Clinical Characteristics of People With Extrapulmonary and Disseminated Tuberculosis: A 23-Year Retrospective Cohort Study in People Without HIV Without Pharmacological Immunosuppression Authors: Brenda Aceves-Sánchez, Sandra Rajme-López, Bernardo A Martínez-Guerra, Hector O Rivera-Villegas, Carla M Román-Montes, Karla M Tamez-Torres, Luz E González-Vázquez, Sebastián Guadarrama-Torres, Oswaldo Lazcano-Delgadillo, Rafael Nares-López, Zurisadai Segura-Ortíz, Karen M Tepo-Ponce, María F González-Lara, Bárbara Chávez-Mazari, Miriam Bobadilla-del-Valle, José Sifuentes-Osornio, Alfredo Ponce-de-León Published: 2026 ABSTRACT AbstractIn this retrospective cohort of non-HIV patients without pharmacological immunosuppression, who presented with extrapulmonary/disseminated tuberculosis, the gastrointestinal tract, lymph nodes, and genitourinary tract were the most commonly affected organs. 24-month all-cause mortality was 18%. Cardiovascular disease, central nervous system involvement, mycobacteremia, and ICU admission were independently associated with death. INTRODUCTION In 2024, there were an estimated 10.7 million new tuberculosis (TB) cases and 1.2 million deaths worldwide [1]. Although pulmonary disease remains the most common clinical presentation, extrapulmonary tuberculosis (EPTB) and disseminated tuberculosis (DTB) represent clinically severe forms of the disease, accounting for approximately 15%–20% and 2% of all TB cases, respectively [2]. The proportion of EPTB varies considerably across geographic regions and patient populations, reflecting differences in host characteristics, human immunodefficiency virus (HIV) prevalence, and access to diagnostic resources [1, 2]. In Latin America and the Caribbean, the incidence rate of all-form TB changed by 0.6% between 2015 and 2023 [3]. Regional data on EPTB remain limited. In Brazil, among 427 548 reported TB cases between 2007 and 2011, 13.4% were classified as EPTB, with HIV infection identified as an important associated factor [4]. Mexico is classified as a moderate-incidence country for TB and remains the country with the highest TB burden in Central Latin America [5]. Tuberculosis transmission in Mexico is heterogeneous across geographic and socioeconomic settings and is influenced by factors including diabetes, HIV, migration, and urban overcrowding [6]. According to the Mexican National Center for the Prevention and Control of Diseases, in 2021, 16.5% of all TB cases were extrapulmonary, and 4.7% were disseminated or miliary disease [7]. However, robust national estimates of the burden of EPTB and DTB remain limited. These data underscore the relevance of nonpulmonary TB in countries such as Mexico and highlight the need for improved characterization of EPTB and DTB in this setting. Extrapulmonary tuberculosis and DTB occur most commonly in immunocompromised patients [8]. However, these forms of TB can also occur in people without recognized immunosuppression. A recent study in China found that 48% of TB cases in a cohort of people who are HIV negative were classified as EPTB or DTB [9]. Despite this, clinical suspicion for EPTB and DTB may remain low among people who are HIV negative without pharmacological immunosuppression, potentially contributing to delays in diagnosis and management [10, 11]. The literature on EPTB and DTB in people who are HIV negative without pharmacological immunosuppression consists mainly of case reports, small case series, and a limited number of larger cohorts [9, 12, 13]. Evidence from Latin America remains particularly scarce, despite the high burden of TB in the region. Therefore, there is a need for larger regional cohorts to better characterize the clinical features and outcomes of EPTB and DTB in these countries. Describing the clinical characteristics and identifying factors associated with mortality may assist clinicians in similar settings in improving early recognition and risk stratification of patients at risk of adverse outcomes. We aimed to describe the clinical characteristics and outcomes of immunocompetent people with EPTB and DTB treated at a referral center from Mexico City. METHODS Study Design and Population This retrospective cohort study was conducted at Instituto Nacional de Ciencias Médicas y Nutrición Salvador Zubirán, a tertiary referral center for adults with complex diseases located in Mexico City. Our institution provides care to patients referred