Nearly 15% of multidrug-resistant (MDR) tuberculosis cases developed resistance to the fluoroquinolones, the second-line injectable drugs, or both during treatment for MDR tuberculosis. The rate of acquired resistance was significantly lower in programs that met specific performance criteria.
Keywords: tuberculosis, multidrug-resistant tuberculosis, extensively drug-resistant tuberculosis, Green Light Committee
Abstract
Background. Increasing access to drugs for the treatment of multidrug-resistant (MDR) tuberculosis is crucial but could lead to increasing resistance to these same drugs. In 2000, the international Green Light Committee (GLC) initiative began to increase access while attempting to prevent acquired resistance.
Methods. To assess the GLC's impact, we followed adults with pulmonary MDR tuberculosis from the start to the end of treatment with monthly sputum cultures, drug susceptibility testing, and genotyping. We compared the frequency and predictors of acquired resistance to second-line drugs (SLDs) in 9 countries that volunteered to participate, 5 countries that met GLC criteria, and 4 countries that did not apply to the GLC.
Results. In total, 832 subjects were enrolled. Of those without baseline resistance to specific SLDs, 68 (8.9%) acquired extensively drug-resistant (XDR) tuberculosis, 79 (11.2%) acquired fluoroquinolone (FQ) resistance, and 56 (7.8%) acquired resistance to second-line injectable drugs (SLIs). The relative risk (95% confidence interval [CI]) of acquired resistance was lower at GLC-approved sites: 0.27 (.16–.47) for XDR tuberculosis, 0.28 (.17–.45) for FQ, and 0.15 (.06–.39) to 0.60 (.34–1.05) for 3 different SLIs. The risk increased as the number of potentially effective drugs decreased. Controlling for baseline drug resistance and differences between sites, the odds ratios (95% CIs) were 0.21 (.07–.62) for acquired XDR tuberculosis and 0.23 (.09–.59) for acquired FQ resistance.
Conclusions. Treatment of MDR tuberculosis involves substantial risk of acquired resistance to SLDs, increasing as baseline drug resistance increases. The risk was significantly lower in programs documented by the GLC to meet specific standards.
(See the Editorial Commentary by Daley and Horsburgh on pages 1064–5.)
Microbial drug resistance has become a major public health concern worldwide. In the case of tuberculosis, the first global drug resistance survey, published in 1997, reported multidrug-resistant (MDR) tuberculosis (ie, resistant to at least isoniazid and rifampicin) in virtually every participating country [1]. Shortly thereafter, the World Health Organization (WHO) and many other partners launched the DOTS (directly observed treatment, short course)–Plus initiative for the programmatic management of MDR tuberculosis in middle- and low-income countries [2–4] because only a small fraction of MDR tuberculosis cases were diagnosed and treated worldwide [5]. Initially, this initiative focused on the availability, affordability, quality, and control of second-line drugs (SLDs) needed to treat MDR tuberculosis [6].
To the extent that the level of MDR tuberculosis in a country was a barometer for the strength of its basic tuberculosis control program, the partners recognized the risk of generating more highly drug-resistant tuberculosis [4]. Suddenly providing access to relatively obscure and toxic SLDs for treatment of MDR tuberculosis would likely lead to the development of widespread resistance to these same drugs. Indeed, in 2010, WHO estimated that 94 000 of 440 000 (21.4%) MDR tuberculosis cases worldwide may be the consequence of recently acquired resistance [5]. Therefore, WHO and its partners planned to introduce second-line treatment for MDR tuberculosis through an expanding series of pilot and demonstration projects [3, 4, 7]. These projects had to meet specific criteria that were developed to maximize cure rates and minimize the extent to which further drug resistance would develop [7, 8]. These criteria included, as examples, robust political commitment and leadership, a strong basic DOTS program, documented proficiency in phenotypic drug susceptibility testing (DST), strong individualized treatment regimens, 100% direct observation of MDR tuberculosis treatment, assiduous management of drug toxicities, and operational infection control measures.
To increase access to SLDs and make them affordable, the Green Light Committee (GLC) established a pooled procurement mechanism consolidating demand into larger, more frequent, more predictable purchases [5, 6]. In turn, major pharmaceutical manufacturers offered 60%–90% price reductions for their brands of quality-assured antituberculosis drugs [6, 7, 9, 10]. Projects procured drugs through this mechanism. To prevent acquired resistance, self-selected programs in middle- and low-income countries applied to the GLC, which evaluated, assisted, approved, and monitored applicants. To evaluate whether programs met the criteria for access to SLDs, the GLC ensured that applicants (1) had strong basic DOTS programs and (2) met the criteria established by WHO's Scientific Panel on MDR Tuberculosis [7, 8]. Under these circumstances, the programs would be granted access to procure high-quality, reduced-price SLDs. From 1 January 2000 to 30 June 2011, the GLC approved 138 projects in 90 countries, representing 131 262 patients with MDR tuberculosis [11].
The GLC Secretariat tracked the procurement, distribution, and utilization of drugs, but there was no built-in mechanism to track acquired SLD resistance. To determine whether this GLC process, in the broad sense, could potentially reduce the frequency of acquired drug resistance, we launched the Preserving Effective TB Treatment Study (PETTS) to quantify the emergence of further resistance to SLD in GLC-approved MDR tuberculosis treatment programs, compared concurrently with non-GLC programs in 2 middle-income and 2 high-income countries.
SUBJECTS AND METHODS
This prospective cohort study included 26 clinical sites in 9 countries. The study was proposed at the International Union Conference on Lung Health in 2003 and 2004 with an open invitation for countries to participate. Nine countries volunteered. Estonia, Latvia, Peru, the Philippines, and Russia had GLC-approved programs. South Africa, South Korea, Thailand, and Taiwan did not participate in the GLC initiative. Details have been published previously [12]. This study was approved by institutional review boards at the US Centers for Disease Control and Prevention (CDC) and in all 9 participating countries.
We enrolled consecutive consenting adults with confirmed pulmonary MDR tuberculosis starting treatment with SLDs from 1 January 2005 through 31 December 2008. Patients had to have a positive culture from a specimen collected within 30 days of starting treatment and receive SLDs for at least 30 days. South Africa restricted enrollment to patients who had not been treated for MDR tuberculosis previously. Standardized information was recorded for each program; demographic, socioeconomic, and clinical information for each subject; treatment and laboratory monitoring details; and, for each medication, the manufacturer and country of manufacture. Duplicates of baseline and monthly sputum cultures were shipped to the CDC for DST as previously described [12–14]. When the paired first and last positive cultures differed in susceptibility to fluoroquinolones (FQs), second-line injectable drugs (SLIs), isoniazid, or rifampicin, the CDC genotyped them with the 24-locus multiple interspersed repetitive unit (MIRU)/variable number tandem repeat method to determine their clonal relatedness [15].
The primary outcome measures were acquired resistance to FQs, SLIs, or both, defined as susceptibility in the baseline isolate, resistance to the same drug in the final isolate, and matching MIRU-24 results. Criteria for matching genotypes were adapted from the US National Genotyping Service (CDC, unpublished data), including isolate pairs that differ by only 1 locus. Patient data were analyzed in relation to the CDC's laboratory results using statistical procedures for complex sample designs in SAS software, version 9.3 (SAS Institute, Cary, North Carolina). The data were weighted to reflect the total number of eligible patients at each site so that comparisons across sites would reflect the actual patient populations, not differences in the proportion of patients included in the study. The weights were calculated as the total number of eligible patients divided by the number actually enrolled. Patients with baseline resistance to a given drug were excluded from the analysis of acquired resistance to that drug. For categorical data, we compared frequency distributions of independent variables vs acquired resistance, calculating P values with the Rao–Scott F test, and reporting unadjusted risk ratios (RRs) with 95% confidence intervals (CIs). We compared continuous variables for 2 groups using Student t test or the Wilcoxon rank-sum test. P < .05 was considered statistically significant.
To assess effect modification and confounding, we stratified the relationships between GLC status and acquired drug resistance by covariates using the Mantel-Haenszel method and multivariable logistic regression. Covariates for which the stratum-specific RRs differed significantly and for which the unadjusted vs adjusted RRs differed substantially were included in multivariable analysis. Because sites were heterogeneous, we controlled for confounding by multiple covariates simultaneously using propensity scores [16–20]. The propensity scores were derived with a logistic regression model with GLC status as the dependent variable and all potential confounders as the independent variables (Supplementary Text Box 1). These propensity scores were included in the final multivariable regression models to estimate adjusted odds ratios (ORs) and 95% CIs.
