Tuberculosis and subsequent risk of lung cancer in Xuanwei, China

Abstract

Tobacco and indoor air pollution from smoky coal are major causes of lung cancer in rural Xuanwei County, China. Tuberculosis has been suggested to increase lung cancer risk, but data from prior studies are limited. We conducted an analysis of data from a retrospective cohort study of 42,422 farmers in Xuanwei. In 1992, interviewers administered a standardized questionnaire that included lifetime medical history, including tuberculosis. Subjects were followed from 1976, with deaths from lung cancer ascertained through 1996. We used proportional hazards regression to assess the association between tuberculosis and subsequent lung cancer mortality. Tuberculosis was reported by 246 subjects (0.6%), and 2459 (5.8%) died from lung cancer during follow-up. Lung cancer mortality was substantially higher in subjects with tuberculosis than in those without (25 vs. 3.1 per 1000 person-years). The association was especially pronounced in the first five years after tuberculosis diagnosis (hazard ratios [HRs] ranging 6.7–13) but remained strong 5–9.9 years (HR 3.4, 95%CI 1.3–9.1) and 10+ years (HR 3.0, 95%CI 1.3–7.3) after tuberculosis. These associations were similar among men and women, and among smoky coal users (70.5% of subjects). Adjustment for demographic characteristics, lung disease, and tobacco use did not affect results. In Xuanwei, China, tuberculosis is an important risk factor for lung cancer. The increased lung cancer risk, persisting years after a tuberculosis diagnosis, could reflect the effects of chronic pulmonary inflammation and scarring arising from tuberculosis.

Introduction

Lung cancer is the most common cancer worldwide and results in 1.2 million deaths each year (1). Tobacco use is responsible for the majority of lung cancer cases, and increasing consumption of cigarettes in the developing world, including in China, has led to a dramatic rise in lung cancer incidence (1,2). Exposure to smoke from poorly vented indoor stoves also plays an important role in causing lung cancer in China and other developing countries (3), particularly among women.

Tuberculosis is another major cause of morbidity and mortality, especially in developing countries (4). Worldwide, approximately one-third of people are infected with Mycobacterium tuberculosis, the causative micro-organism, and with aging or weakening of the immune system, this infection can reactivate, leading to severe and prolonged pneumonia, pulmonary scarring, and wasting. China has an extremely high burden of disease due to tuberculosis, with an estimated 1.3 million incident tuberculosis cases and 205,000 deaths annually (4), with many of these cases occurring in rural areas.

It has been proposed that tuberculosis may increase the risk of lung cancer, due to lung inflammation and fibrosis that could induce genetic damage (5–7). Along these lines, a number of retrospective case-control studies of lung cancer have reported an elevated prevalence of tuberculosis among lung cancer cases compared with controls (8–11). To our knowledge, no prior cohort study has provided an examination of the association between tuberculosis and lung cancer. Such an inquiry would yield valuable data on the absolute risk of lung cancer, as well as the time course relating tuberculosis to the subsequent occurrence of this malignancy.

The rural county of Xuanwei, in Yunnan Province, has had among the highest lung cancer mortality rates in China (12). As shown previously in a large retrospective cohort study of farmers in Xuanwei (3), these high rates are in large part attributable to the high indoor exposure to coal smoke from stoves used for cooking and heating. In the present report, we utilize data from this cohort study to examine in detail the association between tuberculosis and subsequent risk of lung cancer.

Materials and Methods

Study subjects and ascertainment of exposures and outcomes

In 1992, investigators searched local administrative records to identify all farmers who were born during 1917–1951 and who had been living in one of Xuanwei’s four communes (Rong Cheng, Lai Bin, Jing Wan, and Re Shui) as of January 1, 1976. In three communes, most residents used smoky coal, while most used smokeless coal in the other commune (Re Shui). A total of 44,580 potential subjects were identified. Of this total, 2,108 were excluded because they had moved out of the study area subsequent to 1976, and 50 were excluded because their identity could not be verified. The study cohort comprised the remaining 42,422 farmers. The study was approved by the Institutional Review Board of the Chinese Academy of Preventive Medicine. Subjects provided written consent or, if illiterate, oral consent witnessed by a literate relative.

