The possible effect of pulmonary tuberculosis (TB) on subsequent lung cancer development has been suspected, but the evidence remains inconsistent. The purpose of this study was to perform a nationwide population-based cohort study to investigate the risk of lung cancer after pulmonary TB infection.
This nationwide population-based cohort study was based on data obtained from the Taiwan National Health Insurance Database. In total, 5657 TB patients and 23,984 controls matched for age and sex were recruited for the study from 1997 to 2008.
The incidence rate of lung cancer (269 of 100,000 person-years) was significantly higher in the pulmonary TB patients than that in controls (153 of 100,000 person-years) (incidence rate ratio [IRR], 1.76; 95% confidence interval [CI], 1.33-2.32; P < .001). Compared with the controls, the IRRs of lung cancer in the TB cohort were 1.98 at 2 to 4 years, 1.42 at 5 to 7 years, and 1.59 at 8 to 12 years after TB infections. The multivariate Cox proportional hazards model revealed pulmonary TB infections (hazard ratio [HR], 1.64; 95% CI, 1.24-2.15; P < .001) and chronic obstructive pulmonary disease (HR, 1.09; 95% CI, 1.03-1.14; P = .002) to be independent risk factors for lung cancer.
Tuberculosis (TB) poses a global public health threat and remains 1 of the major causes of death among infectious diseases. It is estimated that ⅓ of the human population harbors TB in its latent form.1, 2 With aging or a deteriorated immune system, the causative microorganism, Mycobacterium tuberculosis, can reactivate and cause severe and prolonged pneumonia, pulmonary scarring, and wasting. Pulmonary TB comprises about 85% of clinical TB cases and exists as a chronic inflammation process that may lead to carcinogenesis of the lung tissue.3 Lung cancer is also a major cause of morbidity and mortality; it comprises approximately 12.4% of all new cancer cases, and accounts for 29% of all cancer deaths.4
The possible relationship between pulmonary TB and subsequent lung cancer development has attracted attention for several decades. However, the evidence is inconsistent, with some reporting a positive association, whereas others have reported an insignificant association.5-11 The temporal relationship between TB and lung cancer is also not clear. TB and lung cancer have a large public health impact, and therefore the association between these 2 diseases deserves detailed investigation.
A meta-analysis performed by Liang et al revealed a positive association between pulmonary TB and subsequent lung cancer.12 However, the cited reports were mostly case-control studies. Recall bias of exposure assessment, selection bias of the control group, and the validity of the reported TB diagnosis are all possible limitations in this type of study. In addition, this approach makes it difficult to confirm a temporal relationship between TB and subsequent lung cancer. Some cohort studies have reported a positive association between TB and lung cancer.10, 11 However, these studies might not be able to yield generalizable results, because they were usually conducted in high-risk populations, such as heavy smokers, asbestos-exposed workers, or populations with a high indoor exposure to coal smoke.
We therefore conducted a nationwide population-based cohort study that included a large number of TB cases to investigate the association between pulmonary TB and subsequent risk of lung cancer as well as the temporal relationship between these 2 diseases.
MATERIALS AND METHODS
The National Health Insurance (NHI) is a mandatory universal health insurance program, offering comprehensive medical care coverage to all Taiwanese residents. Up to 96% of the residents in Taiwan have been enrolled in the NHI program since 1996. The NHI sample files, constructed and managed by the National Health Research Institute, consist of comprehensive utilization and enrollment information for a randomly selected sample of 1000,000 NHI beneficiaries, representing approximately 5% of all enrollees in Taiwan in 2000.13 A multistage stratified systematic sampling design was used. There were no statistically significant differences in sex or age between the sample group and all enrollees. Comprehensive healthcare data include the enrollment files, the claims data, the Catastrophic Illness files, and the registry for drug prescriptions.
All information that allows a specific individual patient to be identified has been encrypted. The confidentiality of the data abides by the data regulations of the Bureau of National Health Insurance. This study has been approved by the National Health Research Institutes.