RESULTS
Patient Characteristics
Of 1659 patients who met strict protocol criteria and were enrolled, 1173 (71.4%) had at least 1 subsequent positive culture after the baseline culture (Figure 1). Both first and last positive cultures were tested successfully for 832 (85.1%) individuals—the analysis cohort. DST results of the last isolate differed from the first isolate for the drugs of interest in 250 pairs. Of these, the genotypes matched perfectly in 147 pairs, differed by 1 locus in an additional 18 pairs, and did not match in 85 pairs.
Figure 1.
Derivation of the study population. Abbreviations: CDC, US Centers for Disease Control and Prevention; DST, drug susceptibility testing; SLD, second-line drug.
Patient characteristics and baseline drug resistance were reported previously [12]. Of 832 subjects with paired results, 66 (7.9%) had extensively drug-resistant (XDR) tuberculosis at baseline, and 68 of the remaining 766 subjects (8.9%) acquired XDR tuberculosis during treatment. One hundred twenty-eight (15.4%) had baseline FQ resistance, and 79 of the remaining 704 subjects (11.2%) acquired FQ resistance during treatment; 115 (13.8%) had baseline SLI resistance, and 56 of the remaining 717 subjects (7.8%) acquired SLI resistance.
In general, demographic and socioeconomic characteristics were not associated with acquired drug resistance including age, sex, education, employment, occupation, homelessness, substance abuse, and history of incarceration (Tables 1–3 and Supplementary Tables 1–3). Acquired XDR tuberculosis occurred half as often (RR, 0.48 [95% CI, .24–.97]; P = .03) among patients with any comorbidity, other than human immunodeficiency virus (HIV) infection, compared with patients with no additional comorbidity. With respect to HIV infection, in 1 country, 48.1% of subjects were coinfected with HIV; in 7 other countries, 1.3%–9.4% of patients were coinfected with HIV. Patients were not tested for HIV in 1 country. Within these 3 strata, HIV infection was not associated with acquired resistance (data not shown). Acquired XDR tuberculosis developed in 6.3% of patients without pulmonary cavitation vs 11.6% of patients with pulmonary cavitation (RR, 1.84 [95% CI, 1.04–3.26]; P = .03). Acquired FQ resistance was less frequent in patients being treated for MDR tuberculosis for the first time (RR, 0.52 [95% CI, .31–.89]; P = .02).
Table 2.
Acquired Resistance to Fluoroquinolones Among 832 Adults With Pulmonary Multidrug-Resistant Tuberculosis Starting Treatment With Second-line Drugs, 2005–2010, in 9 Countries
| Characteristics | Category | AR to FQ = Yes No. (%) |
AR to FQ = No No. (%) |
RR (95% CI) | P Value |
|---|---|---|---|---|---|
| Patient characteristics | |||||
| Unweighted total | 832 | 79 (9.5) | 753 (91.5) | ||
| Unweighted total without baseline FQ resistance | 704 | 79 (11.2) | 625 (88.8) | ||
| Weighted total | 1404 | 141 (10.1) | 1262 (89.9) | ||
| Weighted total without baseline FQ resistance | 1169 | 141 (12.1) | 1028 (87.9) | ||
| Sex | Male | 82 (10.9) | 672 (89.1) | 0.78 (.51–1.19) | .27 |
| Female | 58 (14.0) | 354 (86.0) | Ref | ||
| Age group, y | 18–28 | 32 (11.0) | 263 (89.0) | Ref | |
| 29–36 | 38 (13.8) | 239 (86.2) | 1.25 (.68–2.26) | .47 | |
| 37–45 | 33 (11.6) | 248 (88.4) | 1.10 (.59–2.03) | .77 | |
| 46–79 | 36 (11.6) | 276 (88.4) | 1.05 (.57–1.92) | .88 | |
| HIV/AIDS | Negative | 80 (12.0) | 582 (88.0) | Ref | |
| Positive | 51 (21.6) | 184 (78.4) | 1.77 (1.15–2.73) | .01 | |
| ND/Unknown | 9 (3.3) | 259 (96.7) | 0.27 (.12–.63) | .0009 | |
| Diabetes mellitus | Yes | 11 (8.8) | 109 (91.2) | 0.71 (.32–1.59) | .60 |
| No | 129 (12.4) | 910 (87.6) | Ref | ||
| Any comorbidity other than HIV infectiona | Yes | 24 (8.4) | 259 (91.6) | 0.64 (.36–1.15) | .03 |
| No | 116 (13.1) | 766 (86.9) | Ref | ||
| No. of previous tuberculosis episodes | 0 | 3 (4.2) | 78 (95.8) | 0.30 (.07–1.21) | .6 |
| 1 | 66 (14.4) | 393 (85.6) | Ref | ||
| 2 | 39 (10.8) | 322 (89.2) | 0.78 (.48–1.28) | .33 | |
| ≥3 | 31 (11.8) | 233 (88.2) | 0.83 (.48–1.42) | .49 | |
| Prior treatment with any SLD | Yes | 23 (14.3) | 137 (85.7) | 1.23 (.69–2.21) | .51 |
| No | 117 (11.6) | 889 (88.4) | Ref | ||
| Prior treatment with any FQ | Yes | 16 (13.1) | 108 (86.9) | 1.11 (.57–2.16) | .78 |
| No | 123 (11.8) | 917 (88.2) | Ref | ||
| Prior treatment with any SLI | Yes | 14 (15.0) | 77 (85.0) | 1.28 (.62–2.65) | .53 |
| No | 126 (11.7) | 948 (88.3) | Ref | ||
| Class by prior treatment history | New case | 9 (5.6) | 151 (94.4) | 0.45 (.20–1.04) | .48 |
| First-line drugs | 101 (12.3) | 717 (87.7) | Ref | ||
| Second-line drugs | 25 (15.6) | 134 (84.4) | 1.27 (.72–2.23) | .42 | |
| Unilateral vs bilateral pulmonary disease | Unilateral | 13 (6.9) | 174 (93.1) | Ref | |
| Bilateral | 128 (13.1) | 854 (86.9) | 1.96 (1.15–3.33) | .009 | |
| Pulmonary cavity | No | 16 (7.4) | 220 (92.6) | Ref | |
| Yes | 62 (14.5) | 397 (85.5) | 2.04 (1.16–3.59) | .01 | |
| In hospital at time of enrollment | Yes | 117 (15.5) | 639 (84.5) | 2.87 (1.63–5.03) | <.0001 |
| No | 22 (5.4) | 386 (94.6) | Ref | ||
| No. of hospitalizations for MDR tuberculosis | 0 | 15 (5.2) | 268 (94.8) | Ref | .006 |
| 1 | 99 (14.0) | 607 (86.0) | 2.75 (1.45–5.23) | .001 | |
| 2 | 26 (14.9) | 150 (85.1) | 2.85 (1.34–6.08) | .005 | |
| No. of times DST performed on follow-up cultures | 0 | 32 (5.9) | 509 (94.1) | Ref | |
| 1 | 45 (13.8) | 282 (86.2) | 2.48 (1.32–4.68) | .5 | |
| ≥2 | 62 (21.5) | 228 (78.5) | 4.44 (2.44–8.09) | <.0001 | |
| Program characteristics | |||||
| Country | A | 19.8 | 80.2 | ||
| Non-GLC: A–D | B | 18.0 | 82.0 | ||
| C | 16.7 | 83.3 | |||
| D | 13.8 | 86.2 | |||
| GLC-approved: E–I | E | 11.1 | 88.9 | ||
| F | 9.1 | 90.9 | |||
| G | 5.2 | 94.8 | |||
