Abstract
Aim: To explore the association between COX-2 polymorphisms and non-small cell lung cancer (NSCLC) susceptibility. Methods: We collected fasting peripheral venous blood from 60 cases with NSCLC and 62 healthy controls through physical examinations, and applied PCR-RFLP to analyze COX-2 polymorphisms of two groups. Results: With respect to detecting COX-2 rs689466 and rs5275 polymorphisms, the distribution frequency of mutant genotype AA of COX-2 rs689466 in case group was higher than that in control group, which possessed significant difference between two groups (P < 0.05). Carriers with AA genotype were 4.05 times at risk of NSCLC than those with GG genotype (P = 0.04, OR=4.05, 95% CI = 1.14-14.43). The distribution of mutant genotype CC of COX-2 rs5275 was different between two groups, and carriers with genotype CC were at 5.70 times higher risk of NSCLC than those with genotype TT. After corrected by sex, gender, smoking and drinking factors, AA genotype of COX-2 rs689466 and CC genotype of COX-2 rs5275 still contributed to increased risk of NSCLC (OR=4.22, 95% CI=1.10-16.17, OR=6.95, 95% CI=1.27-38.11). After analyzed of linkage disequilibrium (LD) and haplotypes of alleles in two SNPs, the distribution frequency of A-C haplotype in case group was higher than that in control group, with significant difference between two groups (P < 0.05). After corrected by sex, gender, smoking and drinking factors, statistical difference was still found in the total distribution of A-C haplotype between two groups (P = 0.03, OR=6.11, 95% CI=1.16-32.2). Conclusions: COX-2 rs689466 and rs5275 polymorphisms may be related to NSCLC susceptibility. And A-C haplotype might be a susceptibility haplotype for NSCLC.
Keywords: COX-2, non-small cell lung cancer, polymorphism
Introduction
Lung cancer caused by various factors is one of the common malignant tumors in China. With complex biological characteristics and high degree of malignancy, lung cancer has extremely difficult diagnosis at early stage and limited efficacy of the treatment. With increasing cases each year, lung cancer has become one of the major diseases seriously harmful to health and lives of residents in China [1-3]. Based on the pathology, lung cancer can be divided into two types as small cell lung cancer and NSCLC, and patients with NSCLC account for 80% of the total lung cancer cases. The clinical common NSCLC is shown in four types, including squamous cell carcinoma, adenocarcinoma, bronchioloalveolar carcinoma and undifferentiated carcinoma [4,5]. The occurrence of lung cancer is closely related to smoking, air pollution and other environmental factors.
COX is an important rate-limiting enzyme in the process of synthetic prostaglandins (PGs) of arachidonic acid metabolism. So far, there are at least two kinds of enzymes, namely COX-1 and COX-2. As a structural enzyme, COX-1 shows stable expression in most of normal tissues influencing PG synthesis, and further maintains normal physiological functions [6]. As an inducible enzyme, COX-2 is rarely expressed in most of normal tissues. However, when cells are stimulated by various factors such as oncogenes, cytokines, growth factors, endotoxin and tumor promoters, COX-2 enzyme begins to express in most of normal tissues and participates in physiological and pathological processes, such as inflammation or tumors [7]. Furthermore, several studies have testified that overexpression of COX-2 enzyme is associated with metastasis of malignant tumors, and plays an important role in each stage of tumor metastasis, including decreasing cell adhesion by changing margination factors on cell surface and extracellular matrix and further promoting tumor neovascularization. The expression of COX-2 gene is regulated by signaling pathway of mitogen activated protein kinase (MAPK) [8,9]. After some stimulus signals combine with epidermal growth factor receptor (EGFR), a series of phosphorylation cascade reactions occur in the cells, consequently prompting large-scale transcription of COX-2 gene. Afterwards, COX-2 catalyzes and generates plenty of PGE2 which can activate EGFR as a feedback. Therefore, the cycle enhances signaling pathway of MAPK and promotes cell proliferation [10,11]. Our study was to evaluate the association between COX-2 polymorphisms and NSCLC susceptibility by analyzing COX-2 rs689466 and rs5275 polymorphisms.
