Sex differences in risk factors for unsuccessful tuberculosis treatment outcomes in Eastern Europe from 2020 to 2022: a multi-country retrospective cohort study Ole Skouvig Pedersen, a Tetiana Butova, b Valerii Miasoiedov, c Yurii Feshchenko, d Mykhailo Kuzhko, d Stefan Niemann, e,f ,g,h Alex Rosenthal, i Alina Grinev, i Gabriel Rosenfeld, i Michael Drew Hoppes, i Julia Kilmnick, i Valeriu Crudu, j Nelly Ciobanu, j Alexandru Codreanu, j Bekzat Toxanbayeva, k Lyailya Chingissova, l Kateryna Yurko, c Valerii Kucheriavchenko, c Vitalii Vekshyn, c Sergo Vashakidze, l,m Natalia Shubladze, l Zaza Avaliani, l,n Abdullaat Kadyrov, o Gulmira Kalmambetova, o Merbubu Sydykova, o Eugenia Ghita, p Victor Ionel Grecu, q Alina Marinela Miulescu, r Christian Morberg Wejse, s Andreas Fløe, a Victor Naestholt Dahl, s,t,∗ and Dmytro Butov c,e,t a Department of Respiratory Diseases and Allergy, Aarhus University Hospital, Aarhus, Denmark b Outpatient Department, Merefa Central District Hospital, Merefa, Ukraine c Kharkiv National Medical University, Kharkiv, Ukraine d National Scientific Center of Phthisiatry, Pulmonology and Allergology named after F. G. Yanovskyi, National Academy of Medical Sciences of Ukraine, Kyiv, Ukraine e Molecular and Experimental Mycobacteriology Group, Research Center Borstel, Borstel, Germany f German Center for Infection Research (DZIF), Partner Site Hamburg-Lübeck-Borstel-Riems, Borstel, Germany g EPHE, PSL University, Paris, France h Institut de Systématique, Évolution, Biodiversité (ISYEB), Muséum national d’Histoire naturelle, CNRS, Sorbonne Université, EPHE, Université des Antilles, Paris, France i Office of Cyber Infrastructure and Computational Biology, U.S. National Institute of Allergy and Infectious Diseases, Rockville, MD, USA j National TB Reference Laboratory, Institute of Phthisiopneumology, Chisinau, Moldova k National Reference Laboratory, National Scientific Center of Phthisiopulmonology, Almaty, Kazakhstan l National Center for TB and Lung Diseases, Tbilisi, Georgia m The University of Georgia, Tbilisi, Georgia n European University, Tbilisi, Georgia o National Tuberculosis Center, Bishkek, Kyrgyz Republic p Spirans Association, Bucharest, Romania q TB Ambulatory, Victor Babes Clinical Hospital for Infectious Disease and Pneumophthisiology, Craiova, Romania r IIIrd Pneumo-Phthisiology Ward, Leamna Phthisiology Hospital, Bucovat, Romania s Department of Infectious Diseases, Aarhus University Hospital, Aarhus, Denmark Summary Background Addressing the disproportionate representation between sexes is essential for achieving universal health coverage. Studies on the association between sex and unsuccessful tuberculosis treatment outcomes have shown conflicting results. This study examines this association and analyses sex-stratified risk factors associated with unsuccessful outcomes. Methods This retrospective, observational cohort study analysed prospectively collected data from six Eastern Eu- ropean countries from 2020 to 2022. Treatment outcomes were defined using World Health Organization criteria. Uni- and multivariable logistic regression models were used to assess the association between sex and unsuccessful outcomes (‘treatment failure’, ‘lost to follow-up’, ‘died’, or any of these). After propensity score matching females and males, the multivariable analysis was repeated. Risk factors were analysed separately for each sex and compared using interaction terms. Findings Among females, 19⋅5% (n = 290/1490) (95% confidence interval [CI]: 18, 22) achieved an unsuccessful treatment outcome, compared with 30% (n = 1363/4553) (95% CI: 29, 31) among males. In the multivariable an- alyses, female sex was associated with 32% lower odds of any unsuccessful outcome (adjusted odds ratio [aOR] 0⋅68, *Corresponding author. E-mail address: victor.dahl@rm.dk (V.N. Dahl). t Equal contribution. Translation: For the Georgian, Kazakhstan, Kyrgyzstan, Romanian and Moldavian, Ukrainian translations of the summary see Supplementary Materials section. The Lancet Regional Health - Europe 2025;55: 101354 Published Online 25 June 2025 https://doi.org/10. 1016/j.lanepe.2025. 101354 www.thelancet.com Vol 55 August, 2025 1 Articles http://creativecommons.org/licenses/by-nc-nd/4.0/ Delta:1_given-name Delta:1_surname Delta:1_given-name Delta:1_surname Delta:1_given-name Delta:1_surname Delta:1_given-name Delta:1_surname Delta:1_given-name Delta:1_surname Delta:1_given-name Delta:1_surname Delta:1_given-name Delta:1_surname Delta:1_given-name Delta:1_surname Delta:1_given-name Delta:1_surname Delta:1_given-name Delta:1_surname Delta:1_given-name Delta:1_surname Delta:1_given-name Delta:1_surname Delta:1_given-name Delta:1_surname Delta:1_given-name Delta:1_surname Delta:1_given-name Delta:1_surname Delta:1_given-name Delta:1_surname mailto:victor.dahl@rm.dk http://crossmark.crossref.org/dialog/?doi=10.1016/j.lanepe.2025.101354&domain=pdf https://doi.org/10.1016/j.lanepe.2025.101354 https://doi.org/10.1016/j.lanepe.2025.101354 https://doi.org/10.1016/j.lanepe.2025.101354 http://www.thelancet.com 95% CI: 0⋅58, 0⋅80), 36% lower odds of dying (aOR 0⋅64, 95% CI: 0⋅51, 0⋅80), and 37% lower odds of treatment failure (aOR 0⋅63, 95% CI: 0⋅47, 0⋅85). The association between sex and being ‘lost to follow-up’ was not significant. In the propensity score-matched cohort, sex was not associated with unsuccessful outcomes. Risk factors for unsuccessful outcomes were similar for females and males, except that in females aged >65 years, the odds of death were 2⋅2 times higher (95% CI: 1⋅1, 4⋅4). Interpretation Male sex was associated with unsuccessful outcomes, including death and treatment failure, but adjusting for socio-demographic and clinical factors, and matching males to females, attenuated the association, suggesting that sex disparities in tuberculosis outcomes may be driven more by behavioural than biological factors. Longitudinal studies are needed to confirm these findings. Funding The publication fee was funded by the Civilian Research and Development Foundation (CRDF) under grant #G-202407-72538. Copyright © 2025 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Keywords: Tuberculosis; Treatment outcome; Sex; Healthcare; Socioeconomic factors; Gender and health Introduction Addressing disproportionate representation between sexes is essential for achieving universal health coverage, a key component of the World Health Orga- nization’s (WHO) Sustainable Development Goal 3. 1 This is especially important in tuberculosis epidemi- ology, where considerable sex disparities exist. Of the estimated 10⋅8 million individuals with tuberculosis in 2022, 55% were men, 33% were women, and approximately 12% were children. 