| H | 3.9 | 96.1 | |||
| I | 2.8 | 97.2 | |||
| GLC | Yes | 30 (5.0) | 562 (95.0) | 0.28 (.17–.45) | <.0001 |
| No | 110 (19.1) | 463 (80.9) | Ref | ||
| Tuberculosis control program structure | Vertical | 18 (7.5) | 218 (92.5) | 0.57 (.31–1.04) | .06 |
| Integrated | 124 (13.3) | 809 (86.7) | Ref | ||
| Financial sector | Public | 111 (13.5) | 714 (86.5) | 1.54 (.90–2.61) | .11 |
| Public–private | 30 (8.8) | 331 (91.2) | Ref | ||
| Hospitalization | Routine | 103 (15.5) | 562 (84.5) | 3.97 (2.00–7.90) | <.0001 |
| As needed | 24 (17.0) | 119 (83.0) | 4.37 (1.91–10.03) | .0002 | |
| Rarely | 14 (3.9) | 346 (96.1) | Ref | ||
| Treatment strategyb | Individualized | 85 (19.8) | 344 (80.2) | 2.61 (1.70–4.01) | <.0001 |
| Standardized | 56 (7.6) | 683 (92.4) | Ref | ||
| Directly observed treatment | Full | 73 (10.5) | 605 (89.5) | 0.68 (.32–1.42) | .31 |
| Partial | 51 (13.1) | 336 (86.9) | 0.83 (.38–1.78) | .63 | |
| None | 15 (15.9) | 81 (84.1) | Ref | ||
| Mycobacterial characteristics | |||||
| Baseline resistance to SLDs, No. | 0 | 40 (5.6) | 680 (94.4) | Ref | |
| 1 | 18 (7.9) | 208 (92.1) | 1.44 (.72–2.87) | .30 | |
| 2 | 9 (15.0) | 51 (85.0) | 3.18 (1.37–7.35) | .0065 | |
| ≥3 | 73 (45.7) | 86 (54.3) | 8.27 (5.17–13.22) | <.0001 | |
| Baseline resistance to ethambutol | Res | 103 (14.7) | 601 (85.3) | 1.86 (1.14–3.05) | .001 |
| Sus | 36 (7.9) | 425 (92.1) | Ref | ||
| Baseline resistance to streptomycin | Res | 114 (14.5) | 670 (85.5) | 2.18 (1.23–3.86) | <.0001 |
| Sus | 25 (6.7) | 355 (93.3) | Ref | ||
| Baseline resistance to 4 first-line drugs | Res | 89 (17.0) | 434 (83.0) | 2.17 (1.39–3.39) | <.0001 |
| Sus | 50 (7.8) | 592 (92.2) | Ref | ||
| Baseline kanamycin resistance | Res | 83 (36.8) | 143 (63.2) | 6.14 (4.08–9.24) | <.0001 |
| Sus | 56 (6.0) | 882 (94.0) | Ref | ||
| Baseline amikacin resistance | Res | 80 (41.3) | 114 (58.7) | 6.74 (4.53–10.03) | <.0001 |
| Sus | 59 (6.1) | 912 (93.9) | Ref | ||
| Baseline capreomycin resistance | Res | 66 (46.0) | 78 (54.0) | 6.34 (4.32–9.29) | <.0001 |
| Sus | 73 (7.3) | 936 (92.7) | Ref | ||
| Baseline resistance to all available SLIs | Res | 70 (49.9) | 70 (50.1) | 7.36 (5.05–10.73) | <.0001 |
| Sus | 70 (6.8) | 956 (93.2) | Ref | ||
| Baseline XDR tuberculosis | Res | ||||
| Sus | 139 (12.0) | 1026 (88.0) | |||
| Baseline ethionamide resistance | Res | 23 (11.5) | 180 (88.5) | 0.95 (.55–1.63) | .94 |
| Sus | 116 (12.1) | 846 (87.9) | Ref | ||
| Baseline PAS resistance | Res | 11 (10.9) | 89 (89.1) | 0.90 (.40–2.01) | .92 |
| Sus | 129 (12.1) | 937 (87.9) | Ref | ||
Acquired resistance to the FQs was not significantly (P > .1) associated with sex, age group, employment, education, homelessness, occupational risk, smoking, alcohol abuse, prior history of tuberculosis, number of previous tuberculosis episodes, prior treatment with streptomycin, prior treatment with any SLD, prior treatment with any FQ, any SLI, kanamycin, amikacin, capreomycin, thioamides, PAS, or prior treatment with serine analogues, anatomic location of disease, sputum microscopy results at baseline.
Abbreviations: AR, acquired resistance; CI, confidence interval; DST, drug susceptibility testing; FQ, fluoroquinolone; GLC, Green Light Committee; HIV, human immunodeficiency virus; MDR, multidrug-resistant; ND, not done; PAS, para-aminosalicylic acid; Ref, reference group; Res, drug resistant; RR, unadjusted risk ratio; SLD, second-line drug; SLI, second-line injectable drug; Sus, drug susceptible; XDR, extensively drug-resistant.
a Concurrent comorbidities assessed included substance abuse, diabetes mellitus, renal disease, hepatic disease, gastrointestinal disease, seizure disorder, and any other immunosuppressive drugs or conditions.
b Program treatment strategies were classified as “individualized” if treatment regimens were tailored to each patient's own DST results or “standardized” if the same treatment regimen was used for all patients because individual second-line DST was not available.
Table 1.
Acquired Extensive Drug Resistance Among 832 Adults With Pulmonary Multidrug-Resistant Tuberculosis Starting Treatment With Second-Line Drugs, 2005–2010, in 9 Countries
| Characteristic | Category | AR XDR = Yes | AR XDR = No | RR (95% CI) | P Value |
|---|---|---|---|---|---|
| No. (%) | No. (%) | ||||
| Patient characteristics | |||||
| Unweighted total | 832 | 69 (8.3) | 763 (91.7) | ||
| Unweighted total without baseline XDR tuberculosis | 766 | 69 (9.0) | 697 (91.0) | ||
| Weighted total | 1404 | 126 (9.0) | 1277 (91.0) | ||
| Weighted total without baseline XDR tuberculosis | 1280 | 126 (9.9) | 1153 (90.1) | ||
| Sex | Male | 72 (8.7) | 759 (91.3) | 0.79 (.50–1.26) | .32 |
| Female | 51 (11.4) | 393 (88.6) | |||
| Age group, y | 18–28 | 29 (9.1) | 288 (90.9) | Ref | .81 |
| 29–36 | 31 (10.5) | 267 (89.5) | 1.15 (.59–2.23) | ||
| 37–45 | 33 (10.4) | 281 (89.6) | 1.25 (.67–2.35) | ||
| 46–79 | 30 (8.7) | 316 (91.3) | 0.95 (.49–1.84) | ||
| HIV/AIDS | Negative | 70 (9.6) | 659 (90.4) | Ref | |
| Positive | 45 (19.0) | 192 (81.0) | 1.90 (1.19–3.04) | .007 | |
| Not tested | 8 (2.6) | 301 (97.4) | 0.26 (.10–.65) | .002 | |
| Diabetes mellitus | Yes | 7 (4.7) | 132 (95.3) | 0.46 (.17–1.25) | .21 |
| No | 116 (10.3) | 1010 (89.7) | Ref | ||
| Any comorbidity other than HIV infectiona | Yes | 17 (5.3) | 306 (94.7) | 0.48 (.24–.97) | .03 |
| No | 105 (11.1) | 846 (88.9) | Ref | ||
| No. of previous tuberculosis episodes | 0 | 5 (5.5) | 85 (94.5) | 0.50 (.16–1.58) | .21 |
| 1 | 55 (11.4) | 433 (88.6) | Ref | ||
| 2 | 34 (8.8) | 348 (91.2) | 0.85 (.50–1.45) | .55 | |
| ≥3 | 29 (9.2) | 286 (80.8) | 0.81 (.44–1.46) | .48 | |
| Prior treatment with any SLD | Yes | 20 (9.9) | 184 (90.1) | 1.04 (.56–1.95) | .98 |