We conducted a retrospective cohort study from January 1, 1997 to December 31, 2008, based on ambulatory care and inpatient discharge records. We chose all the active pulmonary TB patients during this period as the study cohort. Pulmonary TB was defined by a compatible ICD-9-CM (International Classification of Diseases, ninth revision, clinical modification) code (010-012, 018) or A code (A020, A021), plus the prescriptions of at least 2 antituberculosis medications for >28 days. Those individuals who had lung cancer diagnosed before or within the first year of TB infection were excluded. A total of 5657 pulmonary TB patients were included in this study.
Comorbidities were defined based on the claims data. We included diabetes mellitus (ICD-9-CM code 250 or A code A181), chronic renal failure (ICD-9-CM code 585 or A code A350), autoimmune diseases (ICD-9-CM code 710, 714 or A code A430, A431), and chronic obstructive pulmonary disease (ICD-9-CM code 490-492, 496 or A code A323, A325).
Subjects with no TB records were used as the control group. Four control subjects were selected to match each TB patient by random sampling stratified for age and sex from the database within the same observational period. Similar to the TB patients, those with lung cancer diagnosed before or within the first year of the observation were excluded. A total of 23,984 subjects served as a comparison group.
Incidence of Lung Cancer
Lung cancer was defined by a compatible ICD-9-CM code (162) or A code (A101) from the Registry for Catastrophic Illness Patient Database, which is a separate subpart of the NHI Database. Histological confirmation is needed by the NHI to apply for this Catastrophic Illness Registry. Those who were diagnosed with cancer can apply for a Catastrophic Illness Card in Taiwan. Cardholders are exempted from cost sharing required under the NHI program. The incidence rate of lung cancer was calculated as the number of lung cancer patients divided by the observed person-years.
Lung Cancer Risk Analysis
The 2 cohorts were followed up until the development of lung cancer, death, or the end of the year 2008. Because lung cancer developing in the first year of the index pulmonary TB infection is difficult to differentiate from lung cancer mimicking pulmonary TB, we excluded patients with a diagnosis of lung cancer registered within the first year of the index pulmonary TB infection. Each subject was followed up for a minimum of 1 year and a maximum of 12 years. Incidence rates (per 100,000 person-years), incidence rate ratios (IRRs), and hazard ratios (HRs) of lung cancer were analyzed.
The incidences of lung cancer were compared between pulmonary TB patients and controls by calculating the incidence rates and IRRs. Cumulative incidence analyses were performed by the Kaplan-Meier method, and the differences between the curves were tested with the log-rank test. The Cox proportional hazards model was used to calculate the HRs and 95% confidence interval (CI) to determine whether pulmonary TB is an independent risk factor for lung cancer development. Variables in this model included age, sex, TB infection, diabetes mellitus, chronic renal failure, autoimmune diseases, and chronic obstructive pulmonary disease. The cases of comorbid conditions were taken as time-varying covariates in the Cox model. We used SAS 9.1 (SAS Institute, Cary, NC) to link the data, and Stata 10 (Stata Corporation, College Station, Tex) to perform the statistical analyses.
Between 1997 and 2008, 5657 pulmonary TB patients and 23,984 age- and sex-matched controls were recruited. The mean years of follow-up for pulmonary TB and control cohorts were 5.86 and 6.22, respectively. The individual matching resulted in comparable distributions of cases and controls by age and sex. Other comorbidity demographic data, including diabetes mellitus, chronic renal failure, autoimmune diseases, and chronic obstructive pulmonary disease, are shown in Table 1. Pulmonary TB patients had a higher prevalence of diabetes mellitus, chronic renal failure, and chronic obstructive pulmonary disease.
Table 1. Baseline Characteristics of Pulmonary TB Patients and Age- and Sex-Matched Controls
The TB cohort was associated with a higher risk of lung cancer compared with the controls (1.3% vs 0.8%, P < .001, Table 2). On stratified analysis, in the TB cohort, there were higher risks of lung cancer associated with all age groups and male sex, as well as comorbidity with diabetes mellitus and chronic obstructive pulmonary disease, compared with the controls. Higher IRR trends were also noted for female sex and comorbidity with autoimmune diseases. The observation period was 27,479 person-years for the pulmonary TB cases and 125,105 person-years for the controls. The incidence rate of lung cancer 1 year after pulmonary TB was 269 per 100,000 person-years, which was significantly higher compared with the 153 per 100,000 person-years in the control group. The crude IRR was 1.76 (95% CI, 1.33-2.32; P < .001). The average interval from pulmonary TB infection to diagnosis of lung cancer was 3.95 years (95% CI, 3.31-4.59).
