Lung cancer accounts for approximately 18.4% of the total cancer-related deaths, the highest of all cancer types. The prognosis of lung cancer is relatively unfavorable compared to that of other malignancies, and as a prognosis largely depends on the stage of onset, thus, the early diagnosis of lung cancer is very important.
1. Introduction
Chronic inflammation resulting in pathological changes is a major risk factor in carcinogenesis. Inflammation is known to play a key role in carcinogenesis, such as infection with hepatitis B and C viruses in hepatocellular carcinoma, Helicobacter pylori in gastric cancer, and human papilloma virus in gynecological cancers [7]. Several meta-analyses have shown that previous inflammatory diseases in the lungs, such as pneumonia, chronic bronchitis, and pulmonary tuberculosis (TB), may increase the risk of lung cancer (relative risk ratio 1.36–1.90), independent of cigarette smoking [8,9]. According to forty-nine studies, pulmonary and extra-pulmonary TB infections increase the risk of 10 cancer types, including head and neck cancer, leukemia, lymphoma, gastrointestinal cancer, kidney cancer, bladder cancer, and lung cancer [10]. Thus, TB infection may influence the pathogenesis of lung cancer with or without cigarette smoking. To prevent the emergence of airborne transmittable TB and its progression to cancer, the control and prevention of TB is very important.
2. Pulmonary TB and Risk of Lung Cancer with All Eligible Studies
The overall association between a previous history of pulmonary TB and newly diagnosed lung cancer was statistically significant (odds ratio (OR): 2.09; 95% confidence interval (CI): 1.62–2.69, p < 0.001). There was high heterogeneity (I2 = 95%), no evidence of publication bias, the egger p-value was 0.447, and no visual asymmetry in the funnel plot (Figure 2A and Figure 3A). In the subgroup analysis by TB burden, the high-burden countries showed higher OR (2.57, 95% CI: 1.68–3.93, p < 0.001) than the medium-burden (OR: 2.48, 95% CI: 1.71–3.58, p < 0.001) and low-burden countries (OR: 1.77, 95% CI: 1.22–2.56, p = 0.003). Geographically, East Asia and the Pacific region showed a prominent risk (OR: 2.49, 95% CI: 1.83–3.39, p < 0.001) compared to the Europe and Central Asia (OR: 1.60, 95% CI: 0.80–3.22, p = 0.185) or North America (OR: 1.53, 95% CI: 1.11–2.12, p = 0.010) regions. The economic income statuses of the countries also reflected the characteristics of patients with TB, and the countries with upper-middle incomes (OR: 2.57, 95% CI: 1.68–3.93, p < 0.001) demonstrated a higher risk of lung cancer than high-income status countries (OR: 1.91, 95% CI: 1.41–2.59, p < 0.001). The association between pulmonary TB and newly developed lung cancer was statistically significant regardless of the adjustment for age, sex, smoking status, and cohort type or study design. The magnitude of association was similar regardless of whether pulmonary TB was diagnosed based on medical records (OR: 2.26, 95% CI: 1.29–3.94, p = 0.004), imaging (OR: 2.13, 95% CI: 1.16–3.92, p = 0.015), or self-report/physical examination (OR: 1.96, 95% CI: 1.56–2.47, p < 0.001). The heterogeneity within subgroups remained at a high level in a majority of the subgroup analyses (Table 2).
Figure 2. Forest plots of risk estimates for the association between tuberculosis and lung cancer. (A) Meta-analysis of all eligible studies. (B) Meta-analysis of high-quality studies.
Figure 3. Funnel plot of the study estimates. (A) All eligible studies. (B) High-quality studies.
Table 2. Meta-analysis of 33 eligible cohorts to assess the association between pulmonary tuberculosis and lung cancer.