Materials and methods
Study subjects
We selected 60 patients (42 cases of males and 18 cases of females) with NSCLC diagnosed by surgery and pathology in Department of Thoracic Surgery in the Fourth Hospital of Hebei Medical University from 2009 to 2011. Additionally, mean age of the cases was ranged from 52.10 to 72.82. Meanwhile, 62 healthy adults (mean age: 43.68-68.62) including 37 males and 25 females passed through physical examinations in the same hospital were selected as control group. The subjects with tumors family history and biochemical abnormalities were excluded from the study.
Extraction of genome DNA
A total of 2 ml of fasting venous blood was extracted from each subject in two groups and anticoagulated using EDTA. Then, genome DNA was isolated from leukocytes of peripheral blood according to the instructions of AxyPrep. The genome DNA was stored at -20°C in a refrigerator.
Reaction system of PCR
We detected COX-2 rs689466 and rs5275 polymorphisms by means of PCR-RFLP. The primers sequences of PCR reaction are shown in Table 1. PCR reaction system included 1 µL DNA template, 1 µL forward primer and 1µL reverse primer, 25 µL Master Mix and 21 µL ddH2O. The reaction of PCR was proceeded in PCR instrument at 98°C predegeneration for 2 min, 98°C degeneration for 10 s, 57-60°C reannealing for 10 s, 68°C extension for 30 s (36 cycles), finally conducted at 68°C repair extension for 5 min. The PCR products were detected with agarose gel electrophoresis. 10 µL HhaI of restriction enzyme was added to PCR amplification products, which was incubated at 37°C for 2 h. We uniformly mixed 100 ml 0.5 × TBE with 2.0 g agarose, heated it to boil and then added 4 µL ethidium bromide as developer, followed by cooling the mixture with flat plates. The mixture of 6 µL PCR products and 2 µL bromphenol blue was spotted in sample application holes of electrophoresis tank. Electrophoresis was conducted for 1 hour at 85 V. 100 bp Marker molecular weight of DNA was as the standard. Then we recorded and analyzed COX-2 polymorphisms after photographing products of PCR reaction under the ultraviolet lamp.
Table 1.
Primers sequences
| SNPs | Primers | Sequences |
|---|---|---|
| rs5275 | Forward | 5’-GTTTGAAATTTTAAAGTACTTTGAT-3’ |
| Reverse | 5’-TTTCAAATTATTGTTTCATTGC-3’ | |
| rs689466 | Forward | 5’-CCC TGA GCA CTA CCC ATGAT-3’ |
| Reverse | 5’-GCC CTT CAT AGG AGATAC TGG-3’ |
Genotyping
In terms of COX-2 rs689466, the fragment of PCR products was 273 bp. The PCR amplification products digested by restriction enzyme PvuII were divided into three genotypes with digestion sites, including GG (200 bp, 73 bp), GA (273 bp, 200 bp and 73 bp) and AA (273 bp). With respect to COX-2 rs5257, the PCR amplification products digested by restriction enzyme PvuII were divided into TT, TC and CC genotypes.
Statistical analysis
The results were represented by mean ± standard deviation (x̅ ± s) or rate (%). Statistical analysis was performed using SPSS 18.0. X2 test was applied to compare differences in age, gender, smoking, drinking, and proceed statistical analyses of distribution frequencies of genotypes and alleles between two groups. Additionally, HWE was checked by X2 test. Odds ratio (OR) with 95% confidence interval (CI) was calculated by unconditional logistic regression. The difference was statistically significant when P < 0.05.
Results
General characteristics of study subjects
This study was consisted of 60 cases and 62 controls. As displayed in Table 2, there were no significant differences in gender, age and drinking between two groups (P > 0.05). However, the difference of smoking status in two groups showed statistical significance (P < 0.05).
Table 2.