Table 2. Risk of Lung Cancer for Pulmonary TB Patients and Controls
Table 3 showed the temporal relationship of lung cancer incidence among pulmonary TB and controls. Compared with the control group, the IRRs of the TB cohort were 1.98 (95% CI, 1.37-2.83), 1.42 (95% CI, 0.76-2.53), and 1.59 (95% CI, 0.75-3.11), at 2 to 4 years, 5 to 7 years, and 8 to 12 years, respectively. Higher lung cancer incidence was noted in the TB cohort in the entire 12 years of observation.
Table 3. Temporal Association of Lung Cancer Incidences Among Pulmonary TB Patients and Controls
Cumulative Incidences and Relative Risks of Lung Cancer
Kaplan-Meier estimates of the cumulative incidences of lung cancer for TB patients and controls are shown in Figure 1. The cumulative incidence of lung cancer in the TB patients was significantly higher than that in the controls (P < .0001). On Cox multivariate proportional hazards analysis, TB infection (HR, 1.64; 95% CI, 1.24-2.15; P < .001), being 1 year older (HR, 1.05; 95% CI, 1.04-1.06; P < .001), male sex (HR, 2.30; 95% CI, 1.62-3.28; P < .001), and comorbidity with chronic obstructive pulmonary disease (HR, 1.09; 95% CI, 1.03-1.14; P = .002) were independent risk factors of lung cancer.
This is the first nationwide population-based cohort study to investigate the associations between lung cancer and previous pulmonary TB infection. With the relative low incidence of pulmonary TB, we need to have a large sample size to obtain a sufficient statistical power. Most of the previous reports were hospital-based studies with limited numbers of observation and possible selection bias. The population-based data with age- and sex-matched control group permits us a better external validation than the previous studies. In addition, we used the NHI claims data to create a cohort study, which allowed us to observe the temporal relationship between active pulmonary TB infection and lung cancer.
Pulmonary TB in Taiwan is diagnosed by clinical suspicion, followed by chest x-rays, and confirmation by sputum smear and culture. Combination treatment for TB is used, which is in accordance with the international standard. We identified the TB cases based on their ICD-9-CM codes or A codes, plus the prescription of at least 2 antituberculosis medications for >28 days. TB is a notifiable disease in Taiwan by law. The Taiwan Centers for Disease Control mandates that all patients diagnosed with TB be reported. The incidence of notified pulmonary TB cases during 2002 to 2008 in Taiwan was 58 of 100,000 person-years according to the report from the Taiwan Centers for Disease Control.14 We found a pulmonary TB incidence rate of 55 of 100,000 person-years in the general population in our study, which is quite close to the reported TB cases in Centers for Disease Control data. As for lung cancer diagnosis, we used both diagnostic codes and the Catastrophic Illness Registry. Histological confirmation is needed under the NHI program to apply for a Catastrophic Illness Card, and therefore the TB and lung cancer data should be valid.
Reverse causality is possible, as occult lung cancer may cause TB infection. Because almost ⅓ of the population is latently infected with TB, occult lung cancer may provoke reactivation of latent TB by weakening the local immunity.15, 16 If occult lung cancer really exists, it is more likely to be diagnosed during the intensive and mandatory 6 to 9 months of TB treatment. Previous studies have reported an average of 4 to 11 months delay in lung cancer diagnosis among TB patients.17, 18 In addition, the coexistence of TB and lung cancer has been observed for decades, and poses significant challenges for the diagnosis and treatment of both diseases.19-21 When there is occult lung cancer or the coexistence of TB and lung cancer, this would magnify the lung cancer risk during the first year of TB infection. To focus specifically on the risk of lung cancer after pulmonary TB infection, lung cancer diagnosis before or within the first year of TB infection was excluded to lessen the possibility of reverse causality and coexistence. After excluding lung cancer that occurred during the first year of pulmonary TB infection, we were still able to detect an increased IRR for lung cancer in the TB cohort, when reverse causality is not so likely.
