Articles

Prevalence of and risk factors for resistance to second-line drugs in people with multidrug-resistant tuberculosis in eight countries: a prospective cohort study
Tracy Dalton, Peter Cegielski, Somsak Akksilp, Luis Asencios, Janice Campos Caoili, Sang-Nae Cho, Vladislav V Erokhin, Julia Ershova, Ma Tarcela Gler, Boris Y Kazennyy, Hee Jin Kim, Kai Kliiman, Ekaterina Kurbatova, Charlotte Kvasnovsky, Vaira Leimane, Martie van der Walt, Laura E Via, Grigory V Volchenkov, Martin A Yagui, Hyungseok Kang, and the Global PETTS Investigators*

Summary

Background The prevalence of extensively drug-resistant (XDR) tuberculosis is increasing due to the expanded use of second-line drugs in people with multidrug-resistant (MDR) disease. We prospectively assessed resistance to secondline antituberculosis drugs in eight countries. Methods From Jan 1, 2005, to Dec 31, 2008, we enrolled consecutive adults with locally confirmed pulmonary MDR tuberculosis at the start of second-line treatment in Estonia, Latvia, Peru, Philippines, Russia, South Africa, South Korea, and Thailand. Drug-susceptibility testing for study purposes was done centrally at the Centers for Disease Control and Prevention for 11 first-line and second-line drugs. We compared the results with clinical and epidemiological data to identify risk factors for resistance to second-line drugs and XDR tuberculosis. Findings Among 1278 patients, 437% showed resistance to at least one second-line drug, 200% to at least one second-line injectable drug, and 129% to at least one fluoroquinolone. 67% of patients had XDR tuberculosis (range across study sites 08152%). Previous treatment with second-line drugs was consistently the strongest risk factor for resistance to these drugs, which increased the risk of XDR tuberculosis by more than four times. Fluoroquinolone resistance and XDR tuberculosis were more frequent in women than in men. Unemployment, alcohol abuse, and smoking were associated with resistance to second-line injectable drugs across countries. Other risk factors differed between drugs and countries. Interpretation Previous treatment with second-line drugs is a strong, consistent risk factor for resistance to these drugs, including XDR tuberculosis. Representative drug-susceptibility results could guide in-country policies for laboratory capacity and diagnostic strategies. Funding US Agency for International Development, Centers for Disease Control and Prevention, National Institutes of Health/National Institute of Allergy and Infectious Diseases, and Korean Ministry of Health and Welfare.

Published Online August 30, 2012 http://dx.doi.org/10.1016/ S0140-6736(12)60734-X See Online/Comment http://dx.doi.org/10.1016/ S0140-6736(12)61069-1 *Continued at end of article. Division of Tuberculosis Elimination, Centers for Disease Control and Prevention, Atlanta, GA, USA (T Dalton PhD, P Cegielski MD, J Ershova PhD, E Kurbatova MD, C Kvasnovsky MD); Office of Disease Prevention and Control Region 7, Ubon Ratchatani, Thailand (S Akksilp MD); National Tuberculosis Reference Laboratory, Lima, Peru (L Asencios MPH); Tropical Disease Foundation, Manila, Philippines (J Campos Caoili MD, M T Gler MD); International Tuberculosis Research Center, Masan and Yonsei University College of Medicine, Seoul, South Korea (Prof S-N Cho PhD); Central Tuberculosis Research Institute, Russian Academy of Medical Sciences, Moscow, Russian Federation (V V Erokhin MD); Orel Oblast Tuberculosis Dispensary, Orel, Russian Federation (B Y Kazennyy MD); Korean Institute of Tuberculosis, Seoul, South Korea (H J Kim MD); Tartu University Hospital, Tartu, Estonia (K Kliiman MD); State Agency, Infectology Centre of Latvia, Tuberculosis and Lung Disease Clinic, Riga, Latvia (V Leimane MD); Medical Research Council, Pretoria, South Africa (M van der Walt PhD); National Institute for Allergy and Infectious Disease, National Institutes of Health, Bethesda, MD, USA (L E Via PhD); Vladimir Oblast Tuberculosis Dispensary, Vladimir, Russian Federation (G V Volchenkov MD); National Institute of Health, Lima, Peru (M A Yagui MD); and National

Introduction
Multidrug-resistant (MDR) tuberculosis, defined as tuber culosis caused by Mycobacterium tuberculosis resistant to at least isoniazid and rifampicin, accounts for 3648% of incident cases of tuberculosis worldwide around 440 000 new cases in 2008.1,2 In 2000, the Green Light Committee was formed within the Stop TB Partnership and WHO to increase access to high-quality, second-line antituberculosis drugs at low prices, to prevent additional drug resistance, and to contribute evidence for policy development. By 2011, 255 project applications to the Green Light Committee had been approved that covered more than 130 000 patients with MDR tuberculosis.3,4 The global emergence of extensively drug-resistant (XDR) tuberculosis heralds the advent of widespread, virtually untreatable tuberculosis.2,5 XDR tuber culosis is defined as disease caused by M tuberculosis strains resistant to at least isoniazid, rifamipicin, and one or more drugs within each of the two most important groups of second-line antituberculosis drugs (fluoro

quinolones and injectable drugs).5,6 XDR tuberculosis has been reported in 77 countries, but precise prevalence is unclear. Only two of 27 high-burden MDR tuberculosis countries routinely test for resistance to second-line drugs.2 After an epidemic of XDR tuberculosis in South Africa, the South Africa Medical Research Council convened an emergency consultation in August, 2006, that outlined a global strategy to combat this form of disease.7,8 This strategy was refined in October, 2006, by the WHO Global Task Force on XDR-TB and disseminated widely.7,8 The strategy underscored the urgent need to quantify the extent of XDR tuberculosis with population-based data.7,8 Limited laboratory capacity and inconsistent procedures for testing have hindered understanding of resistance to second-line drugs.2,9 Shortly before these events, we launched a multi national, epidemiological study of MDR tuberculosis, the Preserving Effective TB Treatment Study (PETTS). It focused on the risk factors for and frequency and consequences of acquired resistance to second-line drugs

www.thelancet.com Published online August 30, 2012 http://dx.doi.org/10.1016/S0140-6736(12)60734-X

Articles

Masan Tuberculosis Hospital, Masan, South Korea (H Kang MD) Correspondence to: Dr Tracy Dalton, Centers for Disease Control and Prevention, 1600 Clifton Road Northeast, Mailstop F08, Atlanta, GA 30333, USA tldalton@cdc.gov

in people with MDR tuberculosis. In view of the emergence of XDR tuberculosis, we expanded PETTS in November, 2005, to include additional study sites and participants to provide poulation-based data on the prevalence of second-line-drug resistance in patients with MDR tuberculosis. Here we report the findings of the expanded study in eight countries.

Methods

Participants
The PETTS proposal was presented at an open meeting of the International Working Group on MDR tuberculosis in October, 2003, and included an open invitation to centres to participate. Clinical centres in nine countries (Estonia, Latvia, Peru, the Philippines, Russia, South Africa, South Korea, Taiwan, and Thailand)
3034 eligible patients 1252 excluded 1782 patients enrolled 50 Estonia 106 Latvia 211 Peru 449 Philippines 221 Russia 522 South Africa 118 South Korea 50 Taiwan 55 Thailand 242 excluded 81 isolates not shipped owing to lack of permit 152 no positive follow-up isolate 3 tubes broken 6 not shipped (reasons unknown) 1540 patients isolates shipped to CDC 50 Estonia 106 Latvia 211 Peru 448 Philippines 133 Russia 369 South Africa 118 South Korea 55 Thailand 50 Taiwan 262 excluded 122 no growth 40 contaminated 50 not MDR tuberculosis 50 no drug-susceptibility results* 1278 patients with conrmed MDR tuberculosis assessed 46 Estonia 100 Latvia 177 Peru 397 Philippines 115 Russia 293 South Africa 99 South Korea 51 Thailand

volunteered to participate. The drug-susceptibility data for Taiwan were not available at the time of data analysis and, therefore, were not included in this report (figure). Health-care professionals at these sites provide care for nearly all patients with MDR tuberculosis within their jurisdictions: nationwide in Estonia and Latvia; in two districts in Lima, Peru (Lima Ciudad and Lima Este); in two oblasts (territories) of Russia (Orel and Vladimir oblasts); in greater Manila in the Philippines; in four provinces of South Africa (Eastern Cape, KwaZulu Natal, Mpumalanga, and Northwest); in the two main referral centres for MDR tuberculosis in South Korea (National Masan Tuberculosis Hospital, Masan, and Korean Institute of Tuberculosis, Seoul); and in four provinces in northeast Thailand (Sakon Nakon, Srisaket, Ubon Ratchathani, and Yasothon). In the participating countries, WHO estimated that the prevalence of MDR tuberculosis among patients never previously treated for tuberculosis was 17180%, and among previously treated patients was 67460%.10 Estonia, Latvia, Peru, the Philippines, and Russia were running projects approved by the Green Light Committee at the time of the study. South Africa, South Korea, and Thailand had not submitted project applications to the Green Light Committee before entering the study and, therefore, were deemed non-Green Light Committee countries. All countries had well established tuberculosis programmes, including strategies for MDR tuberculosis, when PETTS started. The study protocol was approved by ethics committees in every participating country and by the Centers for Disease Control and Prevention (CDC) in the USA.

