Time to Sputum Culture Conversion in Multidrug-Resistant Tuberculosis: Predictors and Relationship to Treatment Outcome FREE

—The Editors

The success of anti-tuberculosis (TB) pharmacologic treatment for pulmonary TB is directly linked to the microbiological status of a patient's sputum specimen during treatment. For patients with drug-susceptible TB who are treated in resource-limited settings, diagnosis is based on the presence of acid-fast bacilli identified in sputum through evaluation with smear microscopy. Patient management, including duration of treatment and final treatment success, is based on conversion of sputum smears to acid-fast bacilli-negative status (1). In countries with greater resources, diagnosis is generally based on growth of Mycobacterium tuberculosis from sputum mycobacterial culture, and treatment success is measured by conversion to no growth of M. tuberculosis on sputum culture (2).

For treatment of multidrug-resistant TB or TB with an isolate resistant to at least isoniazid and rifampicin (3), the status of mycobacterial cultures is generally used to guide therapy for patients treated in either resource-limited or resource-replete settings and is considered the most important interim indicator of the efficacy of treatment for multidrug-resistant TB (4–5). For example, in patients with multidrug-resistant TB and a persistently positive culture, treatment with an injected anti-TB medication, such as an aminoglycoside, is continued in addition to the multiple classes of other oral anti-TB medications until the threshold event of sputum culture conversion is achieved. Thereafter, frequency of dosing with the injected medication is often less frequent and ultimately is discontinued, treatment with the remaining oral medications is continued for another 12 to 18 months, and the sputum culture status is regularly monitored to document bacteriologic results indicative of sterilization.

Despite the importance of sputum culture conversion in the management of patients with multidrug-resistant TB, few studies of treatment of these patients have reported sputum culture conversion status or evaluated sputum culture conversion as an interim indicator of final treatment outcome (6–8). No studies of multidrug-resistant TB have reported on predictors of sputum culture conversion or the relationship between time required to achieve sputum culture conversion and final treatment outcomes (9–13).

Since the 1990s, Latvia has had one of the highest levels of multidrug-resistant TB in the world. However, efforts to control this disease, including the introduction of the World Health Organization (WHO)-recommended DOTS strategy for basic TB control in 1993 and the introduction of the DOTS-Plus strategy for multidrug-resistant TB in 1998 (4), resulted in a considerable decline in multidrug-resistant TB among patients with newly diagnosed sputum culture–positive disease (from 14.4% in 1996 to 9.2% in 2003) (14–16). Furthermore, outcomes have been encouraging for patients with multidrug-resistant TB who are treated under Latvia's national DOTS-Plus strategy, which uses regimens tailored to individual patients' anti-TB drug susceptibility profiles. In 2000, the first year of full treatment availability for all patients with multidrug-resistant TB in Latvia, 76% of patients adherent to a full course of therapy achieved successful treatment outcomes (17).

The objectives of this study were to characterize initial time to sputum culture conversion among the 2000 cohort of patients with multidrug-resistant TB who were treated under the DOTS-Plus strategy. The DOTS strategy is a 5-component program for TB control that consists of the following: 1) political commitment, 2) case detection through bacteriologic evaluation, 3) standardized treatment with supervision and patient support, 4) an effective drug supply system, and 5) a reporting and recording system that allows assessment of treatment. The DOTS-Plus strategy, which is undergoing development and testing, is designed to manage patients with multidrug-resistant TB using second-line anti-TB drugs in the context of a functioning DOTS program. This strategy is used in Latvia to determine predictors associated with time required for sputum culture conversion and to compare these predictors with those previously reported for poor treatment outcome (death, default, and failure) among the same cohort (17).

We defined a patient with pulmonary multidrug-resistant TB as having a positive sputum culture for M. tuberculosis with in vitro resistance to at least isoniazid and rifampicin. Our analysis included all new and previously treated patients with multidrug-resistant TB identified from the Latvian national TB registry and from the multidrug-resistant TB database who began treatment with second-line anti-TB drugs using the DOTS-Plus strategy between 1 January and 31 December 2000. These patients were classified into one of the following groups: group 1, new patients who were never treated for TB; group 2, new patients with multidrug-resistant TB who were previously treated for TB according to Latvian guidelines (could include 1 or 2 second-line drugs); or group 3, patients with multidrug-resistant TB who were previously treated for this disease with an individualized regimen. We excluded patients who had a negative culture taken at the time of treatment initiation because they had already had culture conversion.