from Mexico City and other regions of Mexico, including those with suspected or confirmed TB. Using the electronic Clinical Microbiology Laboratory database, we identified all clinical samples with a positive culture or GeneXpert MTB/RIF result for Mycobacterium tuberculosis complex, from January 2000 to December 2023. The corresponding clinical records for all identified patients were subsequently reviewed to determine eligibility for inclusion and to collect clinical and outcome data. We included people with EPTB and DTB who received treatment at our center. We excluded people with isolated pulmonary TB, those with incomplete medical records, and those referred for treatment to other centers. We also excluded people with immunocompromise, defined as having at least one of the following conditions: (1) HIV infection with <200 CD4+ cells/mL, (2) solid organ or bone marrow transplant on systemic immunosuppressants, (3) active malignant neoplasm receiving systemic chemotherapy in the past 6 months, (4) use of systemic steroids (prednisone > 15 mg/day or its equivalent for 2 or more consecutive weeks), or (5) use of other systemic immunosuppressive drugs in the last 6 months (eg, anti-TNF-alfa, anti-metabolites, and anti-CD20 monoclonal antibodies) [14]. People with autoimmune or malignant diseases and solid organ transplant recipients were included only if remission and discontinuation of immunosuppressive therapy for at least 12 months before the diagnosis of TB were documented in the medical records. Information on immunosuppressive treatment was obtained through review of clinical notes in the medical records. When no immunosuppressive therapy was documented in the available medical records, patients were considered not to be receiving such treatment. Definitions Extrapulmonary tuberculosis was defined as a positive Xpert MTB/RIF assay or culture for M. tuberculosis or Mycobacterium bovis from any extrapulmonary site. Disseminated tuberculosis was defined as one of the following: (1) involvement of 2 or more noncontiguous organs; (2) pathogen isolation in blood, bone marrow, or bone; or (3) miliary pattern on chest imaging together with a positive culture from any site other than blood or bone marrow [14]. This criterion combines the characteristic radiographic pattern of miliary TB with microbiological confirmation of M. tuberculosis. Aims and Outcomes We aimed to describe and compare the clinical characteristics of EPTB and DTB in people who are HIV negative without pharmacological immunosuppression. The primary outcome was all-cause 24-month mortality after TB diagnosis. Microbiology Procedures All samples, except cerebrospinal fluid (CSF), processed for mycobacterial cultures underwent initial digestion and decontamination using the N-acetyl-l-cysteine–sodium hydroxide method and were subsequently inoculated into Lowenstein–Jensen, Stonebrink media, and mycobacteria growth indicator tube (MGIT) (Becton-Dickinson, Sparks, MA, USA). The method used for culture confirmation and species identification varied over the study period. From January 2000 to 2018, DNA probe testing (AccuProbe®, San Diego, CA) was used for confirmation. From January 2019 to November 2021, identification was performed using the GenoType® Mycobacterium common mycobacteria/additional species (CM/AS) assay (Hain Lifescience GmbH, Germany). Since November 2021, species complex identification has been performed using matrix-assisted laser desorption/ionization-time of flight (MALDI-TOF) mass spectrometry (BD™ Bruker Maldi Biotyper™, Daltonics, Bremen, Germany). Since the implementation of MALDI-TOF, pyrazinamidase deamidation has been used for identification of M. bovis. Drug susceptibility testing for isoniazid, rifampin, streptomycin, and ethambutol was performed using radiometric methods before 2005 and the BD BACTEC™ MGIT™ 960 system (Becton-Dickinson, Sparks, MA) [15]. Statistical Analysis Baseline characteristics were summarized as percentages and medians with interquartile ranges (IQRs). Comparisons between people with EPTB and DTB were performed using Fisher's exact test or the Wilcoxon rank-sum test, as appropriate. Univariate and multivariate logistic regression analyses were performed to identify factors associated with 24-month mortality; variables with P < .20 in the univariate analysis (to avoid excluding potential confounders) or a false discovery rate of <0.10 (to control