Study population
Adults within the catchment areas with locally con firmed, pulmonary MDR tuberculosis who started treatment with second-line drugs between Jan 1, 2005, and Dec 31, 2008, were eligible for inclusion. Patients must have received treatment with second-line drugs for at least 30 days, had a baseline mycobacterial culture from sputum collected within 30 days before or after the start of second-line treatment, had at least one follow-up positive culture from sputum collected at least 30 days after the baseline sample, and periodically had samples shipped to CDC, Atlanta, GA, USA, for drug-suscept ibility testing. Exclusion criteria were age younger than 18 years, current imprisonment, and pregnancy. After November, 2005, we dropped the requirement for the positive follow-up culture and stratified the analysis according to whether a follow-up isolate was available. Estonia, Latvia, South Africa, and Masan, South Korea, stayed with the two-culture protocol because of their primary interest in acquired drug resistance, but otherwise collected the same data as other regions. South Africa did not enrol patients previously treated for MDR tuberculosis. All patients gave written informed consent.

Figure: Study profile CDC=Centers for Disease Control and Prevention. MDR=multidrug-resistant. *All from Taiwan.

www.thelancet.com Published online August 30, 2012 http://dx.doi.org/10.1016/S0140-6736(12)60734-X

Articles

Procedures
Demographic, socioeconomic, and clinical data were recorded for each patient by trained personnel, including details of previous and current treatments, surgery,
Number of patients per country (%) Estonia (n=46) Sex Male Female Age group (years) 1824 2544 4564 65 Marital status Never married Currently married or cohabitating Previously married Data missing Education Primary or less Secondary Postsecondary Data missing Occupational risk Yes No Data missing Employment status Employed Unemployed Other Data missing History of imprisonment Yes Data missing History of homelessness Yes Data missing Alcohol abuse Yes Data missing Current smoker Yes Data missing HIV infection Yes Unknown or data missing Diabetes Yes Unknown or data missing 4 (87%) 0 2 (20%) 0 2 (44%) 2 (44%) 3 (30%) 0 40 (870%) 0 70 (700%) 1 (10%) 28 (609%) 0 64 (640%) 0 8 (174%) 0 12 (120%) 0 8 (174%) 0 20 (200%) 0 13 (283%) 19 (413%) 14 (304%) 0 35 (350%) 47 (470%) 18 (180%) 0 3 (652%) 41 (891%) 2 (435%) 4 (40%) 96 (960%) 0 15 (326%) 13 (283%) 18 (391%) 0 19 (190%) 41 (410%) 40 (400%) 0 13 (283%) 17 (369%) 15 (326%) 1 (22%) 25 (250%) 40 (400%) 35 (350%) 0 2 (44%) 19 (413%) 22 (478%) 3 (65%) 2 (20%) 47 (470%) 46 (460%) 5 (50%) 36 (783%) 10 (217%) 70 (700%) 30 (300%) Latvia (n=100)

hospital admissions, comorbidities (particularly HIV-1 infection), local microbiology results, baseline chest radiography results, and final treatment outcomes. Cases were classified according to previous treatment and
Total (n=1278)

Peru (n=177) 104 (588%) 73 (412%) 67 (379%) 89 (503%) 16 (90%) 5 (28%) 99 (559%) 59 (333%) 18 (102%) 1 (06%) 17 (96%) 110 (622%) 50 (282%) 0 9 (508%) 167 (944%) 1 (056%) 51 (288%) 75 (424%) 50 (282%) 1 (06%) 6 (34%) 7 (40%) 2 (11%) 0 11 (62%) 2 (11%) 0 (0%) 1 (06%) 6 (34%) 0 7 (40%) 0

Philippines (n=397) 241 (607%) 156 (393%) 55 (139%) 210 (529%) 128 (322%) 4 (10%) 111 (280%) 255 (642%) 31 (78%) 0 67 (169%) 164 (413%) 166 (418%) 0 11 (28%) 386 (972%) 0 274 (690%) 67 (169%) 56 (141%) 0 7 (18%) 2 (05%) 4 (10%) 2 (05%) 11 (28%) 1 (03%) 9 (23%) 1 (03%) 0 396 (998%) 104 (262%) 0

Russia (n=115) 96 (835%) 19 (165%) 14 (122%) 57 (496%) 37 (322%) 7 (61%) 30 (261%) 45 (391%) 39 (339%) 1 (09%) 17 (148%) 48 (417%) 50 (435%) 0 7 (61%) 108 (939%) 0 27 (235%) 60 (522%) 28 (243%) 0 30 (261%) 0 5 (44%) 0 38 (330%) 0 94 (817%) 0 5 (44%) 0 9 (78%) 0

South Africa (n=293) 163 (556%) 130 (444%) 39 (133%) 191 (652%) 63 (215%) 0 166 (567%) 64 (218%) 17 (58%) 46 (157%) 91 (311%) 115 (393%) 11 (38%) 76 (259%) 18 (61%) 208 (71%) 67 (229%) 71 (242%) 198 (676%) 20 (68%) 4 (14%) 16 (55%) 138 (471%) 5 (17%) 100 (341%) 45 (154%) 55 (188%) 52 (178%) 14 (48%) 145 (495%) 24 (82%) 11 (38%) 4 (14%)

South Korea (n=99) 76 (768%) 23 (232%) 7 (71%) 51 (515%) 35 (354%) 6 (61%) 15 (152%) 23 (232%) 7 (71%) 54 (546%) 24 (242%) 46 (465%) 29 (293%) 0 0 48 (485%) 51 (515%) 67 (677%) 23 (232%) 8 (81%) 1 (10%) 1 (10%) 51 (515%) 0 51 (515%) 13 (131%) 1 (10%) 36 (364%) 0 0 45 (455%) 15 (152%) 1 (10%)

Thailand (n=51) 33 (647%) 18 (353%) 4 (78%) 25 (490%) 19 (373%) 3 (59%) 7 (137%) 38 (745%) 6 (118%) 0 37 (726%) 9 (177%) 5 (98%) 0 3 (59%) 48 (941%) 0 43 (843%) 8 (157%) 0 0 6 (118%) 0 0 0 3 (59%) 0 6 (118%) 0 7 (137%) 1 (20%) 10 (196%) 0 819 (641%) 459 (359%) 190 (149%) 689 (539%) 366 (286%) 33 (26%) 466 (365%) 541 (423%) 168 (131%) 103 (81%) 287 (225%) 546 (427%) 369 (289%) 76 (60%) 55 (43%) 1102 (862%) 121 (95%) 581 (455%) 497 (389%) 194 (152%) 6 (05%) 94 (74%) 198 (155%) 36 (28%) 153 (120%) 213 (167%) 59 (46%) 307 (240%) 17 (13%) 168 (132%) 468 (366%) 162 (127%) 5 (04%) (Continues on next page)

www.thelancet.com Published online August 30, 2012 http://dx.doi.org/10.1016/S0140-6736(12)60734-X

Articles

Number of patients per country (%) Estonia (n=46) (Continued from previous page) Contact of tuberculosis case Yes Unknown Contact of MDR tuberculosis case Yes Data missing History of tuberculosis Yes Number of episodes 0 1 2 3 4 Treatment First-line drugs Second-line drugs Fluoroquinolones* Second-line injectable drugs Other oral second-line drugs Data missing Yes Data missing In hospital at enrolment Yes Yes, unilateral Yes, bilateral Unknown or data missing Yes Data missing 38 (826%) 26 (565%) 14 (304%) 0 34 (739%) 0 100 (1000%) 55 (550%) 35 (350%) 0 75 (750%) 0 1 (06%) 60 (339%) 19 (107%) 15 (85%) 116 (655%) 31 (175%) 57 (144%) 125 (315%) 74 (186%) 7 (18%) 380 (957%) 1 (03%) 114 (991%) 62 (539%) 27 (235%) 0 68 (591%) 0 293 (1000%) 113 (386%) 79 (270%) 0 282 (963%) 1 (03%) 51 (515%) 22 (222%) 10 (101%) 41 (414%) 85 (859%) 1 (10%) 5 (98%) 22 (431%) 12 (235%) 0 44 (863%) 2 (39%) 20 (435%) 8 (174%) 4 (87%) 4 (87%) 7 (152%) 0 39 (848%) 1 (22%) 76 (760%) 21 (210%) 14 (140%) 18 (180%) 16 (160%) 0 88 (880%) 0 165 (932%) 21 (119%) 21 (119%) 19 (107%) 20 (113%) 0 157 (887%) 0 397 (1000%) 54 (136%) 53 (134%) 8 (20%) 2 (05%) 0 391 (985%) 0 105 (913%) 27 (235%) 19 (165%) 20 (174%) 20 (174%) 0 109 (948%) 2 (17%) 281 (959%) 8 (27%) 8 (27%) 8 (27%) 8 (27%) 0 282 (963%) 0 91 (919%) 53 (535%) 44 (444%) 21 (212%) 51 (515%) 1 (10%) 33 (333%) 52 (525%) 51 (1000%) 3 (59%) 3 (59%) 1 (20%) 1 (20%) 0 50 (980%) 0 24 (522%) 16 (348%) 4 (87%) 2 (44%) 0 22 (220%) 54 (540%) 22 (220%) 2 (20%) 0 11 (621%) 103 (582%) 45 (254%) 17 (96%) 1 (06%) 0 22 (55%) 153 (385%) 128 (322%) 94 (237%) 9 (78%) 73 (635%) 24 (209%) 5 (44%) 4 (35%) 12 (41%) 145 (495%) 98 (335%) 30 (102%) 8 (27%) 1 (10%) 15 (152%) 30 (303%) 22 (222%) 31 (313%) 0 18 (353%) 24 (471%) 9 (177%) 0 22 (478%) 78 (780%) 166 (938%) 397 (100%) 106 (922%) 281 (959%) 98 (990%) 51 (1000%) 11 (239%) 16 (348%) 28 (280%) 2 (20%) 41 (232%) 19 (107%) 56 (141%) 123 (310%) 20 (174%) 30 (261%) 26 (89%) 157 (536%) 3 (303%) 52 (525%) 2 (39%) 38 (745%) 18 (391%) 13 (283%) 30 (300%) 2 (20%) 102 (576%) 0 239 (602%) 21 (53%) 50 (435%) 7 (61%) 117 (399%) 83 (283%) 25 (253%) 0 17 (333%) 22 (431%) Latvia (n=100) Peru (n=177) Philippines (n=397) Russia (n=115) South Africa (n=293) South Korea (n=99) Thailand (n=51)

