| | Elevated risk of squamous-cell carcinoma of the lung in heavy smokers carrying the variant alleles of the TP53 Arg72Pro and p21 Ser31Arg polymorphismsReceived 11 July 2006; received in revised form 7 September 2006; accepted 14 September 2006. Summary Alterations in cell cycle regulation and apoptosis leading to malignant transformation could be caused by common genetic variants in tumor suppressor genes. The effects of the TP53 polymorphism Arg72Pro on lung cancer risk have been investigated in numerous studies with, however, conflicting results. In many studies, important risk modifiers such as smoking or tumor histology were not taken into account. We therefore investigated the combined effects of polymorphisms in TP53 (Arg72Pro) and p21/CDKN1A (Ser31Arg) and smoking on lung cancer risk. Our case–control study consisted of 405 patients with lung cancer, mainly squamous-cell carcinoma (185) and adenocarcinoma (177) and 404 unmatched tumor-free hospital controls. Multivariate regression analysis showed a moderate but statistically significant risk of lung cancer overall and especially of squamous-cell carcinoma (OR, 1.65; CI, 1.10–2.47) for TP53 72Pro allele carriers. The risk was markedly increased in heavy smokers (>20 pack-years) with squamous-cell carcinoma (OR, 2.80 in patients homozygous for 72Pro; CI, 1.19–6.58), but not in light smokers (≤20 pack-years). The results for the p21 Ser31Arg polymorphism suggested that 31Ser is a moderate-risk allele for squamous-cell carcinoma. Analysis of the combined effects of the two polymorphisms revealed a higher OR for TP53 72Pro carriers homozygous for p21 31Ser than for 72Pro carriers in general; this effect being most pronounced in heavy smokers with squamous-cell carcinoma (OR, 3.84; CI, 1.46–10.1). Our data indicate that the TP53 Arg72Pro polymorphism increases the risk for squamous-cell carcinoma mainly in heavy smokers. The observed interaction with smoking is biologically plausible as, for the 72Pro p53 variant, decreased apoptosis and extended G1 cell cycle arrest is reported after carcinogen exposure. Nevertheless, confirmation by further molecular and epidemiological studies is warranted. 1. Introduction  The tumor suppressor gene TP53 plays a central role in the response to DNA damage owing to its ability to activate DNA repair, arrest cell cycle progression or induce apoptosis [1], [2]. In regulating this response, TP53 efficiently controls genomic stability; it is thus an important factor in obviating malignant transformation of cells after carcinogenic exposure. The p53 protein is activated by and interacts with a large network of other proteins, but many of its effects are mediated by transcriptional activation of a 21-kDa protein, p21/cip1/waf1, which inhibits cyclin-dependent kinase activity and thereby cell division [3], [4]. Many tumors are characterized by early mutations in the TP53 gene, but approximately one-half of tobacco-related cancers, such as lung cancer, do not have mutations in this gene [5]. Thus, it has been postulated that genetic variation in TP53 due to several common polymorphisms could limit efficient growth control and lead to malignant transformation. One of these polymorphisms results in an amino acid exchange from Arg to Pro at codon 72 [6], [7], [8]. Several authors have suggested that this polymorphism affects cancer susceptibility, but the results with regard to lung cancer risk are inconsistent. In a meta-analysis [9] of 13 of the most comprehensive studies, the overall OR was 1.18 (95% CI, 0.99–1.41). This moderate effect might be partly due to the inconsistent results across studies and the lack of a clear consensus on which sequence variant confers an increased risk. Moreover, many of the studies had small sample sizes and limited epidemiological power, and therefore stratification for the main risk factor for lung cancer, exposure to tobacco smoke, could not be performed. Smoking must, however, be considered as a risk modifier, especially in studying TP53, which is involved in the cellular response to DNA damage. In addition, in most studies, no distinction was made between the histological subtypes of non-small-cell lung cancer, such as adenocarcinoma, squamous-cell carcinoma (SCC) and large-cell carcinoma. As adenocarcinoma and SCC occur at different frequencies in smokers and nonsmokers [10], [11] and probably have different causes [11], they should be analyzed separately [12], [13]. A critical downstream mediator of TP53 is the tumor suppressor gene p21, which is transcriptionally activated by p53 to induce cell cycle arrest and DNA repair [3]. One of the single-nucleotide polymorphisms in p21 results in an amino acid change in codon 31 from Ser to Arg. The polymorphism is located in a highly conserved region of p21 and is expected to affect its molecular function [14]. The potential role of the p21 polymorphism in the development of lung cancer has not yet been clarified [15], [16], [17]. To examine whether the differences in biochemical properties between the polymorphic variants affect lung cancer risk, we analyzed the codon 72 polymorphism in TP53 and the codon 31 polymorphism in p21 in a case–control lung cancer study including 405 patients with non-small-cell lung cancer and 404 randomly selected controls. To overcome some of the shortcomings depicted in the meta-analysis [9], we stratified our cohort according to tumor histology, i.e. the 177 cases of adenocarcinoma and the 185 cases of SCCs were evaluated separately. Furthermore, we collected data on tobacco smoking for cases and controls and stratified our analysis for heavy and light smokers. 