All mycobacterial cultures were performed according to international standards at the Latvian national TB reference laboratory in Riga by using conventional Lowenstein–Jensen solid media, as previously described (17). Drug susceptibility testing was performed on M. tuberculosis isolates from sputum at the national laboratory following available international standards; quality was assured by the Swedish Institute for Infectious Disease Control, the WHO-designated supranational reference laboratory for the Baltic region. Laboratory performance, measured through evaluation of control strains provided by the supranational reference laboratory, was high; agreement was 100% for isoniazid and rifampin and greater than 90% for the other anti-TB drugs, including 95% for ofloxacin (18).

Individually tailored regimens are used to treat patients with multidrug-resistant TB in Latvia on the basis of results of in vitro drug susceptibility testing obtained before the treatment initiation. As soon as multidrug-resistant TB is diagnosed, treatment with second-line therapy is started with an empirical individualized regimen considering any previous anti-TB drugs received. This is done because of the 3- to 8-week delay in receiving the results of second-line drug susceptibility testing. The initial regimen consists of between 4 and 8 drugs, including 1 anti-TB drug that is injected. Regimens are modified according to the results of second-line drug susceptibility testing and include at least 5 drugs to which the patient's TB isolate was susceptible. Whenever possible, an injected drug, usually an aminoglycoside, is included in the initial daily treatment regimen until the monthly M. tuberculosis sputum culture converts to negative. Surgery is considered adjuvant therapy for patients with more advanced disease who are not responding well to treatment (delayed conversion) and can tolerate the procedure.

Patients are seen daily by clinicians while hospitalized and are seen monthly while receiving outpatient treatment. Sputum specimens are collected monthly for smear and culture examination and drug-susceptibility testing, and data are obtained on improvement of symptoms, treatment adherence, and side effects. All treatment is directly observed during the entire course of therapy. After M. tuberculosis sputum culture conversion, the injected medication is administered 5 times per week for 2 to 3 additional months and is then administered 3 times per week thereafter, on the basis of the patient's clinical status. Treatment is continued for 12 to 18 months, depending on severity of lung disease, history of treatment for TB, and general response to treatment. After treatment is completed, patients are followed for 2 years with sputum testing done every 6 months.

We used data from treatment charts and bacteriologic laboratory reports that were entered into the Latvian multidrug-resistant TB database. Side effect reports, expert consultation reports, and radiologic reports were also used. We collected baseline demographic characteristics and risk factor information, including age, sex, household location and size, employment status, alcohol use, history of imprisonment, height and weight, HIV status, previous treatment for TB and multidrug-resistant TB, previous contact with a patient with TB, history of surgical procedures for TB, the number of anti-TB drugs to which the M. tuberculosis strain was resistant at treatment initiation, anti-TB drugs used for 3 months or more, treatment duration, drug side effects, and adherence to treatment.

For this analysis, we applied multidrug-resistant TB treatment and outcome definitions recently developed by an international expert consensus group(5, 19). Treatment default was considered to have occurred in patients who interrupted treatment for 2 or more consecutive months. We defined colony count as the following: 1+ growth, less than 100 colonies per slant; 2+ growth, 100 to 200 colonies per slant; 3+ growth, 200 to 500 colonies per slant; and 4+ growth, confluent growth. Low body mass index (BMI), weight in kilograms divided by the square of height in meters, was defined as less than 18.5 kg/m²(20).

