Abstract
Purpose
Evidence on treatment completion among patients with drug-resistant tuberculosis (DR-TB) in the Democratic Republic of the Congo is limited. We estimated the treatment completion rate and identified factors associated with completion among patients with rifampicin-resistant/multidrug-resistant tuberculosis (RR/MDR-TB) in Mbuji-Mayi.
Patients and Methods
This retrospective cohort included all patients with RR/MDR-TB who initiated treatment between January 2019 and December 2023 in the 10 health zones of Mbuji-Mayi. Treatment completion was defined as a documented completed treatment course without evidence of failure; all other recorded outcomes were classified as non-completion. Variables were classified as baseline or treatment-period variables. Modified Poisson regression with cluster-robust variance at the health-zone level estimated adjusted risk ratios (aRRs).
Results
Of 514 patients, 382 (74.3%; 95% CI 70.4–77.9) completed treatment. Completion ranged from 69.7% to 78.8% by year (p = 0.653) and from 60.4% to 87.5% by health zone (p = 0.041). After adjustment, completion was associated with age ≤50 years (aRR 1.21 (1.11–1.31)), normal nutritional status (aRR 1.30 (1.22–1.40)) and directly observed treatment (aRR 1.26 (1.16–1.37)). Among treatment-period variables, negative month-3 culture (aRR 1.36 (1.15–1.62)), absence of treatment interruption (aRR 1.68 (1.30–2.17)) and absence of drug stock-out (aRR 1.84 (1.44–2.34)) were also associated with completion. The age association was driven by patients older than 60 years.
Conclusion
About three quarters of patients completed DR-TB treatment, with variation across health zones. Because month-3 culture, treatment interruption and drug stock-outs were recorded during treatment, temporality cannot be established for these associations; prospective studies with complete outcome reporting are needed.
Keywords: drug-resistant tuberculosis, treatment completion, directly observed treatment, sputum culture conversion, drug stock-outs, Democratic Republic of the Congo
Introduction
Tuberculosis (TB) remains one of the leading causes of death from a single infectious agent worldwide.1 Rifampicin-resistant TB (RR-TB) and multidrug-resistant TB (MDR-TB), defined as resistance to at least isoniazid and rifampicin, require longer and more toxic treatment, are more costly to manage and have poorer outcomes than drug-susceptible TB, particularly in low-resource settings.2,3 In this article, drug-resistant TB (DR-TB) refers to RR/MDR-TB.
Rapid molecular diagnosis, all-oral regimens and newer drugs such as bedaquiline and delamanid have improved DR-TB care, but treatment still lasts 9 to 20 months and requires close clinical and bacteriological monitoring and sustained adherence. Global treatment success remains below international targets, and death, treatment failure and loss to follow-up continue to undermine TB elimination efforts.4
Treatment completion must be distinguished from treatment success. Under World Health Organization (WHO) definitions, treatment success combines patients who are cured and those who complete treatment without evidence of failure.5 Treatment completion, the outcome examined in this study, reflects retention in care until the end of the prescribed regimen and is a practical indicator of programme performance, but it is not equivalent to bacteriologically confirmed cure. Factors reported to be associated with DR-TB treatment outcomes include age, nutritional status or body mass index (BMI), HIV infection, previous TB treatment, education, comorbidities, treatment regimen, adverse drug reactions and patient support.2,3,6
Reported treatment success rates are 71.7% in Ghana,2 71.1% in Shenzhen, China,4 and approximately 69% to 77% in Ethiopia.3,6 In the Democratic Republic of the Congo (DRC), a province-wide cohort in South Kivu reported treatment success of 77.6% among patients treated between 2018 and 2024, although mortality remained substantial (13.4%) despite the introduction of bedaquiline-containing regimens.7 An earlier province-wide cohort in the same province described the predictors of successful outcomes among patients with Xpert-identified RR-TB.8
Evidence on DR-TB outcomes in the DRC therefore comes from a small number of provincial cohorts, and no study has examined treatment completion in Mbuji-Mayi, the capital of Kasai-Oriental Province. Mbuji-Mayi is one of the most populous cities of the DRC, and a published investigation conducted in Mbuji-Mayi Central Prison identified rifampicin resistance in 17 of 199 bacteriologically confirmed tuberculosis cases (8.5%) during 2014–2015, highlighting the documented presence of drug-resistant tuberculosis in this setting.9
DR-TB services in the city are organized and supervised at the level of its 10 health zones, which differ in the facilities and follow-up modalities through which patients are managed. The purpose of focusing on Mbuji-Mayi is therefore not to show that the city differs from the rest of the country, but to produce evidence at the level at which DR-TB services are managed and can be improved. Provincial or national estimates cannot show variation between health zones or identify the programme-level factors on which local managers can act. This retrospective cohort study aimed to estimate the treatment completion rate among patients with DR-TB in Mbuji-Mayi, describe its variation over time and across health zones, and identify baseline and treatment-period factors associated with completion.
Materials and Methods
Study Design and Setting
We conducted a retrospective cohort study of patients with DR-TB managed between January 1, 2019, and December 31, 2023, in the 10 health zones of the city of Mbuji-Mayi (Bipemba, Bonzola, Dibindi, Diulu, Kansele, Lubilanji, Lukelenge, Mpokolo, Muya and Nzamba), Kasai-Oriental Province, DRC. DR-TB diagnosis, treatment and follow-up are provided by 67 DR-TB treatment facilities operating under the National Tuberculosis Control Programme (NTP). Patients are followed either by directly observed treatment (DOT) at the facility or by community-based follow-up. The study is reported according to the STROBE guidelines for cohort studies.