for multiple testing) were included in the multivariate model. Missing data were <5% for all predictor variables; the outcome had 10% missingness due to loss to follow-up, with no significant differences in baseline characteristics between patients with and without outcome data, supporting a missing-at-random assumption, and thus handled using complete-case analysis. Collinearity was assessed using variance inflation factor (VIF), and no variables were excluded because the values ranged from 1.01 to 1.12. A 2-sided P ≤ .05 was considered statistically significant. Analyses were performed using R version 4.3.1 and RStudio 2023.03.2. RESULTS Our search found 5610 positive cultures in 3153 patients. Of these, 2903 were excluded due to samples referred from other centers without available clinical information (2350), isolated pulmonary TB (241), immunocompromised status (154), referrals to other centers for treatment (105), and incomplete records (53) (Supplementary Figure 1). We analyzed 250 patients. Most patients were female (53% [133/250]), and the median age was 50 years (IQR 33–63). Most patients had at least 1 comorbidity (68% [171/250]), with chronic hypertension (26% [63/250]), autoimmune diseases (20% [50/250]), and diabetes (17% [43/250]) as the most frequent. We identified 119 (48%) cases of EPTB and 131 (52%) cases of DTB. The complete baseline characteristics are shown in Table 1. Table 1.Baseline CharacteristicsVariableAlln = 250Extrapulmonaryn = 119Disseminatedn = 131Age, median (IQR)50 (33–63)52 (33–63)49 (32–64)Sex female, n (%)133 (53)74 (62)59 (45)Unemployed, n (%)139 (56)65 (55)74 (56)Alcohol use, n (%)113 (46)52 (44)61 (47)Tobacco use, n (%)92 (37)44 (37)48 (37)Drug use, n (%)5 (2)1 (1)4 (3)Any comorbidity, n (%)171 (68)78 (66)93 (71)Chronic hypertension, n (%)63 (26)39 (33)28 (21)Autoimmune diseasea, n (%)50 (20)20 (17)30 (23) Thyroid disease1569 Rheumatoid arthritis1569 Systemic lupus erythematosus1046 Other1477Diabetes mellitus, n (%)43 (17)18 (15)25 (19) HbA1C, median (IQR)7.2 (6.8–8.3)7.0 (6.6–7.4)7.5 (7.0–8.4)Cardiovascular disease, n (%)23 (9)11 (9)12 (9)Malignancya, n (%)21 (8)10 (8)11 (8) Solid organ1688 Hematologic523Chronic liver disease, n (%)17 (7)7 (6)10 (8) Child A422 Child B220 Child C1138End-stage renal disease, n (%)14 (6)9 (8)5 (4) Hemodialysis844Chronic lung disease, n (%)4 (2)3 (3)1 (1)Chronic neurologic disease6 (2)3 (3)1 (1)Transplant recipienta, n (%)2 (1)1 (1)1 (1) Kidney110 Liver101Microbiological diagnosis, n (%)250 (100)119 (100)131 (100) M. tuberculosis1708585 M. bovis803446Hospitalization, n (%)177 (71)63 (53)114 (87) ICU admission29 (12)9 (8)20 (15)Time to treatment in days, median (IQR)8 (2–23)13 (4–30)5 (2–15)Death at 24 m, n (%) n = 22540 (18)11 (10)29 (25)Abbreviations: Hba1c, glycated hemoglobin, ICU, intensive care unit, IQR, interquartile range. aPeople with these conditions were considered not to be receiving immunosuppressive drugs if no immunosuppressive treatment was documented in the medical record during the 12 m preceding the diagnosis of tuberculosis, based on review of medical records. Baseline Characteristics Abbreviations: Hba1c, glycated hemoglobin, ICU, intensive care unit, IQR, interquartile range. aPeople with these conditions were considered not to be receiving immunosuppressive drugs if no immunosuppressive treatment was documented in the medical record during the 12 m preceding the diagnosis of tuberculosis, based on review of medical records. Microbiological diagnosis was made by culture alone in 74% (186/250) of patients, by GeneXpert alone in 1% (3/250), and by both methods in 24% (61/250). Mycobacterium tuberculosis was the most frequently identified species (68% [170/250]), followed by M. bovis (32% [80/250]). Species identification was established using the microbiological identification methods described above, according to the period in which each sample was processed. The organs and systems affected, in descending order, were the gastrointestinal tract (34% [86/250]), lymph nodes (34% [84/250]), genitourinary system (21% [53/250]), central nervous system (CNS) (12% [30/250]), skin and soft tissues (10% [26/250]), osteoarticular system (8% [21/250]), and pleura (3% [7/250]). Concomitant pulmonary TB was seen in 40% of cases (102/250). Mycobacteremia was identified in 4% of cases (9/250). Resistance to isoniazid was detected in 5% (12/250) of cases; no cases of multidrug resistance