Construction of a risk model and deep learning network based on patients with active pulmonary tuberculosis and pulmonary inflammation
- Authors:
- Published online on: March 28, 2023 https://doi.org/10.3892/br.2023.1616
- Article Number: 34
-
Copyright: © Xu et al. This is an open access article distributed under the terms of Creative Commons Attribution License.
Abstract
Introduction
Tuberculosis (TB) remains a severe global public health problem, especially in developing countries (1). How to reduce the TB epidemic is a major issue hindering economic and social development. Detection of Mycobacterium tuberculosis (Mtb) DNA using Gene Xpert MTB/RIF assay is more sensitive and rapid for diagnosing TB and rifampicin resistance (2). However, due to its costs, environmental limitations, and difficulties in its supply, it is difficult to carry out screening in a low-income rural area (3). Interferon-γ release assay (IGRA) is commonly used in the diagnostic workup of Mtb but distinguishes poorly between TB and latent TB infection (LTBI) (4,5).
In addition, the symptoms of patients with TB are very similar to those of bacterial community-acquired pneumonia (CAP). Both TB and CAP are infections of the lower respiratory tract, but they are often considered as separate entities (6). TB is classically a more indolent disease presenting cavitating lung lesions observed in patients with a history of cough for three months or longer accompanied by weight loss, and, often, is not associated with acute respiratory compromise (7). By contrast, CAP is generally associated with a short history of several days, is rapidly progressive, and is more often associated with respiratory compromise (8). The diagnosis of CAP is based on the detection of a new infiltrate on a chest radiograph or other imaging technique in the presence of recently acquired respiratory signs and symptoms (9). However, clinical findings do not reliably predict radiologically confirmed PN, as features of TB may sometimes be quite similar to those of CAP among patients who experience symptoms at the early stage (10). In addition, the etiology cannot be simply differentiated clinically or radiologically and is undefined in ~50% of patients (8). Because most patients with TB have no sputum specimens or are Mtb smear/culture-negative, this makes diagnosis more intractable (11).
Host markers including secreted molecules in blood have been reported as novel candidate markers to distinguish TB, such as interferon-γ inducible protein 10 kDa (12), interleukin-2 (IL-2) (13), IL-6(14), C-reactive protein (CRP) (15) and vascular endothelial growth factor (14). However, the diagnostic performances of reported biomarkers cannot be applied in low-income areas due to technical, instrumentation or cost limitations. To address this problem, the present study retrospectively analyzed the differences in routinely monitored laboratory variables in blood tests between AFB- IGRA+ TB and PN, and built a risk model to differentiate AFB- IGRA+ TB from PN, and validated its application by an external independent cohort.
Patients and methods
Study design and criteria for study inclusion
The patients provided a full medical history, participated in regular physical examinations, and underwent routine investigations, including acquired immunodeficiency syndrome (AIDS) serology, chest radiography, IGRA, and microbiological sputum examination, where possible.
The inclusion criteria for PN participants were as follows: i) Meeting the diagnostic criteria of ‘The People's Republic of China health industry standard (WS 382-2012; http://www.nhc.gov.cn/wjw/s9494/201209/110b324c465740169a863d57a78c18a6.shtml)’. The clinical diagnosis can be established by any of (a), (b), (c), (d) plus (e) and excluding TB, lung tumor, non-infectious interstitial lung disease, pulmonary edema, pulmonary atelectasis, pulmonary embolism, pulmonary eosinophilic infiltrates, pulmonary vasculitis, etc. The elements of (a) to (e) are as follows: a) Newly developed cough and sputum, or aggravation of existing respiratory symptoms with purulent sputum, with or without chest pain; b) fever; c) solid lung signs and/or wet rales; d) peripheral blood WBC >10x109/l or <4x109/l with or without leftward nuclear shift; and e) chest X-ray showing new lamellar or patchy infiltrative shadows or interstitial changes with or without pleural effusion. ii) Sputum culture with a bacterial pathogenic basis or effective antimicrobial therapy, such as marked improvement in symptoms such as cough, yellow sputum and significant uptake of chest imaging. iii) Exclusion of viral, atypical pathogens, fungal, and Mtb infections. The inclusion criteria for IGRA+ TB participants were as follows: Meeting the diagnostic criteria for TB including positive sputum, bronchoscopic lavage, brush examination, smear microscopy of biopsy specimens for antacid bacilli, isolation and culture of Mtb, or positive nucleic acid test and IGRA for Mtb, or positive pathology of lung tissue biopsy. The inclusion criteria for AFB- TB participants were as follows: There were no sputum or negative bacilli smear and negative Mtb, but there were chest computed tomographic (CT) scans or chest X-ray evidence and symptoms responding to TB treatment.