Study Population and Eligibility Criteria
The source population comprised all patients recorded in the DR-TB treatment registers of the 10 health zones who initiated second-line treatment during the study period. Patients were eligible if they had bacteriologically confirmed RR/MDR-TB using Xpert MTB/RIF, initiated treatment between January 1, 2019, and December 31, 2023, and had a final treatment outcome recorded by the end of data extraction on June 8, 2026. Patients with incomplete data were excluded. Of 550 patients registered during the period, 36 were excluded because of incomplete data, leaving 514 patients for analysis. All 514 patients included in the analysis had pulmonary tuberculosis.
Sampling
All eligible patients were included (exhaustive sampling); no sample size calculation was therefore performed. With 132 non-completion events and 12 parameters in the multivariable model, the study had approximately 11 events per parameter.
Data Collection
Data were extracted from DR-TB treatment registers and individual patient treatment cards using a standardized extraction form covering sociodemographic, clinical, bacteriological, treatment and programmatic variables and the final treatment outcome. Data were entered in KoboCollect. Approximately 10% of extracted records were independently re-extracted by a second investigator, and discrepancies were resolved against the source documents before the database was finalized. No formal pretesting of the data extraction form was conducted.
Variables
Outcome
The primary outcome was treatment completion, defined according to national guidelines as completion of the prescribed DR-TB regimen without evidence of treatment failure, as documented in the register. Only patients recorded as having completed treatment were classified as treatment completers. Patients with any other recorded outcome were classified as not having completed treatment. The extracted dataset recorded the final outcome only as completed or not completed; the specific outcomes of non-completers (death, treatment failure, loss to follow-up, not evaluated) could not be disaggregated.
Exposures
The definitions of the study variables and the time at which each was measured are given in Table 1. Baseline variables were recorded at or before treatment initiation: age, sex, education, HIV status, nutritional status, comorbidities, baseline bacillary load, patient type, diagnosis-to-treatment interval, follow-up modality (DOT or community-based follow-up, assigned at treatment initiation), year of treatment initiation and health zone. Treatment-period variables were recorded after treatment initiation: sputum culture at month 3, adverse drug events, treatment interruption and drug stock-out. Nutritional status (normal, moderate or severe malnutrition) was taken as recorded by clinicians in the patient file; the method of assessment was not documented in the records. Age was dichotomized at 50 years following exploratory analysis of the data; this cut-off was not prespecified.
Table 1.
Operational Definitions and Coding of the Main Study Variables
| Variable | Timing | Operational Definition | Categories/Reference |
|---|---|---|---|
| Treatment completion | End of treatment | Completion of the prescribed DR-TB treatment regimen without evidence of treatment failure, as documented in the patient’s medical record and according to national guidelines | Completed / Not completed |
| Drug-resistant tuberculosis | Diagnosis | Bacteriologically confirmed rifampicin-resistant or multidrug-resistant tuberculosis (resistance to at least isoniazid and rifampicin) | RR/MDR-TB |
| Nutritional status | Baseline | Nutritional status recorded by the clinician in the patient file at treatment initiation (assessment method not documented) | Normal / moderate malnutrition / severe malnutrition |
| Directly observed treatment (DOT) | Baseline | DOT follow-up as documented in the treatment record, assigned at treatment initiation | DOT / community-based |
| Culture at month 3 | Treatment period | Sputum culture result documented at the third month of treatment | Negative / Positive |
| Treatment interruption | Treatment period | Any documented interruption of anti-tuberculosis treatment during follow-up | Yes / No |
| Drug stock-out | Treatment period | Any documented interruption in drug availability during treatment | Yes / No |
| Age | Baseline | Age in completed years at treatment initiation | ≤50 / >50 years (ref); groups ≤20, 21–60, >60 years |
| Sex | Baseline | Sex recorded in the register | Male / Female (ref) |
| Education level | Baseline | Highest level of education recorded | None, primary (ref) / secondary, higher |
| HIV status | Baseline | HIV test result recorded at treatment initiation | Positive / Negative (ref) |
| Comorbidity | Baseline | Any comorbidity recorded at treatment initiation (HIV, malnutrition, diabetes, other) | Yes / No |
| Baseline bacillary load | Baseline | Bacillary load recorded at diagnosis | Low, medium / High (ref) |
| Patient type | Baseline | New or previously treated for TB, as recorded in the register | Previously treated / New (ref) |
| Diagnosis-to-treatment interval | Baseline | Days between diagnostic result and treatment initiation | ≤2 / >2 days (ref) |
| Year of treatment initiation | Baseline | Calendar year of treatment initiation | 2019–2023 (per year) |
| Health zone | Baseline | Health zone of the treatment facility | 10 health zones of Mbuji-Mayi |
| Adverse events | Treatment period | Any adverse drug event recorded during treatment, by type | None / digestive, cutaneous, auditory, neurological |
Abbreviations: DOT, directly observed treatment; DR-TB, drug-resistant tuberculosis; RR/MDR-TB, rifampicin-resistant/multidrug-resistant tuberculosis.