During 1992, investigators administered a standardized questionnaire to all cohort subjects. If direct interview was not feasible (e.g., the subject had died or was unavailable), then a surrogate respondent in the household was interviewed (41% of interviews were with surrogates). The questionnaire included questions on demographic and household characteristics, medical history, and fuel and stove use. For medical conditions subjects were asked whether they had ever had the condition and, if so, the age at diagnosis and type of medical facility where the diagnosis occurred. We considered a subject to have had tuberculosis if the subject or surrogate respondent indicated any prior history of “pulmonary tuberculosis.”

Two follow-ups concerning vital status of study subjects were carried out in 1992 and 1996. Investigators searched death records from hospitals, public security bureaus, and public health bureaus. All subjects lived within 12 miles of one of four large hospitals in Xuanwei. Records of the Qujing District Hospital and Yunnan Province Hospital, which sometimes treat patients from Xuanwei, were also searched to find patients who died outside the county. Information abstracted from death records included date and cause of death.

Statistical analysis

Since the cohort comprised those farmers alive and beginning follow-up on January 1, 1976 (i.e., “baseline”), we distinguish between prevalent tuberculosis (diagnosed prior to or at baseline) and incident tuberculosis (diagnosed after baseline). We used logistic regression to calculate odds ratios (ORs) relating demographic and medical characteristics to the risk of prevalent tuberculosis, although we note that these analyses could be affected by factors related to survival after tuberculosis diagnosis, because subjects with prevalent tuberculosis would need to have survived until entry into the cohort.

For analyses related to tuberculosis incidence, we excluded subjects who had already developed prevalent tuberculosis at baseline, since they were no longer at risk for developing tuberculosis. Proportional hazards regression models were then used to calculate hazard ratios (HRs) associating subject characteristics to tuberculosis incidence. We used age as the time scale in this analysis, to control for possible confounding by age. Subjects were thus considered at risk for incident tuberculosis from their age at cohort entry until the age at which they developed tuberculosis or were censored due to death or end of study follow-up in 1996. As tuberculosis risk factors, only medical conditions that were present at baseline were considered, to ensure that the medical condition preceded development of tuberculosis. For this reason, and because survival following tuberculosis might have affected results of analyses for prevalent tuberculosis but not for incident tuberculosis, the analyses for incident tuberculosis provide a more reliable assessment of the associations than those for prevalent tuberculosis.

As the primary measure of lung cancer risk, our analyses consider death from lung cancer as ascertained from mortality follow-up in 1992 and 1996 (i.e., lung cancer mortality). Lung cancer mortality is largely equivalent to lung cancer incidence, because of the extremely short survival following diagnosis for most lung cancer patients. Based on the 1992 questionnaire, we also had information on diagnosis dates for a subset of lung cancer cases, and in a sensitivity analysis, we analyzed lung cancer risk using these incidence data.

We used proportional hazards models to assess the association between tuberculosis and lung cancer risk, as reflected in death from lung cancer (mortality), or in the sensitivity analysis, diagnosis of lung cancer (incidence). In these models, tuberculosis was considered as a time-dependent variable (i.e., among subjects with tuberculosis, person-time before tuberculosis diagnosis contributed to the estimate of lung cancer risk in the absence of tuberculosis). We evaluated models in which we distinguished between various durations of time since tuberculosis diagnosis (i.e., latency) (13), as well as between prevalent and incident tuberculosis. Models were compared using the likelihood ratio test.

To exclude confounding of the association between tuberculosis and lung cancer, we evaluated proportional hazards models that were adjusted for known or suspected risk factors for lung cancer. We assessed additional models in which subjects were stratified according to gender or use of smoky coal. In a sensitivity analysis to minimize the possibility that study results could be biased by differential recall between lung cancer cases (most of whom were deceased and so had surrogate respondents for the questionnaire) and other study subjects, we included only subjects whose questionnaire data had been provided by a surrogate respondent. In a final sensitivity analysis, we considered only tuberculosis cases that were diagnosed at provincial, district, or county hospitals, which have better diagnostic and treatment resources than commune hospitals, to increase the specificity of the reported diagnoses.

Results

Subject characteristics and associations with tuberculosis

As shown in Table 1, approximately half of the 42,422 subjects were male. The median age at baseline in 1976 was 37 years. The majority (65.1%) of subjects had no education. During their lifetime, most subjects (70.5%) used smoky coal for cooking and heating. Fewer than 1% of women ever smoked tobacco, whereas 92.2% of men were smokers. Other demographic and medical characteristics are shown in Table 1.

Table 1.