2 Recent WHO data from 2021, focussing on 26 tuberculosis high-burden countries and reporting sex-stratified treatment out- comes, found slightly lower treatment success rates among males compared to females (88% vs 90%). 3 It Research in context Evidence before this study A PubMed search on March 26, 2024, using keywords related to tuberculosis, sex, treatment outcomes, and risk factors, yielded 1887 results. After screening of titles and abstracts, 1464 studies were excluded as irrelevant, and 98 were excluded for not reporting relevant association measures or outcomes. Among the remaining 325 studies, 229 included sex as a variable and reported World Health Organization (WHO)-defined outcomes. Of these, 49% (n = 113/229) examined risk factors for the composite outcome of ‘treatment failure’, ‘lost to follow-up’, or ‘died’. Nearly half (46%, n = 52/113) found male sex associated with unsuccessful outcomes, 5% (n = 6/113) found an association with female sex, and 49% (n = 55/113) found no significant association. Fourteen per cent (n = 32/229) of the studies examined sex as a risk factor for being ‘lost to follow-up’; 75% (n = 24/32) found an association with male sex, 3% (1/ 32) with female sex, and 22% (n = 7/32) found no association. Eleven per cent (n = 25/229) examined sex as a risk factor for ‘treatment failure’; with 44% (n = 11/25) finding an association with male sex and 56% (n = 14/25) finding no association with sex. Finally, 34% (n = 78/229) examined sex as a risk factor for ‘died’; 35% (n = 27/78) found male sex to be significantly associated with mortality, 1% (n = 1/78) found female sex to be significantly associated, and 64% (n = 50/78) found no association. Only six studies reported risk factors for unsuccessful outcomes stratified by sex; however, none examined all three unsuccessful outcomes. Added value of this study This study examined risk factors for all three WHO-defined unsuccessful treatment outcomes for males and females combined and stratified by sex. Risk factors were generally similar between the sexes. However, our findings indicate that sex significantly modulates the association between being older than 65 years at diagnosis and the odds of death. Additionally, matching females and males on socio- demographic and clinical characteristics resulted in statistically insignificant associations between sex and unsuccessful outcomes. Implications of all the available evidence Matching females and males weakens the association between sex and unsuccessful outcomes, suggesting that sex disparities in tuberculosis outcomes may be driven more by behavioural than biological factors. Although risk factors are similar between the sexes, being older than 65 years at diagnosis seems to be a stronger predictor of unsuccessful outcomes in females than males. Longitudinal studies are needed to confirm these findings. Articles 2 www.thelancet.com Vol 55 August, 2025 http://creativecommons.org/licenses/by-nc-nd/4.0/ http://www.thelancet.com is well-known that males are much more likely than females to possess risk factors for tuberculosis and to be diagnosed with the disease. 4 Hence, implementing a sex-responsive approach to tuberculosis care is critical to both public health control and clinical management. While numerous studies have examined risk factors for unsuccessful treatment outcomes, including sex, the observed association remains inconsistent. Most studies report that male sex is associated with unsuc- cessful outcomes 5,6 or find no association, 7,8 while some have found that female sex is associated with unsuc- cessful outcomes. 9,10 Additionally, unsuccessful treat- ment outcomes are typically evaluated as a composite measure, grouping ‘failure’, ‘lost to follow-up’ and ‘died’ together. However, these outcomes differ inher- ently, and the same risk factors may not necessarily apply to each. Lastly, risk factor analyses are rarely stratified by sex. In summary, there is limited evidence on the different roles that risk factors play for females and males, as well as how sex may interact with these fac- tors in influencing unsuccessful outcomes. In this study, we aimed to examine the overall association be- tween sex and each unsuccessful treatment outcome (‘failure’, ‘lost to follow-up’, and ‘died’) in tuberculosis patients across six Eastern European countries. Sec- ondly, we aimed to describe sex-stratified risk factors for these outcomes and to assess how sex may modulate their associations. Methods Study design, data collection, and participant eligibility This study was an observational, retrospective cohort study of prospectively collected data from tuberculosis dispensaries in Georgia, Kazakhstan, Kyrgyzstan, Re- public of Moldova, Romania, and Ukraine from January 2020 to December 2022. Data were stored in and extracted from the National Institute of Allergy and Infectious Diseases tuberculosis Portal (NIAID TB Portal), a multinational database containing compre- hensive information on demographics, comorbidities, diagnostics, treatment, and treatment outcomes of bacteriologically confirmed tuberculosis patients. 11–14 The study included patients of all ages with bacterio- logically confirmed tuberculosis and available treatment outcomes. Patients with unreported treatment out- comes were excluded. Only the first enrolment was included in cases where patients were registered more than once. The study was reported following the guidelines provided by Strengthening the Reporting of Observa- tional Studies in Epidemiology (STROBE) (Table 1 in Supplementary Material). 15 Study setting In 2021, the total number of new patients with tuber- culosis ranged from 2100 in Georgia to 42,000 in Ukraine, whereas incidence rates ranged from 55 pa- tients per 100,000 citizens in Romania to 112 in Ukraine. 2 Of the six countries, Kazakhstan, Kyrgyzstan, and Ukraine are included on the WHO’s list of high tuberculosis burden countries due to high prevalences of HIV-associated tuberculosis and multidrug/ rifampicin-resistant tuberculosis. 3 All study sites were included at various dates in early 2020 and 2021, and continued participation throughout the subsequent year. Based on the numbers reported in the 2024 WHO Global TB Report, roughly 156,000 incident tubercu- losis cases were estimated across the six participating countries during the 2020–2021 period. 2 Thus, patients in the NIAID TB Portal represent only a subset of the total TB burden, with inclusion dependent on clini- cians, patient participation, and local capacity. Variables The outcomes of interest were WHO-recommended mutually exclusive treatment outcomes (Table 2 in Supplementary Material). 