Statistical Analysis
Categorical variables are presented as frequencies and percentages and continuous variables as medians with interquartile ranges (IQR). The overall completion rate is reported with a Wilson 95% confidence interval (CI), and completion is reported with numerators and denominators by year of treatment initiation and by health zone. Bivariable comparisons used Pearson’s chi-square test or Fisher’s exact test.
Because completion was a common outcome (74%), associations were estimated as risk ratios (RRs) using modified Poisson regression with robust variance.10 Patients were nested in 10 health zones; standard errors were therefore clustered at the health-zone level with a small-sample correction, and, because of the small number of clusters, 95% CIs and p values were derived from a t distribution with 9 degrees of freedom.11 Variables with p <0.20 in bivariable analysis and variables considered relevant in the literature were entered into the multivariable model.
Baseline and treatment-period variables are presented separately. Associations with treatment-period variables were interpreted as associations observed during treatment, not as baseline predictors, because they are susceptible to reverse causation and immortal-time bias (patients must remain on treatment long enough to have a month-3 culture or to experience an interruption or a stock-out).12 The analysis was explanatory; no predictive model or receiver operating characteristic analysis was developed.
Collinearity was assessed with variance inflation factors (all ≤1.21). One patient with missing sex was excluded from the multivariable model (n = 513). Three sensitivity analyses were performed: (i) generalized estimating equations (GEE) with a Poisson link and an exchangeable working correlation within health zones; (ii) replacement of the recorded nutritional status by BMI (≥18.5 vs <18.5 kg/m2) calculated from recorded weight and height; and (iii) exclusion of patients aged over 60 years. Statistical significance was set at a two-sided p <0.05. Analyses were performed with IBM SPSS Statistics version 26 and Python 3 (NumPy and SciPy).
Ethical Considerations
The study was conducted in accordance with the principles of the Declaration of Helsinki. Ethical approval was obtained from the Ethics Committee of the Official University of Mbuji-Mayi (Approval No. CERUOM/023/2026) before commencement of the study. Administrative authorization was also obtained from the Provincial Health Division and the participating DR-TB treatment facilities.
Because this was a retrospective study based on routinely collected programmatic data, the requirement for informed consent was waived by the ethics committee. Patient confidentiality was ensured by anonymizing all records before analysis, assigning a unique study identification number to each participant, and restricting database access to authorized investigators only. Similar ethical procedures have been reported in previous retrospective MDR-TB cohort studies.
Results
Characteristics of the Patients
Table 2 shows that the median age of the 514 patients was 34.0 years (IQR 23.0–47.8); 296 (57.7%) were male and 217 (42.3%) female, and sex was missing for one patient. Secondary education was the most frequent educational level (38.3%), and 15.2% had higher education. Normal nutritional status was recorded for 54.7% of patients, and 45.3% had moderate or severe malnutrition. HIV infection was recorded in 15.6% of patients, 61.3% were previously treated cases, and 49.0% had at least one comorbidity. During treatment, 65.8% of patients had a negative culture at month 3.
Table 2.
Characteristics of Patients with Drug-Resistant Tuberculosis Included in the Study (N = 514)
| Variable | N = 514 | % |
|---|---|---|
| Baseline characteristics | ||
| Age (years) | ||
| Median (IQR) | 34.0 (23.0–47.8) | — |
| Sex (missing = 1) | ||
| Female | 217 | 42.3 |
| Male | 296 | 57.7 |
| Education level | ||
| None | 52 | 10.1 |
| Primary | 187 | 36.4 |
| Secondary | 197 | 38.3 |
| Higher | 78 | 15.2 |
| Nutritional status | ||
| Normal | 281 | 54.7 |
| Moderate malnutrition | 153 | 29.8 |
| Severe malnutrition | 80 | 15.6 |
| HIV status | ||
| Negative | 434 | 84.4 |
| Positive | 80 | 15.6 |
| Previous TB treatment | ||
| Yes | 258 | 50.2 |
| No | 256 | 49.8 |
| Patient type | ||
| Previously treated | 315 | 61.3 |
| New | 199 | 38.7 |
| Other comorbidities | ||
| Yes | 252 | 49.0 |
| No | 262 | 51.0 |
| Treatment-period characteristic | ||
| Culture at month 3 | ||
| Negative | 338 | 65.8 |
| Positive | 176 | 34.2 |
Treatment Completion
Overall, 382 of 514 patients (74.3%; 95% CI 70.4–77.9) completed treatment and 132 (25.7%) did not. Completion did not differ significantly by year of treatment initiation, ranging from 69.7% (69/99) in 2021 to 78.8% (67/85) in 2019 (p = 0.653; Figure 1). It varied across health zones (p = 0.041), from 60.4% (29/48) in Bonzola and 61.7% (37/60) in Lubilanji to 87.0% (20/23) in Kansele and 87.5% (21/24) in Mpokolo (Figure 2).
Figure 1.
Annual treatment completion among patients with drug-resistant tuberculosis, Mbuji-Mayi, 2019–2023. Bars show the percentage completing treatment, labels give the number completed/number initiated, and error bars show Wilson 95% confidence intervals (p = 0.653, chi-square test).
Figure 2.
Treatment completion by health zone among patients with drug-resistant tuberculosis, Mbuji-Mayi, 2019–2023. Labels give the number completed/number treated in each health zone; error bars show Wilson 95% confidence intervals; the dashed line indicates the overall completion rate (74.3%) (p = 0.041, chi-square test).