Characteristics of 42,422 farmers in the Xuanwei cohort and associations with tuberculosis

Characteristic*	Number of subjects (% of total)	Incident tuberculosis (n=168)		Prevalent tuberculosis (n=78)
		Tuberculosis cases, n (%)	Hazard ratio (95%CI)	Tuberculosis cases, n (%)	Odds ratio (95%CI)
Sex
Male	21,701 (51.2)	71 (0.3)	1.0	36 (0.2)	1.0
Female	20,721 (48.8)	97 (0.5)	1.4 (1.1–2.0)	42 (0.2)	1.2 (0.8–1.9)
Age, years
24–31	13,928 (32.8)	33 (0.2)	1.0	6 (0.0)	1.0
32–44	14,808 (34.9)	57 (0.4)	1.8 (1.1–2.9)	22 (0.2)	3.5 (1.4–8.5)
45–59	13,686 (32.3)	78 (0.6)	5.1 (2.6–9.8)	50 (0.4)	8.5 (3.6–20)
Education level
No education	27,633 (65.1)	126 (0.5)	1.0	62 (0.2)	1.0
Some education	14,789 (34.9)	42 (0.3)	0.7 (0.5–1.0)	16 (0.1)	0.5 (0.3–0.8)
Ever used smoky coal (lifetime)
No	12,532 (29.5)	19 (0.2)	1.0	12 (0.1)	1.0
Yes	29,890 (70.5)	149 (0.5)	3.5 (2.2–5.6)	66 (0.2)	2.3 (1.2–4.3)
Smoking, average cigarette equivalents per day (lifetime, males only)
Non-smoker	1691 (7.8)	7 (0.4)	1.0	8 (0.5)	1.0
0–20	10,263 (47.3)	32 (0.3)	0.7 (0.3–1.7)	15 (0.2)	0.3 (0.1–0.7)
21 or more	9747 (44.9)	32 (0.3)	0.7 (0.3–1.7)	13 (0.1)	0.3 (0.1–0.7)
Asthma
No	42,079 (99.2)	161 (0.4)	1.0	74 (0.2)	1.0
Yes	343 (0.8)	7 (2.1)	10 (4.7–22)	4 (1.2)	6.7 (2.4–18)
Chronic bronchitis
No	41,191 (97.1)	155 (0.4)	1.0	70 (0.2)	1.0
Yes	1231 (2.9)	13 (1.1)	4.4 (2.4–7.8)	8 (0.7)	3.8 (1.8–8.0)
Emphysema
No	42,254 (99.6)	167 (0.4)	1.0	75 (0.2)	1.0
Yes	168 (0.4)	1 (0.6)	4.3 (0.6–31)	3 (1.8)	10 (3.2–33)
Family history of tuberculosis (lifetime)
No	41,980 (99.0)	150 (0.4)	1.0	68 (0.2)	1.0
Yes	442 (1.0)	18 (4.2)	12 (7.2–19)	10 (2.3)	14 (7.3–28)
Average waking hours spent indoors per day (lifetime)
0.0–6.9	25,183 (59.4)	104 (0.4)	1.0	48 (0.2)	1.0
7.0 or more	17,239 (40.6)	64 (0.4)	0.9 (0.7–1.2)	30 (0.2)	0.9 (0.6–1.4)
Average number of rooms in house (lifetime)
1.0–1.9	27,971 (65.9)	138 (0.5)	1.0	64 (0.2)	1.0
2.0–2.9	10,326 (24.3)	19 (0.2)	0.4 (0.2–0.6)	11 (0.1)	0.5 (0.2–0.9)
3.0 or more	4125 (9.7)	11 (0.3)	0.5 (0.3–1.0)	3 (0.1)	0.3 (0.1–1.0)

Abbreviations: CI confidence interval

Hazard ratios and odds ratios were calculated with proportional hazards and logistic regression models, respectively (see Methods).

^*All characteristics are baseline characteristics (i.e., status as of January 1, 1976), unless otherwise noted as lifetime.

Tuberculosis was reported by 246 subjects (0.6%). Of these, 168 were incident cases reported during follow-up (0.4%, incidence 0.2 per 1000 person-years). As shown in Table 1, tuberculosis incidence was significantly elevated among females, older subjects, and those who lacked education or had only a single room in their house. Tuberculosis incidence was higher in individuals with asthma (HR 10) or chronic bronchitis (HR 4.4), tended to be higher with emphysema (HR 4.3), and was strongly elevated in individuals with a family history of tuberculosis (HR 12). Tuberculosis incidence was also elevated among smoky coal users (HR 3.5). Tuberculosis incidence was not associated with smoking (among men) or time spent indoors. Associations were largely similar for prevalent tuberculosis (Table 1). One difference, however, was that, among men, prevalent tuberculosis was less common among smokers than non-smokers.