| No | 102 (9.6) | 968 (90.4) | Ref | ||
| Prior treatment with any FQ | Yes | 12 (7.6) | 150 (92.4) | 0.77 (.36–1.67) | .45 |
| No | 110 (9.9) | 1002 (90.1) | Ref | ||
| Prior treatment with any SLI | Yes | 13 (11.1) | 101 (88.9) | 1.17 (.55–2.49) | .74 |
| No | 110 (9.5) | 1051 (90.5) | Ref | ||
| Class by prior treatment history | New case | 12 (6.6) | 167 (93.4) | 0.63 (.29–1.4) | .25 |
| First-line drugs | 89 (10.3) | 773 (89.7) | Ref | ||
| Second-line drugs | 20 (9.8) | 185 (90.2) | 0.93 (.5–1.76) | .83 | |
| First treatment for MDR tuberculosis | No | 101 (9.2) | 998 (90.8) | 0.75 (.40–1.39) | |
| Yes | 21 (12.4) | 151 (87.6) | Ref | ||
| Unilateral vs bilateral pulmonary disease | Unilateral | 16 (7.4) | 195 (92.6) | Ref | .34 |
| Bilateral | 111 (10.4) | 959 (89.6) | 1.4 (.7–2.82) | ||
| Pulmonary cavity | No | 14 (6.3) | 239 (93.7) | Ref | .03 |
| Yes | 54 (11.6) | 449 (88.4) | 1.84 (1.04–3.26) | ||
| In hospital at time of enrollment | Yes | 110 (13.4) | 709 (86.6) | 4.79 (2.38–9.65) | <.0001 |
| No | 13 (2.8) | 443 (97.2) | Ref | ||
| No. of hospitalizations for this episode of MDR tuberculosis | 0 | 8 (2.6) | 296 (97.4) | Ref | .0009 |
| 1 | 90 (11.6) | 688 (88.4) | 4.48 (1.91–10.51) | ||
| 2 | 25 (12.7) | 168 (87.3) | 5.11 (2.0–13.07) | ||
| No. of times DST performed on follow-up cultures | 0 | 37 (6.5) | 539 (93.5) | Ref | |
| 1 | 38 (10.6) | 322 (89.4) | 1.64 (.87–3.08) | .86 | |
| ≥2 | 47 (14.3) | 284 (85.7) | 2.44 (1.34–4.44) | .01 | |
| Program characteristics | |||||
| Country | A | 18.0 | 82.0 | ||
| Non-GLC: A–D | B | 11.5 | 88.5 | ||
| C | 8.3 | 91.7 | |||
| D | 6.3 | 93.7 | |||
| GLC-approved: E–I | E | 9.5 | 90.5 | ||
| F | 9.8 | 90.2 | |||
| G | 4.1 | 95.9 | |||
| H | 3.8 | 96.2 | |||
| I | 0.6 | 99.4 | |||
| GLC | No | 99 (15.6) | 534 (84.4) | 0.27 (.16–.47) | <.0001 |
| Yes | 24 (3.7) | 618 (96.3) | Ref | ||
| Tuberculosis control program structure | Vertical | 20 (7.5) | 242 (92.5) | 0.72 (.39–1.31) | .27 |
| Horizontal | 107 (10.5) | 912 (89.5) | Ref | ||
| Financial sector | Public | 106 (12.2) | 762 (87.8) | 2.45 (1.28–4.70) | .004 |
| Public–private | 21 (5.0) | 391 (95.0) | Ref | ||
| Hospitalization | Routine | 99 (14.1) | 605 (85.9) | 6.72 (2.7–16.6) | <.0001 |
| As needed | 19 (10.0) | 171 (89.9) | 4.80 (1.65–14.00) | .002 | |
| Rarely | 9 (2.0) | 377 (98.0) | Ref | ||
| Treatment strategyb | Individualized | 47 (5.6) | 791 (94.4) | 0.31 (.20–.50) | <.0001 |
| Standardized | 79 (18.0) | 363 (82.0) | Ref | ||
| Directly observed treatment | Full | 71 (9.5) | 677 (90.5) | 0.82 (.38–1.77) | .61 |
| Partial | 41 (10.2) | 360 (89.8) | 0.88 (.39–1.98) | .75 | |
| None | 14 (11.6) | 109 (88.4) | Ref | ||
| Mycobacterial characteristics | |||||
| Baseline resistance to SLDs, No. | 0 | 17 (2.4) | 706 (97.6) | Ref | |
| 1 | 17 (5.5) | 290 (94.5) | 2.29 (.95–5.54) | .06 | |
| 2 | 20 (23.0) | 66 (77.0) | 9.58 (4.23–21.70) | <.0001 | |
| ≥3 | 73 (44.1) | 91 (55.9) | 18.42 (9.44–35.93) | <0.0001 | |
| Baseline ethambutol resistance | Res | 99 (12.7) | 681 (87.3) | 2.63 (1.43–4.85) | .001 |
| Sus | 24 (4.8) | 471 (95.2) | Ref | ||
| Baseline streptomycin resistance | Res | 111 (12.8) | 757 (87.2) | 4.62 (1.97–10.86) | <.0001 |
| Sus | 11 (2.8) | 395 (97.2) | Ref | ||
| Baseline resistance to 4 first-line drugs | Res | 90 (15.3) | 498 (84.7) | 3.21 (1.87–5.50) | <.0001 |
| Sus | 33 (4.8) | 654 (95.2) | Ref | ||
| Baseline fluoroquinolone resistance | Res | 19 (16.7) | 93 (83.3) | 1.81 (.97–3.40) | .07 |
| Sus | 108 (9.2) | 1061 (90.8) | Ref | ||
| Baseline kanamycin resistance | Res | 83 (36.8) | 143 (63.2) | 9.75 (6.07–15.66) | <.0001 |
| Sus | 40 (3.8) | 1009 (96.2) | Ref | ||
| Baseline amikacin resistance | Res | 80 (41.3) | 114 (58.7) | 10.42 (6.61–16.44) | <.0001 |
| Sus | 43 (4.0) | 1038 (96.0) | Ref | ||
| Baseline capreomycin resistance | Res | 66 (46.0) | 78 (54.0) | 9.10 (5.97–13.87) | <.0001 |
| Sus | 57 (5.1) | 1062 (94.9) | Ref | ||
| Baseline resistance to all available SLIs | Res | 70 (49.9) | 70 (50.1) | 10.72 (7.05–16.31) | <.0001 |
| Sus | 53 (4.7) | 1082 (95.3) | Ref | ||
| Baseline resistance to thioamides | Res | 18 (8.5) | 198 (91.5) | 0.86 (.46–1.61) | .80 |
| Sus | 104 (9.9) | 954 (90.1) | Ref | ||
| Baseline resistance to PAS | Res | 14 (11.4) | 109 (88.6) | 1.20 (.59–2.45) | .41 |
| Sus | 109 (9.4) | 1043 (90.6) | Ref | ||
Acquired resistance to XDR tuberculosis was not significantly (P > .1) associated with employment, education, homelessness, occupational risk, smoking, alcohol abuse, prior treatment with any of the individual antituberculosis drugs (streptomycin, kanamycin, amikacin, capreomycin, thioamides, PAS, serine analogues), baseline sputum microscopy results, bilateral vs unilateral disease radiographically, anatomic site of disease, and body mass index.
Abbreviations: AR, acquired resistance; CI, confidence interval; DST, drug susceptibility testing; FQ, fluoroquinolone; GLC, Green Light Committee; HIV, human immunodeficiency virus; MDR, multidrug-resistant; PAS, para-aminosalicylic acid; Ref, reference group; Res, drug resistant; RR, unadjusted risk ratio; SLD, second-line drug; SLI, second-line injectable drug; Sus, drug susceptible; XDR, extensively drug-resistant.
a Concurrent comorbidities assessed included substance abuse, diabetes mellitus, renal disease, hepatic disease, gastrointestinal disease, seizure disorder, and any other immunosuppressive drugs or conditions.
b Program treatment strategies were classified as “individualized” if treatment regimens were tailored to each patient's own DST results or “standardized” if the same treatment regimen was used for all patients because individual second-line DST was not available.
Table 3.