The exclusion criteria were as follows: i) Use of antimicrobial drugs for >24 h; (2) suffering from diseases that can affect the total number and classification of blood leukocytes, including leukemia, chronic inflammatory states, etc.; iii) having a recent (within 3 months) history of glucocorticoid application or ongoing hormone use; and iv) clearly diagnosed or highly suspected pulmonary edema, pulmonary embolism, pulmonary atelectasis, bronchial asthma, viral PN, fungal PN, atypical pathogenic PN, pulmonary eosinophilic infiltrates and lung cancer.
The involved participants were divided into two groups: The discovery cohort and the external validation cohort. For the discovery cohort, participants were enrolled at Ganzhou Fifth Hospital (Ganzhou, China) between August 2018 and August 2020, including 748 AFB- TB participants and 531 PN participants. As the predefined goal was to assess the ability of laboratory biomarkers to distinguish IGRA+ patients presenting with AFB- TB, 287 participants with IGRA+ TB were subjected for further analysis (AFB- IGRA+ TB). Participants with >50% of laboratory data missing were excluded, thus making a total of 89 TB and 38 PN participants with recorded biomarker values used to construct the risk model. The external validation cohort of 134 participants in the study was collected from Shenzhen Third People's Hospital (Shenzhen, China) from June 2018 to June 2019 (Fig. 1). Among them, 15 were excluded due to missing variables, therefore the external validation cohort consisted of 77 AFB- IGRA+ TB and 42 PN participants.
Ethical approval and patient consent to participate
The study protocol was approved by the Ethics Committee and the Institutional Review Board of Ganzhou Fifth People's Hospital (registration no. 2020-10) to allow retrospective access to the records and files of patients. Written informed consent was waived by the Ethics Committee as this was an observational and retrospective analysis.
Data collection
The medical records of all participants were reviewed by experienced TB clinicians, including medical history, symptoms, clinical signs, microbiological tests, laboratory findings, chest CT chest X-rays, and treatment measures. A total of 41 laboratory biomarkers were assessed by differential statistics and odds ratio (OR) calculation for variable selection.
Statistical analyses
For laboratory results, continuous variables were preprocessed by log2-transformation before analysis. The laboratory data were verified for skewed distribution using the Kolmogorov-Smirnov test. In the present study Wilcoxon rank-sum test was suitable for skewed distribution data (16-18). P-values were adjusted by false discovery rate (FDR). Variables between two conditions were defined as statistically significant when FDR <0.2(19).
A univariate logistic model (glm) was used to calculate the OR for each laboratory biomarker. The regression coefficient of the glm was regarded as the log OR. Variables with an FDR <0.2 or at a statistically significant level (P-value <0.05) in the glm analysis were candidates for the construction of a multivariate risk model (lrm) and nomogram, and the final variables were determined using Akaike's information criterion (AIC) as a stopping rule. The goodness of fit of the lrm model was calculated using Hosmer and Lemeshow C statistics test. The performance of the nomogram was evaluated by the concordance index (C-index) and assessed by comparing nomogram-predicted vs. actual observation, and bootstrapping with 1,000 resamples to decrease the overfit bias was applied for calibration (20,21). The glm was generated by glmnet package (version 4.1-4) (22) and the nomogram was generated by DynNom package (version 5.0.2) (23). All analyses and figures were generated in R version 4.0.3 (https://www.r-project.org/).
Results
Differential laboratory biomarkers between AFB- IGRA+ TB and PN
The characteristics of the AFB- IGRA+ TB and PN participants are shown in Table I. No significant differences in age and sex were found in the first cohort. Males made up the majority in both cohorts. Clinical types of TB participants included 162 infiltrative pulmonary TB, 19 cavitary pulmonary TB, 8 secondary pulmonary TB, and 15 tuberculous pleurisy and empyema. Only 44 and 23 participants had received TB-DNA and TB-antibody examination, yielding 9.09% (4/44) and 13.04% (3/23) positive rates, respectively.