16 An ‘unsuccessful outcome’ was defined as the composite of ‘treatment failure’, ‘lost to follow-up’, and ‘died’, while a ‘successful outcome’ comprised ‘cure’ and ‘treatment completed’. Impor- tantly, ‘lost to follow-up’ referred to patients whose treatments were interrupted for two consecutive months or more, not those who emigrated or trans- ferred out; such cases were classified as not evaluated and excluded from the analyses. Biological sex was the main exposure variable, categorised as either male or female. Sex was also considered an effect modifier in the associations between other covariables and treat- ment outcomes. The NIAID TB Portal database reports various socio- demographic and clinical covariables. 17 In line with a previous study, covariables utilised in this study included socio-demographic and clinical characteristics that are often associated with unsuccessful treatment outcomes (Table 1). 11 For regression analyses, all cova- riables were included as categorical variables, express- ing the presence or absence of each characteristic. Statistical methods Socio-demographic and clinical characteristics were stratified by sex and presented as medians with inter- quartile ranges for continuous variables and as absolute numbers and proportions for categorical variables. Group comparisons were performed using Wilcoxon rank-sum test, Pearson’s Chi-squared test, or Fisher’s exact test, as applicable. False discovery rate corrections were applied to account for multiple testing. The number of patients with each successful and unsuc- cessful outcome was expressed in absolute and relative Articles www.thelancet.com Vol 55 August, 2025 3 http://www.thelancet.com Characteristics Male (n = 4553, 75%) Female (n = 1490, 25%) p-value a q-value b Age at tuberculosis diagnosis in years, median (IQR) 45 (36, 53) 43 (30, 54) <0⋅001 <0⋅001 Body mass index (kg/m 2 ), median (IQR) 20⋅2 (18⋅1, 22⋅0) 19⋅9 (17⋅6, 22⋅2) 0⋅009 0⋅012 Patient country, n (%) <0⋅001 <0⋅001 Georgia 1046 (23) 401 (27) Kazakhstan 332 (7⋅3) 214 (14) Kyrgyzstan 22 (0⋅5) 16 (1⋅1) Moldova 148 (3⋅3) 36 (2⋅4) Romania 107 (2⋅4) 40 (2⋅7) Ukraine 2898 (64) 783 (53) Educational level, n (%) <0⋅001 <0⋅001 College and higher education 772/3238 (24) 336/1011 (33) No education, basic or complete school 2466/3238 (76) 675/1011 (67) Employment group, n (%) <0⋅001 <0⋅001 Employed, student, or homemaker 774/4495 (17) 342/1477 (23) Unemployed, disabled, or retired 3721/4495 (83) 1135/1477 (77) Social risk factors, n (%) A history of tuberculosis 1383/4550 (30) 299/1416 (20) <0⋅001 <0⋅001 Alcohol misuse 1313/4350 (30) 180/1416 (13) <0⋅001 <0⋅001 Drug abuse 293/4350 (6⋅7) 46/1416 (3⋅2) <0⋅001 <0⋅001 Homelessness 143/4350 (3⋅3) 16/1416 (1⋅1) <0⋅001 <0⋅001 Tobacco use c 2554/4350 (59) 393/1416 (28) <0⋅001 <0⋅001 Previous incarceration 443/4350 (10) 8/1416 (0⋅6) <0⋅001 <0⋅001 Comorbidities, n (%) Anaemia 627/4244 (15) 233/1401 (17) 0⋅093 0⋅11 Chronic hepatitis d 438/4244 (10) 84/1401 (6⋅0) <0⋅001 <0⋅001 Diabetes mellitus 185/4244 (4⋅4) 91/1401 (6⋅5) 0⋅001 0⋅002 Living with HIV 543/4244 (13) 193/1401 (14) 0⋅3 0⋅4 Immunosuppression e 16/4244 (0⋅4) 5/1401 (0⋅4) >0⋅9 >0⋅9 Psychiatric illness f 125/4244 (2⋅9) 19/1401 (1⋅4) 0⋅001 0⋅002 Renal insufficiency 44/4244 (1⋅0) 20/1401 (1⋅4) 0⋅2 0⋅3 Biochemical parameters, n (%) Elevated ESR g 1626/2536 (64) 360/708 (51) <0⋅001 <0⋅001 Low total protein h 393/2804 (14) 124/780 (16) 0⋅2 0⋅2 Disease manifestation Pulmonary tuberculosis 4336 (95) 1374 (92) <0⋅001 <0⋅001 Drug resistance pattern <0⋅001 <0⋅001 Drug susceptible 1421 (31) 540 (36) MDR-tuberculosis 1307 (29) 375 (25) Pre-XDR-tuberculosis 835 (18) 229 (15) XDR-tuberculosis 463 (10) 136 (9⋅1) Other types of resistance i 527 (12) 210 (14) Multiple cavities 583/4281 (14) 139/1368 (10) <0⋅001 0⋅001 Smear microscopy <0⋅001 <0⋅001 Negative 1360/4544 (30) 570/1487 (38) Positive 3184/4544 (70) 917/1487 (62) Abbreviations: IQR, interquartile range; HIV, human immunodeficiency virus; ESR, erythrocyte sedimentation rate; MDR-tuberculosis, multidrug-resistant tuberculosis; pre-XDR-tuberculosis, pre-extensively drug-resistant tuberculosis; XDR-tuberculosis, extensively drug-resistant tuberculosis. a Wilcoxon rank sum test; Pearson’s Chi- squared test; Fisher’s exact test. b False discovery rate correction for multiple testing. c Tobacco use was reported as ever having used tobacco. d Defined as a positive hepatitis B surface antigen test and/or a positive hepatitis C antibody test. e Includes patients treated with cytostatics, TNF-alpha inhibitors, and/or corticosteroids. f Most commonly included schizophrenia, anxiety, depression, and bipolar disorder, as diagnosed by psychiatrists at the study sites using ICD-10 diagnostic criteria. g Elevated erythrocyte sedimentation rates were defined as above 20 mm for females aged 0–50 years, above 30 mm for females aged 50 years or older, above 15 mm for males aged 0–50 years, and above 20 mm for males older than 50 years. h Low protein levels were defined as total serum protein below 62 g/L in adults and below 57 g/L for children and adolescents. i These included isoniazid mono resistance, rifampicin mono resistance, other types of mono resistance, and resistance to more than one non- first-line drug. Table 1: Sociodemographic and clinical characteristics of tuberculosis patients stratified by sex. Articles 4 www.thelancet.com Vol 55 August, 2025 http://www.thelancet.com numbers for the whole cohort, stratified by sex, and compared by calculating percentage point differences. In our primary analysis, we examined the overall association between sex and unsuccessful outcomes, both individually and combined, using univariable and multivariable logistic regression models. The depen- dent variable was a binary outcome, with the composite successful outcome (‘treatment completed’ or ‘cure’) as the reference category and unsuccessful outcomes (‘failure’, ‘lost to follow-up’, ‘death’, or any of these) as the alternative. Covariables used for adjustment were selected a priori based on previously identified pre- dictors of unsuccessful outcomes. 11 These included age, BMI, employment level, alcohol misuse, previous incarceration, living with HIV, low total protein, mul- tiple cavities, history of tuberculosis, and drug resis- tance patterns. Results from univariable and multivariable regression models were presented as crude and adjusted odds ratios (cOR and aOR), respectively. To account for missing data, we applied multiple imputations by chained equations (MICE) with 100 imputations and 10 iterations (R package mice v. 3⋅16⋅0). 