2. Materials and methods 2.1. Study population All patients of European origin admitted to the Department of Surgery of the Thoraxklinik in Heidelberg between 1997 and 2000 were potential candidates as cases or controls for this study. The study was approved by the ethical committee of the University of Heidelberg (reference no. 201/98), and all the study participants gave informed consent. Cases (n =405) were patients diagnosed with primary lung cancer without prior malignant disease and without any other treatment before surgery and for whom both a histopathological diagnosis had been performed and genotyping on peripheral blood collected prior to surgery or any other treatment was successfully performed. Controls (n=404) were patients without prior or present malignant disease who were admitted to surgery or the department of internal medicine for other reasons, mainly for suffering from respiratory insufficiency, alveolitis, bronchitis, pneumonia, fibrosis, sarcoidosis, chronic obstructive pulmonary disease and emphysema. Controls were recruited at the same time and with the same criteria as cases but were not matched for age or gender. Recruitment and interviews of cases and controls were prior to surgery or other treatments. The data on histopathology defining cases and controls and also the histopathological subgroups of cases [SCC (n=185), adenocarcinomas (n=177), others (n=43, large cell, carcinoid, mixed cell, etc.)] were extracted from the pathologists records by one of the investigators. Those data were cross-checked with an independently maintained histopathological data base established at the Department of Surgery of the Thoraxklinik. In case of discrepancies an independent examiner rechecked the original pathologists’ records and came to a decision on best available medical evidence. The two histological subgroups of adenocarcinoma and SCC were evaluated by separate statistical analysis. The study participants completed a detailed questionnaire eliciting information on personal history, smoking habits (never smokers (less than 100 cigarettes per life), current smokers, or ex-smokers who quitted smoking more than 1 year before time of interview as well as number of pack-years smoked) and occupation. Therefore, all study participants had to be proficient in the German language. Occupational exposure status was established on the basis of the information given in the questionnaire on job title (e.g. welder, driver, mechanic, industrial worker, painter) or on exposure to known or suspected lung carcinogens including asbestos and mineral fibers, metals, all kinds of dust and fumes, cement, petroleum and coal products, diesel exhaust, solvents and pesticides [18]. The unexposed group consisted of persons who claimed to have had no exposure or having had jobs with no such exposure (e.g. office worker, teacher, and housewife). The demographic data for the cases and controls are shown in Table 1. Briefly, the median age of all cases was higher than that of controls (61 and 55 years, respectively) and there were significantly (p<0.0001) more males (78% versus 59%) among cases. As to be expected for a cohort of lung cancer cases, more smokers (90% versus 65%) and more persons possibly occupationally exposed (72% versus 56%) were found among cases than among controls. Prior to treatment or surgery, all participants gave 5ml of peripheral blood, which was collected in EDTA, and buffy coats were stored at −80°C. Subsequently, genomic DNA was isolated with a QIAamp DNA Blood Kit (Qiagen, Hilden, Germany). 2.2. Genotyping by fluorescence-based melting-curve analysis The polymorphisms detected by amplification with capillary PCR followed by melting-curve analysis with fluorescence-labeled hybridization probes in a LightCycler (Roche Diagnostics, Mannheim, Germany) were TP53 Arg72Pro (rs1042522; exchange of G and C in exon 4, nucleotide position 429, Genebank accession no. U94788) and CDKN1A/p21 Ser31Arg (rs1801270; exchange of C and A in nucleotide position 6829 of the p21 gene, Genebank accession no. AF497972). The oligonucleotides for analysis of the TP53 polymorphism were the PCR primers 5′-gat gct gtc ccc gga cga-3′OH and 5′-agg ggc cgc cgg tgt ag-3′OH, anchor 5′-cca gat gaa gct ccc aga atg cca gag gct-3′-FL and sensor 5′-LC Red640-tcc ccc cgt iic ccc tgc acc-3′P [19], [20]. dInosin bases were introduced to substitute for the guanine bases in order to lower the overall binding strength of the GC-rich target sequence and to achieve a higher temperature difference between the two variants, which were 61°C for C (Pro) and 56°C for G (Arg). The oligonucleotides for analysis of the p21 polymorphism were the PCR primers 5′-gg cgc cat gtc aga acc ggc-3′OH and 5′-cca gac agg tca gcc ctt gg-3′OH, sensor 5′-cat cac agt cgc gtc tca gc-3′-FL and anchor 5′-LC Red640-gct cgc tgt cca ctg ggc cga a-3′P. The PCR primers and probes were designed with the help of Tib Molbiol (Berlin, Germany). The sensor probe was designed to achieve a perfect match with either the wild type or the variant allele sequence. Thus, in the allele in which the sequence deviated from the sensor, a one-nucleotide mismatch between sensor and target DNA sequence was formed, which destabilized the hybrid, causing a melting-point shift. PCR was performed for both polymorphisms in 10-μl volumes in glass capillary tubes (Roche Diagnostics) with Qiagen reagents: 1× PCR buffer, 2.5mmol/l total MgCl2, 0.2mmol/l dNTPs, 0.1% bovine serum albumin in 2mmol/l