The significant variables between the two conditions were further explored. Using unpaired t-test and setting FDR <0.2, 13 variables with marked differences were identified (Table II). Only uric acid (UA) was elevated (fold change >1.2). Notably, five variables, including aspartate aminotransferase (AST), total bile acid (TBA), triglyceride (TG), alanine aminotransferase (ALT), and glutamyl transpeptidase (GGT), were reduced (fold change <0.83) in TB compared with PN (Fig. 2A). These data revealed variable expression differences between the two conditions, and that some variables may be useful to distinguish AFB- IGRA+ TB from PN.
To find the odds that TB would progress or not be given exposure to these laboratory variables, OR was assessed for each variable by univariate logistic model. Notably, 11 variables significantly associated with TB progression (P<0.05; Table II) were identified. Among them, four variables indicated a protective effect in TB progression (OR <1), including UA, red blood cell (RBC), Hb, and ALB; while seven variables including GLU, AST, TBA, TG, ALT, GGT, and ALP, were revealed as risk factors for TB progression (OR >1).
Multivariate risk model to predict TB progression probability
Combined with the aforementioned results, and using AIC as a stopping rule, five laboratory variables (age, UA, ALB, Hb, and WBC) were finally selected to develop a multivariate risk model with 89 AFB- IGRA+ TB and 38 PN participants. Notably, the expression distribution of variables in both groups revealed more unbalance in PN than AFB- IGRA+ TB (Fig. 2B). The risk model yielded a C-index of 0.7 (95% CI: 0.61, 0.8), with P=0.01 (chi-square test) (data not shown). The calibration plot revealed a moderate agreement between the model prediction and the actual observation (Fig. 3A; P=0.18, Hosmer-Lemeshow test). Using the nomogram, the values for each variable were mapped to points on a scale axis ranging from 0 to 10. With a corresponding number of points assigned to given magnitudes of the variables, the risk probability was calculated by the corresponding cumulative point score for all the variables (25). It was revealed that UA had the most protective effect in TB progression, followed by Hb; while age, WBC and ALB were shown to be risk factors (Fig. 3C).
Next, an external validation cohort of 134 participants, consisting of 77 AFB- IGRA+ TB and 42 PN participants were prospectively collected. The C-index of nomogram for predicting the external cohort was 0.77 (95% CI: 0.68, 0.86) (data not shown). The calibration plot also revealed consistent results between the prediction by nomogram and actual observation (Fig. 3B) with P=0.13 (Hosmer-Lemeshow test).
Discussion
In the present study, different profiles were analyzed between AFB- IGRA+ TB and PN, and five laboratory variables (age, UA, ALB, Hb and WBC) were selected to construct a multivariate risk model and nomogram. Internal validation and a calibration plot showed moderate agreement between nomogram probability and actual observation, with a C-index of 0.7 (95% CI: 0.61, 0.8). A similar result in an external validation cohort (C-index: 0.77; 95% CI: 0.68, 0.86) was obtained. These findings indicated that five laboratory variables may be used to predict TB disease probability when a clinical sample is AFB- IGRA+.
It has been reported that patients with TB tend to exhibit increased levels of CRP, erythrocyte sedimentation rate (ESR), and UA, and low levels of Hb (25). An increased UA level was observed in 28.2% of men and 37.5% of women prior to chemotherapy, and more often during the first 2 months of treatment both in men and women, which suffered from multiple drug-resistant pulmonary TB (26). In the present study, serum UA was revealed to be at a significantly higher level in TB (FDR <0.001), with an OR value of 0.36 (P=2.5E-05) compared with PN (Table II), indicating that it may be a specific protective factor in patients with TB. Reduced plasma ALB concentrations have been reported in TB (27) and may be used as a diagnostic and prognostic marker in pretreated HIV and TB patients. WBC was revealed to be significantly increased in patients with TB compared with healthy controls, and the WBC significantly decreased during TB treatment (28,29). In the present study, WBC was statistically significant and a significant risk factor (OR=1.49), but with no higher counts in fold change compared with PN.
To predict the risk of TB for each AFB- IGRA+ patient, a nomogram was used to provide a more accurate profile. With five variables, the nomogram had good predictive accuracy with a C-index of 0.7. External validation was essential to confirm it can be applied to patients outside of the cohort. Thus, a second participant cohort from another center was recruited, and then assessed on the nomogram, and the result obtained was consistent with the actual observation (C-index of 0.77).