18,19 Logistic regression was used to impute cat- egorical variables, and predictive mean matching to impute continuous variables. The set of covariables included in the MICE models closely mirrored those used in the analytical models, with the addition of the outcome variable and a few auxiliary variables to improve the imputation process. A full list of covari- ables used in the imputation models is provided in Table 3 in Supplementary Material. Percentages of missing data and pairwise correlations were calculated for each imputed variable (Figures 2 and 3 in Supplementary Material). Convergence of the MICE algorithm was evaluated by visualising the means and standard deviations of each variable across iterations (Figures 4A–D in Supplementary Material). 20 Estimates from the imputed datasets were pooled using Rubin’s rule and compared to complete-case analyses. 21 We also conducted a propensity score analysis using optimal full matching (R package MatchThem v. 1⋅2⋅1) to estimate the causal effect of sex on treatment outcome, matching males to females based on several socio-demographic and clinical characteristics (Table 1). 22,23 Matching was performed after MICE, following the framework recommended by Pishgar et al. 24 All variables in Table 1 were considered clinically relevant and important to balance. Balance of the matched cohorts was visualised by plotting the stand- ardised mean differences for the matched variables across all imputations, along with those in the un- matched population (Figures 1A–D in Supplementary Material). To guide the selection of a parsimonious set of covariables for the propensity score model, we used p-values to objectively prioritise variables showing the greatest imbalance between sexes. Optimal full matching involved computing subclass-specific weights to estimate a weighted effect of sex on treatment outcome. 23 While this approach retains all individuals in the matched population, the use of weights means that some individuals contribute less to the analysis, which may reduce the effective sample size (Table 4 in Supplementary Material). In our secondary analysis, we examined sex- stratified risk factors for unsuccessful outcomes. As in the primary analysis, covariables used in the multivar- iable models were selected a priori as previously iden- tified significant predictors of unsuccessful outcomes. 11 Models were fitted for males and females separately using the composite successful outcome (‘treatment completed’ or ‘cure’) as the reference category, and each of the unsuccessful outcomes (‘failure’, ‘lost to follow- up’, ‘death’, or any of these) as the alternative category, adjusting for covariables. Regression coefficients for the association between each variable and unsuccessful outcomes were presented separately as aORs for males and females. Finally, interaction terms between sex and each covariable were included in a pooled model to examine whether the strength of associations differed by sex. Interaction estimates >1 indicated a stronger association in females; estimates <1 indicated a stron- ger association in males. Bias and sample size calculations To mitigate potential bias due to sex imbalance and missing data, we used multiple imputation and pro- pensity score matching as described above. As this was a retrospective study using all available data, no formal sample size calculation was performed. Ethics approval Ethical approvals for this study were obtained from the relevant national or institutional ethics committees in all participating countries: Ukraine (Protocol #1, February 5, 2020), Romania (Protocol #1346, February 6, 2015), Georgia (Protocol #796/01–14, March 9, 2015), Kazakhstan (Protocol #5723-1-12247/16-9, December 7, 2017), Moldova (Protocol #2, February 2, 2024), and Kyrgyzstan (Protocol #6, December 21, 2018). Role of the funding source The funder for the publication fee had no role in study design, data collection, data analysis, interpretation, or writing of the report. Results A total of 6134 patients were identified, of whom 91 (1⋅5%) were excluded due to missing outcomes, resulting in a cohort of 6043 patients (Fig. 1a). The overall proportion of unsuccessful outcomes across all six countries was 27% (n = 1653/6043, 95% confidence interval [CI]: 26, 28) (Fig. 1b). For females, the overall Articles www.thelancet.com Vol 55 August, 2025 5 http://www.thelancet.com proportion was 19⋅5% (95% CI: 18, 22), while for males it was 30% (95% CI: 29, 31), with notable country- specific variations (Figure 5 in Supplementary Material). Ukraine contributed the largest number of patients (n = 3681, 61%), while Kyrgyzstan contributed the fewest (n = 38, 0⋅6%) (Table 1). Most patients were male (n = 4553, 75%), and several baseline character- istics differed between female and male patients. In the primary analysis that examined the overall association between sex and unsuccessful outcomes, female sex was associated with 32% lower odds of any unsuccessful outcome, adjusting for covariables (adjusted odds ratio [aOR] 0⋅68, 95% CI: 0⋅58, 0⋅80), 36% lower odds of dying (aOR 0⋅64, 95% CI: 0⋅51, 0⋅80), and 37% lower odds of treatment failure (aOR 0⋅63, 95% CI: 0⋅47, 0⋅85) (Fig. 2). The association between sex and being ‘lost to follow-up’ was not statistically significant in the multivariable model. In the pro- pensity score-matched population, none of the associ- ations remained statistically significant. Regression coefficients were comparable across the adjusted analysis with complete cases, multiple imputations, and multiple imputations with propensity score matching (Tables 5A–D in Supplementary Material). Regardless of sex, the strongest predictor of unsuccessful treat- ment outcomes was extensively drug-resistant tuber- culosis (XDR-tuberculosis) (aOR 2⋅74, 95% CI: 2⋅18, 3⋅46); for death, it was pre-XDR-tuberculosis (aOR 4⋅57, 95% CI: 3⋅34, 6⋅21); for treatment failure, it was alcohol misuse (aOR 2⋅73, 95% CI: 2⋅16, 3.45); and for loss to follow-up, it was XDR-tuberculosis (aOR 4⋅07, 95% CI: 2⋅86, 5⋅79). Several other variables were also significant predictors of unsuccessful treatment outcomes (Tables 5A–D in Supplementary Material). In the secondary analysis that examined sex- stratified risk factors for unsuccessful outcomes, several risk factors were identified for females and males (Figure 6 in Supplementary Material). For fe- males, the strongest risk factors for the composite un- successful outcome were XDR-tuberculosis (aOR 3⋅4, 95% CI: 2⋅0, 5⋅8) and low total protein (aOR 2⋅9, 95% CI: 1⋅9, 4⋅5). For males, the strongest risk factors were Fig. 1: Flowchart of patient inclusion (panel a) and summary of treatment outcomes by sex with percentage point differences (panel b). Articles 6 www.thelancet.com Vol 55 August, 2025 http://www.thelancet.com XDR-tuberculosis (aOR 2⋅6, 95% CI: 2⋅0, 3⋅4) and alcohol misuse (aOR 2⋅4, 95% CI: 2⋅1, 2⋅8). For ‘treat- ment failure’, the strongest risk factors were a history of