We observed elevated rates of lung cancer among previous pulmonary TB cases compared with controls, with the crude IRR of 1.76 (95% CI, 1.33-2.32; P < .001). These results were consistent with many previous studies, which reported a positive association between pulmonary TB and subsequent lung cancer.5-7 After adjusting for possible confounders, the multivariate analysis also revealed pulmonary TB infection to be a significant risk factor for lung cancer (HR, 1.64; 95% CI, 1.24-2.15; P < .001; Table 4).
Table 4. Multivariate Analysis for Prediction of Lung Cancer Development
The risk of lung cancer subsequent to pulmonary TB was significantly higher both for patients younger than 60 years and for those older than 60 years, with IRRs of 3.47 and 1.55, respectively, compared with the controls. This suggests that pulmonary TB plays an important role in the formation of subsequent lung cancer in all age groups, especially in the younger patients.
We also observed a higher lung cancer incidence among males in our study (HR, 2.30; 95% CI, 1.62-3.28; P < .001). Differences in the smoking behavior of men and women were reported in a 2004 cross-sectional survey in Taiwan, with the smoking rate ratio of males/females being 9.5 (45.7% vs 4.8%) and the observation that men smoked significantly more cigarettes per day (18 vs 11).22 Smoking is an important risk factor for lung cancer that could account for the sex discrepancy of the lung cancer incidence in our study.
The association between pulmonary TB and lung cancer has biological plausibility. The respiratory symptoms of TB may persist several months before TB diagnosis, and thereafter treatment typically entails 6 to 9 months of multidrug medication. The TB infection may induce substantial pulmonary inflammation during these extended periods.3 Chronic inflammation orchestrates a tumor-supporting microenvironment that is indispensable to carcinogenesis. The activation of innate immunity and inflammation results in the production of cytokines that can stimulate tumor growth and progression.23 Activated leukocytes that participate in the inflammatory process produce reactive oxygen and nitrogen species, which can bind to DNA and cause genomic alterations.24 Nalbandian et al presented experimental evidence showing that chronic TB infection in the lungs is sufficient to cause the multistep transformation of cells, which can lead to tumor development.25 We observed an increased risk of lung cancer in the TB cohort within the 12 years observation. Further studies should incorporate more experimental and epidemiological research to investigate the mechanisms of lung cancer induced by TB infection.
The results of this research should be viewed in light of its limitations. First, our observations were conducted on a retrospective cohort based on diagnostic codes and prescription history. The TB cases may have been misclassified. The early symptoms of occult lung cancer could have been diagnosed incorrectly as TB before lung cancer diagnosis. To avoid this bias, we excluded the lung cancer cases diagnosed before or within 1 year of TB infection. Second, there may exist ascertainment bias, through which patients with an index diagnosis are more likely to have another disease diagnosed compared with patients without the index diagnosis. This bias could lead to an increase in the diagnosed rate for lung cancer among TB patients compared with the unaffected controls. However, because of the mandatory medical insurance and high accessibility of medical services in Taiwan, the chance of lung cancer being overlooked should be minimal. Third, tobacco smoking is strongly associated with lung cancer, and has increasingly come under scrutiny as a suspected contributor to TB infection. Liang et al have made a meta-analysis of risk estimates of TB to lung cancer.12 They found that TB per se is a risk factor for lung cancer, and smoking can further strengthen the association between TB and lung cancer. Our study contains certain inherent limitations from the use of administrative data in which smoking history was not available. As a result, we are unable to adjust for smoking as a contributing factor in our study. Further research with more detailed smoking history could help advance our knowledge in this respect.
The major strength of our study is its nationwide population-based setting that includes a relative large numbers of active TB cases, and allows a clear observation of the temporal relationship between TB infection and lung cancer. Our findings provide solid epidemiological evidence that pulmonary TB is a risk factor for subsequent lung cancer, and evidence to support that pulmonary TB patients should be monitored more carefully for lung cancer development and targeted with preventive intervention strategies.
CONFLICT OF INTEREST DISCLOSURES
Supported by grant numbers NSC 99-2314-B-254-001 and NSC 98-2314-B-075-029 from the National Science Council, Taiwan; grant number VGH 99C1-107 from the Taipei Veterans General Hospital; and grant number DOH 99-TD-C-111-007 from the Department of Health, Taiwan.