Association between tuberculosis and lung cancer

During follow-up through 1996, 2459 subjects died from lung cancer (5.8%, mortality 3.1 per 1000 person-years). In comparison, 1764 subjects were diagnosed with incident lung cancer through 1992 (incidence 2.7 per 1000 person-years). Among these incident lung cancers, 1694 (96.0%) later died from lung cancer, with a median time from diagnosis to death of 10 months.

Among 246 subjects with tuberculosis, 31 subsequently died from lung cancer (mortality 25 per 1000 person-years). Lung cancer mortality was thus substantially higher in subjects with tuberculosis than among those without tuberculosis (HR 6.1, 95%CI 4.3–8.7; Table 2). Among the 31 subjects with both conditions, the median time from tuberculosis until lung cancer death was 3.6 years (interquartile range 1.7–6.5 years). Notably, lung cancer mortality was higher 0–4.9 years after tuberculosis diagnosis (HRs ranging from 6.7 to 13) than 5 or more years after tuberculosis diagnosis (HRs 3.0–3.4; Table 2). Nonetheless, even HRs in the periods 5–9.9 and 10+ years after tuberculosis diagnosis were significantly elevated.

Table 2.

Tuberculosis and subsequent lung cancer mortality among farmers in Xuanwei, China through 1996

Tuberculosis exposure category	Lung cancer deaths	Lung cancer mortality, per 1000 person-years	Hazard ratio (95%CI)	−2 log likelihood (degrees of freedom)
No tuberculosis	2428	3.1	1.0
Any tuberculosis	31	25	6.1 (4.3–8.7)	47,740.5 (1)
Tuberculosis, by time since tuberculosis				47,729.9 (6)*
0–0.9 years	4	27	7.3 (2.7–19)
1–1.9 years	5	38	10 (4.2–24)
2–2.9 years	3	26	6.7 (2.1–21)
3–4.9 years	10	51	13 (7.1–25)
5–9.9 years	4	14	3.4 (1.3–9.1)
10+ years	5	13	3.0 (1.3–7.3)
Tuberculosis, by time since tuberculosis				47,731.6 (2)†
0–4.9 years	22	37	9.8 (6.4–15)
5+ years	9	13	3.2 (1.7–6.1)
Tuberculosis, by onset relative to baseline				47,733.7 (2)‡
Prevalent tuberculosis	7	13	3.1 (1.5–6.4)
Incident tuberculosis	24	34	8.6 (5.8–13)

Table 2 notes

Abbreviations: HR hazard ratio, CI confidence interval

^*Comparison of this model with the model that has only a single term for “any tuberculosis”: p=0.06 using the likelihood ratio test.

^†Comparison of this model with the model that has only a single term for “any tuberculosis”: p=0.003 using the likelihood ratio test.

^‡Comparison of this model with the model that has only a single term for “any tuberculosis”: p=0.009 using the likelihood ratio test.

Lung cancer mortality was higher following a diagnosis of incident tuberculosis (HR 8.6) than following a diagnosis of prevalent tuberculosis (HR 3.1, Table 2). However, this difference was related to the period of follow-up time after tuberculosis (i.e., short latency intervals could be evaluated effectively only for incident tuberculosis, because prevalent tuberculosis occurred before baseline). Indeed, a model that considered both prevalent/incident status of tuberculosis and time since tuberculosis was not significantly better than the model that only considered time since tuberculosis (p=0.19). Therefore, subsequent analyses considered lung cancer mortality only according to time since tuberculosis (0–4.9 vs. 5+ years). The association between tuberculosis and lung cancer risk was not affected by adjustment for the demographic characteristics and medical conditions shown in Table 1 (see Supplemental Table 1 for adjusted results).