Bivariable Analysis
Table 3 shows treatment completion according to patient characteristics. Completion was 81.6% among patients aged 21–60 years and 73.4% among those aged 20 years or younger, but only 13.6% (6/44) among those older than 60 years (p <0.001). Completion decreased with recorded malnutrition (86.8%, 68.6% and 41.2% for normal nutritional status, moderate and severe malnutrition; p <0.001) and with increasing baseline bacillary load (82.9%, 75.1% and 63.8%; p <0.001), and was higher with DOT than with community-based follow-up (79.5% vs 64.0%; p <0.001). Among treatment-period variables, completion was higher among patients with a negative month-3 culture (82.8% vs 58.0%), without treatment interruption (85.5% vs 40.6%) and without drug stock-out (87.2% vs 39.6%) (all p <0.001).
Table 3.
Clinical and Programmatic Characteristics According to Treatment Completion (N = 514)
| Variable | Category | n | Completed n (%) | Not Completed n (%) | p |
|---|---|---|---|---|---|
| Baseline variables | |||||
| Age group (years) | <0.001 | ||||
| ≤20 | 94 | 69 (73.4) | 25 (26.6) | ||
| 21–60 | 376 | 307 (81.6) | 69 (18.4) | ||
| >60 | 44 | 6 (13.6) | 38 (86.4) | ||
| Sex (missing = 1) | 0.395 | ||||
| Male | 296 | 224 (75.7) | 72 (24.3) | ||
| Female | 217 | 157 (72.4) | 60 (27.6) | ||
| Education level | 0.125 | ||||
| None | 52 | 36 (69.2) | 16 (30.8) | ||
| Primary | 187 | 134 (71.7) | 53 (28.3) | ||
| Secondary | 197 | 146 (74.1) | 51 (25.9) | ||
| Higher | 78 | 66 (84.6) | 12 (15.4) | ||
| HIV status | 0.129 | ||||
| Negative | 434 | 328 (75.6) | 106 (24.4) | ||
| Positive | 80 | 54 (67.5) | 26 (32.5) | ||
| Nutritional status | <0.001 | ||||
| Normal | 281 | 244 (86.8) | 37 (13.2) | ||
| Moderate malnutrition | 153 | 105 (68.6) | 48 (31.4) | ||
| Severe malnutrition | 80 | 33 (41.3) | 47 (58.8) | ||
| Type of comorbidity | 0.043 | ||||
| None | 262 | 193 (73.7) | 69 (26.3) | ||
| HIV | 90 | 62 (68.9) | 28 (31.1) | ||
| Malnutrition | 81 | 70 (86.4) | 11 (13.6) | ||
| Diabetes | 42 | 32 (76.2) | 10 (23.8) | ||
| Others | 39 | 25 (64.1) | 14 (35.9) | ||
| Baseline bacillary load | <0.001 | ||||
| Low | 164 | 136 (82.9) | 28 (17.1) | ||
| Medium | 201 | 151 (75.1) | 50 (24.9) | ||
| High | 149 | 95 (63.8) | 54 (36.2) | ||
| Patient type | 0.419 | ||||
| New | 199 | 144 (72.4) | 55 (27.6) | ||
| Previously treated | 315 | 238 (75.6) | 77 (24.4) | ||
| Diagnosis-to-treatment interval | 0.398 | ||||
| ≤2 days | 233 | 169 (72.5) | 64 (27.5) | ||
| >2 days | 281 | 213 (75.8) | 68 (24.2) | ||
| Follow-up modality | <0.001 | ||||
| DOT | 342 | 272 (79.5) | 70 (20.5) | ||
| Community-based | 172 | 110 (64.0) | 62 (36.0) | ||
| Year of treatment initiation | 0.653 | ||||
| 2019 | 85 | 67 (78.8) | 18 (21.2) | ||
| 2020 | 63 | 48 (76.2) | 15 (23.8) | ||
| 2021 | 99 | 69 (69.7) | 30 (30.3) | ||
| 2022 | 204 | 153 (75.0) | 51 (25.0) | ||
| 2023 | 63 | 45 (71.4) | 18 (28.6) | ||
| Health zone | 0.041 | ||||
| Bipemba | 57 | 45 (78.9) | 12 (21.1) | ||
| Bonzola | 48 | 29 (60.4) | 19 (39.6) | ||
| Dibindi | 47 | 35 (74.5) | 12 (25.5) | ||
| Diulu | 46 | 36 (78.3) | 10 (21.7) | ||
| Kansele | 23 | 20 (87.0) | 3 (13.0) | ||
| Lubilanji | 60 | 37 (61.7) | 23 (38.3) | ||
| Lukelenge | 84 | 65 (77.4) | 19 (22.6) | ||
| Mpokolo | 24 | 21 (87.5) | 3 (12.5) | ||
| Muya | 82 | 65 (79.3) | 17 (20.7) | ||
| Nzamba | 43 | 29 (67.4) | 14 (32.6) | ||
| Treatment-period variables | |||||
| Adverse events | 0.152 | ||||
| None | 319 | 227 (71.2) | 92 (28.8) | ||
| Digestive | 82 | 63 (76.8) | 19 (23.2) | ||
| Cutaneous | 40 | 34 (85.0) | 6 (15.0) | ||
| Auditory | 38 | 28 (73.7) | 10 (26.3) | ||
| Neurological | 35 | 30 (85.7) | 5 (14.3) | ||
| Culture at month 3 | <0.001 | ||||
| Negative | 338 | 280 (82.8) | 58 (17.2) | ||
| Positive | 176 | 102 (58.0) | 74 (42.0) | ||
| Treatment interruption | <0.001 | ||||
| None | 386 | 330 (85.5) | 56 (14.5) | ||
| At least one | 128 | 52 (40.6) | 76 (59.4) | ||
| Drug stock-out | <0.001 | ||||
| No | 375 | 327 (87.2) | 48 (12.8) | ||
| Yes | 139 | 55 (39.6) | 84 (60.4) | ||
| Total | 514 | 382 (74.3) | 132 (25.7) |
In bivariable regression accounting for clustering by health zone (Table 4), completion was associated with age ≤50 years (RR 1.24 (1.06–1.45)), normal nutritional status (RR 1.47 (1.30–1.65)), low or medium baseline bacillary load (RR 1.23 (1.02–1.48)), DOT (RR 1.24 (1.09–1.42)), negative month-3 culture (RR 1.43 (1.13–1.81)), absence of treatment interruption (RR 2.10 (1.41–3.13)) and absence of drug stock-out (RR 2.20 (1.55–3.14)). Sex, education, HIV status, comorbidity, patient type, adverse events and diagnosis-to-treatment interval were not significantly associated with completion.