The present study still had several limitations: First, in low-income and rural settings, not all patients received all routine laboratory tests, leading to numerous missing values in the first cohort of participants. In order to analyze more biomarkers, participants with >50% of missing data were excluded, with 41 laboratory variables and a small number of participants remaining (89 AFB- IGRA+ TBA and 38 PN), resulting in a small sample size. Second, although internal and external validation exhibited good performance, further investigations are required to optimize the nomogram in larger cohorts and more types of pulmonary TB.
In addition, the association between prior TB and lung cancer has been undefined. Several large cohort studies provided evidence supporting an association between prior TB and risk of lung cancer (30,31). In a systematic review and meta-analysis published in 2011, a previous diagnosis of TB was associated with increased lung cancer risk [relative risk (RR)=1.76 (95% CI=1.49 to 2.08)] with little variation by smoking status (32). Similarly, a pooled analysis from the International Lung Cancer Consortium found a lung cancer RR of 1.48 (95% CI=1.17 to 1.87) associated with a history of TB, controlling for smoking status. However, there was no attempt to differentiate TB from lung cancer with the risk model of the present study. In future projects, applicability of this model in other diseases will be further investigated. Following improvement of this model, such as increasing its applicability, it is anticipated that it may help clinicians to reduce the cost and time to diagnose AFB- IGRA+ TB in low-income, high-burdened, and resource-constrained rural area settings.
In conclusion, the present study identified a five-variable signature to distinguish AFB- IGRA+ TB from PN patients. A risk model was built to differentiate AFB- IGRA+ TB from PN, and was validated in an external independent cohort, which could be applied in low-income and resource-constrained rural area settings.
Acknowledgements
Not applicable.
Funding
Funding: The present study was supported by the Natural Science Foundation of Jiangxi Province (grant no. 20202BAB206059) and the Science and Technology Fund of Guangdong Province (grant no. sgybey01).
Availability of data and materials
The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.
Authors' contributions
DX, JZ, FX, QY, KH, WX, HZo and HZh contributed to the study conception and design. Primary clinical case information, data collection, and analysis were performed by DX, JZ, QY and FX. The first draft of the manuscript was written by DX, FX, and HZh. KH, WX and HZo conducted the literature search, as well as the screening and quality assessment of the clinical data. DX and HZh confirm the authenticity of all the raw data. All authors contributed to this manuscript and have consented to its submission. All authors read and approved the final manuscript.
Ethics approval and consent to participate
The present study was approved by the Ethics Committee and the Institutional Review Board of Ganzhou Fifth People's Hospital (registration no. 2020-10). Written informed consent was waived by the Ethics Committee as this was an observational and retrospective analysis study.
Patient consent for publication
Not applicable.
Competing interests
The authors declare that they have no competing interests.
References
|
WHO: Global Tuberculosis Report 2020. World Health Organization WHO, Geneva, 2020. | |
|
Kohli M, Schiller I, Dendukuri N, Yao M, Dheda K, Denkinger CM, Schumacher SG and Steingart KR: Xpert MTB/RIF Ultra and Xpert MTB/RIF assays for extrapulmonary tuberculosis and rifampicin resistance in adults. Cochrane Database Syst Rev. 1(CD012768)2021.PubMed/NCBI View Article : Google Scholar | |
|