Total (n=1278)

598 (468%) 148 (116%) 187 (146%) 437 (342%) 1199 (938%) 79 (62%) 446 (349%) 400 (313%) 215 (168%) 138 (108%) 1186 (928%) 195 (153%) 166 (130%) 99 (78%) 125 (98%) 1 (01%) 1149 (899%) 55 (43%) 659 (516%) 485 (380%) 270 (212) 63 (49%) 1084 (848%) 36 (28%)

First treatment for MDR tuberculosis

Cavitary disease on chest radiograph

Positive sputum-smear test at enrolment

MDR=multidrug-resistant. *Ciprofloxacin, ofloxacin, lomefloxacin, levofloxacin, and moxifloxacin.

Table 1: Characteristics of patients with confirmed MDR tuberculosis

related outcomes. Data abstracted from medical and laboratory records were double entered locally into a customised database (EpiInfo, version 3.3.2). Duplicate databases were compared electronically and discrepancies were resolved from primary sources. Data collection and data entry were supervised by a team of coordinators at country, regional, and international levels for quality assurance and completeness (TD, PC, JCC, JE, MTG, KK, EK, CK, and Melanie Wolfgang, CDC, Atlanta, GA, USA, Carmen Contreras, Socios en Salud Sucursal, Lima, Peru, and Joey Lancaster, Medical Research Council, Pretoria, South Africa). Baseline sputum specimens were tested locally with culture for M tuberculosis complex and for susceptibility to
4

at least isoniazid and rifampicin. Duplicate cultures were inoculated from the same specimens for study purposes. Microbiological methods differed by site, but all lab oratories used internationally recommended media and methods.11,12 Follow-up sputum samples were collected and cultured monthly for the duration of the patients treatment for MDR tuberculosis. Duplicates of positive baseline and follow-up cultures were batched and shipped to CDC. Upon receipt at CDC, isolates were grown in 5 mL Middlebrook 7H9 broth with polysorbate 80 (Remel, Lenexa, KS, USA) at 37C until the turbidity reached roughly a McFarland 10 standard. Baseline isolates were tested for drug susceptibility at CDC, according to the Clinical Laboratory Standards

www.thelancet.com Published online August 30, 2012 http://dx.doi.org/10.1016/S0140-6736(12)60734-X

Articles

Number of patients per country (%) Estonia (n=46) First-line drugs Ethambutol Streptomycin Four first-line drugs* Second-line drugs Any second-line drug At least one fluoroquinolone Injectable drugs Kanamycin Amikacin Capreomycin At least one All Other oral second-line drugs Ethionamide Aminosalicylic acid At least one XDR tuberculosis 5 (109%) 2 (44%) 6 (130%) 3 (65%) 23 (230%) 24 (240%) 38 (380%) 8 (80%) 13 (73%) 23 (130%) 34 (192%) 11 (62%) 121 (305%) 8 (20%) 125 (315%) 3 (08%) 16 (139%) 18 (157%) 30 (261%) 13 (113%) 50 (171%) 23 (79%) 64 (218%) 31 (106%) 11 (111%) 34 (343%) 36 (364%) 15 (152%) 10 (196%) 5 (98%) 13 (255%) 2 (39%) 13 (283%) 11 (239%) 2 (44%) 14 (304%) 2 (435%) 42 (420%) 35 (350%) 15 (150%) 47 (470%) 9 (90%) 30 (170%) 31 (175%) 30 (170%) 34 (192%) 26 (147%) 7 (18%) 7 (18%) 1 (03%) 8 (20%) 1 (03%) 38 (330%) 19 (165%) 8 (70%) 40 (348%) 6 (52%) 80 (273%) 80 (273%) 79 (277%) 84 (287%) 75 (256%) 23 (232%) 18 (182%) 14 (141%) 23 (232%) 13 (131%) 4 (78%) 4 (78%) 3 (59%) 5 (98%) 2 (39%) 24 (522%) 12 (261%) 62 (620%) 14 (140%) 60 (339%) 16 (90%) 153 (385%) 28 (71%) 63 (548%) 21 (183%) 131 (447%) 37 (126%) 49 (495%) 32 (323%) 17 (333%) 5 (98%) 41 (891%) 46 (1000%) 41 (891%) 65 (650%) 96 (960%) 65 (650%) 92 (520%) 99 (559%) 62 (350%) 308 (776%) 239 (604%) 205 (518%) 62 (539%) 110 (965%) 61 (535%) 177 (604%) 210 (717%) 142 (485%) 57 (576%) 46 (465%) 30 (303%) 24 (471%) 35 (686%) 19 (373%) Latvia (n=100) Peru (n=177) Philippines (n=397) Russia (n=115) South Africa (n=293) South Korea (n=99) Thailand (n=51)

Total

826 (646%) 881 (690%) 625 (490%) 559 (437%) 165 (129%) 237 (185%) 205 (160%) 152 (120%) 255 (200%) 134 (105%) 249 (195%) 137 (107%) 346 (271%) 86 (67%)

XDR=extensively drug-resistant. *Isoniazid, rifampicin, ethambutol, and streptomycin.

Table 2: Drug resistance at baseline

Institute standard,11 by the indirect agar proportion method that uses Middlebrook 7H10 agar (BD), supplemented individually with the following drugs: isoniazid 02 g/mL, rifampicin 10 g/mL, etham butol 50 g/mL, strepto mycin 20 g/mL, ofloxacin 20 g/mL, ciprofloxacin 20 g/mL, kanamycin 50 g/mL, capreo mycin 100 g/mL, amikacin 40 g/mL, aminosalicylic acid 20 g/mL, and ethionamide 100 g/mL. Resistance was reported when the proportion of growth on drugcontaining medium was at least 1% of that on drug-free medium. Contamination was defined as colony morph ology inconsistent with M tuberculosis or growth of fungus.

Statistical analysis
All statistical analyses were done with SAS (version 9.2). The incidence of resistance amplification was estimated a priori to be 1030%.13,14 Therefore, we used the test for independent groups and applied the Power and Sample Size Calculation (version 2.0) to calculate group sizes, with an assumed relative risk of at least 20 for resistance amplification in non-Green Light Committee sites (exposed) compared with Green Light Committee projects (unexposed). To achieve 80% power with 95% CI, we estimated that we would need to enrol the following numbers of patients per group, by amplifi cation rate in Green Light Committee sites: 435 at 5%; 199 at 10%; 120 at 15%; and 81 at 20%. Although the primary analysis was stratified by sites to determine an aggregated (pooled) estimate of relative risk, sites were offered the opportunity to increase their own

sample size to enhance the precision of site-specific outcome measures. Drug-susceptibility results were analysed as dependent variables, and clinical, epidemiological, and microbiological information about each patient as independent variables. We analysed risk factors for and prevalence of resistance to each drug separately and in the following groups: four first-line drugs (isoniazid, rifampicin, ethambutol, and streptomycin); at least one second-line fluoroquinolone (ofloxacin or ciprofloxacin); at least one second-line in jectable drug (kanamycin, capreomycin, and amikacin); all second-line injectable drugs; any second-line drug; at least one other oral second-line drug (aminosalicylic acid or ethi onamide); and combinations for XDR tuberculosis. Each drug and combination represents a separate analysis of risk factors. In this report we focus on resistance to any fluoroquinolone, any second-line injectable drug, any other oral second-line drug, and risk of XDR tuberculosis. For continuous variables we calculated means and SDs, medians with ranges and IQRs, and specified percentiles. For categorical data, we tabulated two-way frequency distributions for each characteristic versus each drugsusceptibility result, and used Pearsons statistic or Fishers exact test to test significance, as appropriate. For ordinal variables, the p values for trend were based on the Mantel-Haenszel test. For continuous variables, we compared two groups with Students t test if the data were normally distributed with at least 30 observations per group; otherwise, we used the Wilcoxons rank sum test. For analyses that involved more than two groups, we used
5

www.thelancet.com Published online August 30, 2012 http://dx.doi.org/10.1016/S0140-6736(12)60734-X