tuberculosis (aOR 2⋅6, 95% CI: 2⋅0, 3⋅3 and aOR 2⋅5, 95% CI: 1⋅3, 4⋅6) and alcohol misuse (aOR 2⋅6, 95% CI: 2⋅0, 3⋅4 and aOR 3⋅4, 95% CI: 1⋅7, 6⋅9) for both sexes. For ‘lost to follow-up’, the strongest risk factors for both females and males were XDR-tuberculosis (aOR 3⋅5, 95% CI: 2⋅3, 5⋅2 and aOR 6⋅1, 95% CI: 3⋅0, 12⋅7) and pre-XDR-tuberculosis (aOR 3⋅0, 95% CI: 2⋅1, 4⋅3 and aOR 2⋅6, 95% CI: 1⋅2, 5⋅4). For the outcome ‘died’, the strongest risk factors for females were age >65 years at diagnosis (aOR 6.0, 95% CI: 3⋅3, 10⋅9) and pre-XDR- tuberculosis (aOR 7⋅1, 95% CI: 3⋅5, 14⋅5), while for males, the strongest risk factors were XDR-tuberculosis (aOR 4⋅4 95% CI: 2⋅9, 6⋅5) and pre-XDR-tuberculosis (aOR 4⋅2, 95% CI: 3⋅0, 5⋅9). Regression coefficients were similar for complete case and for all cases with multiple imputations (Tables 6A–H in Supplementary Material). The association between higher age and ‘died’ was significantly modulated by sex (Fig. 3). The odds of dying when aged >65 years at diagnosis were 2⋅2 times higher for females compared with males (95% CI: 1⋅1, 4⋅4). For the association between multiple cavities and ‘died’, sex appeared to modulate the effect, with females having 1⋅8 times higher odds of dying compared to males when cavities were present (95% CI: 1⋅0, 3⋅3), although the interaction was not statistically significant. The interaction terms between sex and the remaining analysed risk factors were not statistically significant. The main effects are available in Tables 7A–E in Supplementary Material. Discussion In this retrospective analysis of prospectively collected data from 6043 tuberculosis patients from across Georgia, Kazakhstan, Kyrgyzstan, Moldova, Romania, and Ukraine in 2020–2022, we examined risk factors for unsuccessful treatment outcomes (both combined and individually) stratified by sex. Overall, the proportion of unsuccessful outcomes was significantly lower among female patients (19⋅5%) than male patients (30%). In unadjusted analyses, female sex was associated with lower odds of all types of unsuccessful treatment out- comes. However, after adjusting for socio-demographic and clinical characteristics, the association between sex and the likelihood of being ‘lost to follow-up’ was no longer statistically significant. In the propensity score- matched cohort, female sex was not associated with any unsuccessful outcome. Risk factors for unsuccess- ful outcomes were similar across both sexes, except for patients aged >65 years at diagnosis, where the odds of death were twice as high for female patients compared with male patients. The lower overall proportions of unsuccessful out- comes among female patients in our cohort align with the existing literature that reports worse outcomes for male tuberculosis patients. Two main hypotheses have been proposed to explain these sex differences. The first hypothesis suggests that biological factors, such as sex hormones and genetic differences, may increase male susceptibility to tuberculosis. The second hypothesis attributes the discrepancy to non-biological factors influenced by sex, including smoking, occupational exposure, and health-care-seeking behaviours. 25 Our findings support the second hypothesis, as socio- demographic risk factors, such as low educational level, unemployment, tobacco usage, substance abuse, homelessness, and previous incarceration, were more prevalent among male patients. These patient-related factors could contribute to health disparities and dif- ferences in healthcare-seeking behaviour. Moreover, our multivariable regression analyses, both with and without propensity score matching, revealed that the association between sex and unsuccessful outcomes weakened after adjusting for these potential con- founders. This suggests that the higher rate of unsuc- cessful outcomes among male patients in our cohort is likely due to indirect effects rather than a direct influ- ence of sex on treatment outcomes. A recent mediation Fig. 2: Association between sex and unsuccessful treatment outcomes with male sex as the reference category: Crude, adjusted, and pro- pensity score-matched odds ratios. Abbreviations: OR, odds ratio; CI, confidence interval; PSM, propensity score matching. 1 Based on multivariable logistic regression adjusted for a priori selected variables: age, body mass index, employment group, alcohol misuse, previous incarceration, HIV status, total protein level, having multiple cavities, history of tuberculosis, and tuberculosis resistance pattern. 2 Females patients were matched with male patients using optimal full matching based on variables that differed significantly between males and females: body mass index, tuberculosis resistance pattern, age, diagnosis year, patient country, employment group, educational level, smear positivity, history of tuberculosis, tuberculosis disease manifestation, chronic hepatitis, diabetes mellitus, psychiatric illness, erythrocyte sedimentation rate, smoking status, alcohol misuse, drug abuse, homelessness, previous incarceration. Following propensity score matching, the multivariable regression was adjusted for the same a priori selected variables as mentioned above. Articles www.thelancet.com Vol 55 August, 2025 7 http://www.thelancet.com Articles 8 www.thelancet.com Vol 55 August, 2025 http://www.thelancet.com analysis found that the link between male sex and un- successful outcomes was mediated by behavioural fac- tors, particularly treatment adherence, rather than biological differences. 6 Nevertheless, regarding the use of propensity score matching in our study, it is important to note that the methodology itself could also influence the weaker association observed in our pro- pensity score-matched cohort, as some researchers have argued that combining this method with multiple im- putations may lead to excessively wide confidence in- tervals for the effect estimate. 26 By examining sex differences in risk factors for un- successful outcomes, we found that sex significantly modified the association between older age and mor- tality. Specifically, females over 65 years had signifi- cantly higher odds of dying compared to males. While previous studies on tuberculosis have shown that fe- males typically present with a lower mycobacterial burden and are less likely to experience composite un- successful outcomes, our findings show that certain groups, particularly older females, may require increased attention in diagnostics or treatment. 27 One possible explanation for the elevated mortality risk in older females may be biological differences in the naturally occurring immunosenescence. Interestingly, a recent study in Nature Communications highlighted pronounced sex differences in immune ageing, with differences accelerating after the age of 65. 