Acquired Resistance to the Second-line Injectable Drugs Kanamycin and Amikacin Among 832 Adults With Pulmonary Multidrug-Resistant Tuberculosis Starting Treatment With Second-line Drugs, 2005–2010, in 9 Countries
| Characteristic | Category | AR SLI = Yes No. (%) |
AR SLI = No No. (%) |
RR (95% CI) | P Value |
|---|---|---|---|---|---|
| Unweighted total | 832 | 51 (6.1) | 781 (93.9) | ||
| Unweighted total without baseline SLI resistance | 673 | 51 (7.6) | 622 (92.4) | ||
| Weighted total | 1404 | 89 (6.3) | 1315 (93.7) | ||
| Weighted total without baseline SLI resistance | 1117 | 89 (7.9) | 1029 (92.1) | ||
| Sex | Male | 65 (9.0) | 663 (91.0) | 1.48 (.79–2.77) | .22 |
| Female | 24 (6.0) | 366 (94.0) | Ref | ||
| Age group, y | 18–28 | 18 (6.6) | 260 (93.4) | Ref | |
| 29–36 | 24 (8.9) | 242 (91.1) | 1.36 (.61–3.04) | .48 | |
| 37–45 | 30 (10.6) | 253 (89.4) | 1.61 (.76–3.43) | .21 | |
| 46–79 | 17 (5.8) | 273 (94.2) | 0.88 (.37–2.09) | .77 | |
| HIV/AIDS | Negative | 53 (8.5) | 577 (91.5) | Ref | |
| Positive | 24 (13.6) | 154 (86.4) | 1.60 (.86–2.98) | .14 | |
| ND/Unknown | 11 (3.5) | 298 (96.5) | 0.41 (.17–1.00) | .04 | |
| Diabetes mellitus | Yes | 6 (4.7) | 124 (95.3) | 0.57 (.21–1.57) | .26 |
| No | 80 (8.2) | 898 (91.8) | Ref | ||
| Any comorbidity other than HIV infectiona | Yes | 19 (6.4) | 279 (93.6) | 0.75 (.39–1.46) | .40 |
| No | 70 (8.5) | 750 (91.5) | Ref | ||
| No. of previous tuberculosis episodes | 0 | 6 (8.6) | 67 (91.4) | 1.33 (.45–3.90) | .61 |
| 1 | 26 (6.5) | 381 (93.5) | Ref | ||
| 2 | 32 (9.3) | 308 (90.7) | 1.44 (.73–2.83) | .29 | |
| ≥3 | 25 (8.3) | 273 (1.7) | 1.28 (.62–2.65) | .51 | |
| Prior treatment with any SLD | Yes | 11 (5.7) | 184 (94.3) | 0.68 (.29–1.62) | .37 |
| No | 78 (8.4) | 845 (91.6) | Ref | ||
| Prior treatment with any FQ | Yes | 10 (5.9) | 154 (94.1) | 0.71 (.28–1.81) | .46 |
| No | 79 (8.4) | 845 (91.6) | Ref | ||
| Prior treatment with any SLI | Yes | 7 (6.9) | 101 (93.1) | 0.86 (.31–2.43) | .78 |
| No | 81 (8.1) | 928 (91.9) | Ref | ||
| Classification by treatment history | New case | 15 (9.9) | 132 (90.1) | 1.34 (.65–2.78) | .43 |
| First-line drugs | 55 (7.4) | 690 (92.6) | Ref | ||
| Second-line drugs | 13 (6.7) | 184 (93.3) | 0.90 (.49–2.04) | .80 | |
| Unilateral vs bilateral pulmonary disease | Unilateral | 71 (7.7) | 856 (92.3) | Ref | .59 |
| Bilateral | 18 (9.2) | 172 (90.8) | 1.20 (.62–2.33) | ||
| Pulmonary cavity | No | 17 (7.0) | 253 (93.0) | Ref | .50 |
| Yes | 32 (5.8) | 519 (94.2) | 0.82 (.45–1.47) | ||
| In hospital at time of enrollment | Yes | 73 (10.7) | 614 (89.3) | 2.97 (1.53–5.78) | .0006 |
| No | 15 (3.6) | 414 (96.4) | Ref | ||
| No. of hospitalizations for MDR tuberculosis | 0 | 9.6 (3.4) | 277 (96.6) | Ref | .03 |
| 1 | 54 (8.1) | 608 (91.9) | 2.40 (1.05–5.49) | ||
| 2 | 26 (15.1) | 143 (84.9) | 4.48 (1.85–10.91) | .0003 | |
| No. of times DST performed on follow-up cultures | 0 | 32 (5.9) | 511 (94.1) | ref | |
| 1 | 21 (6.4) | 303 (93.6) | 1.06 (.50–2.25) | .2 | |
| ≥2 | 39 (12.7) | 270 (87.3) | 2.67 (1.42–5.03) | .0009 | |
| Program characteristics | |||||
| Country | A | 12.0 | 88.0 | ||
| Non-GLC: A–D | B | 6.3 | 93.7 | ||
| C | 8.3 | 91.7 | |||
| D | 9.7 | 90.3 | |||
| GLC-approved: E–I | E | 5.3 | 94.7 | ||
| F | 7.1 | 92.9 | |||
| G | 5.9 | 94.1 | |||
| H | 15.8 | 84.2 | |||
| I | 0.6 | 99.4 | |||
| GLC | Yes | 33 (5.6) | 545 (94.3) | 0.54 (.31–.94) | .02 |
| No | 56 (10.4) | 483 (89.6) | Ref | ||
| Tuberculosis control program structure | Vertical | 23 (10.7) | 192 (89.3) | 1.46 (.81–2.64) | .21 |
| Integrated | 66 (7.3) | 836 (92.3) | Ref | ||
| Financial sector | Public | 74 (10.4) | 642 (89.6) | 2.87 (1.36–6.07) | .003 |
| Public–private | 15 (3.6) | 387 (96.4) | Ref | ||
| Hospitalization | Routine | 66 (11.5) | 508 (88.5) | 4.27 (1.82–10.0) | .0002 |
| As needed | 13 (7.3) | 168 (92.7) | 2.7 (.92–7.91) | .06 | |
| Rarely | 10 (2.7) | 353 (97.3) | Ref | ||
| Treatment strategyb | Individualized | 45.8 (6.0) | 713 (94.0) | 0.50 (.29–.86) | |
| Standardized | 43 (12.0) | 316 (88.0) | Ref | .01 | |
| Directly observed treatment | Full | 46 (7.2) | 594 (92.8) | 1.13 (.39–3.27) | .81 |
| Partial | 36 (9.9) | 319 (90.1) | 1.55 (.53–4.56) | .41 | |
| None | 7 (6.4) | 108 (93.6) | Ref | ||
| Mycobacterial characteristics | |||||
| Baseline resistance to SLDs, No. | 0 | 47 (6.6) | 676 (93.4) | Ref | |
| 1 | 28 (9.0) | 279 (91.0) | 1.37 (.74–2.52) | .31 | |
| ≥2 | 14 (15.7) | 74 (8.3) | 2.40 (1.09–5.27) | .03 | |
| Baseline ethambutol resistance | Res | 61 (9.3) | 602 (90.7) | 1.54 (.86–2.77) | .14 |
| Sus | 27 (6.0) | 426 (94.0) | Ref | ||
| Baseline streptomycin resistance | Res | 66 (9.1) | 657 (90.9) | 1.57 (.84–2.95) | .15 |
| Sus | 23 (5.8) | 372 (94.2) | Ref | ||
| Baseline resistance to 4 first-line drugs | Res | 47 (9.9) | 434 (90.1) | 1.52 (.88–2.62) | .13 |
| Sus | 41 (6.5) | 595 (93.5) | Ref | ||
| Baseline fluoroquinolone resistance | Res | 22 (15.7) | 116 (84.3) | 2.29 (1.21–4.32) | .01 |
| Sus | 67 (6.9) | 913 (93.1) | Ref | ||
| Baseline ethionamide resistance | Res | 20 (10.9) | 165 (89.1) | 1.48 (.77–2.86) | .24 |
| Sus | 69 (7.4) | 864 (92.6) | Ref | ||
| Baseline PAS resistance | Res | 13 (11.5) | 104 (88.5) | 1.53 (.70–3.32) | .29 |
| Sus | 75 (7.5) | 925 (92.5) | Ref | ||
Acquired resistance to the SLIs was not significantly (P > .1) associated with employment, education, homelessness, occupational risk, smoking, alcohol abuse, prior treatment with any of the individual antituberculosis drugs (streptomycin, kanamycin, amikacin, capreomycin, thioamides, PAS, serine analogues), baseline sputum microscopy results, bilateral vs unilateral disease radiographically, anatomic location of disease, or body mass index.
Abbreviations: AR, acquired resistance; CI, confidence interval; DST, drug susceptibility testing; FQ, fluoroquinolone; GLC, Green Light Committee; HIV, human immunodeficiency virus; MDR, multidrug-resistant; ND, not done; PAS, para-aminosalicylic acid; Ref, reference group; Res, drug resistant; RR, unadjusted risk ratio; SLD, second-line drug; SLI, second-line injectable drug; Sus, drug susceptible.
a Concurrent comorbidities assessed included substance abuse, diabetes mellitus, renal disease, hepatic disease, gastrointestinal disease, seizure disorder, and any other immunosuppressive drugs or conditions.
b Program treatment strategies were classified as “individualized” if treatment regimens were tailored to each patient's own DST results or “standardized” if the same treatment regimen was used for all patients because individual second-line DST was not available.
Program Characteristics
The GLC group included 3 upper-middle, 1 lower-middle, and 1 low-income country, whereas the comparison group included 2 high-income and 2 upper-middle-income countries. GLC approval was associated with a reduced risk of acquired resistance (Tables 1–3). Compared with non-GLC sites, the relative risk of acquired XDR tuberculosis was 73% lower (RR, 0.27 [95% CI, .16–.47]; P < .0001) in the GLC-approved sites. FQ resistance was 72% lower (RR, 0.28 [95% CI, .17–.45]; P < .0001). Acquired FQ resistance was lower in all 5 GLC-approved countries than in all 4 non-GLC countries. For the SLIs, the unadjusted relative risks for acquired resistance to kanamycin (KAN), amikacin (AMK), and capreomycin, respectively, were 0.60 (95% CI, .34–1.05; P = .07), 0.15 (95% CI, .06–.39; P < .0001), and 0.24 (95% CI, .16–.47; P = .002). The risk of acquired SLI resistance was lower in 4 GLC-approved countries than in all 4 non-GLC countries.