were observed. Twenty-four-month mortality was assessed among the 225 patients (90%) with available follow-up data. The remaining 25 patients (10%) were lost to follow-up and excluded from the mortality analysis because their vital status at 24 months could not be ascertained. Forty deaths occurred, corresponding to a 24-month mortality of 18%. People with DTB had significantly higher mortality (25% vs 10%, P = .007) but significantly shorter time-to-treatment initiation (median 5 vs 13 days, P = .002). Multivariate analysis identified cardiovascular disease (OR 4.30 [1.37–13.11], P = .01), cirrhosis (OR 5.07 [1.37–18.06], P = .01), CNS involvement (OR 3.33 [1.13–9.58], P = .03), mycobacteremia (OR 14.14 [2.80–82.56], P = .002), and intensive care unit (ICU) admission (OR 7.68 [2.92–20.83], P < .001) as independent risk factors for 24-month all-cause mortality (Table 2). The multivariate logistic regression model demonstrated excellent discriminative ability, with an area under the receiver operating characteristic curve (AUC) of 0.84. Assessment of multicollinearity among predictor variables revealed no concerning correlations, with all VIFs ranging from 1.01 to 1.12. The Akaike information criterion for the final model was 167.63. Table 2.Univariate and Multivariate Analysis of Risk Factors Associated With Death 24 Months After DiagnosisUnivariateMultivariateaOR95% CI P valueOR95% CI P valueAge ≥ 50 y1.150.99–1.001.00………Sex female1.280.64–2.60.50………Unemployment2.361.14–5.21.032.140.90–5.40.09Alcohol use0.890.44–1.79.75………Tobacco use1.180.57–2.39.65………Cardiovascular disease3.601.32–9.44.014.301.37–13.11.01Cirrhosis3.451.10–10.22.035.071.37–18.06.01Chronic hypertension0.700.30–1.53.40………Malignancy0.520.08–1.92.40………Diabetes mellitus1.400.55–3.24.45………End-stage renal disease0.450.02–2.45.45………Autoimmune disease0.790.30–1.84.61………Mycobacterial species M. tuberculosisRef…………… M. bovis1.070.50–2.19.86………Resistance to isoniazid1.340.19–5.80.72………CNS involvement3.291.34–7.80.0073.331.13–9.58.03Mycobacteremia8.672.03–43.83.00414.142.80–82.56.002ICU admission9.614.10–23.25<.0017.682.92–20.83<.001Disease presentation ExtrapulmonaryRef………… Disseminated2.911.41–6.40.0052.060.86–5.20.112 aVariables with P < .20 (to avoid excluding potential confounders) in the univariate analysis or a false discovery rate of <0.10 (to control for multiple testing) were included in the multivariate model. Missing data were handled using complete-case analysis; collinearity was assessed using variance inflation factor (no variables were excluded as values ranged from 1.01 to 1.12). Multivariate model included 224 patients, 39 events, and 7 predictors (events per variable = 6). Univariate and Multivariate Analysis of Risk Factors Associated With Death 24 Months After Diagnosis aVariables with P < .20 (to avoid excluding potential confounders) in the univariate analysis or a false discovery rate of <0.10 (to control for multiple testing) were included in the multivariate model. Missing data were handled using complete-case analysis; collinearity was assessed using variance inflation factor (no variables were excluded as values ranged from 1.01 to 1.12). Multivariate model included 224 patients, 39 events, and 7 predictors (events per variable = 6). DISCUSSION This study highlights meaningful characteristics of EPTB and DTB in people who are HIV negative without pharmacological immunosuppression, providing valuable information to support clinical suspicion and reduce diagnostic delays associated with their often-mimicking presentations. Disseminated tuberculosis accounted for 52% of cases in our study, a strikingly high proportion. This may be attributed to referral patterns at a tertiary care center, where people with advanced or atypical disease are overrepresented, as well as to the availability of diagnostic methods that are not accessible in most healthcare facilities nationwide. These factors may also contribute to the shorter time-to-treatment observed among people with DTB, although this finding may partly reflect greater disease severity and the need for prompt empiric treatment rather than more rapid diagnostic confirmation alone. The median age was 50 years, consistent with national reports indicating that 77% of TB cases occur among individuals aged 20–64 years in Mexico [7]. This contrasts with a Mexican study evaluating factors associated with EPTB in which younger age emerged