Table 4.
Bivariable Analysis of Factors Associated with Treatment Completion
| Variable | Completed, n/N (%) | Completed in Reference Group, n/N (%) | Crude RR (95% CI) | p-value |
|---|---|---|---|---|
| Age ≤50 years (ref: >50) | 319/413 (77.2) | 63/101 (62.4) | 1.24 (1.06–1.45) | 0.013 |
| Male sex (ref: female) | 224/296 (75.7) | 157/217 (72.4) | 1.05 (0.85–1.29) | 0.635 |
| Secondary/higher education (ref: none/primary) | 212/275 (77.1) | 170/239 (71.1) | 1.08 (0.94–1.25) | 0.231 |
| HIV negative (ref: positive) | 328/434 (75.6) | 54/80 (67.5) | 1.12 (0.93–1.34) | 0.190 |
| Low/medium bacillary load (ref: high) | 287/365 (78.6) | 95/149 (63.8) | 1.23 (1.02–1.48) | 0.031 |
| Normal nutritional status (ref: malnutrition) | 244/281 (86.8) | 138/233 (59.2) | 1.47 (1.30–1.65) | <0.001 |
| No comorbidity (ref: ≥1 comorbidity) | 193/262 (73.7) | 189/252 (75.0) | 0.98 (0.84–1.15) | 0.804 |
| Negative culture at month 3 (ref: positive)a | 280/338 (82.8) | 102/176 (58.0) | 1.43 (1.13–1.81) | 0.008 |
| Previously treated (ref: new) | 238/315 (75.6) | 144/199 (72.4) | 1.04 (0.87–1.26) | 0.609 |
| No adverse event (ref: ≥1 adverse event)a | 227/319 (71.2) | 155/195 (79.5) | 0.90 (0.76–1.06) | 0.170 |
| No drug stock-out (ref: stock-out)a | 327/375 (87.2) | 55/139 (39.6) | 2.20 (1.55–3.14) | <0.001 |
| No treatment interruption (ref: ≥1 interruption)a | 330/386 (85.5) | 52/128 (40.6) | 2.10 (1.41–3.13) | 0.002 |
| DOT follow-up (ref: community-based) | 272/342 (79.5) | 110/172 (64.0) | 1.24 (1.09–1.42) | 0.005 |
| Diagnosis-to-treatment interval ≤2 days (ref: >2 days) | 169/233 (72.5) | 213/281 (75.8) | 0.96 (0.87–1.05) | 0.317 |
Multivariable Analysis
After adjustment (Table 5), three baseline characteristics were associated with completion: age ≤50 years (aRR 1.21 (1.11–1.31); p <0.001), normal nutritional status (aRR 1.30 (1.22–1.40); p <0.001) and DOT (aRR 1.26 (1.16–1.37); p <0.001). Among treatment-period variables, a negative month-3 culture (aRR 1.36 (1.15–1.62); p = 0.003), absence of treatment interruption (aRR 1.68 (1.30–2.17); p = 0.001) and absence of drug stock-out (aRR 1.84 (1.44–2.34); p <0.001) were also associated with completion. The bivariable association with baseline bacillary load did not persist after adjustment (aRR 0.98 (0.89–1.08)). Sex, HIV status, education, year of treatment initiation and adverse events were not significantly associated with completion.
Table 5.