Associations between tuberculosis and lung cancer were similar for men (N=1233 lung cancer deaths) and women (N=1226). Specifically, among men, lung cancer HRs (95%CIs) were 9.7 (4.8–19) and 4.3 (1.8–10) in the 0–4.9 years and 5+ years after tuberculosis diagnosis. Among women, the corresponding lung cancer HRs (95%CIs) were 9.7 (5.7–16) and 2.4 (0.9–6.4), respectively. Results were similar among men after adjustment for smoking status (HRs 9.7 and 4.3 in the in the 0–4.9 years and 5+ years after tuberculosis, respectively) and among women after adjustment for smoky coal use (HRs 7.5 and 2.5, respectively). Among ever-users of smoky coal (N=2430 lung cancer deaths), tuberculosis was associated with increased lung cancer mortality in the 0–4.9 years after tuberculosis diagnosis (HR 7.5, 95%CI 4.9–11) and in the 5+ years after tuberculosis diagnosis (HR 2.5, 95%CI 1.2–5.0).

In a sensitivity analysis utilizing lung cancer incidence data, the association between tuberculosis and lung cancer risk remained similar to the primary analyses, i.e., HRs (95%CIs) were 7.9 (4.7–13) and 4.0 (2.0–8.0) in the 0–4.9 and 5+ years after tuberculosis diagnosis, respectively. In another sensitivity analysis restricted to subjects whose questionnaire information had been provided by a surrogate respondent, the associations with lung cancer mortality persisted, although they appeared somewhat attenuated: HRs (95%CIs) were 4.7 (3.1–7.1) and 2.3 (1.2–4.5) in the 0–4.9 and 5+ years after tuberculosis diagnosis, respectively.

In a final sensitivity analysis that included only the 204 tuberculosis cases diagnosed in provincial, district, or county hospitals, lung cancer HRs (95%CIs) were 11 (7.4–17) and 3.2 (1.5–6.8) in the same time intervals. Lastly, we were able to validate questionnaire reports of tuberculosis for subjects who had the condition listed as a cause of death on their death certificate. Of 150 subjects with tuberculosis listed as a cause of death, 136 (90.7%) had tuberculosis reported on the study questionnaire.

Discussion

The present study demonstrates a strongly increased risk of lung cancer following a diagnosis of tuberculosis among farmers in Xuanwei, China. The increased lung cancer risk appeared greatest within the first five years after tuberculosis diagnosis, but the risk for lung cancer remained three-fold elevated for more than ten years following tuberculosis diagnosis. Importantly, the association between tuberculosis and lung cancer was observed in both men and women, and was not due to confounding by the effects of tobacco use or indoor environmental smoke, because adjusted results were similar to unadjusted results.

These results from our cohort study extend upon those from prior case-control studies (8–11,14,15), most of which, although not all, have demonstrated an association between lung cancer and tuberculosis. Three of the studies (8,11,14), like our study conducted in China, reported odds ratios for lung cancer ranging from 1.3 to 3.8.

Like the prior studies, our study also relied on subject recall for the diagnosis of tuberculosis. The population of Xuanwei were largely uneducated, and subjects and their families could have mis-reported this diagnosis. However, the incidence of tuberculosis reported by our subjects was 0.2 per 1000 person-years, which is similar to that estimated for China by the World Health Organization (0.1 per 1000 person-years) (4). Further evidence for the accuracy of tuberculosis diagnoses in our study is provided by our observation that most subjects who had tuberculosis listed on their death certificates had this diagnosis reported on the questionnaire. While it is possible that subjects or family members may have confused a diagnosis of lung cancer with tuberculosis, this misdiagnosis of tuberculosis would most likely have been attributed to the several years around the onset of lung cancer, and so would not have led to the observed associations between tuberculosis and lung cancer over longer intervals. Additionally, we note that the association between tuberculosis and lung cancer in our study persisted when we considered only tuberculosis diagnoses from hospitals at or above the county level, where medical evaluation would have been more sophisticated and diagnoses more accurate than at local hospitals. Thus, the reported diagnoses of tuberculosis seem largely supported, and it is unlikely that the association that we observed between tuberculosis and lung cancer is entirely due to confusion of these two diagnoses.