Articles

Weighted frequency of resistance Yes No Sex Male Female Yes No Yes No Yes No Yes No Yes No No Yes Yes No Unilateral Bilateral Unemployed 167 140 87 242 131 222 42 264 62 256 291 83 60 314 202 172 28 248 191 183

Weighted percentage for resistance (95% CI) 120 (94145) 178 (139216) 120 (96145) 182 (142223) 121 (50192) 131 (106155) 146 (111180) 129 (98161) 149 (96202) 130 (107153) 165 (121210) 124 (101148) 100 (52148) 175 (143208)

p value

Risk ratio (95% CI)

Weighted frequency of resistance

Weighted percentage for resistance (95% CI)

p value

Risk ratio (95% CI)

Green Light Committee approval 001 068 (050091)

(Continued from previous column) Cavitary disease on chest radiograph Unilateral Bilateral 00072 066 (049089) 080 049 049 009 003 092 (050171) 113 (080158) 115 (077171) 133 (096185) 057 (034096) No cavity Positive Negative Yes No Yes No Yes No Yes No 159 71 95 302 66 100 240 134 205 108 231 32 307 153 (117189) 119 (76162) 114 (81147) 136 (113158) 199 (123274) 408 (301514) 111 (90131) 387 (305469) 100 (79120) 012 085 008 134 (092195) 104 (066167) 068 (045103)

Sputum-smear test results at enrolment

History of imprisonment

Previous treatment with second-line injectable drugs <00001 368 (267507)

Previous treatment with fluoroquinolones <00001 389 (289523)

Current alcohol abuse

Previous treatment with another oral second-line drug 358 (267449) <00001 327 (238449) 110 (89131) 626 (417835) 130 (108152)

Current tobacco use

Previous treatment with a third-line drug <00001 481 (331698)

HIV infection

First time treated for MDR tuberculosis 370 (246494) <00001 333 (227489) 111 (91131) 169 (139200) 92 (69115) 134 (85183) 147 (123172)

In hospital at enrolment <00001 184 (135250) 064 091 (061136)

MDR=multidrug-resistant. *Age, marital status, education, occupation risk, homelessness, contact with a tuberculosis or MDR tuberculosis patient, previous surgery for tuberculosis, diabetes mellitus, and comorbidities were not significantly associated with fluoroquinolone resistance (p>01; data not shown). Site-specific sampling weights were calculated as the total number of eligible cases during the enrolment period divided by the number of patients enrolled.

Table 3: Risk factors for resistance to fluoroquinolones at baseline*

Pulmonary radiographic abnormality

Role of the funding source

The US Agency for International Development had no role in the design, implementation, analysis, and interpretation of results. CDC Division of Tuberculosis Elimination led the study design, training for data collection and monitoring, data analysis, data inter pretation, and writing of the report. Other sponsors had no roles in these activities. The corresponding author had full access to the data in the study and had final responsibility for the decision to submit for publication.

(Continues in next column)

one-way analysis of variance and the Kruskal-Wallis test. Site-specific selec tion fractions were calculated as the number of patients enrolled divided by total number patients eligible during the enrolment period. Statistical comparisons across countries were based on data weighted by the reciprocal of the selection fraction.15,16 We took p values of 005 or less to be significant, and those of 0001 or less to be highly significant. We report actual numbers and percentages for descriptions of patients characteristics and drug resistance for each country, and weighted numbers and percentages for results from statistical analyses of risk factors across sites. To assess whether the prevalence of drug resistance and relations with risk factors and drug-susceptibility results differed in the two sampling protocols, we stratified the analysis by whether a follow-up positive culture was available or not. We used the Breslow-Day test for homo geneity to determine whether relative-risk estimates across strata were significantly different from each other.15
6

Results
Of 3034 eligible patients, 1782 (587%) were enrolled (figure). The most common reasons for patients not to be enrolled were no positive follow-up culture, no consent, previous treatment with second-line drugs for MDR tuberculosis, and staff turnover, although exact numbers for the different reasons are not available. Baseline isolates were not shipped to CDC for 242 (135%) patients. The centres in Estonia, Latvia, South Africa, and Masan, South Korea, maintained the original follow-up culture protocol throughout the study. Of the 1540 baseline isolates received and recovered successfully at CDC, MDR tuber culosis was confirmed in 1278 (830%, figure, table 1).

www.thelancet.com Published online August 30, 2012 http://dx.doi.org/10.1016/S0140-6736(12)60734-X

Articles

The characteristics of patients with MDR tuberculosis isolates assessed at CDC are shown in table 1. In every country, more patients were male than female. The highest proportion of patients were in the 2544-year age group, and the overall median age was 37 years (range 1881). Social characteristics of patients differed between countries. Unemployment was lowest in Thai land and highest in South Africa. History of imprison ment was lowest in South Korea and highest in Russia. Homelessness was uncommon, but reached 120% in Latvia and 174% in Estonia. Alcohol abuse varied widely (range 11 [28%] of 397 to 64 [640%] of 100), as did tobacco use (range 0 to 40 [870%] of 46). Patients infected with HIV were enrolled in all countries except the Philippines and South Korea, where HIV testing was not routine but those tested were negative. 145 (863%) of 168 patients with HIV infections lived in South Africa. In Estonia, Latvia, Russia, South Africa, and Masan, South Korea, most patients were in hospital
Weighted frequency of resistance Yes No Sex Male Female Yes No Yes No Yes No Yes No Yes No No Yes Yes No Unilateral Bilateral 418 261 108 458 368 216 248 381 296 364 147 507 67 562 596 83 89 583 249 (214283) 278 (230325) 467 (355580) 241 (210272) 320 (274366) 200 (161239) 425 (352499) 204 (175233) 374 (314435) 204 (174234) 350 (273426) 336 (295377) 403 (275531) 243 (214272) 347 (308387) 91 (68115) 198 (139257) 273 (242305) 033 090 (072112) 372 307 Weighted percentage for resistance (95% CI) 234 (199269) 298 (251344) p value Risk ratio (95% CI)

at the time of enrolment, but in the other countries very few had been admitted. 1199 (938%) of 1278 patients had a history of tuberculosis, with percentages ranging from 478% to 1000% across countries. Of the 1199 patients, most (706%) had had one or two previous tuberculosis episodes. 1186 (928%) of 1278 patients had received first-line antituberculosis drugs before the study. By contrast, only 195 (153%) had received second-line drugs, with the lowest percentage being 27% in South Africa and the highest 535% in South Korea. The prevalence of resistance varied substantially between countries (table 2). 625 (490%) of 1278 M tuber culosis isolates were resistant to ethambutol and streptomycin as well as isoniazid and rifampicin. Resist ance to any second-line drug was 437% and ranged from 333% in Thailand to 620% in Latvia. The prevalence of resistance to fluoroquinolones was 129% and was lowest in the Philippines and highest in South Korea. Resistance to at least one second-line injectable drug was 200% overall, with the lower prevalence being in the Philippines and the highest in Latvia. Resistance

Green Light Committee approval 003 079 (063098)

Weighted frequency of resistance Cavitary disesae Unilateral Bilateral No cavity Positive 321 173 141 566 105 110 499 124 485 128 480 26 583

Weighted percentage for resistance (95% CI)

p value

Risk ratio (95% CI)

(Continued from previous column) 308 (261356) 290 (226354) 169 (129209) 254 (224284) 318 (232404) <00001 182 (137243) 00012 172 (124238) 015 080 (059107)

History of imprisonment <00001 194 (147255)

Sputum-smear test result at enrolment Negative Yes

Unemployed 00001 160 (125204)

Previous treatment with second-line injectable drugs 449 (342556) <00001 195 (149256) 230 (201259) 358 (275441) 235 (205265) 423 (329518) 228 (199257) 501 (283719) 247 (219276) No Yes

Current alcohol abuse <00001 208 (166260)

Previous treatment with fluoroquinolones 00031 153 (117199) No Yes

Current tobacco use <00001 184 (148228) 076 104 (081134)

Previous treatment with another oral second-line drug <00001 186 (143241) No Yes

HIV infection

Previous treatment with a third-line drug 00096 203 (129319) No

First time treated for MDR tuberculosis 00076 166 (118233)

In hospital at enrolment <00001 380 (286504) 004 073 (053100)

Pulmonary radiographic abnormality

MDR=multidrug-resistant. *Age, marital status, education, occupation risk, homelessness, contact with a tuberculosis or MDR tuberculosis patient, previous surgery for tuberculosis, diabetes mellitus, and comorbidities were not significantly associated with resistance to second-line injectable drugs (diabetes mellitus p=002, other characteristics p>01; data not shown). Site-specific sampling weights were calculated as the total number of eligible cases during the enrolment period divided by the number of patients enrolled.