as a significant predictor of extrapulmonary disease [16]. Notably, HIV infection was a common comorbidity in that study. Mycobacterium tuberculosis was the most frequently isolated species in our study, consistent with findings from another Mexican cohort [17]. The substantial proportion of M. bovis observed in these and other studies from Latin America and Africa [18, 19] highlights its persistence as a relevant etiologic agent in low- and middle-income countries (LMICs). The presence of underlying comorbidities has been reported as a predictor of poor outcomes in DTB. Factors most consistently associated with mortality include cirrhosis, CNS involvement, and advanced age [20]. In our study, the comorbidities associated with increased mortality were cirrhosis and cardiovascular disease. Most people with cirrhosis were classified as Child–Pugh C, suggesting advanced liver disease and potential functional immunosuppression, as well as potentially reduced treatment tolerance; however, these factors were not specifically assessed. The association between cardiovascular disease and increased mortality should be interpreted with caution, as residual confounding from unmeasured markers of clinical severity, comorbid conditions, and treatment delays cannot be excluded. Interestingly, we did not find an association between diabetes and mortality. However, median glycated hemoglobin (HbA1c) levels were within recommended targets. Regarding organ involvement and disease extent, CNS was the only site associated with higher mortality, an observation reported by other authors, including in different geographical settings, such as Ghana [9, 11, 21, 22]. Consistent with previous reports, mycobacteremia has been associated with increased mortality [23]. However, it is more frequently described in people with HIV, and reports in people who are immunocompetent remain scarce [23]. In our study, mycobacteremia was uncommon, and the corresponding estimates had wide confidence intervals, suggesting limited precision. Therefore, this association should be interpreted with caution given the small number of cases and potential instability of the estimate. Limitations of this study include its retrospective single-center design conducted in a tertiary referral hospital, which may introduce bias by overrepresenting people with more complex comorbidities and severe disease, potentially limiting generalizability to other healthcare settings. Additionally, mortality at 24 months could only be assessed in 90% of the cohort, resulting in potential attrition bias; patients lost to follow-up may have differed systematically in disease severity and outcomes compared with those included in the final analysis, which could have influenced the observed mortality estimates. The long study period (2000–2023) encompasses substantial changes in diagnostic techniques, microbiological identification methods, and clinical management, which may have affected time to diagnosis and treatment initiation. However, all included patients fulfilled predefined microbiological case definitions throughout the study period, likely minimizing the impact of these advances on case classification, although diagnostic timing may have improved over time. Finally, the multivariate model was derived and evaluated in a single-center cohort without internal validation; thus, the reported AUC reflects apparent performance and may be optimistic, and the results should be considered exploratory until externally validated. Strengths include that only microbiologically confirmed cases were analyzed and that this study provides a large cohort of HIV negative without pharmacological immunosuppression with EPTB and DTB from Latin America. Despite these limitations, our study provides valuable insights into a large, microbiologically confirmed cohort of EPTB and DTB in people who are HIV negative without pharmacological immunosuppression, a population that remains underrepresented in the literature. CONCLUSION Although the 24-month mortality of DTB/EPTB was 18%, it was significantly higher in the DTB group than in the EPTB group, aligning with other reports and possibly due to disease severity. Risk factors associated with mortality were cardiovascular disease, cirrhosis, mycobacteremia, ICU admission, and CNS involvement, which highlight the importance of early acknowledgment of severe disease in these patients. Supplementary Material