General characteristics in case and control group
| General characteristics | Case (n=60) | Control (n=62) | P | OR (95% CI) |
|---|---|---|---|---|
| Gender | ||||
| Male | 42 (70) | 37 (59.6) | 0.260 | 1.577 (0.745-3.337) |
| Female | 18 (30) | 25 (40.4) | ||
| Age | ||||
| ≤50 | 14 (23.3) | 21 (33.9) | 0.233 | 0.594 (0.268-1.318) |
| >50 | 46 (76.7) | 41 (66.1) | ||
| Smoking status | ||||
| Smoking | 40 (66.7) | 30 (48.4) | 0.046 | 2.133 (1.026-4.437) |
| Not smoking | 20 (33.3) | 32 (51.6) | ||
| Drinking status | ||||
| Drinking | 45 (75) | 42 (67.7) | 0.427 | 1.429 (0.648-3.150) |
| Not drinking | 15 (25) | 20 (32.3) |
Association between COX-2 polymorphisms and NSCLC risk
All the genotypes of COX-2 polymorphisms in two groups were in accordance with HWE (P > 0.05), suggesting these samples were representative in the total population (Table 3). The distribution frequency of mutant genotype AA of COX-2 rs689466 in case group was higher than that of control group, which presented significant difference between two groups (P < 0.05). Carriers with genotype AA were at 3.05 times higher risk of NSCLC than those with genotype GG (P = 0.04, OR=4.05, 95% CI=1.14-14.43). The distribution of mutant genotype CC of COX-2 rs5275 was different in two groups, and carriers with genotype CC were at 6.70 times the risk of NSCLC than those with genotype TT (OR=6.70, 95% CI=1.35-33.31). After corrected by sex, gender, smoking and drinking factors, AA genotype of COX-2 rs689466 and CC genotype of COX-2 rs5275 still resulted in increased risk of NSCLC (OR=4.22, 95% CI=1.10-16.17; OR=6.95, 95% CI=1.27-38.11).
Table 3.
Distributional comparisons of genotypes and alleles of COX-2 rs689466 and rs5275 polymorphisms
| Genotype/Allele | Case n= 60 (%) | Control n= 62 (%) | P | OR (95% CI) | P a | ORa (95% CI) |
|---|---|---|---|---|---|---|
| rs689466 | ||||||
| GG | 20 (33.3) | 27 (43.5) | 1.00 | 1.00 | ||
| GA | 28 (46.7) | 31 (50) | 0.70 | 1.22 (0.56-2.64) | 0.62 | 1.23 (0.54-2.79) |
| AA | 12 (20) | 4 (6.5) | 0.04 | 4.05 (1.14-14.43) | 0.04 | 4.22 (1.10-16.17) |
| G | 68 (56.7) | 85 (68.5) | 1.00 | |||
| A | 52 (43.3) | 39 (31.5) | 0.06 | 1.67 (0.99-2.81) | 0.13 | 1.3 (0.83-4.45) |
| rs5275 | ||||||
| TT | 23 (38.3) | 28 (45.2) | 1.00 | 1.00 | ||
| TC | 26 (43.3) | 32 (51.6) | 1.00 | 0.99 (0.46-2.11) | 0.69 | 0.85 (0.38-1.91) |
| CC | 11 (18.4) | 2 (3.2) | 0.01 | 6.70 (1.35-33.31) | 0.03 | 6.95 (1.27-38.11) |
| T | 72 (60) | 88 (71) | 1.00 | |||
| C | 48 (40) | 36 (29) | 0.08 | 1.63 (0.96-2.78) | 0.52 | 0.31 (0.58-2.95) |
Indicates data after corrected by sex, gender, smoking and drinking.
Analyses of LD and haplotypes
LD refers to obvious difference between frequency of two alleles at different sites occurring in the same haplotype and the expected random frequency. The value of Lewontin’ s D’ is one of the standards in judging LD. SHEsis online software was applied to analyze relevance of alleles in all the study subjects, with significance of D’> 0.7. Our study showed that D’ values of alleles between two SNPs were higher than 0.7. Therefore, obvious LD existed between COX-2 rs689466 and rs5275, which constituted a haplotype field.
After analyzing LD and haplotypes of alleles in two SNPs, the distribution frequency of A-C haplotype in case group was higher than that in control group, with significant difference between two groups (P < 0.05). After corrected by sex, gender, smoking and drinking factors, statistical difference was still found in the total distribution of A-C haplotype between two groups (P = 0.03, OR=6.11, 95% CI=1.16-32.2). However, the distribution frequencies of other three haplotypes between two groups showed no statistical significance (P > 0.05). Thus, A-C haplotype might be a susceptibility haplotype for NSCLC, as shown in Table 4.
Table 4.