Trébucq A, Enarson DA, Chiang CY, Van Deun A, Harries AD, Boillot F, Detjen A, Fujiwara PI, Graham SM, Monedero I, et al: Xpert® MTB/RIF for national tuberculosis programmes in low-income countries: When, where and how? Int J Tuberc Lung Dis. 15:1567–1572. 2011.PubMed/NCBI View Article : Google Scholar | |
|
Lim WS: From latent to active TB: Are IGRAs of any use? Thorax. 71:585–586. 2016.PubMed/NCBI View Article : Google Scholar | |
|
Auguste P, Tsertsvadze A, Pink J, Court R, McCarthy N, Sutcliffe P and Clarke A: Comparing interferon-gamma release assays with tuberculin skin test for identifying latent tuberculosis infection that progresses to active tuberculosis: Systematic review and meta-analysis. BMC Infect Dis. 17(200)2017.PubMed/NCBI View Article : Google Scholar | |
|
Dheda K, Makambwa E and Esmail A: The great masquerader: Tuberculosis presenting as community-acquired pneumonia. Semin Respir Crit Care Med. 41:592–604. 2020.PubMed/NCBI View Article : Google Scholar | |
|
Vessière A, Font H, Gabillard D, Adonis-Koffi L, Borand L, Chabala C, Khosa C, Mavale S, Moh R, Mulenga V, et al: Impact of systematic early tuberculosis detection using Xpert MTB/RIF Ultra in children with severe pneumonia in high tuberculosis burden countries (TB-Speed pneumonia): A stepped wedge cluster randomized trial. BMC Pediatr. 21(136)2021.PubMed/NCBI View Article : Google Scholar | |
|
Grossman RF, Hsueh PR, Gillespie SH and Blasi F: Community-acquired pneumonia and tuberculosis: Differential diagnosis and the use of fluoroquinolones. Int J Infect Dis. 18:14–21. 2014.PubMed/NCBI View Article : Google Scholar | |
|
Williams DJ, Creech CB, Walter EB, Martin JM, Gerber JS, Newland JG, Howard L, Hofto ME, Staat MA, Oler RE, et al: Short-vs Standard-course outpatient antibiotic therapy for community-acquired pneumonia in children: The SCOUT-CAP randomized clinical trial. JAMA Pediatr. 176:253–261. 2022.PubMed/NCBI View Article : Google Scholar | |
|
Hopstaken RM, Muris JW, Knottnerus JA, Kester AD, Rinkens PE and Dinant GJ: Contributions of symptoms, signs, erythrocyte sedimentation rate, and C-reactive protein to a diagnosis of pneumonia in acute lower respiratory tract infection. Br J Gen Pract. 53:358–364. 2003.PubMed/NCBI | |
|
Peters JS, McIvor A, Papadopoulos AO, Masangana T, Gordhan BG, Waja Z, Otwombe K, Letutu M, Kamariza M, Sterling TR, et al: Differentially culturable tubercle bacteria as a measure of tuberculosis treatment response. Front Cell Infect Microbiol. 12(1064148)2023.PubMed/NCBI View Article : Google Scholar | |
|
Blauenfeldt T, Villar-Hernández R, García-García E, Latorre I, Holm LL, Muriel-Moreno B, De Souza-Galvão ML, Millet JP, Sabriá F, Sánchez-Montalva A, et al: Diagnostic accuracy of interferon gamma-induced protein 10 mRNA release assay for tuberculosis. J Clin Microbiol. 58:e00848–20. 2020.PubMed/NCBI View Article : Google Scholar | |
|
Qiu X, Wang H, Tang Y, Su X, Ge L, Qu Y and Mu D: Is interleukin-2 an optimal marker for diagnosing tuberculosis infection? A systematic review and meta-analysis. Ann Med. 52:376–385. 2020.PubMed/NCBI View Article : Google Scholar | |
|
Ahmad R, Xie L, Pyle M, Suarez MF, Broger T, Steinberg D, Ame SM, Lucero MG, Szucs MJ, MacMullan M, et al: A rapid triage test for active pulmonary tuberculosis in adult patients with persistent cough. Sci Transl Med. 11(eaaz9925)2019.PubMed/NCBI View Article : Google Scholar | |
|
Yoon C, Semitala FC, Atuhumuza E, Katende J, Mwebe S, Asege L, Armstrong DT, Andama AO, Dowdy DW, Davis JL, et al: Point-of-care C-reactive protein-based tuberculosis screening for people living with HIV: A diagnostic accuracy study. Lancet Infect Dis. 17:1285–1292. 2017.PubMed/NCBI View Article : Google Scholar | |
|
Smyth GK: Linear models and empirical bayes methods for assessing differential expression in microarray experiments. Stat Appl Genet Mol Biol. 3(Article3)2004.PubMed/NCBI View Article : Google Scholar | |
