In this large, population-based cohort study conducted in Brazil, we found a significantly elevated risk of mortality among individuals diagnosed and treated for TB compared with TB-free participants with similar socioeconomic characteristics. Over 14 years of follow-up, patients diagnosed with TB had 15,168 more deaths per 100,000 persons compared with TB-free participants. After treatment, we observed a decrease in the RD, but it remained substantial at 8,206 more deaths per 100,000 persons. Our findings reveal that individuals successfully treated for TB still experience excess mortality across multiple organ systems and causes, highlighting the lasting impact of TB on overall health.
Previous studies have also evaluated mortality following TB6,7,8,9,10,11,12. However, our study provides finer control over socioeconomic characteristics, particularly for having detailed information for the TB-free participants. Controlling for socioeconomic variables is extremely important when evaluating the residual burden of TB, as the heightened vulnerability of individuals who develop TB also contributes to the increased risk of mortality compared with the general population, partly due to poverty13. We have also conducted a comprehensive evaluation of mortality risks using a robust methodology, which includes exact matching of multiple characteristics within a competing risks framework for individuals with a TB diagnosis and those who have completed TB treatment. Second, we have presented both absolute and relative measures for risks and the rate ratios by year of follow-up in all comparisons. These measures complement one another, offering a thorough view of the TB burden. Third, we have assessed the risks after confirmed treatment completion, with clinical or microbiological confirmation of treatment success, rather than defining the post-TB period as 12 months following diagnosis, as in previous studies6,7. Relying solely on time after diagnosis can group individuals who abandoned treatment, experienced treatment failure, who are classified as having drug-resistant TB, or who were lost to follow-up with those who completed treatment, likely biasing the results6. In our study, half of the patients completed their treatment at 199 days post-diagnosis, with most finishing treatment before 8 months. Finally, we used natural causes deaths as the main outcome and excluded TB and HIV related deaths. The curated outcome definition and use of TB cases with proof of treatment completion allow us to quantify the residual burden of TB with greater accuracy.
Our findings showed that diagnosed and treated TB cases had a higher risk of mortality across a broad range of causes, including respiratory, cardiovascular, endocrine and cancer. Our results complement previous evidence showing that TB is associated with an increased risk of respiratory mortality, mainly due to direct lung damage caused by TB, increasing the risk of recurrent pneumonia and chronic obstructive pulmonary disease, bronchiectasis and other specific infections, such as aspergillomas3,14. Our findings of sustained increased risk of cardiovascular deaths post-TB diagnosis for more than a decade complements previous epidemiological studies assessing this association on the short-term scale7,15,16.
Additionally, endocrine-related deaths were notably increased, potentially reflecting a bidirectional relationship between diabetes mellitus and TB17,18. Diabetes can predispose individuals to TB through impaired immune responses, whereas chronic inflammation from TB can exacerbate insulin resistance and metabolic dysregulation, thereby increasing the risk of diabetes-related mortality18,19. Shared inflammatory mechanisms and dysregulated immune responses likely underpin the mutual exacerbation observed between these two conditions18,20.
Our findings showed that TB is associated with an increased risk of deaths from cancer, including cancer in the digestive organs. This finding is consistent with a growing body of evidence; notably, a prior meta-analysis of 11 studies showed an elevated risk of cancer in patients with TB for more than 5 years after diagnosis21. Although the mechanisms underlying this association are not fully understood, several hypotheses have been proposed, such as chronic systemic inflammation that can promote carcinogenesis through the promotion of reactive oxygen species and DNA damage22,23,24. The increased risk of deaths from cancer may also be partially explained by a higher prevalence of shared lifestyle risk factors for cancer and TB, such as smoking and alcohol use8,25,26,27. Our work extends this body of evidence by providing detailed, year-by-year risk estimates, highlighting the significant long-term outcomes of TB and supporting the need for continued patient care for TB-related sequelae after successful treatment.
We also observed an increased risk of deaths due to external causes, with similar magnitudes in the diagnosed and treated TB cases. Although we cannot dismiss the possibility that this finding is due to residual confounding, as there is no biological plausibility for this relationship, it may also reflect the social stigma experienced by patients with TB. This stigma can lead to social isolation, limited economic opportunities, and potentially worsening underlying or new-onset mental health issues such as depression or anxiety, which, in turn, could promote riskier behaviors28,29,30. However, there is also the possibility that poor mental health or living in contexts of violence could increase the chances of TB reactivation31 and further lead to increased external causes-related mortality after TB diagnosis or treatment. However, similarities in the RR between the treated and diagnosed cohorts reinforce the plausibility of these increased risks being due to a societal rather than biological phenomenon. In addition, our direct comparison of TB cases to household contacts of the same sex and similar age also showed elevated risk of deaths from external causes in the TB group. This finding indicates that, even under similar socioeconomic conditions, patients with TB experience an increased risk of death from external causes. Although some of this association can be related to residual confounding, it is unlikely to account for all of it. In this context, increased awareness of TB stigma and interventions to address it, such as TB support groups, training healthcare workers to provide nonjudgmental care, and community-wide educational campaigns to dispel myths and misinformation about TB can improve the lives of TB survivors32.
In our study, we observed a slight increase in the risk of death among household TB contacts compared with unexposed individuals, suggesting some degree of residual confounding likely due to heightened social vulnerability in households with a TB case. However, the excess risk of death among patients diagnosed with, and even those treated for, TB far exceeds the risk found among household contacts. This suggests that although there are indeed connections between poverty and TB13the increased mortality after TB treatment cannot be attributed solely to poverty, as indicated in previous studies4.