Tris–HCl/2.5% glycerol, 1× Q solution, 0.1U Taq DNA polymerase, 0.5μmol/l of each primer, 0.15μmol/l of each probe and 10ng of DNA. The reaction conditions were as follows: initial denaturation at 95°C for 3min, then 40 cycles of denaturation at 95°C for 15s, annealing at 57°C (TP53) or 62°C (p21) for 10s and elongation at 72°C for 12s. Melting-curve analysis was performed with an initial denaturation step at 95°C, 15s at 40°C, slow heating to 75°C at a ramping rate of 0.2°C/s and continuous fluorescence detection. A negative control containing all the reagents but with water instead of the DNA template was included in each amplification set. The melting curves were evaluated by two independent observers who were unaware of the clinical diagnosis. In addition, 100 randomly selected samples were analyzed by conventional PCR-RFLP to verify the LightCycler results; 100% concordance was found. For each polymorphism, PCR fragments of the homozygous wild-type allele, the homozygous variant allele, and one heterozygous sample were sequenced. 2.3. Statistical analysis Each polymorphism was tested for deviation from Hardy–Weinberg equilibrium in cases and controls by the χ2 test of goodness of fit with one degree of freedom, with respect to the distribution of genotypes using the ‘ALLELE Procedure’ from SAS statistical evaluation system, Genetics 9.1. Lung cancer risk was analyzed for association with polymorphisms by multivariate unconditional logistic regression, as described elsewhere [21], [22]. In detail, the more frequent allele was used as the reference. ORs were adjusted for variables showing a significantly different distribution in cases and controls, i.e. age, gender, smoking, and occupation as follows: Age at diagnosis was stratified into four age groups of ≤50, 51–60, 61–70, and ≥71 years. Smoking was quantified by the amount of tobacco consumption, categorized into ≤20 and >20 pack-years (1 pack-year=20 cigarettes/day for 1 year). The cut point for pack-years corresponds to the mean value calculated for tobacco consumption of the control group. Exposure by occupation (see Section 2.1) was categorized as “no exposure” or “possible exposure”. Possible differences between light (≤20 pack-years) and heavy (>20 pack-years) smokers in the association between lung cancer and polymorphisms was assessed by introducing an interaction term (polymorphism×smoking) into the logistic regression model. ORs were examined for trends in the number of variant alleles by ordinal regression according to McCullagh [23]. Lung cancer risk was also analyzed for the combined effect of TP53 and p21 polymorphisms. As the rare variant p21 31Arg allele had a protective effect also in studies about other cancers [24], [25] we selected the 31Ser allele of p21 as the risk allele for this gene–gene interaction analysis. Calculations and statistical tests for crude ORs and ordinal regression were performed with the ADAM statistical analysis system of the Biostatistics Unit of the German Cancer Research Center. Multivariate logistic regression analysis and calculation of confidence intervals were performed with the SAS statistical evaluation system, release 8.1 (SAS Institute Inc., Cary, NC, USA). 3. Results The frequencies of the rare alleles were 0.23 for the 72Pro allele of TP53 and 0.07 for the 31Arg allele of p21 in the control group and 0.29 and 0.07 in cases. The allele frequencies for both polymorphisms were in line with those found for the control population in a German breast cancer study [26] and a cohort of German prostate cancer patients (unpublished data) as well as with those detected for white control populations in other studies [27], [28], [29], [30]. Genotype distribution was consistent with the Hardy–Weinberg equilibrium in cases and in controls for the TP53 polymorphism (p value controls=0.06; p value cases=0.18) and the p21 polymorphism (p values=0.93 and 0.58, respectively). The genotype distribution (see Table 2) was different between all cases and controls (p=0.04) and between SCC cases and controls for TP53 (p=0.004) but was similar in adenocarcinoma cases and controls and for all p21 genotypes. | | |  | Genotype | Casesa, n (%) | Controlsb, n (%) | Adjustedc OR | 95% CI | p value | Interaction polymorphism×smoking p value | |
|---|
| All persons | 405 (100) | 404 (100) | | | | | | | TP53, codon 72 | | | Arg/Arg | 209 (52) | 244 (60) | 1.0 | | | | | | Arg/Pro | 156 (38) | 131 (33) | 1.29 | 0.93–1.79 | 0.13 | 0.27 | | | Pro/Pro | 40 (10) | 29 (7) | 1.74 | 0.98–3.09 | 0.057 | 0.59 | | | Pro carriers | 196 (48) | 160 (40) | 1.37 | 1.00–1.86 | 0.049 | 0.27 | | | | | | p21, codon 31 | | | Ser/Ser | 349 (87) | 346 (86) | 1.0 | | | | | | Ser/Arg | 51 (12.5) | 54 (13) | 0.83 | 0.52–1.31 | 0.42 | 0.46 | | | Arg/Arg | 2 (0.5) | 3 (1) | 1.04 | 0.15–7.26 | 0.97 | 0.64 | | | Arg carriers | 53 (13) | 57 (14) | 0.84 | 0.53–1.31 | 0.43 | 0.41 | | | | | | SCC | 185 (100) | | | | | | | | TP53, codon 72 | | | Arg/Arg | 86 (46.5) | 244 (60) | 1.0 | | | | | | Arg/Pro | 76 (41) | 131 (33) | 1.54 | 1.00–2.38 | 0.049 | 0.43 | | | Pro/Pro | 23 (12.5) | 29 (7) | 2.13 | 1.05–4.33 | 0.037 | 0.25 | | | Pro carriers | 99 (53.5) | 160 (40) | 1.65 | 1.10–2.47 | 0.016 | 0.27 | | | | | | p21, codon 31 | | | Ser/Ser | 163 (89.5) | 346 (86) | 1.0 | | | | | | Ser/Arg | 18 (10) | 54 (13) | 0.59 | 0.31–1.13 | 0.11 | 0.42 | | | Arg/Arg | 1 (0.5) | 3 (1) | 1.28 | 0.08–19.4 | 0.86 | 0.99 | | | Arg carriers | 19 (10.5) | 57 (14) | 0.61 | 0.32–1.16 | 0.13 | 0.19 | | | | | | Adenocarcinoma | 177 (100) | | | | | | | | TP53, codon 72 | | | Arg/Arg | 100 (56) | 244 (60) | 1.0 | | | | | | Arg/Pro | 63 (36) | 131 (33) | 1.14 | 0.76–1.72 | 0.53 | 0.21 | | | Pro/Pro | 14 (8) | 29 (7) | 1.31 | 0.64–2.69 | 0.46 | 0.85 | | | Pro carriers | 77 (44) | 160 (40) | 1.17 | 0.80–1.72 | 0.42 | 0.26 | | | | | | p21, codon 31 | | | Ser/Ser | 148 (83.5) | 346 (86) | 1.0 | | | | | | Ser/Arg | 28 (16) | 54 (13) | 1.07 | 0.63–1.83 | 0.81 | 0.79 | | | Arg/Arg | 1 (0.5) | 3 (1) | 0.85 | 0.08–9.18 | 0.87 | 0.98 | | | Arg carriers | 29 (16.5) | 57 (14) | 1.06 | 0.63–1.79 | 0.83 | 0.90 | | | | |