Five countries routinely hospitalized patients at the start of treatment. In 2 countries, hospitalization depended on clinical indications. Two countries provided ambulatory care to virtually all MDR tuberculosis patients. Hospitalization at the time of enrollment increased the risk 4.80-fold for acquiring XDR tuberculosis and 2.90-fold for FQs and SLIs (all P < .001). Three of 5 GLC-approved programs and 2 of 4 non-GLC programs routinely performed culture and DST on all patients with suspected tuberculosis, whereas the other programs reserved DST only for previously treated patients. One non-GLC program used a standardized treatment strategy, which was associated with an elevated frequency of acquired drug resistance. In contrast to the GLC-approved sites, at the time of this study, the non-GLC programs did not administer 100% of treatment under direct observation and not all of the drugs were manufactured under stringent drug regulatory authorities.
Mycobacterial Characteristics
The strongest, most consistent risk factor was pretreatment (baseline) resistance to other drugs—that is, having fewer effective drugs available. The frequency of acquired resistance to each drug and combination of drugs increased with increasing baseline resistance (Tables 1–3) to other drugs in the regimen. Among patients with no baseline SLD resistance, 2.4% acquired XDR tuberculosis, increasing to 44.1% among those with baseline resistance to at least 3 SLDs (RR, 18.42 [95% CI, 9.44–35.93]; P < .0001). Acquired FQ resistance increased 8.30-fold (95% CI, 5.17- to 13.22-fold), from 5.6% among patients with baseline resistance to only first-line drugs to 45.7% among patients with baseline resistance to at least 3 SLDs (P < .0001). Acquired SLI resistance also increased with increasing baseline resistance. With baseline resistance to only first-line drugs, the risk of acquired resistance to KAN or AMK, for example, was 6.6%, increasing 2.40-fold (95% CI, 1.09- to 5.27-fold) to 15.7% with resistance to ≥2 other SLDs (P = .03).
In terms of specific drugs, baseline resistance to the SLI had the greatest impact on the risk of acquired XDR tuberculosis. With baseline resistance limited to first-line drugs, the risk of acquired XDR tuberculosis was 2.4%. With baseline resistance to an FQ, the risk of acquired XDR was 16.7%. With baseline SLI resistance, the risk of acquired XDR was 36.8% to 46.0%, depending on the specific drug (all P < .0001). For example, with baseline resistance to KAN, the risk of acquired XDR tuberculosis was 36.8%, 6.10-fold (95% CI, 4.08- to 9.24-fold) higher than the risk of 6.0% among isolates susceptible to KAN at baseline (P < .0001). Baseline resistance to FQs had the greatest impact on the risk of acquired SLI resistance. For example, acquired resistance to KAN or AMK was observed in 15.7% of those with baseline FQ resistance, 2.30-fold (95% CI, 1.21- to 4.32-fold) higher than the 6.9% among those with no baseline FQ resistance (P = .01).
The findings were not explained by differences in microbiological monitoring. The median duration of follow-up, the median number of follow-up sputum specimens examined, and the default rates did not differ significantly between those who acquired resistance and those who did not (Table 4).
Table 4.
Local Microbiological Monitoring and Acquired Resistance (AR) to Fluoroquinolones, AR to at Least 1 Second-line Injectable Drug, and Acquired Extensively Drug-Resistant Tuberculosis
| Characteristic | Category | AR XDR, Yes | AR XDR, No | RR (95% CI) | P Value | AR FQ, Yes | AR FQ, No | RR (95% CI) | P Value | AR SLI, Yes | AR SLI, No | RR (95% CI) | P Value |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| No. of follow-up sputum specimens examined (quintiles) | 1–4 | 23 (18.7) | 205 (17.9) | Ref | Ref | 26 (18.6) | 185 (18.2) | Ref | Ref | 15 (16.1) | 192 (17.7) | Ref | Ref |
| 5–9 | 42 (34.1) | 209 (18.3) | 0.43 (.09–2.04) | .3 | 44 (31.4) | 187 (18.4) | 0.53 (.13–2.09) | .1 | 27 (29.0) | 201 (18.5) | 0.65 (.10–4.27) | .4 | |
| 10–15 | 27 (22.0) | 237 (20.7) | 1.13 (.34–3.76) | .3 | 28 (7.1) | 219 (21.5) | 1.24 (.38–4.06) | .6 | 20 (21.5) | 220 (20.3) | 1.28 (.22–7.36) | .4 | |
| 16–19 | 18 (14.6) | 206 (18.0) | 0.92 (.30–2.85) | .6 | 21 (15.0) | 188 (18.5) | 1.08 (.35–3.35) | .9 | 13 (14.0) | 195 (18.0) | 1.05 (.16–6.81) | .8 | |
| 20–51 | 13 (10.6) | 287 (25.1) | 0.65 (.17–2.53) | .7 | 21 (15.0) | 239 (23.5) | 1.57 (.45–5.50) | .3 | 18 (19.4) | 276 (25.5) | 0.79 (.12–4.97) | .7 |
Data are presented as No. (column %).
Abbreviations: AR, acquired resistance; CI, confidence interval; FQ, fluoroquinolone; Ref, reference group; RR, relative risk; SLI, second-line injectable drug; XDR, extensively drug-resistant.
Stratified and Multivariable Analysis
The disparity in rates of acquired resistance between GLC and comparison programs depended on the baseline drug resistance profile (Figure 2). When the patient's baseline isolate was susceptible to all SLDs, the risk of acquired XDR tuberculosis was not significantly lower in the GLC-approved programs (1.4% vs 3.4%; risk difference, 2.0%; P = .18). As the number of ineffective SLDs increased, the risk of acquiring XDR tuberculosis increased in both groups, but it increased faster in the non-GLC group, the difference becoming much larger and highly significant (for baseline resistance to at least 3 SLDs, 15.9% vs 60.9%; risk difference 45.0%; P < .0001; Figure 2). Similarly, stratifying by previous treatment, among new, treatment-naive MDR tuberculosis patients, the risk of acquired resistance in the non-GLC group was not significantly higher than in the GLC group (data not shown).
Figure 2.
Risk of acquired drug resistance in Green Light Committee (GLC)–approved and non-GLC programs, stratified by baseline resistance to second-line drugs, in 9 countries, 2005–2010. Abbreviations: CI, confidence interval; GLC, Green Light Committee.
The regression model for the propensity score yielded a c-statistic of 0.93. The adjusted ORs for the effects of baseline drug resistance and for the difference between GLC and non-GLC sites remained significant, controlling for differences between the 2 groups (Table 5). For each additional drug to which the isolate was resistant, the odds of acquired XDR tuberculosis more than doubled (OR, 2.39 [95% CI, 1.89–3.03]; P < .0001), the odds of acquired FQ resistance nearly doubled (OR, 1.98 [95% CI, 1.60–2.45]; P < .0001), and the odds of acquired KAN resistance increased by two-thirds (OR, 1.67 [95% CI, 1.15–2.43]; P = .007), after controlling for GLC status and other covariates. At the same time, the odds of acquired XDR tuberculosis were significantly lower (OR, 0.21 [95% CI, .07–.62]; P = .004) in the GLC-approved programs, as were the odds of acquired FQ resistance (OR, 0.23 [95% CI, .09–.59]; P = .001), independent of baseline resistance to other drugs and other covariates.
Table 5.