Multivariable Poisson Regression Analysis of Factors Associated with Treatment Completion (n = 513)
| Variable | aRR (95% CI) | p-value | GEE aRR (95% CI)a | p-valuea |
|---|---|---|---|---|
| Baseline variables | ||||
| Age ≤50 years | 1.21 (1.11–1.31) | <0.001 | 1.21 (1.12–1.30) | <0.001 |
| Male sex | 0.98 (0.87–1.10) | 0.674 | 0.99 (0.88–1.10) | 0.822 |
| HIV positive | 0.92 (0.82–1.02) | 0.090 | 0.90 (0.81–1.01) | 0.060 |
| Secondary/Higher education | 0.99 (0.89–1.09) | 0.794 | 0.99 (0.90–1.09) | 0.844 |
| Low/medium bacillary load | 0.98 (0.89–1.08) | 0.658 | 0.97 (0.88–1.07) | 0.530 |
| Normal nutritional status | 1.30 (1.22–1.40) | <0.001 | 1.31 (1.23–1.40) | <0.001 |
| DOT follow-up | 1.26 (1.16–1.37) | <0.001 | 1.26 (1.17–1.36) | <0.001 |
| Treatment year (per year) | 0.99 (0.95–1.04) | 0.789 | 0.99 (0.94–1.03) | 0.528 |
| Treatment-period variables | ||||
| Negative culture at month 3 | 1.36 (1.15–1.62) | 0.003 | 1.36 (1.16–1.59) | 0.002 |
| Adverse events | 1.05 (0.94–1.18) | 0.342 | 1.05 (0.93–1.18) | 0.368 |
| No treatment interruption | 1.68 (1.30–2.17) | 0.001 | 1.67 (1.32–2.10) | <0.001 |
| No drug stock-out | 1.84 (1.44–2.34) | <0.001 | 1.83 (1.45–2.32) | <0.001 |
Note: ªGeneralized estimating equations (GEE) Poisson model with robust standard errors and an exchangeable working correlation structure, accounting for clustering of patients within treatment centres.
Abbreviations: aRR, adjusted risk ratio; CI, confidence interval; DOT, directly observed treatment; GEE, generalized estimating equations.
Sensitivity Analyses
GEE estimates were virtually identical to the cluster-robust estimates (within-zone correlation 0.03; Table 5). When the recorded nutritional status was replaced by BMI, BMI ≥18.5 kg/m2 was not associated with completion (aRR 1.03 (0.95–1.12)), and the aRR for age ≤50 years decreased to 1.11 (1.03–1.20); the other associations were similar (DOT 1.30 (1.19–1.41); negative month-3 culture 1.35 (1.16–1.59); no interruption 1.78 (1.35–2.34); no stock-out 1.85 (1.43–2.40)). After exclusion of the 44 patients aged over 60 years (n = 470), age ≤50 years was no longer associated with completion (aRR 0.92 (0.85–1.00)), whereas the associations with normal nutritional status (1.18 (1.11–1.25)), DOT (1.28 (1.19–1.37)), negative month-3 culture (1.34 (1.13–1.60)), absence of interruption (1.43 (1.20–1.71)) and absence of stock-out (1.51 (1.24–1.85)) persisted.
Discussion
Three quarters of patients with DR-TB in Mbuji-Mayi completed treatment, a level close to that reported elsewhere in the DRC and in the region, but with marked variation between health zones. The factors associated with completion relate mainly to programme delivery (supervision, continuity of treatment and drug supply) and to the bacteriological response, whereas the associations with age and recorded nutritional status are less robust.
The completion rate is comparable to the treatment success reported in Ethiopia,13 South Kivu7,8 and Guinea among patients receiving a shorter regimen,14 and slightly higher than in Ghana and China2,4 and the pooled estimate of a meta-analysis.15 These comparisons are only indicative: treatment success combines cure and completion and is a stricter end point,5 whereas our outcome was completion as recorded in the registers, and the composition of non-completion (death, failure, loss to follow-up) could not be described. The value of the local analysis lies less in the overall rate than in the variation it reveals within the city: completion ranged from about 60% to 88% across health zones, a variation that provincial or national estimates cannot show. We did not collect facility-level characteristics and therefore cannot explain this variation; it identifies health zones in which service delivery warrants review.
Patients followed by DOT had a 26% higher probability of completion than those followed in the community (aRR 1.26 (1.16–1.37)). This is consistent with observations from Shenzhen, where favourable outcomes were reported with supervised treatment combined with patient support and management of adverse events.4 Regular contact with providers may help detect missed doses and adverse effects early. However, follow-up modality was not randomly allocated; patients living closer to facilities or in better clinical or social conditions may have been preferentially assigned to DOT, and residual confounding cannot be excluded.
Normal nutritional status was associated with a 30% higher probability of completion (aRR 1.30 (1.22–1.40)). Low BMI has been associated with unfavourable outcomes in Ghana and Ethiopia.2,6 This finding should nevertheless be interpreted with caution. Nutritional status was recorded by clinicians without a documented assessment method, it did not agree with the BMI calculated from recorded weight and height, and BMI itself was not associated with completion in the sensitivity analysis (aRR 1.03 (0.95–1.12)). The recorded category may reflect a global clinical assessment, possibly including deterioration during follow-up, and may therefore be affected by reverse causation. Standardized measurement of nutritional status at treatment initiation is needed to clarify this association.
A negative month-3 culture was associated with completion (aRR 1.36 (1.15–1.62)), in line with evidence that early culture conversion is associated with favourable end-of-treatment outcomes.16,17 Absence of treatment interruption (aRR 1.68 (1.30–2.17)) and absence of drug stock-out (aRR 1.84 (1.44–2.34)) were associated with 68% and 84% higher probabilities of completion, respectively. These three variables were documented during treatment, and their associations must be interpreted with caution. Patients had to remain in care long enough to have a month-3 culture result or to experience an interruption or a stock-out (immortal-time bias),12 and treatment interruption partly overlaps with the process that leads to loss to follow-up (reverse causation). Drug stock-outs, by contrast, are a health-system factor that is largely independent of patient behaviour; their association with completion is consistent with the importance of continuous access to second-line medicines reported elsewhere,4 and identifies supply-chain performance as an area for programme review.