| a No PCR products were obtained for three SCC (p21). bNo PCR products were obtained for one (p21) controls. cAdjusted for gender, age group, pack-year group and occupation; OR values with p<0.05 are shown in bold. |
ORs were calculated for both polymorphisms for the heterozygous and homozygous variants and for carriers of the variant allele combining heterozygous and homozygous variants. Adjustment of ORs for the possible confounders (i.e. gender, age, smoking, and exposure by occupation) resulted in only minor changes of the ORs. When all cases were evaluated (Table 2), the variant allele of the TP53 polymorphism, 72Pro, was associated with an increased risk for lung cancer, which was statistically significant for 72Pro allele carriers (OR=1.37; CI 1.001–1.86). The separate analysis of SCC and adenocarcinoma patients revealed a significantly increased OR for SCC patients with one or more Pro alleles (OR=1.65 in Pro allele carriers; CI, 1.10–2.47), whereas no significant increase in OR was found for adenocarcinoma patients (OR, Pro carriers=1.17; 0.80–1.72). For the p21 Ser31Arg polymorphism, no significant effect on lung cancer risk was found for all lung cancers or for the two histological subtypes; yet carriage of the 31Arg allele appeared to have a protective effect for all cases (OR, 0.84; CI, 0.53–1.31) and for SCC cases (0.61; CI, 0.32–1.16). To investigate modification of the effect of the TP53 polymorphism on lung cancer risk by smoking, the study population was subdivided into light or nonsmokers, with a cigarette consumption of up to 20 pack-years, and heavy smokers, with more than 20 pack-years (Table 3). The TP53 polymorphism had different effects on lung cancer risk in the two subgroups: heavy smokers had ORs of 1.55 for 72Pro carriers (CI, 1.04–2.33), whereas in light smokers, no effect on risk was detectable (OR, 1.10; CI, 0.66–1.81). In patients with SCC, the increase in risk was more pronounced in heavy smokers with increasing numbers of Pro alleles (Table 3), reaching an OR of 2.80 (CI, 1.19–6.58) for persons homozygous for 72Pro. The increase in risk with the number of 72Pro alleles showed a significant trend (p for trend=0.001). Again, no effect was found for light smokers. No significant association was observed in heavy or light smokers for TP53 genotype and risk of adenocarcinoma. When data for the p21 Ser31Arg polymorphism was analyzed for heavy and light smokers, a significant protective effect of the Arg allele was determined in heavy smokers with SCC (OR, 0.48; CI, 0.23–0.996). p Values obtained from interaction analysis (polymorphism×smoking) were not significant (p>0.05) in the evaluation of both polymorphisms in the overall population and in SCC and adenocarcinoma (Table 2). | | | | Genotype | Heavy smokers: >20 pack-yearsa | Light smokers: ≤20 pack-yearsa | |
|---|
| | Cases, n (%) | Controls, n (%) | OR | 95% CI | p value | Cases, n (%) | Controls, n (%) | OR | 95% CI | p value | |
|---|
| All persons | | | TP53, codon 72 | | | Arg/Arg | 151 (50) | 93 (62) | 1.0 | | | 52 (55) | 146 (59.5) | 1.0 | | | | | Arg/Pro | 122 (40) | 49 (32) | 1.47 | 0.96–2.26 | 0.075 | 32 (34) | 80 (33) | 0.99 | 0.58–1.71 | 0.98 | | | Pro/Pro | 30 (10) | 9 (6) | 1.99 | 0.90–4.43 | 0.091 | 10 (11) | 19 (7.5) | 1.56 | 0.66–3.72 | 0.31 | | | | | | p for trend: 0.015 | | | | p for trend: 0.4 | | | | Pro carriers | 152 (50) | 58 (38) | 1.55 | 1.04–2.33 | 0.032 | 42 (45) | 99 (40.5) | 1.10 | 0.66–1.81 | 0.72 | | | | | | p21, codon 31 | | | Ser/Ser | 261 (87) | 126 (83) | 1.0 | | | 82 (87) | 212 (87) | 1.0 | | | | | Arg carriers | 39 (13) | 25 (17) | 0.70 | 0.40–1.23 | 0.21 | 12 (13) | 32 (13) | 1.09 | 0.52–2.29 | 0.83 | | | | | | SCC | | | TP53, codon 72 | | | Arg/Arg | 68 (44) | 93 (62) | 1.0 | | | 15 (53) | 146 (59.5) | 1.0 | | | | | Arg/Pro | 64 (42) | 49 (32) | 1.71 | 1.04–2.81 | 0.035 | 11 (40) | 80 (33) | 1.09 | 0.46–2.63 | 0.84 | | | Pro/Pro | 21 (14) | 9 (6) | 2.80 | 1.19–6.58 | 0.019 | 2 (7) | 19 (7.5) | 1.09 | 0.21–5.61 | 0.92 | | | | | | p for trend: 0.001 | | | | p for trend: 0.7 | | | | Pro carriers | 85 (56) | 58 (38) | 1.88 | 1.18–3.00 | 0.008 | 13 (47) | 99 (40.5) | 1.09 | 0.47–2.53 | 0.83 | | | | | | p21, codon 31 | | | Ser/Ser | 136 (91) | 126 (83) | 1.0 | | | 24 (86) | 212 (87) | 1.0 | | | | | Arg carriers | 14 (9) | 25 (17) | 0.48 | 0.23–0.996 | 0.049 | 4 (14) | 32 (13) | 1.21 | 0.36–4.05 | 0.75 | | | | | | Adenocarcinoma | | | TP53, codon 72 | | | Arg/Arg | 63 (54) | 93 (62) | 1.0 | | | 34 (60) | 146 (59.5) | 1.0 | | | | | Arg/Pro | 46 (39) | 49 (32) | 1.43 | 0.84–2.44 | 0.19 | 16 (29) | 80 (33) | 0.78 | 0.39–1.55 | 0.48 | | | Pro/Pro | 8 (7) | 9 (6) | 1.46 | 0.52–4.09 | 0.47 | 6 (11) | 19 (7.5) | 1.22 | 0.43–3.44 | 0.70 | | | | | | p for trend: 0.254 | | | | p for trend: 0.8 | | | | Pro carriers | 54 (46) | 58 (38) | 1.44 | 0.86–2.39 | 0.16 | 22 (40) | 99 (40.5) | 0.87 | 0.47–1.61 | 0.66 | | | | | | p21, codon 31 | | | Ser/Ser | 96 (82) | 126 (83) | 1.0 | | | 49 (87) | 212 (87) | 1.0 | | | | | Arg carriers | 21 (18) | 25 (17) | 0.99 | 0.52–1.91 | 0.99 | 7 (13) | 32 (13) | 1.21 | 0.49–3.00 | 0.26 | | | | |