Subgroup |
No. of Cohorts * |
OR (95% CI) |
p-Value |
I2 Value (%) |
I2 between Subgroups (%) |
All cohorts |
33 |
2.09 (1.62–2.69) |
<0.001 |
95 |
|
TB burden of country |
Low |
18 |
1.77 (1.22–2.56) |
0.003 |
97 |
12 |
Medium |
6 |
2.48 (1.71–3.58) |
<0.001 |
75 |
High |
9 |
2.57 (1.68–3.93) |
<0.001 |
81 |
Region of country |
East Asia and Pacific |
19 |
2.49 (1.83–3.39) |
<0.001 |
93 |
58 |
Europe and Central Asia |
7 |
1.60 (0.80–3.22) |
0.185 |
98 |
North America |
7 |
1.53 (1.11–2.12) |
0.010 |
0 |
Economic status of country |
High-income |
24 |
1.91 (1.41–2.59) |
<0.001 |
96 |
20 |
Upper-middle-income |
9 |
2.57 (1.68–3.93) |
<0.001 |
81 |
Age |
Adjusted |
29 |
2.00 (1.54–2.61) |
<0.001 |
95 |
14 |
Not adjusted |
4 |
3.84 (1.21–12.15) |
0.022 |
82 |
Sex |
Adjusted |
22 |
2.23 (1.60–3.11) |
<0.001 |
96 |
0 |
Not adjusted |
11 |
1.90 (1.47–2.46) |
<0.001 |
61 |
Smoking |
Adjusted |
22 |
2.03 (1.51–2.73) |
<0.001 |
90 |
0 |
Not adjusted |
11 |
2.19 (1.34–3.59) |
0.002 |
98 |
Hypertension |
Adjusted |
2 |
1.92 (0.66–5.57) |
0.230 |
99 |
0 |
Not adjusted |
31 |
2.10 (1.62–2.73) |
<0.001 |
92 |
Diabetes |
Adjusted |
2 |
1.72 (0.48–6.20) |
0.404 |
99 |
0 |
Not adjusted |
31 |
2.13 (1.63–2.77) |
<0.001 |
94 |
Respiratory comorbidities |
Adjusted |
8 |
1.32 (0.93–1.86) |
0.121 |
94 |
90 |
Not adjusted |
25 |
2.51 (2.04–3.08) |
<0.001 |
78 |
Cohort of the study |
Population-based |
23 |
1.95 (1.41–2.68) |
<0.001 |
96 |
0 |
Hospital-based |
10 |
2.36 (1.85–3.01) |
<0.001 |
49 |
Study design |
Prospective cohort study |
4 |
1.96 (1.22–3.15) |
0.005 |
84 |
94 |
Retrospective cohort study |
2 |
3.95 (3.58–4.36) |
<0.001 |
0 |
Case-control study |
27 |
1.99 (1.56–2.53) |
<0.001 |
89 |
Diagnostic method of pulmonary TB |
Medical record |
8 |
2.26 (1.29–3.94) |
0.004 |
99 |
0 |
Imaging |
3 |
2.13 (1.16–3.92) |
0.015 |
80 |
Self-report or physical examination |
22 |
1.96 (1.56–2.47) |
<0.001 |
66 |
* Since two separate cohorts were reported in one article, a total of 33 eligible cohorts were extracted and analyzed from 32 enrolled studies. Abbreviations: CI, confidence interval; No, Number; OR, odds ratio; TB, tuberculosis.
3. Pulmonary TB and Risk of Lung Cancer with High-Quality Studies
The analysis of eight high-quality studies showed a higher OR (2.26, 95% CI: 1.29–3.94, p = 0.004) than the analysis of all the studies. There was a high heterogeneity (I2 = 99%) with no publication bias, with Egger p = 0.621, and no visual asymmetry in the funnel plot (Figure 2B and Figure 3B). Of the eight articles, seven had cohorts from countries with a low TB burden, and only one had a cohort from a country with a medium TB burden. In the subgroup analysis with a TB burden, the medium-burden countries showed higher OR (4.18, 95% CI: 3.15–5.55, p < 0.001) than the low-burden countries (OR: 2.04, 95% CI: 1.12–3.73, p = 0.020). Geographically, the East Asia and the Pacific region showed a more prominent risk (OR: 2.79, 95% CI: 1.21–6.39, p = 0.016) compared to the Europe and Central Asia regions (OR: 1.79, 95% CI: 0.67–4.77, p = 0.244) (Table 3).
Table 3. Meta-analysis of high-quality studies to assess the association between TB and lung cancer.
Subgroup |
No. of Studies |
OR (95% CI) |
p-Value |
I2 Value (%) |
I2 between Subgroups (%) |
All studies |
8 |
2.26 (1.29–3.94) |
0.004 |
99 |
|
Country of TB burden |
|
|
|
|
|
Low |
7 |
2.04 (1.12–3.73) |
0.020 |
99 |
78 |
Medium |
1 |
4.18 (3.15–5.55) |
<0.001 |
- |
High |
0 |
- |
- |
- |
Region of country |
|
|
|
|
|
East Asia and Pacific |
4 |
2.79 (1.21–6.39) |
0.016 |
98 |
0 |
Europe and Central Asia |
4 |
1.79 (0.67–4.77) |
0.244 |
99 |
North America |
0 |
- |
- |
- |
Abbreviations: CI, confidence interval; No, Number; OR, odds ratio; TB, tuberculosis.
4. Stratified and Sensitivity Analysis
The quality of the 33 included articles was evaluated using the NOS. The quality assessment of 27 case–control studies is shown in
Table 4 and that of six retrospective cohort studies is demonstrated in
Table 5. Meta-regression analyses were performed with continuous variables, such as the mean age at diagnosis of pulmonary TB, baseline characteristics including comorbidity, and pathological cell type of lung cancer. All the results are shown in
Supplementary Table S2. Of these, patients with a low mean age at diagnosis of pulmonary TB showed a significant association between pulmonary TB and lung cancer. The primary analysis with all 32 articles estimated a regression coefficient of 0.949 (
p < 0.001). The secondary analysis with eight high-quality studies with stringent TB diagnostic methods showed similar results (regression coefficient = 0.945,
p < 0.001) (
Figure 4).
Figure 4. Meta-regression analysis of the mean patient age and association between tuberculosis and lung cancer. (A) All eligible studies. (B) High-quality studies.