(Continues in next column)

Table 4: Risk factors for resistance to second-line injectable drugs at baseline*

www.thelancet.com Published online August 30, 2012 http://dx.doi.org/10.1016/S0140-6736(12)60734-X

Articles

to all three second-line injectable drugs was significantly more frequent in the Eastern Cape province, South Africa, than in the other South African provinces (65 [489%] of 133 vs 10 [63%] of 160, p<00001). Resistance to other oral second-line drugs was seen in all countries, with the aggregate prevalence being 271% (range 130380%). XDR tuberculosis was seen in 86 (67%) of 1278 patients overall, and was least prevalent in the Philippines and most prevalent in South Korea (table 2). Findings for potential risk factors are shown in tables 36. The strongest, most-consistent risk factor was previous treatment for MDR tuberculosis with any second-line drug, and the risk remained significant when fluoroquinolones, second-line injectable drugs and other oral second-line drugs were assessed separately. Resistance to fluoroquino lones and secondline injectable drugs and XDR tuberculosis, but not resistance to other oral second-line drugs, were
Weighted frequency of resistance Yes No Sex Male Female Yes No Yes No Yes No Yes No Yes No No Yes Yes No Unilateral Bilateral 96 118 20 132 110 70 54 133 71 135 39 161 38 145 185 29 31 183 57 (3975) 125 (89161) 87 (27146) 70 (5189) 95 (66124) 65 (4188) 93 (48137) 71 (5390) 90 (55125) 75 (5695) 93 (47139) 107 (80134) 00002 046 (030070) 057 010 035 047 061 124 (059261) 148 (092237) 130 (075223) 119 (074191) 087 (050151) 87 126 Weighted percentage for resistance (95% CI) 55 (3674) 123 (89156) p value Risk ratio (95% CI)

significantly less prevalent in countries with than in those without Green Light Committee approved projects. This difference was due to the very low prevalence of resistance to second-line drugs in the Philippines, which had the largest Green Light Committee project. Being in hospital at the time of enrolment was a strong risk factor for resistance to fluoroquinolones and second-line injectable drugs, and for XDR tuberculosis, but not for resistance to other oral second-line drugs. Other, risk factors differed between drugs and countries (tables 36). For instance, fluoro quinolone resistance and XDR tuberculosis were more frequent in women than in men. Patients with HIV infection had significantly less fluoroquinolone resist ance than those not infected with HIV. Cavitary lung disease nearly doubled the risk of resistance to second-line injectable agents. Other risk factors for resistance to second-line injectable drugs included unemployment, a history of imprisonment, alcohol abuse, and tobacco use (table 4). Little difference was seen between patients with one or at least two positive cultures in terms of association of specific risks with specific drug resistance.
Weighted frequency of resistance Cavitary disease Unilateral Bilateral No Cavity Positive Negative Yes No Yes No Yes No Yes No 89 46 53 173 39 69 129 82 116 73 125 16 182 86 (57114) 76 (40113) 63 (3789) 78 (6096) 118 (55181) 026 055 016 136 (080233) 122 (064230) 066 (037118) Weighted percentage for resistance (95% CI) p value Risk ratio (95% CI)

Green Light Committee approval 00002 045 (029069)

(Continued from previous column)

History of imprisonment

Unemployed

Sputum-smear test result at enrolment

Current alcohol abuse

Previous treatment with second-line injectable drugs 283 (183384) <00001 475 (305742) 60 (4475) 237 (161313) 56 (4072) 241 (157324) 59 (4376)

Current tobacco use

Previous treatment with fluoroquinolones <00001 421 (273649)

HIV infection

Previous treatment with another oral second-line drug <00001 405 (260631)

First time treated for MDR tuberculosis 230 (114346) <00001 366 (208643) 63 (4779) 108 (82133) 32 (1846) 69 (33106) 86 (66106)

Previous treatment with a third-line drug 322 (118526) <00001 418 (213821) 77 (5995)

In hospital at enrolment <00001 336 (203557) 047 081 (045144)

Pulmonary radiographic abnormality

MDR=multidrug-resistant. *Age, marital status, education, occupation risk, homelessness, contact with a tuberculosis or MDR tuberculosis patient, previous surgery for tuberculosis, diabetes mellitus, and comorbidities were not significantly associated with extensively drug-resistant tuberculosis (p>01 ; data not shown). Site-specific sampling weights were calculated as the total number of eligible cases during the enrolment period divided by the number of patients enrolled.

(Continues in next column)

Table 5: Risk factors for extensively drug-resistant tuberculosis at baseline*

www.thelancet.com Published online August 30, 2012 http://dx.doi.org/10.1016/S0140-6736(12)60734-X

Articles

Weighted frequency of resistance Yes No Sex Male Female Yes No Yes No Yes No Yes No Yes No No Yes Yes No 494 230 67 500 285 312 156 518 229 478 95 417 62 592 485 239 467 257

Weighted percentage for resistance (95% CI) 293 (258329) 249 (207291) 294 (259329) 244 (202287) 290 (183398) 264 (233294) 249 (207290) 290 (249331) 267 (201334) 278 (248308) 290 (233347) 268 (238299) 226 (160292) 276 (238315) 370 (245496) 257 (229284) 282 (245319) 264 (229299) 241 (178303) 284 (253314)

p value

Risk ratio (95% CI)

Weighted frequency of resistance Cavitary disease Unilateral 265 Bilateral 179 215 624 84 105 570 142 533 136 539 32 643 None Positive

Weighted percentage for resistance (95% CI)

p value

Risk ratio (95% CI)

Green Light Committee approval 011 008 063 017 078 050 022 006 048 024 118 (096145) 120 (097148) 110 (075162) 086 (069107) 096 (073126) 108 (086136) 082 (059113) 144 (101206) 107 (089129) 085 (064112)

(Continued from previous column) 255 (212298) 300 (238361) 257 (214300) 280 (251309) 255 (173336) 428 (321535) 263 (235292) 409 (326492) 258 (229288) 095 026 058 099 (078126) 117 (089152) 110 (078154)

Sputum-smear test results at enrolment Negative Yes No Yes No Yes No Yes No

History of imprisonment

Previous treatment with second-line injectable drugs 00013 163 (124213)

Unemployed

Previous treatment with fluoroquinolones 00003 158 (126200)

Current alcohol abuse

Previous treatment with another oral second-line drug 448 (354543) <00001 175 (138222) 256 (227284) 628 (432825) 272 (244300)

Current tobacco use

Previous treatment with a third-line drug 00001 231 (166321)

HIV infection

First time treated for MDR tuberculosis

In hospital at enrolment

MDR=multidrug-resistant. *Age, marital status, education, occupation risk, homelessness, contact with a tuberculosis or MDR tuberculosis patient, previous surgery for tuberculosis, diabetes mellitus, and comorbidities were not significantly associated with resistance to other oral second-line drugs (diabetes mellitus p=006, other characteristics p>01; data not shown). Site-specific sampling weights were calculated as the total number of eligible cases during the enrolment period divided by the number of patients enrolled.

Pulmonary radiographic abnormality Unilateral 108 Bilateral 605

Table 6: Risk factors for resistance to other oral second-line drugs at baseline*

(Continues in next column)

Discussion
This large, prospective study of resistance to second-line drugs for MDR tuberculosis shows comprehensively that the prevalence of resistance is high (437%), and that the risk of XDR tuberculosis (67%) in the eight countries studied is worrying. The prevalence of drug resistance correlates with the time that second-line drugs have been available in each country. They had been available for 10 years or less in Thailand (7 years), the Philippines (9 years), and Peru (10 years), and these countries had the lowest rates of resistance. By contrast, South Korea and Russia had the longest histories of availability (more than 20 years) and the highest rates of resistance. Other practices, including criteria for treat ment, admission to hospital, directly observed therapy, and drug procurement, should be assessed to find out whether they affect resistance rates. WHO data show that 54% of patients with MDR tuberculosis have XDR tuberculosis.2 In our population,

67% had XDR tuberculosis. This higher rate might be due at least partly to differences in laboratory procedures. We tested all three second-line injectable agents for this study, but most countries test one or two, which could underestimate the burden of XDR tuberculosis. The same may be said for fluoroquinolones. The prevalence values we found show some differences from and similarities to country-specific surveillance data from WHO. Fluoroquinolone resistance was 261% in Estonia, 140% in Latvia, and 126% in South Africa, and 306%, 156%, and 142% in WHO data. For XDR tuberculosis, however, although the rates for South Africa are similar (106% vs 105%), those for Estonia and Latvia were lower in our study (65% vs 125% and 80% vs 148%, respectively).2 Thus, prevalence of resistance to second-line drugs in the three countries was not likely to be artificially increased because of the two-culture criterion. Another study of MDR tuberculosis in Estonia showed a similar rate of 52% for XDR tuberculosis.17 Few studies have been done of resistance to secondline drugs, probably because of low capacity for laboratory testing.1 Previous studies have reported prevalence of
9

www.thelancet.com Published online August 30, 2012 http://dx.doi.org/10.1016/S0140-6736(12)60734-X