| a Odds ratios were adjusted for gender, age groups and occupation; TP53. Data on smoking are missing for eight cases and eight controls; p21: data on genotype or smoking are missing for 11 cases and nine controls OR values with p<0.05 are shown in bold. |
The combined effects of the two risk alleles TP53 72Pro and p21 31Ser were investigated, with Arg/Arg homozygotes in TP53 and Arg carriers in p21 as the reference (Table 4). ORs were calculated for potentially minor-risk individuals, with one risk allele (72Pro or 31Ser), and for high-risk individuals, with the risk variants at both gene loci (72Pro and 31Ser). An increase in risk was found in the analysis of all lung cancer cases. In the group of SCC patients, the crude (data not shown) and adjusted ORs (2.49; CI, 1.08–5.72) for high-risk individuals were significantly and considerably higher than those for persons with the TP53 72Pro allele only. In addition, the trend for an increase in risk with number of risk alleles was p=0.002. In adenocarcinoma cases, no effect of the combined risk alleles was detected (OR, 1.13; CI, 0.56–2.28). | | | | Genotypea | Cases, n (%) | Controls, n (%) | Adjusted OR | Interaction polymorphisms×smoking p value | |
|---|
| | | | ORb | 95% CI | p value | | |
|---|
| All persons | 405 (100) | 404 (100) | | | | | | | Reference | 29 (7) | 37 (9) | 1.0 | | | | | | Moderate risk | 206 (51) | 227 (56) | 1.19 | 0.66–2.15 | 0.56 | 0.006 | | | High risk | 170 (42) | 140 (35) | 1.59 | 0.87–2.90 | 0.13 | 0.016 | | | | | | p for trend: 0.030 | | | | | | | | SCC | 185 (100) | 404 (100) | | | | | | | Reference | 11 (6) | 37 (9) | 1.0 | | | | | | Moderate risk | 85 (46) | 227 (56) | 1.62 | 0.71–3.71 | 0.25 | 0.10 | | | High risk | 89 (48) | 140 (35) | 2.49 | 1.08–5.72 | 0.032 | 0.059 | | | | | | p for trend: 0.002 | | | | | | | | Adenocarcinoma | 177 (100) | 404 (100) | | | | | | | Reference | 17 (9) | 37 (9) | 1.0 | | | | | | Moderate risk | 95 (54) | 227 (56) | 1.01 | 0.51–1.99 | 0.98 | 0.036 | | | High risk | 65 (37) | 140 (35) | 1.13 | 0.56–2.28 | 0.73 | 0.08 | | | | | | p for trend: 0.8 | | | | | | |
| a Genotype combinations tested: reference (no risk allele at either site): TP53, codon 72: Arg/Arg/p21, codon 31: Arg carriers; moderate risk (one risk allele at one of the two sites): TP53, codon 72: Pro carriers/p21, codon 31: Arg carriers or TP53, codon 72: Arg/Arg/p21, codon 31: Ser/Ser; high risk (risk alleles at both sites): TP53, codon 72: Pro carriers/p21, codon 31: Ser/Ser. bOdds ratios were adjusted for gender, age group, pack-year group and occupation; OR values with p<0.05 are shown in bold. |
When the combined effects of the two risk alleles were evaluated separately for heavy and light smokers, a statistically significant interaction (polymorphism×smoking) was found for the analysis of all persons with the minor and high-risk allele combination and for adenocarcinoma with the minor-risk combination (Table 4). In the subgroup analysis (Table 5), a statistically significantly increased risk was found for heavy smokers for all lung cancer cases and especially in SCC cases. This increase was nearly fourfold for SCC patients with both risk alleles (TP53 72Pro and p21 31 Ser; OR, 3.84; CI, 1.46–10.1). In the group of light smokers, no effect modification by these two alleles was observed for all cases of lung cancer or SCC. In adenocarcinoma, neither the minor nor the high-risk allele combination showed a significant effect on lung cancer risk. | | | | Genotypea | Heavy smokers: >20 pack-yearsb | Light smokers: ≤20 pack-yearsb | |
|---|
| | Cases, n (%) | Controls, n (%) | OR | 95% CI | p value | Cases, n (%) | Controls, n (%) | OR | 95% CI | p value | |
|---|
| All cases | | | Reference | 16 (5) | 18 (12) | 1.0 | | | 11 (12) | 19 (8) | 1.0 | | | | | Moderate risk | 160 (53) | 82 (54) | 2.41 | 1.14–5.12 | 0.022 | 42 (45) | 140 (57) | 0.50 | 0.21–1.18 | 0.11 | | | High risk | 127 (42) | 51 (34) | 2.96 | 1.37–6.40 | 0.006 | 41 (43) | 86 (35) | 0.71 | 0.30–1.71 | 0.45 | | | | | | p for trend: 0.014 | | | | p for trend: 0.5 | | | | | | | SCC | | | Reference | 7 (4) | 18 (12) | 1.0 | | | 3 (11) | 19 (8) | 1.0 | | | | | Moderate risk | 70 (46) | 82 (54) | 2.36 | 0.90–6.19 | 0.080 | 13 (46) | 140 (57) | 0.61 | 0.15–2.56 | 0.50 | | | High risk | 76 (50) | 51 (34) | 3.84 | 1.46–10.1 | 0.007 | 12 (43) | 86 (35) | 0.75 | 0.17–3.30 | 0.70 | | | | | | p for trend: 0.001 | | | | p for trend: 0.7 | | | | | | | Adenocarcinoma | | | Reference | 9 (8) | 18 (12) | 1.0 | | | 7 (13) | 19 (8) | 1.0 | | | | | Moderate risk | 66 (56) | 82 (54) | 1.90 | 0.78–4.65 | 0.16 | 27 (48) | 140 (57) | 0.43 | 0.16–1.16 | 0.20 | | | High risk | 42 (36) | 51 (34) | 1.98 | 0.78–5.04 | 0.15 | 22 (39) | 86 (35) | 0.51 | 0.18–1.44 | 0.09 | | | | | | p for trend: 0.4 | | | | p for trend: 0.9 | | | | | |
| a Combinations tested: reference genotype (no risk allele at either site): TP53, codon 72: Arg/Arg/p21, codon 31: Arg carriers; moderate-risk genotype (one risk allele at one of the two sites): TP53, codon 72: Pro carrier/p21, codon 31: Arg carriers or TP53, codon 72: Arg/Arg/p21, codon 31: Ser/Ser; high-risk genotype (risk alleles at both sites): TP53, codon 72: Pro carriers/p21, codon 31: Ser/Ser. bOdds ratios were adjusted for gender, age group and occupation; data on smoking are missing for 17 cases and 20 controls, OR values with p<0.05 are shown in bold. |
