We conducted a retrospective study of patients with pulmonary MDR TB treated through the National Center for Tuberculosis and Lung Diseases (NCTLD) in Tbilisi, Georgia. All patients were sputum smear–positive for acid-fast bacilli (AFB) and culture-positive for M. tuberculosis at baseline and had MDR TB confirmed by conventional drug susceptibility testing (DST). Patients with MDR TB were given a diagnosis during March 2009–October 2012, as described in a study that evaluated the clinical effect of a rapid diagnostic test for detection of MDR TB (15). Approval for this study was obtained from the Georgia NCTLD and Emory University (Atlanta, GA, USA) Institutional Review Boards.

Direct sputum smears with Ziehl-Neelsen staining were examined by light microscopy at a sputum microscopy center, and 1 sputum smear sample positive for acid-fast bacilli (AFB) was sent to the National Reference Laboratory (NRL) at NCTLD, where it was processed as reported (16). Cultures were prepared by using Löwenstein-Jensen–based solid medium or the MGIT 960 broth culture system (Becton Dickinson, Franklin Lakes, NJ, USA). Cultures with positive results by either method were confirmed to be M. tuberculosis complex by using the MTBDRplus assay (Hain Lifescience, Nehren, Germany) and the Capilia TB assay (Tauns Laboratories, Inc., Shizuoka, Japan) (16).

DST for first-line drugs was performed as described (16). DST for SLDs was performed by using the proportion method and Löwenstein-Jensen medium with the following drug concentrations: ethionamide, 40.0 μg/mL; ofloxacin, 2.0 μg/mL; p-aminosalicylic acid, 0.5 μg/mL; capreomycin, 40.0 μg/mL; and kanamycin, 30.0 μg/mL (17). The NRL has undergone external quality assessment by the WHO Supranational TB Reference Laboratory (Antwerp, Belgium) annually since 2005. In 2012, a certificate from the WHO Supranational TB Reference Laboratory was received by the NRL for successfully passing DST quality assurance testing for isoniazid, rifampin, ethambutol, kanamycin, capreomycin, and ofloxacin.

As per standard of care, follow-up sputum smears and cultures were obtained monthly during the intensive phase of MDR TB treatment (minimum 6 months). Second-line DST was performed at baseline and was recommended at 3 and 6 months if culture results remained positive. During the continuation phase, sputum smears and cultures were obtained every 3 months, and second-line DST was recommended for all positive cultures.

A standardized data form was used to abstract data from medical charts, patient treatment cards, the NCTLD surveillance database, and the NRL database. Information was collected about sociodemographic characteristics, TB history, signs and symptoms, treatment regimens, outcomes; and all sputum smear, culture, and DST results. Data were entered into a REDCap database (18).

Acquired resistance was defined as any SLD that was susceptible on baseline DST and resistant on any subsequent DST result. Time to MDR TB treatment initiation was defined as time from initial diagnostic sputum collection to start of SLD therapy. Initial MDR TB treatment was defined as any drug received <30 days of starting an SLD regimen. Treatment interruption was defined as a continuous interruption for >1 SLDs for >1 week. Final treatment outcomes were determined using WHO criteria (19). A favorable outcome was defined as cure or treatment completion; a poor outcome was defined as treatment failure, death during treatment, or loss to follow up (LFU) (formerly known as default). Two alternative classifications were used in secondary analyses: 1) patients with a negative culture result at time of LFU were included as a favorable outcome, and 2) patients with a poor outcome were defined as treatment failure or death. LFU patients were excluded from analysis.

The NCTLD Drug Resistance TB Treatment Committee provided initial guidance on choosing an empiric SLD regimen for patients initiating MDR TB treatment. After second-line DST results were available, treatment regimens were individualized if needed and guided by WHO recommendations. When possible, regimens were designed to include at least 4 drugs to which an M. tuberculosis isolate was susceptible. All treatment regimens included a fluoroquinolone, pyrazinamide, and capreomycin or kanamycin for >6 months. All patients received directly observed therapy. Patients initiated therapy as inpatients before transitioning to outpatient treatment. Patients were recommended to remain hospitalized until showing sputum smear or culture conversion and clinical improvement.

Data were analyzed by using SAS software version 9.3 (SAS Institute, Cary, NC, USA). For descriptive statistics, differences in categorical variables were tested by using the χ² or Fisher exact tests. The Wilcoxon-Mann-Whitney test was used for comparing non–normally distributed continuous variables. A 2-sided p value <0.05 was considered significant. A logistic regression model was used to estimate the independent association of potential risk factors with acquired resistance and the adjusted association of acquired resistance with a poor outcome. Logistic model building and covariate selection was based on purposeful selection of patient-level factors as described (20). Additional logistic regression models using our alternative definitions of poor outcome as defined above were used.

A total of 158 patients with pulmonary MDR TB were included in the study. For analysis of acquired resistance, 17 patients were excluded because of XDR TB at baseline or absence of a 2-month or later follow-up sputum examination (Figure 1). The remaining 141 patients had a mean age of 37.9 years; most (73%) were men. Less than half (44%) had a history of TB treatment. M. tuberculosis isolates with baseline resistance to capreomycin or kanamycin were found in 33% of patients, and isolates with baseline resistance to ofloxacin were found in 6%. Other patient characteristics are shown in Table 1.