The association with age requires particular caution. Completion was 73.4% among patients aged 20 years or younger and 81.6% among those aged 21–60 years, but 13.6% (6/44) among those older than 60 years. Within the 21–60 year group, all 57 patients aged 51–60 years were recorded as having completed treatment, a pattern that is unlikely to reflect a biological gradient and suggests possible misclassification of age or outcome in the source records. The association with age ≤50 years was driven by patients older than 60 years and was no longer significant when they were excluded (aRR 0.92 (0.85–1.00)). Older age has been associated with unsuccessful outcomes in Ethiopia,18 whereas no independent association was found in Uganda or in another Ethiopian cohort.19,20
Sex, education, HIV status, adverse events and year of treatment initiation were not significantly associated with completion. Some confidence intervals were compatible with modest associations (for example, HIV infection, aRR 0.92 (0.82–1.02)), and these null findings should not be interpreted as evidence of no association. HIV infection and male sex have been associated with unfavourable outcomes in Ethiopia and Vietnam.13,21 The bivariable association with baseline bacillary load did not persist after adjustment; this attenuation is expected if early bacteriological response, which was included in the model, lies on the pathway between baseline bacillary load and completion.
For programme managers and clinicians in Mbuji-Mayi, these findings point to areas warranting attention rather than to proven interventions: monitoring of second-line drug stocks and redistribution between facilities to prevent stock-outs; early identification and tracing of patients who interrupt treatment; systematic month-3 bacteriological follow-up; nutritional assessment and support at treatment initiation; review of service delivery in the lower-performing health zones; and support for community-based follow-up, whose patients completed treatment less often. Registers should also record all treatment outcomes according to the WHO 2021 definitions5 so that deaths, failures and losses to follow-up can be monitored separately.
Limitations
This study has several limitations. First, it relied on routinely collected programme data, which may be incomplete or inaccurately recorded; internal inconsistencies were found in the registers, notably between recorded nutritional status and BMI, between the history of previous TB treatment and patient type, and between HIV serology and the type of comorbidity recorded, and the age distribution of completion suggests possible misclassification. In addition, the data extraction form was not formally pretested, although approximately 10% of records were independently re-extracted. Second, the outcome was available only as completed or not completed, so the specific outcomes of non-completers could not be analysed and the outcome cannot be equated with treatment success. Third, 36 of the 550 registered patients (6.5%) were excluded because of incomplete data, and no information was available to compare them with the included patients; this may have introduced selection bias if their characteristics differed from those of the included patients. Fourth, treatment-period variables are subject to immortal-time bias and reverse causation, and their associations do not establish temporality. Fifth, treatment regimen, socioeconomic status, distance to the treatment facility, social support, substance use and facility characteristics were not available, so residual confounding is likely. Sixth, with only 10 health zones, cluster-robust inference is approximate, although GEE gave similar results. Seventh, the age cut-off of 50 years was selected following exploratory analysis rather than being prespecified. Moreover, the association between age and treatment completion was not robust in the sensitivity analysis excluding patients older than 60 years, suggesting that the observed age association should be interpreted cautiously. Finally, the study was conducted in one city, and the findings may not apply to other settings in the DRC.
Conclusion
In Mbuji-Mayi, one in four patients with DR-TB did not complete treatment, and completion differed by up to 27 percentage points between health zones. For the National Tuberculosis Control Programme and provincial health authorities, these findings point to areas warranting priority attention: an uninterrupted supply of second-line drugs, early detection and tracing of treatment interruptions, supervised follow-up with systematic month-3 bacteriological monitoring, and nutritional assessment at treatment initiation, with particular attention to the lower-performing health zones. Because the data are retrospective and several of these factors were recorded during treatment, the associations do not establish causality; prospective cohorts that record all WHO treatment outcomes and use standardized nutritional assessment are needed to establish temporality and to reduce the risks of reverse causation and immortal-time bias.
Abbreviations
aRR, adjusted risk ratio; BMI, body mass index; CI, confidence interval; DOT, directly observed treatment; DR-TB, drug-resistant tuberculosis; DRC, Democratic Republic of the Congo; GEE, generalized estimating equations; IQR, interquartile range; MDR-TB, multidrug-resistant tuberculosis; NTP, National Tuberculosis Control Programme; RR, risk ratio; RR-TB, rifampicin-resistant tuberculosis; TB, tuberculosis; WHO, World Health Organization.
Disclosure
The authors declare that they have no conflicts of interest in relation to this work.