Analyses of LD and haplotypes of alleles in COX-2 rs689466 and rs5275 polymorphisms
| Haplotype Locus1-2 Locus 2 | Case 2n=120 (%) | Control 2n=124 (%) | P | OR (95% CI) | P a | ORa (95% CI) |
|---|---|---|---|---|---|---|
| G-T | 40 (33.3) | 53 (42.7) | - | 1.00 | - | 1.00 |
| G-C | 28 (23.3) | 32 (25.8) | 0.62 | 1.18 (0.62-2.27) | 0.78 | 1.15 (0.43-3.06) |
| A-T | 32 (26.7) | 35 (28.2) | 0.53 | 1.23 (0.66-2.32) | 0.75 | 1.16 (0.46-2.95) |
| A-C | 20 (16.4) | 4 (3.3) | 0.00 | 9.00 (2.5-32.4) | 0.03 | 6.11 (1.16-32.2) |
| Total | 120 (100.0) | 124 (100.0) | - | - |
Indicates data after corrected by sex, gender, smoking and drinking.
Discussion
The human COX-2 gene with the length of 8 kb is located in 1q25.2-25.3 and consisted of forward region of transcriptional start site at 5’ end, protein coding region including 10 exons and 9 introns, and non-coding region at 3’ end. In addition, the 5’ flanking region contains some possible transcriptional regulatory sequences, including TATA box sequence at 25 bp site in forward transcription region, one motif of CCAAT enhancer binding protein (C/EBP), two AP-2 sites, three SP1 sites, two NF-kβ sites, one motif of cAMP response elements and one Est-1 site [12,13]. Stimulated cells are transcribed at 5’ end after a series of signal transduction, and further induce the expression of COX-2. Existing data confirm that the genetic variations of above regulatory elements play important roles in transcriptional control and expression of COX-2 [14]. PGs catalyzed by COX-2 are not only involved in inflammation, differentiation and conduction of normal keratinocytes, neural conduction and body heat regulation, but also closely associated with tumors in the process of occurrence, development, and metastasis [15].
The COX-2 rs5257 is located at the reverse site of terminator in 3’ non-coding region which contains abundant AU sequences. These AU sequences can achieve regulation of post-transcriptional level and mediate the degradation of mRNA by interactions with trans-acting RNA conjugated protein or miRNAs. Certainly, different intensity of combination can result in distinct mRNA degradation, ultimately leading to different susceptibility to tumors [16,17]. Thus, COX-2 rs5277 polymorphisms can influence genetic expression levels and change individual susceptibility to tumors. COX-2 rs689466 is located in promoter region of COX-2 gene which can activate special transcription factors and further regulate COX-2 expression with various enhancers and transcriptional regulatory elements. And polymorphisms in promoter region are crucial genetic factors influencing genetic transcription and regulation [18]. Zhang et al., demonstrated COX-2 rs689466 polymorphisms which produced a combined site of c-MYB transcription factor, enhanced the transcriptional activity of COX-2 gene and was in association with susceptibility to cancer [19,20].
Lung cancer is one of the diseases with increasing high morbidity and mortality in the world at present. The incidence of lung cancer is considered to be associated with environmental and genetic factors. Based on the study of Campa et al., in European Caucasians, compared with TT genotype of COX-2 rs5275, TC and CC genotypes contributed to significantly increased risk of NSCLC [21]. Subsequently, Campa et al., carried out a large-scale study and indicated no relationship between COX-2 rs5257 polymorphisms and lung cancer susceptibility [22]. However, Hu et al., pointed out CT and CC genotypes could decrease risk of lung cancer, and showed a more significant correlation of COX-2 rs5275 polymorphisms with NSCLC risk in stratified analysis in Chinese population, with TT genotype of COX-2 rs5257 as a control [23].
Few studies reported the association between COX-2 polymorphisms and NSCLC susceptibility. In our study, the distribution frequency of mutant genotype AA of COX-2 rs689466 in case group was higher than that in control group, with significant difference between two groups. Carriers with AA genotype were 4.05 times at risk of NSCLC than those with GG genotype. The distribution of mutant genotype CC of COX-2 rs5275 was different in two groups, and carriers with CC genotype were 5.70 times higher risk of NSCLC than those with genotype TT. Thus, AA genotype in COX-2 rs689466 and CC genotype in COX-2 rs5275 might be susceptibility genotypes. After corrected by sex, gender, smoking and drinking factors, AA genotype of COX-2 rs689466 and CC genotype of COX-2 rs5275 still contributed to increased risk of NSCLC. Based on analyses of haplotypes in two SNPs, A-C haplotype in two SNPs could increase risk of NSCLC. In conclusion, mutant genotypes of COX-2 rs689466 and rs5275 polymorphisms might be associated with NSCLC susceptibility and COX-2 gene might be a susceptibility gene.