4. Discussion A moderate increase in risk was observed in our study for carriers of the TP53 72Pro allele, indicating that this allele is a risk factor for non-small-cell lung cancer. Our odds ratio is higher than that calculated for 72Pro homozygotes in a meta-analysis of 13 studies [9] but similar to those in the two largest studies in the meta-analysis, comprising 635 [30] and 1168 [13] cases of non-small-cell lung cancer, mainly in whites. For 72Pro homozygote carriers, these studies showed adjusted ORs of 1.67 (95% CI 0.94–2.98) [30] and 1.32 (95% CI 1.1–1.6) [13]. In addition, it was reported that the TP53 72Pro allele was associated with other cancers such as gastric adenocarcinoma [31], urinary bladder cancer [32], familial breast cancer [33] or thyroid cancer and endometrial cancer [19], [24]. The protein variants of the Arg72Pro polymorphism of TP53 differ in several biochemical properties after p53-inducing treatment: (i) the Arg allele induces apoptosis more efficiently than the Pro allele because it has a greater ability to localize to mitochondria and to release cytochrome c into the cytosol [8], [34]; (ii) the Pro allele induces higher levels of G1 arrest [35]; (iii) the Pro allele is more efficient in specifically activating p53-dependent DNA repair target genes and cells carrying the TP53 Pro allele have a significantly higher DNA repair capacity than cells with the TP53 Arg allele [36]; (iv) the Pro allele is associated with an increased frequency of TP53 mutations in non-small-cell lung cancers [37], [38], [39]. These findings indicate that the 72Pro variant of p53 tends to prevent removal of heavily damaged cells by apoptosis and to favor their DNA repair during G1 arrest. The 31Arg variant of p21 showed a non-significant protective effect against lung cancer in our study, supporting the findings of earlier studies showing no effect of the p21 polymorphism on lung cancer risk [15], [17]. For our analysis of the combined effects of the two polymorphisms, we defined the 31Ser variant as the risk form, as it has recently been found to be a risk allele in endometrial and gastric cancers [24], [25], and we obtained increased ORs with increasing number of TP53 and p21 risk alleles. The p21 protein protects cells from apoptosis induced by radiation and chemotherapy by imposing cell cycle arrest in G1 phase [3], [40]. Thus, the protein appears to be an opponent of p53 in apoptosis induction but an intensifier of cell cycle arrest. There is little experimental evidence for the phenotypic consequences of the amino acid change in codon 31 of p21, but in peripheral leukocytes the 31Ser variant is associated with significantly more p21 mRNA expression than the 31Arg variant [41], which might result in larger amounts of p21 Ser protein and more G1 arrest after DNA damage. The combined effects of the p21 Ser variant and the p53 72Pro variant would be to intensify cell cycle arrest as the cellular response to DNA damage, thereby suppressing the apoptosis response. In assessing the consequences of this risk genotype, it must be noted that both polymorphisms are defined as low penetrance mutations that alter protein functions only slightly. Normally, the balance between cell cycle arrest and apoptosis is adequate to allow error-free repair or apoptotic removal of heavily damaged genomes. In TP53 72Pro carriers and in persons with the TP53/p21 high-risk genotype, however, this balance might be shifted to cell cycle arrest and might lead to more mutations and genomic instability in persons heavily exposed to carcinogens, as cells with incompletely repaired genomes might not be removed by apoptosis. Based on this hypothesis, we investigated risk modification by tobacco smoking, a carcinogenic exposure that induces a variety of DNA lesions. We found that light and nonsmokers who carry the 72Pro variant of p53 and the 31Ser variant of p21 had no increase in risk. In heavy smokers with more than 20 pack-years, the ORs were considerably increased, by nearly four-fold for persons carrying the high-risk variants of both proteins, indicating a strong gene-smoking interaction for persons with both risk alleles. Our risk estimations might be biased by the fact that data on carcinogen exposure by smoking were derived from a questionnaire, and self-reports of lifestyle factors are subject to different recall bias for cases and controls. However, gene×smoking interaction analysis supports the data of the subgroup analysis when calculated for all persons and adenocarcinoma cases carrying minor and high-risk allele combinations. Statistical significance was not achieved in the other interaction tests. The power of our study might still be too low to measure the effects obtained in the subgroup analysis although our number of cases and controls is relatively large, as compared to previous studies in the literature. Furthermore, our results are supported by a similarly strong effect of tobacco exposure on lung cancer risk due to TP53 genotype described by other authors [28]. To the best of our knowledge, this is the first report about the effect of the combined risk alleles. We did not stratify our study group for environmental exposure due to occupation because our exposure data, which were based on job titles and self-reported exposure, were not detailed enough for a separate analysis. When we adjusted the ORs for occupational exposure as a possible confounder, however, the observed gene–smoking interaction was confirmed, i.e. the association of TP53 and p21 variant genotypes with lung cancer risk remained strongest in heavy smokers. This supports that carrying the TP53/p21 risk allele combination