Our study has several limitations. First, like any observational study, it is susceptible to residual confounding. We also could not account for potential time-varying confounders after the index date, such as loss or decrease in family income. Our attempt to quantify residual confounding involved conducting an additional analysis to evaluate mortality among household contacts compared to people without TB using the same approach as the main analysis. Although our intent was not to estimate the risk of death in contacts itself, contacts share similar socioeconomic and housing conditions to people with TB, and we do not expect a strong causal link between being a TB contact and an increased risk of death. This analysis revealed a slightly elevated risk of death in this group. However, the excess risk of death among patients diagnosed with and even those treated for TB far exceeds the risk found among household contacts.
We also estimated E-values, which quantify how strong unmeasured confounding would need to be to explain away an observed association, to assess the robustness of our findings, the E-value for the RR at 10 years was 4.50 for the diagnosed cohort and 3.91 for the treated cohort, considering the strength of confounders in previous studies, the only confounder with effect measures associations greater than 4.00 was age, which was accounted in our study, indicating that the TB-mortality association found in our study is unlikely to be fully explained by unmeasured confounders6,8,33.
Additionally, our analysis is restricted to the poorer half of Brazil. Although this may limit generalizability to wealthier populations, it likely enhances control for confounding factors due to socioeconomic characteristics by studying a more homogenous group and focusing on a highly vulnerable population where the impact of TB is often most severe. Nevertheless, even in this large cohort, we were unable to match all participants, with the unmatched group exhibiting worse socioeconomic conditions than the matched sample. Our decision to match increases internal validity, providing adjustment by design without making any assumptions about the exposure or outcome model. However, even with a higher percentage of matched cases (>88%), matching may lead to an underestimation of TB effects, considering that the unmatched group had poorer socioeconomic conditions, and likely a higher baseline mortality risk. The presence of unmatched cases also changes the estimate of the average exposure effect in the exposed to average exposure effect in the matched sample34.
Second, our results may be subject to reverse causality, particularly for cancer and endocrine-related mortality. Given that active malignancies and diabetes are known risk factors for TB and that we were unable to adjust for pre-exposure comorbidity, it is possible that an undiagnosed underlying condition precipitated the onset of active TB in some individuals14,19. In such cases, TB would be a marker of underlying vulnerability or an exacerbating factor, rather than the primary cause of development and death from cancer or endocrine diseases. Although our long-term follow-up and analysis of the treated cohort mitigate this concern for deaths occurring many years after diagnosis, the possibility cannot be fully excluded, especially for mortality observed in the early follow-up period. Third, we only had comorbidity data on HIV and DM for the TB group, which were assessed at the time of TB diagnosis via the SINAN system. This information was absent for the unexposed participants in the CadÚnico database. Consequently, our subgroup analyses estimate the joint effect of TB plus comorbidities, rather than the effect of TB alone. This prevents us from disentangling the independent effect of TB from the effect of these pre-existing comorbidities. Fourth, only 15% of the TB cases had sputum culture positive, which is considered the gold standard for the diagnosis of TB. This proportion reflects the Brazilian guidelines, which prioritizes sputum culture for suspected TB cases that present a negative bacilloscopy, and specific cases such as suspected cases of resistance, retreatment cases, in prison populations35. In addition, due to the required infrastructure and time for diagnosis, the use of sputum culture as primary diagnosis for TB remains limited in low-and-middle income countries36,37. The inclusion of nonconfirmed TB cases in the TB group likely underestimates the risk in this group.
Fifth, we lack detailed information (type of resistance) about the TB cases that were identified as drug-resistant (n = 945; 0.5%) after completing the treatment scheme, considering the very low numbers it is unlikely to significant bias the analysis of diagnosed cases, and it has no effect in the analysis of treated cases, as these TB cases were not included. Sixth, we lack information about latent TB infection (LTBI), which can result in some individuals with LTBI being classified as TB-free unexposed participants, biasing our estimates downwards as previous studies have shown that LTBI slightly increases the risk of death compared to healthy individuals38. This also represents that our results reflect excess mortality following a TB disease episode, not the isolated biological effect of TB. Seventh, after 11 years of follow-up, we had a relatively small number of participants at risk, resulting in a limited number of events and imprecise estimates by risk period. This issue also occurred in the subgroup analysis, requiring caution when interpreting the point estimates from those analyses. The subgroup and cause-specific analyses were not adjusted for multiplicity adjustment, as specified in the analysis plan, and should not be used in place of hypothesis testing. Eighth, we relied on International Classification of Diseases, 10th edition (ICD-10th) codes from the mortality information system to classify the underlying cause of death, whereas there is no comprehensive study evaluating the quality of this codification in Brazil. The quality of the death certificate information assessed by the percentage of garbage codes has improved nationwide, with a reduction of up to 90% of garbage codes between 2000 and 2015 (ref. 39). Lastly, we were unable to censor individuals when they emigrated from Brazil. However, emigration rates for Brazil are low, estimated at 0.8% in 2019 (ref. 40), making it unlikely to substantially bias our estimates.
One fundamental limitation about the management of TB worldwide is the sole focus on diagnosing and curing active disease, overlooking long-term health consequences. For decades, WHO guidelines have appropriately emphasized the diagnosis and bacteriological cure of active disease, considering this a complete return to health, without considering possible post-TB complications. Our findings strongly support the need for long-term clinical follow-up into routine TB care. Integrating post-TB assessments, such as lung function testing, cardiovascular risk screening, and cancer surveillance, into national guidelines for post-TB management is essential. Such measures will increase clinician awareness of post-TB complications, ensure timely management, and direct resources towards truly comprehensive, patient-centered care.