Multivariable Analysis of Risk Factors for Acquired Resistance to Crucial Second-line Drugs, Controlling for Demographic, Social, and Clinical Characteristics, Hospitalizations, and Baseline Drug Resistance, 2005–2010, in 9 Countries
| Characteristic | Category | XDR: aOR (95% CI) | P Value | FQ: aOR (95% CI) | P Value | SLI: aOR (95% CI) | P Value |
|---|---|---|---|---|---|---|---|
| Green Light Committee | Yes | 0.21 (.07–.62) | .004 | 0.23 (.09–.59) | .001 | 0.74 (.29–1.87) | .53 |
| No | Ref | Ref | Ref | ||||
| No. of hospitalizations for this episode of tuberculosis | 0 | Ref | Ref | Ref | |||
| 1 | 1.36 (.43–4.25) | .77 | 0.78 (.31–1.99) | .35 | 1.48 (.54–4.06) | .50 | |
| 2 | 2.34 (.69–7.91) | .12 | 1.19 (.45–3.13) | .43 | 3.73 (1.44–9.69) | .002 | |
| No. of SLDs to which the isolate was susceptible | 0–2 | 32.03 (14.47–70.89) | <.0001 | 13.95 (7.42–26.22) | <.0001 | 15.53 (2.84–84.79) | .005 |
| 3–4 | 4.47 (2.12–10.62) | .58 | 2.25 (1.13–4.48) | .10 | 1.97 (1.09–3.57) | .16 | |
| 5 | Ref | Ref | Ref | ||||
| Baseline ethambutol resistance | Res | 1.98 (.95–4.13) | .07 | 1.54 (.84–2.81) | .16 | 1.40 (.69–2.84) | .35 |
| Sus | Ref | Ref | Ref | ||||
| Baseline streptomycin resistance | Res | 3.26 (1.25–8.48) | .02 | 1.54 (.75–3.15) | .24 | 1.33 (.65–2.73) | .43 |
| Sus | Ref | Ref | Ref | ||||
| Baseline fluoroquinolone resistance | Res | 7.51 (2.83–19.94) | <.0001 | … | 2.40 (1.11–5.20) | .02 | |
| Sus | Ref | … | |||||
| Baseline KAN or AMK resistance | Res | 13.59 (4.72–39.09) | <.0001 | 6.32 (2.58–15.49) | <.0001 | … | |
| Sus | Ref | Ref | … | ||||
| Baseline capreomycin resistance | Res | 1.65 (.62–4.39) | .32 | 1.65 (0.68–4.02) | .27 | 1.15 (.23–5.67) | .87 |
| Sus | Ref | Ref | Ref | ||||
| Baseline ethionamide resistance | Res | 0.89 (.37–2.09) | .78 | 1.09 (.52–2.30) | .82 | 1.45 (.71–2.96) | .30 |
| Sus | Ref | Ref | Ref | ||||
| Baseline PAS resistance | Res | 0.76 (.28–2.07) | .60 | 0.60 (.19–1.91) | .39 | 0.98 (.36–2.63) | .96 |
| Sus | Ref | Ref | Ref | ||||
| Propensity score | 0.98 (.70–1.38) | P = .93 | 0.932 (.70–1.25) | P = .63 | 0.85 (.62–1.16) | .30 |
Propensity scores included terms for the following variables: human immunodeficiency virus status, employment, marital status, education, high-risk occupation, alcohol abuse, diabetes, chronic renal insufficiency, history of treatment with any first-line drugs, history of treatment with ethambutol or streptomycin, history of treatment with SLIs, history of treatment with PAS or cycloserine, outcome of previous treatment, unilateral vs bilateral disease and cavitary vs non-cavitary disease radiographically, sputum microscopy results for acid-fast bacilli, number of drugs tested locally for drug susceptibility before treatment, and hospitalization at enrollment. Coefficients and tests of statistical significance for these variables do not appear individually in the final multivariable models for acquired resistance.
Abbreviations: AMK, amikacin; aOR, adjusted odds ratio; CI, confidence interval; FQ, fluoroquinolone; KAN, kanamycin; PAS, para-aminosalicylic acid; Ref, reference group; Res, resistant; SLD, second-line drug; SLI, second-line injectable drug; Sus, susceptible; XDR, extensively drug-resistant.
DISCUSSION
We measured the frequency of and risk factors for acquired resistance in MDR tuberculosis patients at 26 clinical facilities in 9 countries. Overall, 14.8% (95% CI, 12.2%–17.3%) of patients with MDR tuberculosis patients starting treatment with SLD acquired resistance to an FQ, an SLI, or both during treatment. The observation that treatment in a GLC-approved program was associated with lower risk of acquired resistance raises questions about the specific factors that may have been responsible. The criteria for GLC approval of these demonstration projects were disseminated by WHO [7, 8] and focused specifically on minimizing the risk of acquired drug resistance. All of the non-GLC programs met most of these criteria, but none of them met all of the criteria. For example, they did not use 100% directly observed treatment, not all drugs were manufactured under stringent drug regulatory authorities, they performed less DST, and not all had DOTS programs, but it was not the effect of a single poorly performing program. All 5 GLC-approved countries had lower rates of acquired FQ resistance than all 4 comparison countries. For acquired SLI resistance, 4 GLC-approved countries had lower rates of acquired resistance than all 4 comparison countries. Because programs that had more of these features had lower rates of drug resistance, incorporating these features into an MDR tuberculosis treatment program is likely to be associated with decreased emergence of drug resistance.
Resistance to specific SLD at the start of treatment was a critical risk factor for the subsequent development of acquired resistance to other drugs in the regimen. Because baseline drug resistance increased the probability of acquiring further drug resistance, we stratified the analysis by resistance to each of the major drugs. The stratified analysis revealed that (1) the absolute risk of acquired resistance increased substantially as the amount of baseline resistance to other drugs in the regimen increased in both GLC- and non-GLC programs, but (2) the difference between GLC- and non-GLC programs in the degree of risk became substantially more pronounced as baseline resistance increased. In GLC-approved projects, for example, the risk of acquired XDR tuberculosis increased from 1.4% to 15.9% as the extent of baseline drug resistance increased, whereas in non-GLC projects the risk increased from 3.4% to 60.9%.
This study was not a controlled trial. The groups differed as much as these countries differ from one another. We measured and controlled for >90% of the potential confounding using propensity scores and multivariable statistical methods [16–20]. Most of the demographic, social, and economic differences between patients in different countries were not associated significantly with acquired drug resistance and, therefore, could not confound the association. Another limitation is the potential for sampling bias. Overall, we enrolled approximately 60% of the total number of eligible patients. However, this figure varied across sites from 10% to 90%; hence, we weighted the analysis to reflect the actual number of patients at each site, not each site's success at consenting and enrolling patients. Although not quantified, in our experience, the main reasons for nonenrollment were related not to the study but to startup delays and turnover of personnel at study sites, which is why we believe the samples are representative. However, nonconsenting patients may differ from consenting patients.
These results have important implications for policy regarding scale-up of services for MDR tuberculosis, including molecular diagnostics and new antituberculosis drugs [21–23]. First, DST should be repeated during treatment. Second, as new antituberculosis drugs are introduced, there may be a tendency to use them to treat patients with otherwise inadequate or marginally adequate regimens. If the results of PETTS can be generalized, then policies promoting the use of new drugs in otherwise marginally effective treatment regimens, without at least 2 or 3 other highly effective drugs, will lead rapidly to widespread acquired resistance to the new drug.
Although <20% of patients with MDR tuberculosis have access to adequate diagnosis and treatment, global interest in scaling up MDR tuberculosis treatment programs is unprecedented [5, 24]. Under these conditions, programs should first treat patients with the lowest risk of developing further resistance. When comparing new patients without additional drug resistance and previously treated patients with further resistance, treating new patients generally led to successful outcomes, whereas treating the previously treated patients, who had worse drug resistance at the start of retreatment, led to acquired drug resistance.
Policy guidelines for scaling up access to rapid molecular drug resistance testing focus on previously treated patients because the prevalence of resistance is greater. However, based on this study, targeting previously treated patients selects for precisely those in whom acquired resistance and poor treatment outcomes are most likely. The magnitude of acquired resistance may be even greater in programs without the socioeconomic resources of the countries participating in this study. Therefore, the evidence suggests that, in settings where drug-resistant tuberculosis has already become endemic and rapid molecular tests are used selectively due to cost constraints, new patients should be prioritized for rapid screening with these tests. The challenge related to predictive value can be substantially reduced by repeating positive tests and comparing with phenotypic DST results. We encourage regulatory authorities, policy makers, and drug manufacturers to consider these findings when introducing new diagnostics and new drugs for MDR tuberculosis.
Supplementary Data
Supplementary materials are available at Clinical Infectious Diseases online (http://cid.oxfordjournals.org). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author.
Notes
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 enrollment sites for their contributions to this work. We thank the people in the following organizations for support and contributions: Division of Tuberculosis Elimination, Centers for Disease Control and Prevention (CDC); 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; Riga East University Clinic Centre of Tuberculosis and Lung Diseases; 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 Specialized Tuberculosis Hospital; Jose Pearson Hospital; Fort Grey Tuberculosis Hospital; Korean Institute of Tuberculosis; National Masan Hospital and the Korean Ministry of Health and Welfare; Thailand Office of Disease Prevention and Control Region 7; and the World Health Organization in Switzerland, Denmark, Peru, and Russia. We also thank medical and nursing staff in the Ubon Ratchathani, Srisaket, Sakon Nakon, and Yasothon provinces, Thailand.