Articles

23% in South Korea and of 6% in Peru for XDR tuberculosis among patients with MDR tuber culosis treated at tertiary referral hospitals.6,18 In these countries we showed 15% and 6% prevalence, respect ively. A study done in Thailand before XDR tuberculosis was defined found that 9% of patients with MDR tuberculosis had resistance to a fluoroquinolone and 5% to kanamycin.19 These values are similar to those in our study. Previous treatment for MDR tuberculosis with a secondline drug was the strongest risk factor for resistance, which is consistent with previous reports.6,20 Patients being in hospital at enrolment was also strongly associated with resistance, possibly because of noso comial transmission or disease severity. Women had greater prevalence of fluoroquinolone resistance than men, and thereby greater risk of XDR tuberculosis, which is consistent with the findings of a study done in South Korea.6 By contrast, HIVinfected patients were less likely than other patients to have resistance to fluoroquinolones, but in other studies HIV infection has been a strong risk factor for XDR tuberculosis.20 Unlike fluoroquinolones, resistance to second-line in jectables was associated with social factors,
Panel: Research in context Systematic review We searched PubMed with the search term (tuberculosis OR TB) AND (extensive drug resistance OR XDR OR second-line drug resistance OR fluoroquinolone resistance OR kanamycin resistance OR amikacin resistance OR capreomycin resistance) AND (epidemiology OR prevalence OR risk factors). The search identified 568 publications. Of these, 85 articles contained original data on the epidemiology of drug-resistant tuberculosis, including information on resistance to second-line drugs. The remainder were reviews, editorials, letters, studies focused on treatment and treatment outcomes, phylogenetic and transmission studies, and case reports or small case series. Interpretation Of 85 articles on the epidemiology of resistance to second-line drugs, 60 were retrospective reviews based on medical records or laboratory records at tertiary referral hospitals, specialised tuberculosis hospitals, and mycobacteriology reference laboratories, and data had been recorded previously for other purposes. Thus, they reflected highly selected groups of patients and did not represent the general population. Seven reports based on national or multinational surveillance systems included little information about risk factors because they were limited to routinely captured data. With one exception, South Korea, susceptibility testing for second-line drugs is not done routinely, and, therefore, resistance data are not routinely captured by surveillance systems. 16 studies focused on specific classes of second-line drugs (usually fluoroquinolones) or on extensively drug-resistant tuberculosis, but did not assess individual drugs. Seven publications focused narrowly on specific groups such as prisoners, miners, health-care workers, people with HIV infection, and migrants and another seven phylogenetic analyses focused on transmission dynamics, including contact investigations, and on molecular characterisation of specific DNA mutations associated with phenotypic resistance to specific individual drugs. These 85 papers represented little geographical overlap with our study 72 (85%) reported data from countries or regions not included in this study, and design limitations in ten of the remaining 13 reports meant little population crossover. Thus, our report adds prospective, population-based data from many locations not previously studied that include detailed information on risk factors related to resistance to individual drugs as well as drug combinations, according to centralised laboratory testing.

including imprisonment, unemployment, alcohol abuse, and smoking. Social factors should be taken into account in the management of tuberculosis. PETTS had important limitations. The prospective gathering of data under programmatic conditions led to some variability between sites in the information avail able. Differences in demographic, social, and clinical risk factors might be related to the extent of missing data for specific variables. Data collection was based on medical records, where some features are not routinely recorded and we could not acquire the data. However, of the variables that applied to all patients, only six had more than 10% of data missing. Data from Masan, South Korea, were extracted from a separate study that was being done in collaboration with the US National Institutes of Health and, therefore, we used their data collection instrument, which included all variables except years of education and number of children.6 When we expanded the enrolment criteria in November 2005, Estonia, Latvia, and Masan, South Korea, were close to their enrolment targets and maintained the original protocol of requiring a second positive culture per patient. South Africa began enrolling patients with one culture, but only shipped samples for those with a second positive culture to CDC. The rest of the sites changed protocols and required only a baseline positive culture. This difference might have contributed to country-specific differences. The patients tested might not have been representative of the larger populations of adults with pulmonary MDR tuberculosis in the study countries to the extent that the prevalence of drug resistance among enrolled patients differed from that in patients who were not enrolled. We could not assess how representative our patient cohorts were because we did not collect demographic and medical information for eligible patients who were not enrolled. The enrolment rates at some centres were low owing to circumstances not related to the study, such as changes in personnel. Finally, the results are not generalisable to the world as a whole because India and Chinathe countries with the highest numbers of tuberculosis casesdid not participate. India and China had pilot projects for MDR tuberculosis approved by the Green Light Committee, but not until 3 and 4 years, respectively, after PETTS started. Population-based data on resistance to second-line drugs in these two countries are limited. In Chinas 2007 national drug resistance survey, 274% (95% CI 231321) of cases of MDR tuberculosis tested for resistance to secondline drugs showed fluoro quinolone resistance and 72% of patients (49102) had XDR tuberculosis.2 In India, a 2006 population-based survey of tuberculosis drug resistance in Gujarat State reported fluoroquinolone resistance in 241% of cases (185303) and XDR tuberculosis in 32% (1266).2 In both surveys, fluoro quinolone resistance was near the high end of the range of values in our study. The prevalence of XDR tuberculosis in China was slightly higher than the average value in our study, but in Gujarat State, India, it was similar to the

10

www.thelancet.com Published online August 30, 2012 http://dx.doi.org/10.1016/S0140-6736(12)60734-X

Articles

lower values in our study. Other reports from China and India have been based on retrospective reviews of cases at specialised tuberculosis referral centres that happened to have drug-susceptibility results for second-line drugs and are not representative of the general population.2124 As a large, multicentre, collaborative study, PETTS also has strengths. All drug-susceptibility testing was done in one reference laboratory with standard, quality-controlled methods, which eliminated variability in trinsic to pheno typic testing in multiple laboratories. Additionally, the study was designed to provide data within defined criteria that were representative for the populations served by the participating programmes. Five of the participating countries had projects approved by the Green Light Committee in place at the time of the study and, there fore, were representative of other countries with approved projects. Of the countries without approved projects, one was a high-income and two were upper-middle-income countries and would not be representative of the world wide situation, especially for low-income countries. Nevertheless, our country-specific results can be extrapo lated to guide in-country policy for laboratory capacity and for designing effective treatment recommendations for MDR tuberculosis. PETTS continues, and follow-up isolates are being tested to investigate the frequency of and risk factors for acquired resistance to second-line drugs in patients with MDR tuberculosis. The effect of the Green Light Committee initiative in combating acquired resistance to second-line drugs, the timing of acquired resistance, and the role of specific genetic mutations in different regions of the world are also being assessed.
Contributors PC and Katherine Tan designed the initial study proposal. TD, PC, JCC, JE, Alison Taylor, Katherine Tan, EK, CK, Melanie Wolfgang, Carmen Contreras, and Joey Lancaster contributed to the study design, database development, data collection, monitoring, training, data analysis, and data interpretation. TD and Lois Diem did the drug-susceptibility testing at the Centers for Disease Control and Prevention. Michael Chen provided statistical consultation. SA, JCC, MTG, BYK, HJK, KK,VL, GVV, MAY, Rattanawadee Akkslip, Wanpen Wattanaamornkiet, Jaime Bayona, Carmen Contreras, Seonyoung Min, Tatiana Khorosheva, Elena Kyryanova, Thelma Tupasi, Ingrida Sture, Tiina Kummik, Tatiana Kuznetsova, and Tatiana Somova were responsible for patients enrolment and treatment at the study centres. LA, LEV, Wanlaya Sitti, Sofia Andreevskaya, Larisa Chernousova, Elea Larionova, Tatyana Smirnova, Alena Vorobyeva, Isdore Shamputa, Jeanette Brand, Eunjin Cho, Seok Yong Eum, Hyun Kyung Kwak, Jongseok Lee, Evgenia Nemtsova, Grace Egos, Chang-ki Kim, Inga Norvaisa, Girts Skenders, Klavdia Levina, Gloria Yale, Gustavo Pariona, Carmen Suarez, and Eddy Valencia were responsible for local laboratory testing and shipping cultures to the Centers for Disease Control and Prevention. All authors contributed to data collection, study coordination, and critical revision of the paper. Conflicts of interest We declare that we have no conflicts of interest. Global PETTS Investigators and coauthors (continued from page 1) Rattanawadee Akksilp, Wanlaya Sitti, and Wanpen Wattanaamornkiet, Office of Disease Prevention and Control Region 7, Ubon Ratchatani, Thailand; Sofia N Andreevskaya, Larisa N Chernousova, Olga V Demikhova, Elena E Larionova, Tatyana G Smirnova, Irina A Vasilieva, and Alena V Vorobyeva, Central Tuberculosis

