment of patients with chronic disease with second-line drugs is, on average, less cost-effective than the basic DOTS programme, which relies on first-line drugs. The DOTS programme is much more effective in terms of the cure rate 8,9 and previous estimates of averted deaths per patient treated, 26 and is substantially less expensive. Nevertheless, the costeffectiveness figures for all the strategies that we assessed are much lower than per capita Gross National Income, which is a more general benchmark for assessing whether or not an intervention is cost-effective, 34,35 and around or below the US$ per DALY gained (converted to year 2000 US$ prices) that was estimated for the limited care component of an essential health package for middle-income countries by the World Bank in The figures for the standardised programme implemented are similar to the US$176 (converted to year 2000 US$ prices) per DALY gained that was suggested as a threshold for defining an attractive investment in lowincome countries in The threshold should be higher in Peru, in view of its middle-income status. Our study has several limitations. The treatment outcome data are not the result of a trial, so we were not able to assess efficacy. Although patients are being followed up for 2 or more years to assess the relapse rate, long-term follow up data are not yet available. There were no data with respect to susceptibility to second-line drugs, since the necessary infrastructure and logistics had not been developed in Peru during the period of enrolment and follow-up. A quarter of the patients enrolled for treatment did not have susceptibility results for first-line drugs. In the cost-effectiveness analysis, we had no data on the long-term outcomes and costs associated with isoniazid monotherapy, or comprehensive and detailed cost data for individualised treatment in a low-income or middle-income setting. Several assumptions were required to estimate effectiveness, and costs from the perspective of patients were not assessed. Nonetheless, our assumptions were conservative, and the exclusion of costs borne by patients is consistent with the cost, budget, and cost-effectiveness figures with which we compared our results. It is noteworthy that 83% of enrolled patients were from Lima. There is evidence to suggest that Lima has high rates of MDR tuberculosis, 38 but these data could indicate selection bias, resulting from the special committee s location in Lima. Moreover, some of the patients enrolled in this programme as exceptions had a history of only one previous treatment, suggesting that they did not meet the inclusion criteria. The data have important programmatic policy implications, and Peru s National Tuberculosis Programme adopted three new policies in First, a patient will be declared as not responding to treatment if they are culture positive after 6 months of treatment (or 150 doses) with a standardised second-line drug regimen. Second, patients who do not respond to the standardised second-line drug regimen will have access to a secondline drug regimen tailored to their drug susceptibility pattern. This regimen will be provided through an international partnership collaboration that will include upgrading the capabilities of the national reference laboratory to enable testing for susceptibility to secondline drugs. Third, patients who do not respond to the first-line treatment regimen will be treated with the standardised second-line drug regimen if they are diagnosed with MDR ie, the first-line re-treatment regimen will not be used for these patients. 39 To enhance access to second-line drugs outside Lima, special subcommittees to approve or reject requests for access to standardised second-line drug treatment have been established in each of the country s 34 health departments. Our findings are probably applicable to middle-income countries where high cure and compliance rates with firstline drugs are achieved, prevalence of HIV-1 is low, and ambulatory care is used. Tuberculosis programmes that are not able to manage first-line drugs properly are unlikely to be successful with second-line drugs. Effectiveness in terms of DALYs gained would be lower if rates of HIV-1 infection among tuberculosis patients are high, due to a smaller number of years of life being gained per death averted. If admission to hospital is relied on, costs would be higher in both low-income and middleincome settings. Our findings indicate that use of second-line drugs to treat chronic tuberculosis patients with high levels of MDR can be feasible and cost-effective in middle-income countries with strong tuberculosis control programmes. Widespread use of second-line drugs in other countries will depend on efforts to reduce costs and increase access under tightly controlled approaches, 20 and on success in building strong tuberculosis control programmes. Contributors M Espinal and K Floyd wrote the report; P Suárez, E Alarcón, J Portocarrero, I Sabogal, and M Espinal designed and planned the study; J Portocarrero was the study coordinator; C Bonilla and G Ramos coordinated doctors at the MDR tuberculosis unit; I Aranda was the head nurse coordinating patient follow-up and data management; E Rapiti and M Espinal were responsible for data analysis with respect to feasibility and treatment outcomes; K Floyd was responsible for cost and costeffectiveness analysis; P Suárez and E Alarcón assisted with the collection of cost data; C Dye assisted with the transmission component of the costeffectiveness analysis; M Raviglione assisted with analysis of treatment outcomes; and all investigators contributed to the interpretation of the data and revisions to the report. Conflict of interest statement None declared. THE LANCET Published online May 14,