Two additional biases, which could have partly affected both our cohort study and prior case-control studies, deserve mention. First, ascertainment bias arises when exposed individuals are more likely to be diagnosed with cancer. This bias could have arisen in the studies under consideration, since patients with tuberculosis may have been more likely to be diagnosed with lung cancer than unaffected individuals, due to the use of chest x-rays obtained in the evaluation of the infection. A second bias, reverse causality, would have occurred if early-stage lung cancer, otherwise occult, caused a weakening of the immune system, leading in turn to reactivation of latent M. tuberculosis infection. The clinical infection would then appear to precede the cancer, when the cancer actually preceded the clinical infection. Both of these biases would be reflected in strong associations between tuberculosis and cancer over short time intervals (i.e., short latencies). Indeed, our study and some of the case-control studies (8,10,11) observed the strongest associations in the few years after tuberculosis diagnosis. However, the associations observed over much longer periods (8,11) (persisting for more than a decade in our study) are unlikely to be explained by such artifacts. Also, the similarity of the results of our sensitivity analysis, in which we only considered incident lung cancer cases, further supports that the observed associations are attributable to tuberculosis preceding the onset of cancer.

An increased risk for lung cancer following a diagnosis of tuberculosis is biologically plausible (5–7). Respiratory symptoms can last several months before the clinical diagnosis of tuberculosis, and treatment typically entails 6–9 months of multi-drug treatment. During this extended time, the infection induces substantial pulmonary inflammation (16), with production of tumor necrosis factor linked to prolonged fever and wasting. Reactive oxygen and nitrogen species produced by activated leukocytes participating in the inflammatory response can bind to DNA, leading to genomic alterations (17). Tuberculosis also causes extensive pulmonary fibrosis, perhaps related to production of transforming growth factor beta, interleukin 4, and interleukin 13 (16). Since the 1930s, scars such as those caused by tuberculosis have been postulated to play an etiologic role in lung cancer (18), based on the observation that lung cancers frequently arise in proximity to scar tissue. Tissue repair is associated with cellular proliferation, during which errors in chromosomal replication might lead to further DNA mutations.

Our results highlight the importance of poverty and pre-existing lung disease in promoting acquisition and clinical reactivation of tuberculosis (19). In the rural Chinese population that we studied, tuberculosis risk was significantly elevated among persons with low socioeconomic status, as indicated by lack of education and residence in a house with only a single room. The strong association that we observed with a family history of tuberculosis likely reflects transmission of M. tuberculosis within households, although the association could also partly reflect environment or genetic predisposition to tuberculosis. We also found substantially increased tuberculosis risk among individuals with asthma, chronic bronchitis, and emphysema. We did not observe an increased risk of tuberculosis related to smoking, which has been reported in previous studies (20). Indoor exposure to smoke from coal used in cooking and heating is an important cause of chronic obstructive pulmonary disease (21), which could explain the strongly increased risk of tuberculosis among smoky coal users (HR 3.5, 95%CI 2.2–5.6), although we did not see associations with tuberculosis risk for waking hours spent indoors (Table 1) or hours spent near the stove (data not shown). Prior studies have demonstrated elevated risk of tuberculosis among persons exposed to indoor smoke from stoves that use biomass as fuel (22,23). Programs that assist communities in mitigating indoor air pollution levels may facilitate reductions in the burden of tuberculosis (24).

Strengths of our study include its large size and setting in an area where both tuberculosis and lung cancer are major public health problems. These advantages provided a sufficient number of individuals with tuberculosis for us to examine the association between tuberculosis and subsequent lung cancer risk. We conducted extensive analyses regarding the risk of lung cancer over time in relation to tuberculosis diagnosis, and ruled out major confounding or effect modification by other environmental causes of lung cancer. A limitation of the study was that ascertainment of both tuberculosis and lung cancer deaths was retrospective. In particular, we were unable to perform a comprehensive review of medical records to validate the medical events or obtain additional details (e.g., regarding tuberculosis treatment or lung cancer histologic subtype). We conducted a sensitivity analysis restricted to subjects whose data were provided by surrogates, in order to minimize any difference in recall between subjects who died from lung cancer and those who did not. Although the associations appeared attenuated in this analysis, it was based on substantially fewer subjects than the primary analysis, and the associations between tuberculosis and lung cancer remained significant.

In conclusion, our study supports the possibility that tuberculosis may act through a process of localized pulmonary inflammation and fibrosis to initiate or promote the development of lung cancer, particularly in conjunction with other carcinogenic exposures. Additional research is needed to better understand the biological mechanisms of this association. The potential contribution of tuberculosis to the development of lung cancer points to the continuing need to develop and implement improved tuberculosis prevention and treatment programs, especially in the developing world. Our findings also raise the possibility that periodic lung cancer screening of patients with a history of tuberculosis may be an effective strategy for early detection, which might ultimately improve cancer outcomes.