Research Institute, Russian Academy of Medical Sciences, Moscow, Russian Federation; Clifton E Barry III, Ying Cai, and Isdore C Shamputa, National Institute for Allergy and Infectious Disease, National Institutes of Health, Bethesda, MD, USA; Jaime Bayona and Carmen Contreras, Socios en Salud Sucursal, Lima, Peru; Cesar Bonilla and Oswaldo Jave, Ministry of Health, National TB Strategy, Lima, Peru; Jeannette Brand, Joey Lancaster, and Ronel Odendaal, Medical Research Council, Pretoria, South Africa; Michael P Chen, Lois Diem, Beverly Metchock, Kathrine Tan, Allison Taylor and Melanie Wolfgang, Centers for Disease Control and Prevention, Division of TB Elimination, Atlanta, GA, USA; Eunjin Cho, Seok Yong Eum, Hyun Kyung Kwak, Jiim Lee, Jongseok Lee, and Seonyeong Min, International Tuberculosis Research Center, Masan and Yonsei University College of Medicine, Seoul, South Korea; Irina Degtyareva, Evgenia S Nemtsova, Tatiana Khorosheva, and Elena V Kyryanova, Orel Oblast Tuberculosis Dispensary, Orel, Russian Federation; Grace Egos, Ma Therese C Perez, and Thelma Tupasi, Tropical Disease Foundation, Manila, Philippines; Soo Hee Hwang, National Masan Tuberculosis Hospital, Masan, South Korea; Chang-ki Kim, Su Young Kim, and Hee Jeong Lee, Korean Institute of Tuberculosis, Seoul, South Korea; Liga Kuksa, Inga Norvaisha, Girts Skenders, and Ingrida Sture, State Agency, Infectology Centre of Latvia, Tuberculosis and Lung Disease Clinic, Riga, Latvia; Tiina Kummik, Tartu University Hospital, Tartu, Estonia; Tatiana Kuznetsova and Tatiana Somova, Vladimir Oblast Tuberculosis Dispensary, Vladimir, Russian Federation; Klavdia Levina, North Estonia Regional Hospital, Tallinn, Estonia; Gustavo Pariona and Gloria Yale, Lima City Health District Reference Laboratory, Lima, Peru; Carmen Suarez, Lima East Health District Reference Laboratory, Lima, Peru; Eddy Valencia, National Tuberculosis Reference Laboratory, Lima, Peru; and Piret Viiklepp, National Tuberculosis Registry, National Institute for Health Development, Tallinn, Estonia. Acknowledgments We thank the patients who gave their time and energy to contribute to this study and the doctors, nurses, and microbiologists at each of the enrolment sites for their contributions to this work. We thank the people in the following countries and organisations for support and contributions: Centers for Disease Control and Prevention Division of Tuberculosis Elimination, US Agency for International Development, National Institute of Allergy and Infectious Diseases, North Estonia Regional Hospital, Tartu University Hospital, Estonia National Tuberculosis Registry, Estonia National Institute for Health Development, State Agency Infectology Centre of Latvia, Riga Tuberculosis and Lung Disease Clinic, Lima Ciudad and Lima Este Health Districts and reference laboratories, Philippines Tropical Disease Foundation, Orel and Vladimir Oblast Tuberculosis Dispensaries, Central Tuberculosis Research Institute of the Russian Academy of Medical Sciences, South Africa Medical Research Council, KwaZulu-Natal King George V Hospital, Klerksdorp Hospital, Witbank Specialised Tuberculosis Hospital, Jose Pearson Hospital, Korean Institute of Tuberculosis, National Masan Tuberculosis Hospital, Thai Office of Disease Prevention and Control, and WHO in Switzerland, Denmark, Peru, and Russia. We also thank medical and nursing staff in the Ubon Ratchathani, Srisaket, Sakon Nakon, Yasothon Provinces, Thailand. References 1 Wright A, Zignol M, Van Deun A, et al. Epidemiology of antituberculosis drug resistance 200207: an updated analysis of the Global Project on Anti-Tuberculosis Drug Resistance Surveillance. Lancet 2009; 373: 186173. 2 WHO. Multidrug and extensively drug-resistant TB (M/XDR-TB). 2010 Global Report on Surveillance and response. Geneva: World Health Organization, 2010. 3 GLC Program applications. http://www.who.int/tb/challenges/ mdr/greenlightcommittee/report_glc_applications_apr2011rev1.pdf (accessed June 1, 2012). 4 Gupta R, Cegielski JP, Espinal MA, et al. Increasing transparency in partnerships for healthintroducing the Green Light Committee. Trop Med Int Health 2002; 7: 97076. 5 Shah NS, Wright A, Bai G-H, et al. Worldwide emergence of extensively drug-resistant tuberculosis. Emerg Infect Dis 2007; 13: 38087.

www.thelancet.com Published online August 30, 2012 http://dx.doi.org/10.1016/S0140-6736(12)60734-X

11

Articles

7 8 9 10 11 12

13 14

15

Jeon C, Hwang S, Min J, et al. Extensively drug-resistant tuberculosis in South Korea: risk factors and treatment outcomes among patients at a tertiary referral hospital. Clin Infect Dis 2008; 46: 4249. WHO. Extensively drug-resistant tuberculosis (XDR-TB): recommendations for prevention and control. Wkly Epidemiol Rec 2006; 81: 43032. WHO. Addressing the threat of tuberculosis caused by extensively drug-resistant Mycobacterium tuberculosis. Wkly Epidemiol Rec 2006; 81: 38690. WHO. Anti-tuberculosis drug resistance in the world: fourth global report. Geneva: World Health Organization, 2008. WHO. Global tuberculosis control: WHO report 2011. Geneva: World Health Organization, 2011. CLSI. Susceptibility testing of Mycobacteria, Nocardiae, and other aerobic actinomycetes: approved standard, 2nd edn. Wayne, PA: Clinical Laboratory and Standards Institute, 2011. Kent PT, Kubica GP. Public health mycobacteriology: a guide for the level III laboratory Atlanta, GA: US Department of Health and Human Services, Public Health Service, Centers for Disease Control and Prevention, 1985. Post F, Wilcox P, Mathema B, et al. Genetic polymorphism in M tuberculosis isolates from patients with chronic multidrug-resistant tuberculosis. J Infect Dis 2004; 190: 99106. Han LL, Sloutsky A, Canales R, et al. Acquisition of drug resistance in multidrug-resistance Mycobacterium tueruclosis during directly observed empiric retreatment with standardized regimens. Int J Tuberc Lung Dis 2005; 9: 81821. Fleiss JL, Levin B, Paik MC. Statistical methods for rates and proportions, 3rd edn. Hoboken, NJ: Wiley-Interscience, 2003.

16 Stokes ME, Davis CS, Koch GG. Categorical data analysis using the SAS System, 2nd edn. Cary, NC: SAS Publishing, 2009. 17 Kliiman K, Altraja A. Predictors of poor treatment outcome in multi- and extensively drug-resistant pulmonary TB. Eur Respir J 2009; 33: 108594. 18 Bonilla C, Crossa A, Jave H, et al. Management of extensively drug-resistant tuberculosis in Peru: cure is possible. PLoS One 2008; 3: e2957. 19 Prammananan T, Arjratanakool W, Chaiprasert A, et al. Second-line drug susceptibilities of Thai multidrug-resistant Mycobacterium tuberculosis isolates. Int J Tuberc Lung Dis 2005; 9: 21619. 20 Andrews J, Shah NS, Weissman D, Moll A, Friedland G, Gandhi N. Predictors of multidrug- and extensively drug-resistant tuberculosis in a high HIV prevalence community. PLoS One 2010; 5: e15735. 21 Liu CH, Yang N, Wang Q, et al. Risk factors associated with fluoroquinolone-resistant tuberculosis in a Beijing tuberculosis referral hospital. Respirology 2011; 16: 91825. 22 Yu HT, Wang Q, Yang N, Li HM, Liu CH. Risk factors associated with kanamycin-resistant tuberculosis in a Beijing tuberculosis referral hospital. J Med Microbiol 2012; published online March 15. DOI:10.1099/jmm.0.042655-0. 23 Paramasivan CN, Rehman F, Wares F, et al. First- and second-line drug resistance patterns among previously treated tuberculosis patients in India. Int J Tuberc Lung Dis 2010; 14: 24346. 24 James P, Gupta R, Christopher DJ, Thankagunam B, Veeraraghavan B. MDR- and XDR-TB among suspected drug-resistant TB patients in a tertiary care hospital in India. Clin Respir J 2011; 5: 1925.

12

www.thelancet.com Published online August 30, 2012 http://dx.doi.org/10.1016/S0140-6736(12)60734-X