predisposes mainly those persons for lung cancer who are heavily exposed to exogenous carcinogens. The effects of the TP53 and p21 polymorphisms were different for the different histological subtypes of lung cancer, adenocarcinoma and SCC. In cases of SCC, we observed a statistically significant increase in OR with increasing number of TP53 72Pro alleles or TP53 and p21 risk alleles, which was strongest in heavy smokers. In cases of adenocarcinoma, the effects of the TP53 72Pro allele alone and in combination with the p21 31Ser allele were smaller and not statistically significant. Similarly to our data, Sakiyama et al. found that the TP53 72 Pro allele was a strong risk allele in SCC, even in smokers, but not in adenocarcinoma [42]. In contrast to our results, Liu et al., found that the TP53 Pro allele increased lung cancer risk more strongly in adenocarcinoma than in SCC cases [13]. This difference in results might be due to the fact that our study had less statistical power and that significantly more adenocarcinoma cases than SCC cases occurred among persons who have never smoked or were light smokers. The different effects of the combined risk genotype related to smoking in adenocarcinoma and SCC cases might reflect the different etiologies of these two subtypes [11], [43]. Risk alleles of both polymorphisms might confer different susceptibility to different histological types of cancer and to different pathogenic mechanisms. Our study design might have several limitations. Controls were younger than cases, but a stronger matching for age and gender could not be achieved without losing a considerable number of cases and controls thereby reducing statistical power. After adjustment for age and gender in multivariate logistic regression analysis, however, the differences between the crude and adjusted ORs were small, indicating that age and gender did not have a strong influence on the risk estimates of the polymorphisms. In addition, recruitment of hospital controls with chronic inflammatory lung diseases could lead to selection bias because inflammation can induce p53-dependent apoptosis which might be modulated by the TP53 polymorphism we analyzed. Data about a genetic trait in inflammatory lung diseases are, however, rare and not conclusive [44]. In addition, chronic inflammatory processes were suspected to be predisposing factors for lung cancer [45], [46]. This finding is in contrast to those reviewed by Goode et al. [47] where hospital controls were suggested to be unlikely to represent a major source of confounding as ORs determined in repair polymorphism studies with population-based controls were generally not higher than the ORs obtained with hospital-based controls. Similarly, no differences between hospital- and population-based controls (∼15,000) were found in case–control studies of polymorphic metabolism genes [48]. If some of the nonmalignant lung diseases in our hospital-based control population were predisposing conditions for lung cancer, we would suggest that our risk estimates must therefore underestimate the risk compared to a normal control population without these diseases. It is nevertheless conceivable that the genetic contribution would be more pronounced if controls were selected from a healthy population and the degree of matching would be higher. In conclusion, our data indicate that the TP53 72Pro allele contributes to lung cancer risk in persons with heavy exposure to tobacco carcinogens. This risk is increased by the simultaneous presence of the p21 31Ser risk allele. Interpretation of our results might be limited, as we concentrated on only one functionally active polymorphism per gene. As both proteins have several active domains and sub-functions, other variants might also be relevant for risk evaluation. In addition, interactions with the numerous templates upstream and downstream of p53 were not considered. Therefore, more studies of biochemical functions and interactions and of epidemiological associations are necessary to elucidate further the impact of polymorphisms in cell cycle regulatory genes on lung cancer risk. Acknowledgments The authors wish to thank Birgit Jäger (Division of Toxicology and Cancer Risk Factors, DKFZ) for excellent technical assistance and Renate Rausch (Division of Biostatistics, DKFZ) for substantial help with the statistical analysis. We thank O. Landt (Tib Molbiol, Berlin, Germany) for help with the LightCycler probe design and E. Heseltine for critically reading the manuscript and editorial help. We are grateful to all the patients and staff involved in sample and data collection. This work was partly supported by a scholarship granted by the Deutsches Krebsforschungszentrum (D.B.) and by funding from the Verein zur Förderung der Krebsforschung in Deutschland e.V. (A.R.) and the Deutsche Krebshilfe (for sample collection). References [1]. [1]Vogelstein B, Lane D, Levine AJ. Surfing the p53 network. Nature. 2000;408:307–310. MEDLINE |
CrossRef
[2]. [2]Woods DB, Vousden KH. Regulation of p53 function. Exp Cell Res. 2001;264:56–66. MEDLINE |
CrossRef
[3]. [3]Harada K, Ogden GR. An overview of the cell cycle arrest protein, p21(WAF1). Oral Oncol. 2000;36:3–7. Abstract | Full Text |
Full-Text PDF (120 KB)
|
CrossRef
[4]. [4]Robles AI, Linke SP, Harris CC. The p53 network in lung carcinogenesis. Oncogene. 2002;21:6898–6907. MEDLINE |
CrossRef
[5]. [5]Wistuba II, Gazdar AF, Minna JD. Molecular genetics of small cell lung carcinoma. Semin Oncol. 2001;28:3–13. Abstract |
Full-Text PDF (1071 KB)
|
CrossRef