28 Older fe- males exhibited enhanced adaptive immune activity, including more active B- and T-cell responses, while males demonstrated heightened innate and pro- inflammatory activity alongside reduced adaptive re- sponses. 28 This difference might contribute to the increased mortality in females over 65 years, as their comparatively lower innate activity could impair their ability to contain new infections. Conversely, an enhanced adaptive immunity in females over 65 years could also lead to increased immunopathology in tuberculosis. However, these hypotheses are purely speculative and require further investigation. Finally, it is important to note that although we observed female sex being a statistically significant modifier of the as- sociation between being older than 65 years and having the outcome ‘died’, the significance could be a result of multiple testing. This study has limitations. Despite using multivar- iable regression and propensity score matching, the retrospective design carries an inherent risk of selection bias. The NIAID TB Portal is not fully representative and includes only a subset of the tuberculosis patients in the participating countries. Because patients were prospectively included by treating clinicians, sampling bias is possible. This may be augmented if individuals with poorer outcomes were less likely to seek care, or if one sex was more or less likely to be diagnosed or included in the dataset. While propensity score matching was applied to increase the comparability between female and male patients, this method may have led to excessively wide confidence intervals, potentially diluting any true effects. Moreover, residual confounding is a concern, as unmeasured confounders may still influence the relationship between sex and treatment outcomes. Notably, information on factors that could differentially affect the risk of unsuccessful outcomes between males and females, such as symp- tom duration before healthcare contact or treatment adherence, was not available. Furthermore, while the covariables used to match female and male patients were clinically relevant and important to balance, they were partly selected based on observed imbalances in this specific sample, potentially introducing sample- specific bias. Additionally, although the overall num- ber of patients was high, the sex-stratified risk factor analysis resulted in some cohorts, particularly those with female patients, having relatively few patients with the analysed outcomes, which may affect the validity of regression coefficients. There are also limitations to the application of MICE. For most imputed variables, missing data was below ten per cent. Still, for low total protein, erythrocyte sedimentation rate, education group, and BMI, missing data were close to 50%, potentially affecting the validity of the imputations. Furthermore, correlations were observed between some variables, particularly social risk factors, country, and resistance patterns. Nevertheless, regression co- efficients based on MICE were generally comparable to those based on complete case analysis. Finally, direct biological markers, such as hormonal or genetic profiles, which could have provided more insight into sex-related differences in outcomes, were not avail- able. Hence, our results were limited to socio- demographic and behavioural factors, and we could not explore potential biological mechanisms. Regarding representativeness, only biological sex was recorded in the database, limiting our ability to examine how these factors may influence outcomes and reducing the generalisability of our findings to minority populations. This study also has strengths. The study is multi- national and includes a large sample size of patients with tuberculosis, increasing external validity. Data were collected prospectively and are publicly available, Fig. 3: Interaction effects between sex and risk factors on unfavourable tuberculosis treatment outcomes. aOR > 1 indicates a stronger association in females compared to males; aOR < 1 indicates a stronger association in males. Abbreviations: aOR, adjusted odds ratio; CI, confidence interval; LTFU, lost to follow-up; TB, tuberculosis; BMI, body mass index; HIV, human immunodeficiency virus; MDR-TB, multi- drug-resistant tuberculosis; pre-XDR-TB, pre-extensively drug-resistant tuberculosis; XDR-TB, extensively drug-resistant tuberculosis. 1 Compared to drug-susceptible tuberculosis. Articles www.thelancet.com Vol 55 August, 2025 9 http://www.thelancet.com ensuring transparency. Comprehensive statistical ap- proaches were applied to examine the effect of sex on treatment outcomes. Additionally, we analysed risk factors for all three WHO-defined unsuccessful out- comes, unlike most published risk factor studies, providing a more detailed examination. Our findings suggest that sex disparities in tubercu- losis outcomes may be driven more by behavioural fac- tors rather than biological ones. This highlights the clinical importance of considering patient sex when identifying individuals at higher risk of unsuccessful outcomes. To validate these findings, longitudinal studies, particularly those exploring biological mecha- nisms, are needed. At the policy-making level, our re- sults emphasise the need for tuberculosis programmes to adopt a more individualised, patient-centred approach that considers the unique risk factors and treatment needs of different populations. Addressing these dis- parities can contribute to reducing disproportionate representation between sexes in tuberculosis care. Contributors Ole Skouvig Pedersen: Conceptualisation, Methodology, Formal Anal- ysis, Data Curation, Data Visualisation, Writing - Original Draft, Writing - Review & Editing, Visualisation; Tetiana Butova: Investiga- tion, Writing - Review & Editing; Valerii Miasoiedov: Investigation, Writing - Review & Editing; Yurii Feshchenko: Investigation, Writing - Review & Editing; Mykhailo Kuzhko: Investigation, Writing - Review & Editing; Stefan Niemann: Investigation, Writing - Review & Editing; Alex Rosenthal: Investigation, Writing - Review & Editing; Alina Grinev: Investigation, Writing - Review & Editing; Gabriel Rosenfeld: Investi- gation, Writing - Review & Editing; Michael Drew Hoppes: Investiga- tion, Writing - Review & Editing; Julia Kilmnick: Investigation, Writing - Review & Editing; Valeriu Crudu: Investigation, Writing - Review & Editing; Nelly Ciobanu: Investigation, Writing - Review & Editing; Alexandru Codreanu: Investigation, Writing - Review & Edit- ing; Bekzat Toxanbayeva: Investigation, Writing - Review & Editing; Lyailya Chingissova: Investigation, Writing - Review & Editing; Kater- yna