Patients who began multidrug-resistant TB treatment with a positive sputum culture and had 2 negative consecutive sputum cultures taken at least 30 days apart after initiation of treatment were considered to have sputum culture conversion. Among those with documented conversion, initial time to sputum culture conversion was calculated as the interval in days between the date of treatment initiation for multidrug-resistant TB and the collection date of the first of 2 consecutive negative sputum cultures. Time to sputum smear conversion was calculated similarly. In addition, we recognized that patients could achieve sputum culture conversion and become culture-positive again. The time to sputum culture reconversion was calculated as the interval in days between the date of initial conversion and the collection date of the first of 2 negative consecutive cultures that were not followed by another positive sputum culture during that treatment episode.

We evaluated predictor variables associated with time to initial sputum culture conversion. Categorical variables are reported as proportions, and the reported P values are for comparisons between groups by using the chi-square test, chi-square test for trend, or Fisher exact test. Continuous variables are reported as medians and were compared by using the Wilcoxon rank-sum test. All statistical tests were 2-sided, no multiple comparison adjustment was made, and a P value less than 0.05 was considered statistically significant.

Cumulative survival curves were constructed by using Kaplan–Meier estimates. From these survival estimates, the log cumulative hazards were inspected as part of the initial model selection diagnostics. We inspected the proportional hazards assumption and whether specific parametric models better suited the description of the behavior of the log cumulative hazards. On the basis of these results, an accelerated failure time model was chosen over the traditional Cox proportional hazards model because the assumption of proportional hazards was not regularly met. An accelerated failure time model does not necessarily rely on the proportional hazards assumption, although an underlying probability distribution must be assumed for the time to initial conversion. Although this distribution is not known, the best-fitting parametric model was chosen by using a combination of diagnostics, such as the plot of the log cumulative hazards, probability plots, and comparison of the Akaike information criterion between the various models with different distributional assumptions and no covariates. These results suggested that the 2 best-fitting models were the generalized γ distribution (Figure 1) and the log-normal model (Appendix Figure). There was little statistically significant difference between the 2 models, so the results from the generalized γ model are presented. Of note, neither of these distributions produces accelerated failure time models with proportional hazards.

The interpretation of the parameter coefficients in an accelerated failure time model is such that the parameters associated with a specific explanatory variable act multiplicatively on the time scale, so that the parameter reflects the speed or delay at which conversion occurs. For a dichotomous variable, the exponential of the parameter estimate from an accelerated failure time model gives an estimate of the ratio of the median conversion times of both the categories. The parameter coefficients in the accelerated failure time model were converted to the percentage difference in conversion time through the equation ([e^β− 1] × 100%). For continuous variables, this value indicates the percentage difference in conversion time associated with a 1-unit increment in the explanatory variable. For categorical variables, this value indicates the percentage difference in conversion time by comparing 1 level with the reference level. Positive values imply longer time to conversion, and negative values imply shorter time.

We identified a list of potential predictors guided by published papers, biological plausibility, and our previous experience, which included a list of approximately 30 covariates. Variables for which more than 20% of data were missing or those that were not routinely collected were not considered. Variables were created to indicate whether there was a specific drug resistance to antibiotics available in Latvia to treat multidrug-resistant TB and to indicate whether each of those drugs had been used for 3 months or more in the treatment regimen. We also created aggregate variables for the total number of drugs to which a patient's isolate was resistant at treatment initiation and for the total number of drugs given for 3 months or more. Because multicollinearity was a concern when including the individual variables and the aggregate totals, we considered variables separately when developing the model using the eliminative approach.

We performed a univariate analysis on each of the 30 covariates and considered those with a P value less than 0.20 for the multivariate model. We considered surgical intervention to be a secondary outcome in the model of conversion time and therefore did not include it in our modeling. We eliminated the variables with the largest P values individually until all the remaining variables had a P value less than 0.05. None of the variables that were eliminated produced changes of more than 30% in the parameter estimates of the previous model. Finally, we investigated all pairwise interactions among the remaining variables in the final model. Data were analyzed by using Epi Info, version 6.04d (U.S. Centers for Disease Control and Prevention, Atlanta, Georgia), and SAS for Windows, version 8.02 (SAS Institute, Inc., Cary, North Carolina). Missing data were not replaced or imputed. Graphics were produced by using R 1.8.1 (Computer-Aided Engineering Center, University of Wisconsin, Madison, Wisconsin).