Of the 141 patients evaluated for acquired resistance, 32 patients had ≥1 follow-up DST performed, including 40% of patients with a positive 4-month culture and 52% with a positive 6-month culture. A total of 22 patients had different follow-up DST results that showed a change in resistance pattern, including 19 (13.5%) with acquired resistance and 3 (2.1%) with follow-up DST showing a reversion to susceptibility for 1 SLD. Median time to initial development of acquired resistance was 142 days (range 85–480 days), including 200 days for capreomycin or kanamycin and 149 days for ofloxacin. Of 132 patients infected with M. tuberculosis that had baseline ofloxacin susceptibility, ofloxacin resistance developed in isolates from 12 (9.1%) patients. Among 123 patients infected with isolates that had baseline susceptibility to capreomycin, kanamycin, or both, resistance to capreomycin or kanamycin developed in 12 (9.8%) (Figure 1). Of 12 patients infected with isolates that had acquired resistance to an injectable drug, resistance to capreomycin and kanamycin developed in isolates from 5 patients, resistance to capreomycin developed in isolates with baseline kanamycin resistance from 3 patients, and susceptibility to capreomycin or kanamycin remained in isolates from 4 patients.

Among 19 patients infected with isolates that had acquired resistance, increasing resistance to 1 drug developed in isolates from 9 patients, to 2 drugs in isolates from 8 patients, and to 3 and 4 additional drugs in isolates from 1 patient each. Almost all acquired resistance was to ofloxacin, capreomycin, or kanamycin; 2 patients each had isolates with acquired resistance to ethionamide and p-aminosalicylic acid. Acquired resistance led to emergence of XDR TB in 14 (74%) of 19 patients.

Patients with and without isolates that had acquired resistance were similar in regards to age, sex, prior TB, and other characteristics (Table 1). In contrast, patients who had isolates with acquired resistance had a lower mean baseline body mass index (19.1 vs. 20.7 kg/m²; p = 0.02) and a higher prevalence of baseline cavitary disease (58% vs. 16%; p<0.01) and were more likely to have isolates resistant to >6 drugs at baseline DST (74% vs. 38%; p = 0.01) than patients who had isolates without acquired resistance. Patients with isolates that had acquired resistance were less likely to receive kanamycin (32% vs. 62%; p = 0.01) as part of their initial MDR TB treatment regimen. Regarding sputum examinations, patients with isolates that had acquired resistance were more likely to have baseline AFB sputum smear values >3+ (63% vs. 28%; p<0.01) and to be sputum smear and culture positive at 4 and 6 months (Table 1) than patients without isolates that had acquired resistance.

Factors associated with acquired resistance by univariate analysis were baseline cavitary disease, high baseline drug resistance, baseline ofloxacin resistance, number of known active drugs in the initial MDR TB treatment regimen, initial AFB sputum smear result >3+, and sputum smear or culture positivity at 4 and 6 months (Table 2). By multivariate analysis, factors associated with acquired resistance were baseline cavitary disease (adjusted odds ratio [aOR] 5.21, 95% CI 1.56–17.38); resistance to >6 drugs at baseline DST (aOR 5.31, 95% CI 1.50–18.77); and sputum smear positivity at 4 months (aOR 6.54, 95% CI 1.23–34.88).

A total of 140 patients were evaluated for final treatment outcomes; 1 patient was excluded because he was still receiving treatment (Figure 2). Of the remaining 139 patients, 61 (44%) had a poor outcome. Poor outcomes were more frequent among patients with isolates that had acquired resistance than patients without these isolates (89% vs. 36%; p<0.01). Of 2 patients who had favorable outcomes and isolates that had acquired resistance, 1 underwent adjunctive surgery and 1 had an isolate with acquired resistance to ethionamide; these isolates remained susceptible to ofloxacin, kanamycin, and capreomycin. Most (44 of 61) poor outcomes were caused by LFU, and the remaining poor outcomes were caused by deaths (11) and treatment failure (6). Of LFU patients, 15 (34%) of 44 had positive sputum cultures at the time of LFU, including 6 with isolates that had acquired resistance. The 29 patients with negative cultures at the time of LFU received a median of 111 days of treatment for MDR TB (range 42–325 days), and the 15 patients with positive cultures received a median of 91 days of treatment (range 80–360 days).

Patients with a poor outcome were significantly more likely to have isolates with acquired resistance (28% vs. 3%; p<0.01) and be sputum smear positive at 4 and 6 months and sputum culture positive at 2, 4, and 6 months (Table 3) than patients with a favorable outcome. There were no other differences between groups (Table 3).

Risk factors associated with a poor outcome by univariate analysis were acquired resistance, high baseline drug resistance, and sputum smear or culture positivity at 4 and 6 months (Table 4). Multivariate analysis showed that acquired resistance (aOR 8.05, 95% CI 1.56–41.66) and sputum smear positivity at 6 months (aOR 3.43, 95% CI 1.36–8.63) remained associated with a poor outcome. In our first alternative multivariate model, when we classified patients with a negative culture at time of LFU as a favorable outcome, acquired resistance was associated with a poor outcome (aOR 24.91, 95% CI 4.21–147.48). In a second alternative multivariate model, in which poor outcome was defined as treatment failure or death and LFU patients were excluded, acquired resistance remained associated with a poor outcome (aOR 38.44, 95% CI 5.96–247.73).