Yurko: Investigation, Writing - Review & Editing; Valerii Kucher- iavchenko: Investigation, Writing - Review & Editing; Vitalii Vekshyn: Investigation, Writing - Review & Editing; Sergo Vashakidze: Investi- gation, Writing - Review & Editing; Natalia Shubladze: Investigation, Writing - Review & Editing; Zaza Avaliani: Investigation, Writing - Review & Editing; Abdullaat Kadyrov: Investigation, Writing - Review & Editing; Gulmira Kalmambetova: Investigation, Writing - Review & Editing; Merbubu Sydykova: Investigation, Writing - Review & Editing; Eugenia Ghita: Investigation, Writing - Review & Editing; Victor Ionel Grecu: Investigation, Writing - Review & Editing; Alina Marinela Miulescu: Investigation, Writing - Review & Editing; Christian Morberg Wejse: Supervision, Writing - Review & Editing; Andreas Fløe: Super- vision, Writing - Review & Editing; Victor Naestholt Dahl: Con- ceptualisation, Methodology, Writing - Original Draft, Writing - Review & Editing, Supervision, Project Administration; Dmytro Butov: Inves- tigation, Resources, Writing - Review & Editing, Funding Acquisition, Project Administration. Data sharing statement Data is publicly available (https://datasharing.tbportals.niaid.nih.gov/ #/home), and all coding is available upon reasonable request. Declaration of interests CMW reports being a co-inventor on a pending patent application related to diagnosing tuberculosis in saliva (PCT/EP2023/080683; TECH-2022-631-459) (outside the scope of this work); serves as chair- person of the Danish Society for Migrant Health (since 2023) and as Chair of ESGITM (ESCMID Study Group for Infections in Migrants and Travellers) (outside of the scope of this work). This had no influ- ence on the design, analysis, interpretation, or reporting of the current study. All other authors declare no conflicts of interest. Acknowledgements We sincerely thank all the patients who shared their personal infor- mation for this research. We also extend our appreciation to the members of the NIAID TB Portals Study Group for their invaluable contributions to project administration and data collection. This study received no specific funding and was made possible through access to the freely available data provided by the NIAID TB Portals Program, for which we are grateful. Finally, we thank the Civilian Research and Development Foundation (CRDF) for supporting the publication fee. Appendix A. Supplementary data Supplementary data related to this article can be found at https://doi. org/10.1016/j.lanepe.2025.101354. References 1 Achrekar A, Akselrod S, Clark H, et al. Delivering health for all: the critical role of gender-responsive health systems. Lancet Glob Health. 2024;12(5):e733–e734. 2 Global Tuberculosis Report 2024. Geneva: World Health Organiza- tion; 2024. Licence: CC BY-NC-SA 3.0 IGO. 3 Global Tuberculosis Report 2023. Geneva: World Health Organiza- tion; 2023. Licence: CC BY-NC-SA 3.0 IGO. 4 Sex matters for tuberculosis control. Lancet Infect Dis. 2002;2 (6):317. 5 Bogale L, Tsegaye T, Abdulkadir M, Akalu TY. Unfavorable treat- ment outcome and its predictors among patients with multidrug- resistance tuberculosis in Southern Ethiopia in 2014 to 2019: a multi-center retrospective follow-up study. Infect Drug Resist. 2021;14:1343–1355. 6 Barathi A, Krishnamoorthy Y, Sinha P, et al. Effect of treatment adherence on the association between sex and unfavourable treat- ment outcomes among tuberculosis patients in Puducherry, India: a mediation analysis. J Public Health (Oxf). 2023;45(2):304–311. 7 Massud A, Khan AH, Syed Sulaiman SA, et al. Unsuccessful treatment outcome and associated risk factors. A prospective study of DR-TB patients from a high burden country, Pakistan. PLoS One. 2023;18(8):e0287966. 8 Ngah VD, Rangoanana M, Fwemba I, et al. Evaluating determinants of treatment outcomes among tuberculosis patients in the mining district of Butha Buthe, Lesotho. IJID Reg. 2023;6:62–67. 9 Matulyte E, Davidaviciene E, Kancauskiene Z, et al. The socio- demographic, clinical characteristics and outcomes of tubercu- losis among HIV infected adults in Lithuania: a thirteen-year analysis. PLoS One. 2023;18(3):e0282046. 10 Simieneh A, Gashaneh S, Dereje R. Nutritional status and treat- ment outcomes of tuberculosis in Mizan Tepi University Teaching Hospital, a five -year retrospective study. PLoS One. 2024;19(2): e0298244. 11 Pedersen OS, Butova T, Kapustnyk V, et al. Treatment outcomes and risk factors for an unsuccessful outcome among patients with highly drug-resistant tuberculosis in Ukraine. Clin Microbiol Infect. 2024;30(3):360–367. 12 Dahl VN, Butova T, Rosenthal A, et al. Drug-resistant tuberculosis, Georgia, Kazakhstan, Kyrgyzstan, Moldova, and Ukraine, 2017- 2022. Emerg Infect Dis. 2024;30(4):831–833. 13 Skouvig Pedersen O, Butova T, Borovok N, et al. Chest X-ray alone is insufficient for predicting drug-resistant pulmonary tubercu- losis. Clin Microbiol Infect. 2024;30(8):1075–1077. 14 Rosenthal A, Gabrielian A, Engle E, et al. The TB Portals: an open-access, web-based platform for global drug-resistant- tuberculosis data sharing and analysis. J Clin Microbiol. 2017;55 (11):3267–3282. 15 von Elm E, Altman DG, Egger M, et al. Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. BMJ. 2007;335(7624):806–808. 16 Definitions and reporting Framework for Tuberculosis – 2013 Revision (updated December 2014 and January 2020). World Health Organization; 2020. Available at: https://iris.who.int/bitstream/handle/10665/79199/ 9789241505345_eng.pdf. Articles 10 www.thelancet.com Vol 55 August, 2025 https://datasharing.tbportals.niaid.nih.gov/#/home https://datasharing.tbportals.niaid.nih.gov/#/home https://doi.org/10.1016/j.lanepe.2025.101354 https://doi.org/10.1016/j.lanepe.2025.101354 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref1 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref1 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref1 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref2 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref2 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref3 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref3 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref4 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref4 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref5 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref5 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref5 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref5 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref5 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref6 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref6 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref6 