|
Smyth GK: limma: Linear Models for Microarray Data. In: Bioinformatics and Computational Biology Solutions Using R and Bioconductor. Gentleman R., Carey V.J., Huber W., Irizarry R.A., Dudoit S (eds). Springer, New York, NY, pp 397-420 (, 2005). | |
|
Lim WK and Lim AW: A Comparison Of Usual t-Test Statistic and Modified t-Test Statistics on Skewed Distribution Functions. J modern applied statistical methods: JMASM. 15:67–89. 2016. | |
|
Capanu M and Seshan VE: False discovery rates for rare variants from sequenced data. Genet Epidemiol. 39:65–76. 2015.PubMed/NCBI View Article : Google Scholar | |
|
Tang LQ, Li CF, Li J, Chen WH, Chen QY, Yuan LX, Lai XP, He Y, Xu YX, Hu DP, et al: Establishment and validation of prognostic nomograms for endemic nasopharyngeal carcinoma. J Natl Cancer Inst. 108(djv291)2015.PubMed/NCBI View Article : Google Scholar | |
|
Jang JY, Park T, Lee S, Kim Y, Lee SY, Kim SW, Kim SC, Song KB, Yamamoto M, Hatori T, et al: Proposed nomogram predicting the individual risk of malignancy in the patients with branch duct type intraductal papillary mucinous neoplasms of the pancreas. Ann Surg. 266:1062–1068. 2017.PubMed/NCBI View Article : Google Scholar | |
|
Friedman J, Hastie T and Tibshirani R: Regularization paths for generalized linear models via coordinate descent. J Stat Softw. 33:1–22. 2010.PubMed/NCBI | |
|
Jalali A, Alvarez-Iglesias A, Roshan D and Newell J: Visualising statistical models using dynamic nomograms. PLoS One. 14(e0225253)2019.PubMed/NCBI View Article : Google Scholar | |
|
Balachandran VP, Gonen M, Smith JJ and DeMatteo RP: Nomograms in oncology: More than meets the eye. Lancet Oncol. 16:e173–e180. 2015.PubMed/NCBI View Article : Google Scholar | |
|
Gil-Santana L, Cruz LAB, Arriaga MB, Miranda PFC, Fukutani KF, Silveira-Mattos PS, Silva EC, Oliveira MG, Mesquita EDD, Rauwerdink A, et al: Tuberculosis-associated anemia is linked to a distinct inflammatory profile that persists after initiation of antitubercular therapy. Sci Rep. 9(1381)2019.PubMed/NCBI View Article : Google Scholar | |
|
Аbdullaev RY, Komissarova OG, Chumakova ES and Odinet VS: Level of uric acid in blood serum of new pulmonary tuberculosis patients with multiple drug resistance. Tuberculosis and Lung Diseases. 95:31–36. 2017. | |
|
Bisaso KR, Owen JS, Ojara FW, Namuwenge PM, Mugisha A, Mbuagbaw L, Luboobi LS and Mukonzo JK: Characterizing plasma albumin concentration changes in TB/HIV patients on anti retroviral and anti-tuberculosis therapy. In Silico Pharmacol. 2(3)2014.PubMed/NCBI View Article : Google Scholar | |
|
Rohini K, Surekha Bhat M, Srikumar PS and Mahesh Kumar A: Assessment of hematological parameters in pulmonary tuberculosis patients. Indian J Clin Biochem. 31:332–335. 2016.PubMed/NCBI View Article : Google Scholar | |
|
Carole C, Kokhreidze E, Tukvadze N, Banu S, Uddin MKM, Biswas S, Russomando G, Acosta CCD, Arenas R, Ranaivomanana PP, et al: Association of baseline white blood cell counts with tuberculosis treatment outcome: A prospective multicentered cohort study. Int J Infect Dis. 100:199–206. 2020.PubMed/NCBI View Article : Google Scholar | |
|
Yu YH, Liao CC, Hsu WH, Chen HJ, Liao WC, Muo CH, Sung FC and Chen CY: Increased lung cancer risk among patients with pulmonary tuberculosis: A population cohort study. J Thorac Oncol. 6:32–37. 2011.PubMed/NCBI View Article : Google Scholar | |
|
Wu CY, Hu HY, Pu CY, Huang N, Shen HC, Li CP and Chou YJ: Pulmonary tuberculosis increases the risk of lung cancer: A population-based cohort study. Cancer. 117:618–624. 2011.PubMed/NCBI View Article : Google Scholar | |
|
Brenner DR, McLaughlin JR and Hung RJ: Previous lung diseases and lung cancer risk: A systematic review and meta-analysis. PLoS One. 6(e17479)2011.PubMed/NCBI View Article : Google Scholar |