Comment

Unexpected high levels of multidrug-resistant tuberculosis present new challenges for tuberculosis control
Most international recommendations for tuberculosis control have been developed for multidrug-resistant (MDR) tuberculosis prevalence of up to around 5%. Yet we now face prevalence up to ten times higher in some places, where almost half of the patients with infectious disease are transmitting MDR strains of Mycobacterium tuberculosis.1 The highest prevalence of MDR tuber culosis documented to date, 478%, was reported in 2011, in Minsk, Belarus.1 Among patients with infectious (smear-positive) pulmonary disease, MDR tuberculosis was seen in 353% of newly detected cases and in no less than 765% of previously treated patients.1 However, MDR tuberculosis is not an issue isolated in one city or country, but reflects a wider public health threat resulting from severely resistant forms of M tuberculosis. To adequately address MDR tuberculosis, more solid epidemiological infor mation is needed to increase overall understanding of disease development and transmission. In The Lancet, Tracy Dalton and colleagues2 report data on the prevalence of, and risk factors for, resistance to second-line drugs in 1278 patients with MDR tuberculosis from Estonia, Latvia, Peru, the Philippines, Russia, South Africa, South Korea, Thailand, and Taiwanalthough drug susceptibility data from Taiwan were not available at the time of the data analysis and are not included in the report. Isolates from all patients were shipped to the tuberculosis laboratory at the US Centers for Disease Control and Prevention and tested for susceptibility to 11 antituberculosis drugs. Risk factors for being infected with a strain resistant to second-line drugs as well as for having extensively drug resistant (XDR) tuberculosis were assessed. Differences between the countries in the use of antituberculosis drugs, and their combinations, as well as the use of some of these agents for treatment of other bacterial infections, could be expected to lead to different drug-resistance profiles in isolates in the various countries studied. Such a difference was seen, for example, for fluoroquinolones: overall 129% of patients had resistant strains and the prevalence ranged from 71% in the Philippines to 323% in South Korea. For resistance to at least one second-line injectable drug, the corresponding values were 200% overall, ranging from 20% in the Philippines to 470% in Latvia. Similar differences were also seen for XDR tuberculosis. Some notable differences in risk factors were reported. As would be expected, the strongest risk factor for MDR tuberculosis was a history of tuber culosis, for which the overall value was 938% and ranged from 922% to 1000% in six of the countries in Dalton and colleagues study. In Estonia and Latvia, however, the values were only 478% and 780%, respectively. This difference might indicate greater current transmission of MDR tuberculosis in these two countries than in the others.3,4 A wide variation was also seen in the percentage of patients with HIV infection, ranging from 495% in South Africa to zero in South Korea and the Philippines. Such local differences in drugresistance and epidemiological characteristics should be taken into account when international guidelines for diagnostic algorithms, infection-control measures, and treatment recommendations for drug-resistant tuber culosis are adapted. The findings of Dalton and colleagues add to the worldwide data on MDR tuberculosis from WHOs 2008 report, which provided data for 200207 from about 80 countries.5 In this period, the overall worldwide proportion of MDR tuberculosis was 53%, and among new cases 29% were MDR strains. The values varied strikingly in different regions. For instance, eastern Europe had the highest proportion of new cases of tuberculosis that were MDR cases (generally more than 10%), but in most other regions the percentage was lower than 20%. The top ten locations for primary MDR tuberculosis were in the former Soviet Union region. In an update, WHO concluded that rates of MDR tuberculosis had risen in 2009 and 2010, but that trends were still unclear in most areas.6 The WHO update also showed that 94% of all cases of MDR tuberculosis were XDR tuberculosis. In 2009, the Global Project on Anti-Tuberculosis Drug Resistance Surveillance showed that several countries in the former Soviet Union region reported significant numbers of patients with XDR tuberculosis; the prevalence of XDR tuberculosis among patients with MDR tuberculosis ranged from 66% to 237%.7 Similar rates were seen by Dalton and colleagues2 who reported that the proportion of XDR tuberculosis among cases of MDR tuberculosis
Karin Schermbrucker/AP/Press Association Images

Published Online August 30, 2012 http://dx.doi.org/10.1016/ S0140-6736(12)61069-1 See Online/Articles http://dx.doi.org/10.1016/ S0140-6736(12)60734-X and http://dx.doi.org/10.1016/ S0140-6736(12)61080-0

www.thelancet.com Published online August 30, 2012 http://dx.doi.org/10.1016/S0140-6736(12)61069-1

Comment

ranged from 65% to 113% in the three countries from the former Soviet Union. The overall prevalence of XDR tuberculosis in Dalton and colleagues study varied substantially, from less than 08% in the Philippines up to 152% in South Korea. These results show that XDR tuberculosis is increasingly a cause for concern, especially in areas where prevalence of MDR tuberculosis is high. Nevertheless, information remains insufficient to give a clear view of the worldwide distribution and true mag nitude of XDR tuberculosis, even more so for the most severely resistant cases, sometimes called totally drug resistant tuberculosis.8,9 Whether the transmissibility and virulence differ between MDR, XDR, and totally drugresistant isolates and their pan-susceptible counterparts is unclear, and studies like that by Dalton and colleagues2 are important to increase our understanding. Updated information on MDR tuberculosis and investigation of the trends are urgently needed, especially since the true scale of the burden of MDR and XDR tuberculosis might be underestimated and seem to be rapidly increasing. Sensitive, specific, and timely identification of MDR and XDR tuberculosis is crucial to the management of drug-resistant M tuberculosis. Knowledge of the risk factors in specific settings will help to direct the use of new diagnostic tools. In a study in four eastern European settings, the line probe diagnostic assay was assessed for rapid identification of XDR tuberculosis and was shown to be useful in screening resistance to individual secondline drugs.10 On the basis of those results, the authors recommended the assay for use in settings with high burdens of MDR and XDR tuberculosis.10 Quality-assured susceptibility data for the important first-line drug, pyrazinamide, are almost totally lacking for areas with a high prevalence of MDR tuberculosis. Drug-susceptibility testing of pyrazinamide is unfor tunately difficult because no reliable test is generally available. Moreover, unlike the other first-line drugs,11 there is no global system for external quality assurance of the susceptibility data for pyrazinamide. This lack of data is especially unlucky, since evidence shows that combinations of pyrazinamide with moxifloxacin and either of the experimental compounds bedaquiline or PA-824 could be interesting alternatives in the treatment of MDR tuberculosis.12 As for detection of resistance to other anti tubercu losis compounds, molecular testing might play a part. Detection of mutations in the M tuberculosis pncA gene
2

as indicators of resistance to pyrazinamide has been reported in a meta-analysis to have specificity of 93% and sensitivity of 87%.13 Additional studies are, however, needed to understand fully the genetic background of pyrazinamide resistance. A reliable and affordable suscep tibility test method for pyrazinamide is needed, especially for settings where MDR tuberculosis is prevalent. Dalton and colleagues study2 increases awareness of the clinical and public health issues caused by resistant M tuberculosis and reveals differences in prevalence and risk factors between countries and settings. Hopefully, these findings will contribute to the identification of the tools needed for optimum control of MDR tuberculosis in specific epidemiological settings. Sven Hoffner
Department of Preparedness, Swedish Institute for Communicable Disease Control, SE 171 82 Solna, Sweden sven.hoffner@smi.se
I declare that I have no conflicts of interest. 1 2 Skrahina A, Hurevich H, Zalutskaya A, et al. Alarming levels of drug-resistant tuberculosis in Belarus: results of a survey in Minsk. Eur Respir J 2012; 39: 142531. Dalton T, Cegielski P, Akksilp S et al, and the Global PETTS Investigators. Prevalence of and risk factors for resistance to second-line drugs in people with multidrug-resistant tuberculosis in eight countries: a prospective study. Lancet 2012; published online Aug 30. http://dx.doi.org/10.1016/ S0140-6736(12)60734-X Kruuner A, Hoffner SE, Sillastu H, et al. Spread of drug-resistant pulmonary tuberculosis in Estonia. J Clin Microbiol 2001; 39: 333945. Nodieva A, Jansone I, Broka L, Pole I, Skenders G, Baumanis V. Recent nosocomial transmission and genotypes of multidrug-resistant Mycobacterium tuberculosis. Int J Tuberc Lung Dis 2010; 14: 42733. WHO. Anti-tuberculosis drug resistance in the world, report no. 4: The WHO/IUALTD Global Project on Anti-tuberculosis Drug Resistance Surveillance 20022007. Geneva: World Health Organization, 2008. Zignol M, van Gemert W, Falzon D, et al. Surveillance of anti-tuberculosis drug resistance in the world: an updated analysis, 20072010. Bull World Health Organ 2012; 90: 11119. Wright A, Zignol M, Van Deun A, et al, for the Global Project on Anti-Tuberculosis Drug Resistance Surveillance. Epidemiology of antituberculosis drug resistance 200207: an updated analysis of the Global Project on Anti-Tuberculosis Drug Resistance Surveillance. Lancet 2009; 373: 186173. Velayati AA, Masjedi MR, Farnia P, et al. Emergence of new forms of totally drug-resistant tuberculosis bacilli: super extensively drug-resistant tuberculosis or totally drug-resistant strains in Iran. Chest 2009; 136: 42025. Udwadia ZF, Amale RA, Ajbani KK, Rodrigues C. Totally drug-resistant tuberculosis in India. Clin Infect Dis 2012; 54: 57981. Ignatyeva O, Kontsevaya I, Kovalyov A, et al. Detection of resistance to second-line antituberculosis drugs by use of the genotype MTBDRsl assay: a multicenter evaluation and feasibility study. J Clin Microbiol 2012; 50: 159397. Van Deun A, Wright A, Zignol M, Weyer K, Rieder HL. Drug susceptibility testing proficiency in the network of supranational tuberculosis reference laboratories. Int J Tuberc Lung Dis 2011; 15: 11624. Diacon AH, Dawson R, von Groote-Bidlingmaier F, et al. 14-day bactericidal activity of PA-824, bedaquiline, pyrazinamide, and moxifloxacin combinations: a randomised trial. Lancet 2012; published online July 23. http://dx.doi.org/10.1016/S0140-6736(12)61080-0. Chang KC, Yew WW, Zhang Y. Pyrazinamide susceptibility testing in Mycobacterium tuberculosis: a systematic review with meta-analyses. Antimicrob Agents Chemother 2011; 55: 4499505.

3 4 5 6 7

8 9 10

11 12

13

www.thelancet.com Published online August 30, 2012 http://dx.doi.org/10.1016/S0140-6736(12)61069-1