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref6 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref7 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref7 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref7 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref7 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref8 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref8 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref8 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref9 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref9 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref9 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref9 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref10 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref10 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref10 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref10 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref11 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref11 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref11 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref11 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref12 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref12 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref12 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref13 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref13 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref13 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref14 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref14 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref14 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref14 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref15 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref15 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref15 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref15 https://iris.who.int/bitstream/handle/10665/79199/9789241505345_eng.pdf https://iris.who.int/bitstream/handle/10665/79199/9789241505345_eng.pdf http://www.thelancet.com 17 NIAID TB Portals data dictionary (n.d.). Available at: https://data. tbportals.niaid.nih.gov/dictionary. 18 White IR, Royston P, Wood AM. Multiple imputation using chained equations: issues and guidance for practice. Stat Med. 2011;30(4):377–399. 19 Austin PC, White IR, Lee DS, van Buuren S. Missing data in clinical research: a tutorial on multiple imputation. Can J Cardiol. 2021;37(9):1322–1331. 20 Woods AD, Gerasimova D, Van Dusen B, et al. Best practices for addressing missing data through multiple imputation. Infant Child Dev. 2024;33(1):e2407. 21 Schafer JL, Kang J. Average causal effects from nonrandomized studies: a practical guide and simulated example. Psychol Methods. 2008;13(4):279–313. 22 Austin PC, Stuart EA. Estimating the effect of treatment on binary outcomes using full matching on the propensity score. Stat Methods Med Res. 2017;26(6):2505–2525. 23 Greifer N. Matching methods. MatchIt vignette. R package version 4.5.2. Available at: https://cran.r-project.org/web/packages/ MatchIt/vignettes/matching-methods.html; 2025. 24 Pishgar F, Greifer N, Leyrat C, Stuart E. MatchThem: matching and weighting after multiple imputation. R J. 2020;13(2):292–305. 25 Dabitao D, Bishai WR. Sex and gender differences in tuberculosis pathogenesis and treatment outcomes. Sex and Gender Differ- ences in Infection and Treatments for Infectious Diseases. Curr Top Microbiol Immunol. 2023;441:139–183. 26 Ségalas C, Leyrat C, Carpenter JR, Williamson E. Propensity score matching after multiple imputation when a confounder has missing data. Stat Med. 2023;42(7):1082–1095. 27 Deshmukh S, Sane M, Gaikwad S, et al. Sex differences in TB clinical presentation, drug exposure, and treatment outcomes in India. Chest. 2023;163(4):778–789. 28 Márquez EJ, Chung CH, Marches R, et al. Sexual-dimorphism in human immune system aging. Nat Commun. 2020;11(1):751. Articles www.thelancet.com Vol 55 August, 2025 11 https://data.tbportals.niaid.nih.gov/dictionary https://data.tbportals.niaid.nih.gov/dictionary http://refhub.elsevier.com/S2666-7762(25)00146-2/sref18 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref18 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref18 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref19 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref19 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref19 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref20 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref20 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref20 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref21 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref21 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref21 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref22 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref22 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref22 https://cran.r-project.org/web/packages/MatchIt/vignettes/matching-methods.html https://cran.r-project.org/web/packages/MatchIt/vignettes/matching-methods.html http://refhub.elsevier.com/S2666-7762(25)00146-2/sref24 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref24 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref25 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref25 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref25 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref25 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref26 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref26 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref26 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref27 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref27 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref27 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref28 http://refhub.elsevier.com/S2666-7762(25)00146-2/sref28 http://www.thelancet.com Sex differences in risk factors for unsuccessful tuberculosis treatment outcomes in Eastern Europe from 2020 to 2022: a mul ... Introduction Methods Study design, data collection, and participant eligibility Study setting Variables Statistical methods Bias and sample size calculations Ethics approval Role of the funding source Results Discussion ContributorsOle Skouvig Pedersen: Conceptualisation, Methodology, Formal Analysis, Data Curation, Data Visualisation, Writi ... Data sharing statementData is publicly available (https://datasharing.tbportals.niaid.nih.gov/#/home), and all coding is av ... Declaration of interests Acknowledgements Appendix A. Supplementary data References