This analysis was determined to be programmatic evaluation and was approved as nonhuman subjects research by the Office of the Associate Director for Science at the U.S. Centers for Disease Control and Prevention. The views expressed in this paper are solely those of the authors and do not reflect those of the U.S. Public Health Service or the government of Latvia.

Funding for this project was provided by the United States Agency for International Development (USAID), through Centers for Disease Control and Prevention Cooperative Agreement U23 CCU021873. The USAID did not have a role in the design or conduct of the study; in the collection, analysis, or interpretation of the data; in the preparation of the manuscript; or in the decision to submit the manuscript for publication.

The cohort consisted of 204 patients who began therapy for multidrug-resistant TB in 2000. Of these, 37 had negative sputum cultures taken at initiation of second-line treatment because they had conversion while taking first-line drugs. These patients were statistically significantly less likely to live in the capital city of Riga, to live alone, to have bilateral cavitations, and to have an isolate resistant to 5 or more drugs. Of the 167 patients who were sputum culture–positive at initiation of second-line therapy, 130 (78%) were men. The median patient age was 43 years (range, 17 to 70 years) for men and 39 years (range, 21 to 78 years) for women (P = 0.39). Most patients lived in their own home or apartment (n = 142 [85%]), more than three quarters (n = 137 [82%]) were born in Latvia, and approximately three quarters (n = 124 [74%]) had received first-line or second-line drugs for TB before the current episode of multidrug-resistant TB. Bilateral cavitations on initial chest radiography were present in 54 (32%) patients. Initial M. tuberculosis isolates were resistant to a median of 5 drugs (range, 2 to 10 drugs).

Among these 167 patients, 90 (53%) were also sputum smear–positive. In 18 (20%) of these 90 patients, the sputum smear did not convert to negative. Of the remaining 72 patients, the median time to initial sputum smear conversion was 83 days (range, 1 to 698 days). Forty-seven percent of patients missed 1 or 2 nonconsecutive monthly sputum collections, and 75% missed 4 or fewer collections.

In our cohort, 129 patients (77%) achieved sputum culture conversion and 38 (23%) did not. Among all patients, the median initial sputum culture conversion time was 83 days (Figure 2). Those who did not convert were more likely to have a history of treatment for multidrug-resistant TB (29% vs. 11%; P = 0.004), to have a history of incarceration (45% vs. 26%; P = 0.003), and to have resistance to a greater number of drugs at treatment initiation (6 vs. 5; P = 0.008). Among the 129 patients who converted, 16% did so after 1 month of treatment, 39% did so after 2 months of treatment, and 60% had converted after 4 months of treatment. For these patients, the median initial sputum culture conversion time was 60 days (range, 4 to 462 days).

We performed univariate analysis of median initial sputum culture conversion times among the 129 patients who converted, stratified by selected demographic and clinical variables. A high colony count (3+ or 4+) on initial culture (83 days vs. 47 days; P = 0.010); requiring surgical management during treatment (173 days vs. 56 days; P < 0.001); and the use of thiacetazone (69 days vs. 42 days; P = 0.047), para-aminosalicylic acid (71 days vs. 56 days; P = 0.008), or cycloserine for 3 months or more (68 days vs. 47 days; P = 0.02) were statistically significantly associated with longer initial sputum culture conversion times.

Fourteen (11%) of 129 patients achieved initial sputum culture conversion, subsequently had positive sputum cultures, and then achieved reconversion, according to our definition criteria. The median interval between initial conversion and reconversion for these 14 patients was 197 days (range, 70 to 746 days). Twelve patients had only 1 or 2 intervening positive sputum cultures (after 3 to 4 negative cultures), and 2 had 7 or more positive sputum cultures. All but 1 patient who ultimately experienced treatment failure had a successful treatment outcome. The median sputum culture conversion time after including reconversion was 69 days (range, 4 to 770 days).