Disclaimer. The findings and conclusions in this article are those of the authors and do not necessarily represent the views of the US Agency for International Development or the US CDC.
Financial support. This work was supported by the US Agency for International Development (IAA: OGH09-011:CGH10-1010196) and the US CDC, Division of Tuberculosis Elimination (CDC-OGH11-12021).
Potential conflicts of interest. All authors: No potential conflicts.
All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.
APPENDIX
Global PETTS Investigators. Gloria Yale Calahuanca, Lima City Health District Reference Laboratory, Lima, Peru; Melanie Wolfgang, Allison Taylor, Kathrine Tan, Erika Sigman, and Beverly Metchock, CDC, Atlanta, Georgia; Piret Viiklepp, National Tuberculosis Registry, National Institute for Health Development, Tallinn, Estonia; Irina A. Vasilyeva, Elena E. Larionova, Olga V. Demikhova, and Sofia N. Andreevskaya, Central Tuberculosis Research Institute, Russian Academy of Medical Sciences, Moscow; Carmen Suarez, Lima East Health District Reference Laboratory, Lima, Peru; Ingrida Sture, Girts Skenders, Vija Riekstina, Inga Norvaisha, Gunta Dravniece, and Andra Cirula, Riga East University Hospital Centre of Tuberculosis and Lung Diseases, Riga, Latvia; Tatiana Somova, Tatyana G. Smirnova, Vladimir Oblast Tuberculosis Dispensary, Vladimir, Russian Federation; Evgenia S. Nemtsova, Irina Degtyareva, and Tatiana Khorosheva, Orel Oblast Tuberculosis Dispensary, Orel, Russian Federation; Wanlaya Sitti and Rattanawadee Akksilp, Office of Disease Prevention and Control Region 7, Ubon Ratchatani, Thailand; Isdore C. Shamputa, Clifton E. Barry III, and Ying Cai, National Institute for Allergy and Infectious Diseases, National Institutes of Health, Bethesda, Maryland; Kristine Rose Pua, M. Therese C. Perez, and M. Tarcela Gler, Tropical Disease Foundation, Manila, Philippines; Seungkyu Park, Seonyeong Min, and Soo Hee Hwang, National Masan Hospital, Changwon, Republic of Korea; Klavdia Levina, North Estonia Regional Hospital, Tallinn; Yung-Chao Lei and Angela Song-En Huang, Taiwan Centers for Disease Control, Taipei; Jongseok Lee, Doosoo Jeon, Seokyong Eum, and Eunjin Cho, International Tuberculosis Research Center, Changwon, Korea; Joey Lancaster and Jeanette Brand, Medical Research Council, Pretoria, South Africa; Cesar Bonilla and Oswaldo Jave, National Tuberculosis Strategy, Ministry of Health, Lima, Peru; Manfred Danilovits, Tartu University Hospital, Tartu, Estonia; Luis Asencios, National Institute of Health, Lima, Peru.
References
- 1.World Health Organization. The World Health Organization/International Union Against Tuberculosis and Lung Disease Global Project on Anti-Tuberculosis Drug Resistance Surveillance: anti-tuberculosis drug resistance in the world. Geneva, Switzerland: WHO; 1997. [Google Scholar]
- 2.World Health Organization. Basis for the development of an evidence-based case-management strategy for MDR TB within the WHO's DOTS Strategy. Proceedings of 1998 Meeting and Protocol Recommendations; Available at: http://www.who.int/tb/publications/1999/en/ Accessed 6 June 2014. [Google Scholar]
- 3.World Health Organization. WHO meeting to co-ordinate the DOTS-Plus workplan on pilot projects for the management of multidrug resistant (MDR) tuberculosis, 29 January 1999. Available at: http://www.who.int/tb/publications/1999/en/ Accessed 6 June 2014.
- 4.World Health Organization. DOTS-Plus and the Green Light Committee: improving access to second-line anti-TB drugs. Available at: http://www.who.int/tb/publications/2000/en/ Accessed 6 June 2014.
- 5.World Health Organization. Geneva, Switzerland: WHO; 2012. Global tuberculosis report 2012. [Google Scholar]
- 6.World Health Organization Working Group on DOTS-Plus for MDR-TB. Procurement of second-line anti-tuberculosis drugs for DOTS-Plus pilot projects. Cambridge, MA: 5–6 July 1999 Available at: http://www.who.int/tb/publications/2000/en/ Accessed 20 March 2014. [Google Scholar]
- 7.World Health Organization. Geneva, Switzerland: WHO; 2000. Guidelines for establishing DOTS-Plus pilot projects for the management of multidrug-resistant tuberculosis (MDR-TB) [Google Scholar]
- 8.World Health Organization. Instructions for applying to the Green Light Committee for access to second-line anti-tuberculosis drugs. Geneva, Switzerland: WHO; 2002. Available at: http://www.who.int/tb/publications/2001/en/ Accessed 27 March 2014. [Google Scholar]
- 9.Gupta R, Cegielski JP, Espinal MA, et al. Increasing transparency in partnerships for health—introducing the Green Light Committee. Trop Med Int Health. 2002;7:970–6. doi: 10.1046/j.1365-3156.2002.00960.x. [DOI] [PubMed] [Google Scholar]
- 10.Gupta R, Kim JY, Espinal MA, et al. Responding to market failures in tuberculosis control. Science. 2001;293:1049–51. doi: 10.1126/science.1061861. [DOI] [PubMed] [Google Scholar]
- 11.World Health Organization. GLC programmes applications. Available at: http://www.who.int/tb/challenges/mdr/greenlightcommittee/report_glc_applications_apr2011rev1.pdf. Accessed 23 June 2011.
- 12.Dalton T, Cegielski P, Akksilp S, et al. Prevalence of and risk factors for resistance to second-line drugs in people with multidrug-resistant tuberculosis in eight countries: a prospective cohort study. Lancet. 2012;380:1406–17. doi: 10.1016/S0140-6736(12)60734-X. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Clinical and Laboratory Standards Institute. Susceptibility testing of mycobacteria, nocardiae, and other aerobic actinomycetes; approved standard. 2nd ed. Wayne, PA: CLSI; 2011. CLSI document M24-A2. [PubMed] [Google Scholar]
- 14.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; 1985. [Google Scholar]
- 15.Cowan LS, Mosher L, Diem L, Massey JP, Crawford JT. Variable-number tandem repeat typing of Mycobacterium tuberculosis isolates with low copy numbers of IS6110 by using mycobacterial interspersed repetitive units. J Clin Microbiol. 2002;40:1592–602. doi: 10.1128/JCM.40.5.1592-1602.2002. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Rosenbaum PR, Rubin DB. Reducing bias in observational studies using subclassification on the propensity score. J Am Stat Assoc. 1984;79:516–24. [Google Scholar]
- 17.Rosenbaum PR, Rubin DB. The central role of the propensity score in observational studies for causal effects. Biometrika. 1983;70:41–55. [Google Scholar]
- 18.Brookhart MA, Schneeweis S, Rothman KJ, Glynn RJ, Avorn J, Sturmer T. Variable selection for propensity score models. Am J Epi. 2006;163:1149–56. doi: 10.1093/aje/kwj149. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Williamson E, Morley R, Lucas A, Carpenter J. Propensity scores: from naıve enthusiasm to intuitive understanding. Stat Methods Med Research. 2011;21:273–93. doi: 10.1177/0962280210394483. [DOI] [PubMed] [Google Scholar]
- 20.Austin PC. An introduction to propensity score methods for reducing the effects of confounding in observational studies. Multivariate Behav Res. 2011;46:399–424. doi: 10.1080/00273171.2011.568786. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.World Health Organization. TB diagnostics and laboratory strengthening. Available at: http://who.int/tb/laboratory/mtbrifrollout/en/ Accessed 15 July 2013.
- 22.World Health Organization. Interim guidance on the use of bedaquiline to treat MDR-TB. Available at: http://www.who.int/tb/challenges/mdr/bedaquiline/en/index.html. Accessed 15 July 2013.
- 23.Food and Drug Administration. FDA news release. Available at: http://www.fda.gov/NewsEvents/Newsroom/PressAnnouncements/ucm333695.htm. Accessed 15 July 2013.
- 24.World Health Organization. Largest ever roll-out of GeneXpert(R) TB test machines, in 21 countries. Available at: http://www.who.int/tb/features_archive/xpertprojectlaunch/en/ Accessed 21 March 2014.
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.