The occurrence and development of lung cancer are influenced by various factors, and the pathogenesis is relatively complex. With certain limitations of sample size and individual differences in samples, large-scale studies are needed to further evaluate the association between COX-2 polymorphisms and NSCLC susceptibility.
Disclosure of conflict of interest
None.
References
- 1.Ulukaya E, Acilan C, Yilmaz M, Yilmaztepe-Oral A, Ari F, Zik B, Ursavas A, Tokullugil AH. sFas levels increase in response to cisplatin-based chemotherapy in lung cancer patients. Cell Biochem Funct. 2010;28:565–570. doi: 10.1002/cbf.1689. [DOI] [PubMed] [Google Scholar]
- 2.Nakamoto Y, Ogata M, Yamamoto M. [A long-term survivor treated with S-1 and gemcitabine for recurrence following an operation for advanced gall bladder cancer] . Gan To Kagaku Ryoho. 2010;37:1979–1981. [PubMed] [Google Scholar]
- 3.Ikeda S, Takabe K, Inagaki M, Suzuki K. Expression of phosphorylated Akt in patients with small cell carcinoma of the lung indicates good prognosis. Pathol Int. 2010;60:714–719. doi: 10.1111/j.1440-1827.2010.02594.x. [DOI] [PubMed] [Google Scholar]
- 4.Takahashi H, Kurishima K, Ishikawa H, Kagohashi K, Kawaguchi M, Satoh H. Optimal cutoff points of CYFRA21-1 for survival prediction in non-small cell lung cancer patients based on running statistical analysis. Anticancer Res. 2010;30:3833–3837. [PubMed] [Google Scholar]
- 5.Batus M, Fidler MJ, Bonomi PD. Primary and secondary therapeutic strategies for EGF receptor pathway inhibition in non-small-cell lung cancer. Expert Rev Anticancer Ther. 2010;10:1589–1599. doi: 10.1586/era.10.104. [DOI] [PubMed] [Google Scholar]
- 6.Coffey RJ, Hawkey CJ, Damstrup L, Graves-Deal R, Daniel VC, Dempsey PJ, Chinery R, Kirkland SC, DuBois RN, Jetton TL, Morrow JD. Epidermal growth factor receptor activation induces nuclear targeting of cyclooxygenase-2, basolateral release of prostaglandins, and mitogenesis in polarizing colon cancer cells. Proc Natl Acad Sci U S A. 1997;94:657–662. doi: 10.1073/pnas.94.2.657. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Tsujii M, Kawano S, Tsuji S, Sawaoka H, Hori M, DuBois RN. Cyclooxygenase regulates angiogenesis induced by colon cancer cells. Cell. 1998;93:705–716. doi: 10.1016/s0092-8674(00)81433-6. [DOI] [PubMed] [Google Scholar]
- 8.Dannenberg AJ, Altorki NK, Boyle JO, Dang C, Howe LR, Weksler BB, Subbaramaiah K. Cyclo-oxygenase 2: a pharmacological target for the prevention of cancer. Lancet Oncol. 2001;2:544–551. doi: 10.1016/S1470-2045(01)00488-0. [DOI] [PubMed] [Google Scholar]
- 9.Howe LR, Dannenberg AJ. A role for cyclooxygenase-2 inhibitors in the prevention and treatment of cancer. Semin Oncol. 2002;29:111–119. doi: 10.1053/sonc.2002.34063. [DOI] [PubMed] [Google Scholar]
- 10.Eberhart CE, Coffey RJ, Radhika A, Giardiello FM, Ferrenbach S, DuBois RN. Up-regulation of cyclooxygenase 2 gene expression in human colorectal adenomas and adenocarcinomas. Gastroenterology. 1994;107:1183–1188. doi: 10.1016/0016-5085(94)90246-1. [DOI] [PubMed] [Google Scholar]
- 11.Wolff H, Saukkonen K, Anttila S, Karjalainen A, Vainio H, Ristimaki A. Expression of cyclooxygenase-2 in human lung carcinoma. Cancer Res. 1998;58:4997–5001. [PubMed] [Google Scholar]