Our univariate regression analysis (Table 1) showed that certain demographic and clinical characteristics, such as concurrent diabetes, category of previous treatment for multidrug-resistant TB (Figure 3), bilateral cavitations on initial chest radiography, high colony count on the initial sputum culture, resistance to more drugs at treatment initiation, and resistance to pyrazinamide or kanamycin, were statistically significantly associated with a longer time to initial sputum culture conversion. In addition, the use of prothionamide for 3 months or more was associated with a shorter time to initial sputum culture conversion.

In our multivariate regression model, several factors remained independently associated with time to sputum culture conversion (Table 2). The initial conversion time in patients previously treated for multidrug-resistant TB was 169% (95% CI, 49% to 384%) longer than that in those who had never been treated for TB. Patients with colony counts of 3+ or 4+ took 49% (CI, 5% to 111%) longer to achieve initial sputum culture conversion than those with colony counts of 1+ or 2+. The presence of bilateral cavitations on chest radiography was associated with a 47% (CI, 2% to 113%) longer time to initial sputum culture conversion. The addition of each additional drug to the drug resistance profile of the initial multidrug-resistant TB isolate was associated with a 16% (CI, 3% to 30%) longer time to initial sputum culture conversion.

The median initial sputum culture conversion time was 48 days among those who were cured or completed treatment versus 169 days among those with a poor outcome (P < 0.001). Of the 129 patients who ultimately had sputum culture conversion, 108 (84%) achieved a good outcome, 5 died (4%), 14 had treatment default (11%), and 2 had treatment failure (1%) (Table 3). Among the 38 patients who never had sputum culture conversion, 8 died (21%), 4 had treatment default (11%), and 26 had treatment failure (68%). In 2 years of follow-up after treatment completion, 3 (2%) patients required retreatment for multidrug-resistant TB. Of patients with initial sputum culture conversion within 60 days (2 months) of treatment, 86% achieved a good treatment outcome versus 51% of those who did not achieve conversion within 60 days (Table 3) (P < 0.001). For those not achieving conversion by 120 days (4 months), only 33% had a good treatment outcome.

Overall, we found that more than 75% of the 167 Latvian patients with pulmonary multidrug-resistant TB and sputum mycobacterial cultures positive for M. tuberculosis who initiated treatment in 2000 under the DOTS-Plus strategy achieved sputum culture conversion. Half of the patients converted within 12 weeks; among those who converted, the median time to conversion was 8 weeks. Despite the high rate of drug resistance in this population of patients with multidrug-resistant TB, this time frame is only twice as long as that observed for most patients with drug-susceptible TB.

From the survival analysis of demographic and clinical characteristics of the 167 patients with multidrug-resistant TB who were sputum culture–positive, we identified several independent predictors of longer time to sputum culture conversion while receiving treatment. Specifically, patients with a history of treatment for multidrug-resistant TB, those with bilateral cavitations on initial chest radiography, and those with a high colony count of M. tuberculosis on sputum culture at treatment initiation had prolonged sputum culture conversion. Furthermore, conversion was prolonged from 3% to 30% for each additional drug documented in a patient's drug resistance profile.

The aforementioned characteristics are generally identifiable before multidrug-resistant TB is diagnosed or early in the course of treatment. They give clinicians an opportunity to identify patients who might benefit from more aggressive diagnosis, treatment, and management, including earlier initiation of therapy using rapid testing methods (21); more aggressive initial regimens using more drugs (22); more timely use of adjunctive therapies, such as surgery (8); changes in a regimen already in progress; or treatment with more or the maximum number of drugs to which the patient's disease is susceptible(7). More frequent monitoring of sputum cultures in patients at risk for prolonged conversion in low-resource settings where sputum cultures are not done as often will allow earlier detection of those who are not experiencing conversion. The cost-effectiveness of more frequent sputum cultures, however, requires additional study.

Previous reports on time to sputum culture conversion have lacked data on total drug resistance and resistance to specific drugs (23–25), important information that our study adds to the literature. The finding that resistance to and use of certain drugs can prolong sputum culture conversion time is clinically significant. In Latvia, resistance to kanamycin and pyrazinamide and use of prothionamide seem to be the most important factors that prolong culture conversion time. Patients with these characteristics should be also targeted for aggressive management.

