The characteristics of the 200 patients enrolled in the DOTS-Plus program are displayed in Table 1. In total, 165 (85%) were male and the mean age was 42 years; 144 (72%) patients were previously treated for TB, 102 with second-line agents; 133 patients (67%) had isolates that exhibited resistance to EMB, including 30/54 (56%) new patients and 101/144 (70%) previously treated patients. Sixty-nine patients (35%) had isolates resistant to a second-line injectable agent, including 34 to KM alone, two to CPM alone and 33 to both agents. Two patients (1%), both previously treated, had initial isolates resistant to OFX; 198 patients had initial resistance to both INH and RMP and are included in subsequent analyses.

Among the 198 MDR-TB patients, cure was documented in 108 (55%) and 10 (5%) completed treatment; 18 (9%) patients died during treatment and treatment failed in 30 (15%); 26 (13%) defaulted and 6 (3%) transferred out of the region. Treatment was stopped for six patients for >2 months; these patients were clinically monitored, resumed treatment and had a reported outcome. We analyzed their outcomes as reported.

Of the 30 patients in whom treatment failed, 29 underwent DST during DOTS-Plus. By the end of treatment, the proportion of patients with resistance to SLDs had increased considerably: 20/29 (69%) had TB resistant to either KM or CPM and 9/29 (31%) had fluoroquinolone-resistant TB (Figure).

Among those with favorable outcomes, the mean total treatment duration was 621 days (standard deviation [SD] 109), the mean intensive phase regimen was 242 days (SD 89) and the mean continuation phase regimen was 377 days (SD 71). The median number of treatment interruptions during the intensive phase was one (range 0–14), and the median number of treatment interruptions during the continuation phase was two (range 0–17).

Among patients in whom treatment failed, the mean total treatment duration was 465 days (SD 128). The median number of treatment interruptions during the intensive phase was two (range 0–26) and the median number of treatment interruptions during the continuation phase was four (range 0–12).

Excluding patients who transferred out of the region, factors significantly associated with poor outcome in univariate analyses of the remaining 192 patients were previous treatment, bilateral disease on initial chest radiograph, positive sputum smear at enrollment, three or more treatment interruptions during the intensive phase of treatment and addiction to alcohol or illicit drugs (Table 2). In multivariate analysis, smear status at enrollment modified the effect of previous treatment on outcome; the stratified results of the final model with the interaction are presented in Table 2. The strongest association with default was alcohol or drug addiction (OR 2.28, 95%CI 0.99–5.33, adjusted OR [aOR] 2.12, 95%CI 0.90–5.03).

Ten patients (5%) developed XDR-TB during the course of treatment. Of these, 4 died, treatment failed in 4, 1 transferred out and 1 was reported as cured. Among the 115 remaining patients for whom there were at least 6 months of follow-up culture data after the reported treatment outcome, an additional 13 (11%) developed XDR-TB. Of these, 10 (77%) used illicit drugs or alcohol excessively, 5 (38%) had defaulted during treatment, treatment had failed in 6 (46%) and 2 (15%) had been reported as cured (Table 3). Among 70 patients who had a favorable outcome and for whom there were at least 12 months of follow-up culture data after treatment outcome, 7 (10%) had a subsequent positive culture after treatment outcome was documented.

Our analysis suggests that previous treatment, smear positivity at initiation of MDR-TB treatment and treatment interruptions during the intensive phase were associated with poor outcome. Alcohol or drug addiction also appeared to be associated with both poor outcome, which included defaulting, as well as with defaulting alone, but these associations did not reach conventional statistical significance in our analysis.

Previous research has described factors associated with poor outcome in patients treated for MDR-TB in DOTS-Plus programs in countries of the former Soviet Union,^17,18,20,21 and the outcomes for this cohort were similar to descriptions in other treatment cohorts.^9,17,22 Default, while arguably distinct from death or treatment failure, was included in our definition of poor outcome to allow for a more complete program evaluation, consistent with previous analyses.^18,20,22 As this was the first cohort of DOTS-Plus patients in the region, it is likely that this group was more sick than subsequent cohorts, with longer durations of disease and higher rates of previous treatment with, and consequent resistance to, second-line agents—both of which were associated with poor treatment outcome in univariate analysis. In this cohort, over half of the patients had previously been treated with second-line agents prior to enrollment, over a third of whom had resistance to a second-line agent. Equally worrying is the fact that 12/54 newly treated patients had resistance to second-line agents, suggesting transmission of resistant organisms in the community. Our data also suggest that previously treated patients with positive sputum smears at treatment initiation may have a particularly high association with poor outcomes, even after adjusting for initial resistance to an injectable agent, but the lack of precision limits our conclusions. These patients may harbor a higher burden of organisms or may be a population that is more heterogeneous and with variable drug susceptibility, and may require more deliberately aggressive therapy during the intensive phase of treatment.

Treatment interruptions and alcohol or drug addiction were identified risks for poor outcome that are potentially modifiable. A predictive rule to identify patients at risk for treatment interruption has been developed,²³ and focused interventions to prevent interruptions should be developed and evaluated. Alcoholism and drug addiction are prevalent among patients with TB disease in Russia,^24,25 and while not statistically significant, our analysis suggests association with both poor outcome and default. This has been previously documented in other Russian TB programs.^18,22,26,27 Alcohol and drug addiction can be treated in the context of anti-tuberculosis treatment in many ways using pharmaceutical, behavioral and other interventions,²⁸ and these interventions urgently need evaluation.

Our data demonstrate that 5% of MDR-TB patients in this cohort developed XDR-TB during the course of treatment, a proportion similar to that described in another review of outcomes in Russia.¹⁴ The number of patients who developed XDR-TB after completion of treatment is worrying. This concern is amplified by the fact that 13% of the patients in this cohort defaulted, and for many of them, outcomes and follow-up are not known. Given the potentially catastrophic consequence of XDR-TB transmission in a community,²⁹ more detailed research into risk factors for increased resistance should be performed and interventions to lower default rates should be evaluated and implemented. The high recurrence among patients who were successfully treated and for whom follow-up data were available, raises concerns and questions about current definitions of treatment outcomes. Longer term and more complete follow-up of patients after treatment for MDR-TB is needed, and the use of genotyping to distinguish between re-infection and relapse may help better characterize these findings.

Our analysis has several limitations. First, this was a retrospective review with a limited number of participants, and we were unable to control for additional factors that may have confounded or modified our findings. Alcohol and drug addiction were identified by experienced clinicians, but the reliability of these determinations is not certain. The reliability and reproducibility of DST against SLDs is questionable, and our conclusions about DST may not be fully generalizable. The relatively high rate of patients with unknown outcomes may bias the association between these risk factors and outcome, and this limits the confidence of these findings. The limited follow-up data on patients from this cohort, and the significant proportion of patients who were lost to follow-up after outcomes were recorded, restricts the conclusions that can be drawn from these findings and introduces the possibility of bias. Finally, the precision of the findings and conclusions were limited by the relatively small size of this cohort. Data collection, however, was thorough and documentation on a wide variety of variables was available. We believe that in spite of these limitations, this analysis expands upon previous reports about MDR-TB treatment in this region and highlights areas where further research is needed to inform programmatic policy.