References
- 1.World Health Organization. Global tuberculosis report 2024. World Health Organization; 2024. [Google Scholar]
- 2.Panford V, Kumah E, Kokuro C, et al. Treatment outcomes and associated factors among patients with multidrug-resistant tuberculosis in Ashanti Region, Ghana: a retrospective, cross-sectional study. BMJ Open. 2022;12(7):e062857. doi: 10.1136/bmjopen-2022-062857 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Belachew T, Yaheya S, Tilahun N, et al. Multidrug-resistant tuberculosis treatment outcome and associated factors at the university of Gondar comprehensive specialized hospital: a ten-year retrospective study. Infect Drug Resist. 2022;15:2891–12. doi: 10.2147/IDR.S365394 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Ji L, Mijiti P, Hong C, et al. Treatment outcomes of multidrug-resistant tuberculosis patients receiving ambulatory treatment in Shenzhen, China: a retrospective cohort study. Front Public Health. 2023;11:1134938. doi: 10.3389/fpubh.2023.1134938 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.World Health Organization. Meeting report of the WHO expert consultation on drug-resistant tuberculosis treatment outcome definitions, 17–19 November 2020. World Health Organization; 2021. [Google Scholar]
- 6.Wakjira MK, Sandy PT, Mavhandu-Mudzusi AH. Treatment outcomes of patients with MDR-TB and its determinants at referral hospitals in Ethiopia. PLoS One. 2022;17(2):e0262318. doi: 10.1371/journal.pone.0262318 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Musafiri EM, Kabesha BT, Birembano F, et al. Drug-resistant tuberculosis outcomes in South Kivu Province, Democratic Republic of Congo, during a humanitarian crisis (2018–2024). Open Forum Infect Dis. 2026;13(5). doi: 10.1093/ofid/ofag303 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Bulabula ANH, Nelson JA, Musafiri EM, et al. Prevalence, predictors, and successful treatment outcomes of Xpert MTB/RIF-identified rifampicin-resistant tuberculosis in post-conflict eastern Democratic Republic of the Congo, 2012–2017: a retrospective province-wide cohort study. Clin Infect Dis. 2019;69(8):1278–1287. doi: 10.1093/cid/ciy1105 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Kayomo MK, Hasker E, Aloni M, et al. Outbreak of tuberculosis and multidrug-resistant tuberculosis, Mbuji-Mayi Central Prison, Democratic Republic of the Congo. Emerg Infect Dis. 2018;24(11):2029–2035. doi: 10.3201/eid2411.180769 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Zou G. A modified poisson regression approach to prospective studies with binary data. Am J Epidemiol. 2004;159(7):702–706. doi: 10.1093/aje/kwh090 [DOI] [PubMed] [Google Scholar]
- 11.Cameron AC, Miller DL. A practitioner’s guide to cluster-robust inference. J Hum Resour. 2015;50(2):317–372. doi: 10.3368/jhr.50.2.317 [DOI] [Google Scholar]
- 12.Suissa S. Immortal time bias in pharmaco-epidemiology. Am J Epidemiol. 2008;167(4):492–499. doi: 10.1093/aje/kwm324 [DOI] [PubMed] [Google Scholar]
- 13.Tola H, Holakouie-Naieni K, Mansournia MA, et al. National treatment outcome and predictors of death and treatment failure in multidrug-resistant tuberculosis in Ethiopia: a 10-year retrospective cohort study. BMJ Open. 2021;11(8):e040862. doi: 10.1136/bmjopen-2020-040862 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Hassane-Harouna S, Cherif GF, Ortuno-Gutierrez N, et al. Better programmatic outcome with the shorter regimen for the treatment of multidrug-resistant tuberculosis (MDR-TB) in Guinea: a retrospective cohort study. PLoS One. 2020;15(8):e0237355. doi: 10.1371/journal.pone.0237355 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Kibret KT, Moges Y, Memiah P, Biadgilign S. Treatment outcomes for multidrug-resistant tuberculosis under DOTS-Plus: a systematic review and meta-analysis of published studies. Infect Dis Poverty. 2017;6(1):7. doi: 10.1186/s40249-016-0214-x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Kurbatova EV, Cegielski JP, Lienhardt C, et al. Sputum culture conversion as a prognostic marker for end-of-treatment outcome in patients with multidrug-resistant tuberculosis: a secondary analysis of data from two observational cohort studies. Lancet Respir Med. 2015;3(3):201–209. doi: 10.1016/S2213-2600(15)00036-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Chen Q, He W, Du J, et al. Predictors of early and interim culture un-conversion in multidrug-resistant/rifampicin-resistant tuberculosis: a retrospective multi-center cohort study in China. Antimicrob Resist Infect Control. 2024;13(1):126. doi: 10.1186/s13756-024-01480-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Anito AA, Abdissa G, Gezahegn Y, Muhammed EU, Lemi H. Treatment outcomes and associated factors among patients with drug-resistant tuberculosis in Ethiopia: a retrospective cohort study. IJID Reg. 2026;19:100905. doi: 10.1016/j.ijregi.2026.100905 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Barteka G, Bwayo D, Matovu JKB, et al. Treatment outcomes and predictors of success for multidrug resistant tuberculosis (MDR TB) in Ugandan regional referral hospitals. Sci Rep. 2025;15(1):14144. doi: 10.1038/s41598-025-97027-x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Achalu DL, Kiltu AB, Teferi M, et al. Treatment outcomes of standardized injectable shorter regimen for multi-drugs resistant tuberculosis in Ethiopia: a retrospective cohort study. BMC Infect Dis. 2024;24(1):837. doi: 10.1186/s12879-024-09745-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Wrohan I, Nguyen TA, Nguyen VN, et al. Predictors of treatment outcomes among patients with multidrug-resistant tuberculosis in Vietnam: a retrospective cohort study. BMC Infect Dis. 2022;22(1):68. doi: 10.1186/s12879-021-06992-x [DOI] [PMC free article] [PubMed] [Google Scholar]