Achieving more rapid sputum culture conversion can simplify a patient's therapy and increase comfort by reducing the amount of time he or she is given an injectable drug. In addition, from the public health perspective, reducing the time to sputum culture conversion is an important infection control measure because patients with multidrug-resistant TB and positive sputum cultures are infectious and may transmit the disease to other persons, family members, and health care providers. This is especially true in settings of limited resources, such as in Latvia, where infection control capacity is less adequate.

Sputum culture conversion is used routinely as an indicator of the progress of treatment of multidrug-resistant TB despite little evidence to justify its use or to provide benchmarks against which progress can be measured. Our analysis confirmed that sputum culture conversion is a useful interim indicator of treatment outcome for patients with multidrug-resistant TB and that sputum culture conversion is an appropriate first goal of therapy. Specifically, we found that patients who achieved earlier sputum culture conversion were more likely to have successful treatment outcomes. Treatment outcomes were statistically significantly worse for patients who did not have sputum culture conversion within 2 months. A previous analysis of this cohort of patients showed that previous treatment for multidrug-resistant TB, use of 5 or fewer anti-TB drugs for extended treatment, resistance to ofloxacin, and a BMI less than 18.5 kg/m² at the time of treatment initiation were all independent predictors of poor treatment outcome (death and treatment failure)(17). Only 1 predictor of longer duration for sputum culture conversion, specifically previous treatment for multidrug-resistant TB, was common in patients with poor treatment outcomes. This is not surprising, because sputum culture conversion occurs more quickly than final treatment outcome and is a function of the pathogenesis of the disease. Thus, after conversion is achieved, other factors predictive of a good treatment outcome but not predictive of conversion, such as the number of drugs used for 3 months or more, a higher BMI, and sensitivity to ofloxacin, should guide clinicians' therapy decisions. The outcome data suggest that more than 5 drugs should be used for treating patients with multidrug-resistant TB in countries, such as Latvia, where there is extensive resistance to first-line and second-line drugs(17). Because fluoroquinolones, with their bactericidal and sterilizing effect(26–27), are integral to successful treatment of patients with multidrug-resistant TB, it is surprising that ofloxacin resistance did not also predict delayed sputum culture conversion.

Bilateral cavitations on chest radiography and a high colony count of M. tuberculosis, which are statistically significant risk factors for delayed conversion, were not identified as predictors of poor treatment outcome in the previous study of treatment outcomes. This suggests that even patients with fairly advanced disease and high colony counts of M. tuberculosis can be successfully treated with aggressive management (possibly more drugs) despite delayed conversion times. In fact, among the 54 patients in our study with multidrug-resistant TB who were sputum culture–positive at treatment initiation and had bilateral cavitations, 36 (67%) ultimately achieved sputum culture conversion.

Our study has several limitations. Although it is the policy of the DOTS-Plus program in Latvia to do sputum cultures monthly, they were not performed in every patient. Three quarters of the patients, however, missed only 4 or fewer nonconsecutive monthly sputum culture collections. Second, because cultures were done monthly, the actual number of days to conversion was not observed. Using our methods, we were not able to account for the interval censoring in this analysis. As with any model, our accelerated failure time model should be interpreted cautiously because the purpose of our multivariate analysis was to develop an initial descriptive model of some of the predictors of conversion time on the basis of the available data. We recognize that there are limitations in developing descriptive models with an eliminative approach. Finally, these findings of sputum culture conversion are true for patients in Latvia and may not be generalizable to patients with multidrug-resistant TB who are enrolled in other DOTS-Plus programs.

In conclusion, our findings build on limited previous data of sputum culture conversion in multidrug-resistant TB and show that under DOTS-Plus program conditions in Latvia, most patients with this disease achieved sputum culture conversion within 12 weeks of starting treatment. Our analysis found specific predictors for sputum culture conversion and found that chest radiography and drug susceptibility testing can assist physicians in predicting which patients will convert more slowly. We found a strong relationship between the interim indicator of sputum culture conversion and final treatment outcome. Our findings offer the possibility that early intervention will help achieve sputum culture conversion and ultimately a successful treatment outcome.