- 12.Appleby SB, Ristimaki A, Neilson K, Narko K, Hla T. Structure of the human cyclo-oxygenase-2 gene. Biochem J. 1994;302:723–727. doi: 10.1042/bj3020723. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Dixon DA. Regulation of COX-2 expression in human cancers. Prog Exp Tumor Res. 2003;37:52–71. doi: 10.1159/000071363. [DOI] [PubMed] [Google Scholar]
- 14.Masferrer JL, Leahy KM, Koki AT, Zweifel BS, Settle SL, Woerner BM, Edwards DA, Flickinger AG, Moore RJ, Seibert K. Antiangiogenic and antitumor activities of cyclooxygenase-2 inhibitors. Cancer Res. 2000;60:1306–1311. [PubMed] [Google Scholar]
- 15.Akre K, Ekstrom AM, Signorello LB, Hansson LE, Nyren O. Aspirin and risk for gastric cancer: a population-based case-control study in Sweden. Br J Cancer. 2001;84:965–968. doi: 10.1054/bjoc.2001.1702. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Young LE, Dixon DA. Posttranscriptional Regulation of Cyclooxygenase 2 Expression in Colorectal Cancer. Curr Colorectal Cancer Rep. 2010;6:60–67. doi: 10.1007/s11888-010-0044-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Shankar A, Yuan JM, Koh WP, Lee HP, Yu MC. Morbidity and mortality in relation to smoking among women and men of Chinese ethnicity: the Singapore Chinese Health Study. Eur J Cancer. 2008;44:100–109. doi: 10.1016/j.ejca.2007.10.015. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Zhao D, Xu D, Zhang X, Wang L, Tan W, Guo Y, Yu D, Li H, Zhao P, Lin D. Interaction of cyclooxygenase-2 variants and smoking in pancreatic cancer: a possible role of nucleophosmin. Gastroenterology. 2009;136:1659–1668. doi: 10.1053/j.gastro.2009.01.071. [DOI] [PubMed] [Google Scholar]
- 19.Zhang X, Miao X, Tan W, Ning B, Liu Z, Hong Y, Song W, Guo Y, Shen Y, Qiang B, Kadlubar FF, Lin D. Identification of functional genetic variants in cyclooxygenase-2 and their association with risk of esophageal cancer. Gastroenterology. 2005;129:565–576. doi: 10.1016/j.gastro.2005.05.003. [DOI] [PubMed] [Google Scholar]
- 20.Altorki N. COX-2: a target for prevention and treatment of esophageal cancer. J Surg Res. 2004;117:114–120. doi: 10.1016/j.jss.2003.12.005. [DOI] [PubMed] [Google Scholar]
- 21.Campa D, Zienolddiny S, Maggini V, Skaug V, Haugen A, Canzian F. Association of a common polymorphism in the cyclooxygenase 2 gene with risk of non-small cell lung cancer. Carcinogenesis. 2004;25:229–235. doi: 10.1093/carcin/bgh008. [DOI] [PubMed] [Google Scholar]
- 22.Campa D, Hung RJ, Mates D, Zaridze D, Szeszenia-Dabrowska N, Rudnai P, Lissowska J, Fabianova E, Bencko V, Foretova L, Janout V, Boffetta P, Brennan P, Canzian F. Lack of association between polymorphisms in inflammatory genes and lung cancer risk. Cancer Epidemiol Biomarkers Prev. 2005;14:538–539. doi: 10.1158/1055-9965.EPI-04-0513. [DOI] [PubMed] [Google Scholar]
- 23.Hu Z, Miao X, Ma H, Wang X, Tan W, Wei Q, Lin D, Shen H. A common polymorphism in the 3’UTR of cyclooxygenase 2/prostaglandin synthase 2 gene and risk of lung cancer in a Chinese population. Lung Cancer. 2005;48:11–17. doi: 10.1016/j.lungcan.2004.09.004. [DOI] [PubMed] [Google Scholar]