Table 4. Quality assessment of the included case–control studies using the Newcastle–Ottawa Scale.
Study |
Selection |
Comparability |
Outcome |
Quality Score |
Adequacy of Case Definition |
Degree of Representation of Cases |
Selection of Controls |
Definition of Controls |
Comparability of Cases and Controls on the Basis of Design or Analysis |
Confirmation of Exposure |
Same Method of Confirmation for Cases and Controls |
Non-Response Rate |
An et al. 2020 [20] |
* |
* |
* |
* |
** |
* |
* |
* |
9 |
Yang et al. 2015 [22] |
* |
* |
* |
* |
** |
|
* |
* |
8 |
Yang et al. 2015 [23] |
* |
* |
* |
* |
* |
* |
* |
* |
8 |
Hosgood et al. 2013 [25] |
* |
* |
* |
* |
* |
|
* |
* |
7 |
Lo et al. 2013 [27] |
* |
* |
* |
* |
** |
|
* |
* |
8 |
Bodmer et al. 2012 [28] |
* |
* |
* |
* |
** |
* |
* |
* |
9 |
Koshiol et al. 2010 [31] |
* |
* |
* |
* |
** |
|
* |
* |
8 |
Park et al. 2010 [32] |
* |
* |
* |
* |
** |
|
* |
|
7 |
Liang et al. 2009 [33] |
* |
* |
* |
* |
** |
|
* |
* |
8 |
Wang et al. 2009 [34] |
* |
* |
* |
* |
* |
|
* |
* |
7 |
Galeone et al. 2008 [35] |
* |
* |
* |
* |
** |
|
* |
* |
8 |
Ramanakumar et al. 2006 [36] a |
* |
* |
* |
* |
** |
|
* |
* |
8 |
Ramanakumar et al. 2006 [36] b |
* |
* |
* |
* |
** |
|
* |
* |
8 |
Zatloukal et al. 2003 [37] |
* |
* |
* |
* |
** |
|
* |
* |
8 |
Chan-Yeung et al. 2003 [38] |
* |
* |
* |
* |
* |
|
* |
* |
7 |
Kreuzer et al. 2002 [39] |
* |
* |
* |
* |
* |
|
* |
* |
7 |
Brenner et al. 2001 [40] |
* |
* |
* |
* |
** |
|
* |
* |
8 |
Kreuzer et al. 2001 [41] |
* |
* |
* |
* |
* |
* |
* |
* |
8 |
Zhou et al. 2000 [42] |
* |
* |
* |
* |
* |
|
* |
* |
7 |
Osann et al. 2000 [43] |
* |
* |
* |
* |
** |
|
* |
* |
8 |
Mayne et al. 1999 [44] |
* |
* |
* |
* |
* |
|
* |
* |
7 |
Ko et al. 1997 [45] |
* |
* |
* |
* |
|
* |
* |
* |
7 |
Schwartz et al. 1996 [46] |
* |
* |
* |
* |
** |
|
* |
* |
8 |
Luo et al. 1996 [47] |
* |
* |
* |
* |
* |
|
* |
* |
7 |
Wu et al. 1995 [48] |
* |
* |
* |
* |
** |
|
* |
* |
8 |
Alavanja et al. 1992 [49] |
* |
* |
* |
* |
** |
|
* |
* |
8 |
Wu-Williams et al. 1990 [50] |
* |
* |
* |
* |
** |
|
* |
* |
8 |
a,b: Two separate cohorts reported in one article. Study a was conducted in 1979–1986 (755 cases and 512 controls); study b was conducted in 1996–2001 (1205 cases and 1541 controls). A study can be awarded a maximum of one star for each numbered item within the Selection and Outcome categories. A maximum of two stars can be given for Comparability.
Table 5. Quality assessment of the included retrospective cohort studies using the Newcastle–Ottawa Scale.
|
Selection |
Comparability |
Outcome |
Quality Score |
Study |
Degree of Representation of the Exposed Cohort |
Selection of the Non-Exposed Cohort |
Confirmation of Exposure |
Demonstration That the Current Outcome of Interest Is Absent at the Start of the Study |
Comparability of Cohorts Based on Design or Analysis |
Assessment of Outcome |
Sufficiency of Follow-Up to Detect Outcomes |
Adequacy of Follow-Up of Cohorts |
Kim et al. 2020 [19] |
* |
* |
|
* |
** |
* |
* |
|
7 |
Oh et al. 2020 [21] |
* |
* |
|
* |
** |
* |
* |
* |
8 |
Simonsen et al. 2014 [24] |
* |
* |
* |
* |
* |
* |
* |
* |
8 |
Bae et al. 2013 [26] |
* |
* |
|
* |
** |
* |
* |
|
7 |
Shiels et al. 2011 [29] |
* |
* |
* |
* |
** |
* |
* |
* |
9 |
Yu et al. 2011 [30] |
* |
* |
* |
* |
* |
* |
* |
|
7 |
A study can be awarded a maximum of one star for each numbered item within the Selection and Outcome categories. A maximum of two stars can be given for Comparability.
This entry is adapted from the peer-reviewed paper 10.3390/jcm11030765