[6]. [6]Matlashewski GJ, Tuck S, Pim D, Lamb P, Schneider J, Crawford LV. Primary structure polymorphism at amino acid residue 72 of human p53. Mol Cell Biol. 1987;7:961–963. MEDLINE [7]. [7]Pietsch EC, Humbey O, Murphy ME. Polymorphisms in the p53 pathway. Oncogene. 2006;25:1602–1611. MEDLINE |
CrossRef
[8]. [8]Thomas M, Kalita A, Labrecque S, Pim D, Banks L, Matlashewski G. Two polymorphic variants of wild-type p53 differ biochemically and biologically. Mol Cell Biol. 1999;19:1092–1100. MEDLINE [9]. [9]Matakidou A, Eisen T, Houlston RS. TP53 polymorphisms and lung cancer risk: a systematic review and meta-analysis. Mutagenesis. 2003;18:377–385. MEDLINE |
CrossRef
[10]. [10]Charloux A, Quoix E, Wolkove N, Small D, Pauli G, Kreisman H. The increasing incidence of lung adenocarcinoma: reality or artefact? A review of the epidemiology of lung adenocarcinoma. Int J Epidemiol. 1997;26:14–23. MEDLINE |
CrossRef
[11]. [11]Devesa SS, Shaw GL, Blot WJ. Changing patterns of lung cancer incidence by histological type. Cancer Epidemiol Biomarkers Prev. 1991;1:29–34. MEDLINE [12]. [12]Butkiewicz D, Popanda O, Risch A, et al. Association between the risk for lung adenocarcinoma and a (-4) G-to-A polymorphism in the XPA gene. Cancer Epidemiol Biomarkers Prev. 2004;13:2242–2246. MEDLINE [13]. [13]Liu G, Miller DP, Zhou W, et al. Differential association of the codon 72 p53 and GSTM1 polymorphisms on histological subtype of non-small cell lung carcinoma. Cancer Res. 2001;61:8718–8722. MEDLINE [14]. [14]Chedid M, Michieli P, Lengel C, Huppi K, Givol D. A single nucleotide substitution at codon 31 (Ser/Arg) defines a polymorphism in a highly conserved region of the p53-inducible gene WAF1/CIP1. Oncogene. 1994;9:3021–3024. MEDLINE [15]. [15]Shih CM, Lin PT, Wang HC, Huang WC, Wang YC. Lack of evidence of association of p21WAF1/CIP1 polymorphism with lung cancer susceptibility and prognosis in Taiwan. Jpn J Cancer Res. 2000;91:9–15. MEDLINE [16]. [16]Sjalander A, Birgander R, Rannug A, Alexandrie AK, Tornling G, Beckman G. Association between the p21 codon 31 A1 (arg) allele and lung cancer. Hum Hered. 1996;46:221–225. MEDLINE |
CrossRef
[17]. [17]Su L, Liu G, Zhou W, et al. No association between the p21 codon 31 serine–arginine polymorphism and lung cancer risk. Cancer Epidemiol Biomarkers Prev. 2003;12:174–175. MEDLINE [18]. [18]Sun Y, Taeger D, Weiland SK, Keil U, Straif K. Job titles and work areas as surrogate indicators of occupational exposure. Epidemiology. 2003;14:361–367. MEDLINE |
CrossRef
[19]. [19]Boltze C, Roessner A, Landt O, Szibor R, Peters B, Schneider-Stock R. Homozygous proline at codon 72 of p53 as a potential risk factor favoring the development of undifferentiated thyroid carcinoma. Int J Oncol. 2002;21:1151–1154. MEDLINE [20]. [20]Tan XL, Popanda O, Ambrosone CB, et al. Association between TP53 and p21 genetic polymorphisms and acute side effects of radiotherapy in breast cancer patients. Breast Cancer Res Treat. 2006;97:255–262. MEDLINE |
CrossRef
[21]. [21]Risch A, Dally H, Edler L. Genetic polymorphisms in metabolising enzymes as lung cancer risk factors. In: Edler L, Kitsos C editor. Recent Advances in Quantitative Methods in Cancer and Human Health Risk Assessment. John Wiley & Sons Ltd.; 2005;p. 43–62. [22]. [22]Breslow NE, Day NE. Statistical methods in cancer research. Volume I. The analysis of case–control studies. IARC Sci Publ. 1980;5–338. [23]. [23]McCullagh P. Regression models for ordinal data (with discussion). J R Stat Soc: B (Stat Meth) B. 1980;42:109–142. [24]. [24]Roh JW, Kim JW, Park NH, et al. p53 and p21 genetic polymorphisms and susceptibility to endometrial cancer. Gynecol Oncol. 2004;93:499–505. MEDLINE |
CrossRef
[25]. [25]Xi YG, Ding KY, Su XL, et al. p53 polymorphism and p21WAF1/CIP1 haplotype in the intestinal gastric cancer and the precancerous lesions. Carcinogenesis. 2004;25:2201–2206. MEDLINE |
CrossRef
[26]. [26]Wang-Gohrke S, Rebbeck TR, Besenfelder W, Kreienberg R, Runnebaum IB. p53 germline polymorphisms are associated with an increased risk for breast cancer in German women. Anticancer Res. 1998;18:2095–2099. MEDLINE [27]. [27]Birgander R, Sjalander A, Saha N, Spitsyn V, Beckman L, Beckman G. The codon 31 polymorphism of the p53-inducible gene p21 shows distinct differences between major ethnic groups. Hum Hered. 1996;46:148–154. MEDLINE |
CrossRef
[28]. [28]Fan R, Wu MT, Miller D, et al. The p53 codon 72 polymorphism and lung cancer risk. Cancer Epidemiol Biomarkers Prev. 2000;9:1037–1042. MEDLINE [29]. [29]Jones JS, Chi X, Gu X, Lynch PM, Amos CI, Frazier ML. p53 polymorphism and age of onset of hereditary nonpolyposis colorectal cancer in a Caucasian population. Clin Cancer Res. 2004;10:5845–5849. MEDLINE |
CrossRef
[30]. [30]Wu X, Zhao H, Amos CI, et al. p53 genotypes and haplotypes associated with lung cancer susceptibility and ethnicity. J Natl Cancer Inst. 2002;94:681–690. MEDLINE [31]. [31]Lai KC, Chen WC, Tsai FJ, Li SY, Jeng LB. Arginine and proline alleles of the p53 gene are associated with different locations of gastric cancer. Hepatogastroenterology. 2005;52:944–948. MEDLINE [32]. [32]Chen YC, Xu L, Guo YL, et al. Polymorphisms in GSTT1 and p53 and urinary transitional cell carcinoma in south-western Taiwan: a preliminary study. Biomarkers. 2004;9:386–394.
CrossRef
[33]. [33]Ohayon T, Gershoni-Baruch R, Papa MZ, Distelman MT, Eisenberg BS, Friedman E. The R72P P53 mutation is associated with familial breast cancer in Jewish women. Br J Cancer. 2005;92:1144–1148. MEDLINE |
CrossRef
[34]. [34]Dumont P, Leu JI, Della PA, George DL, Murphy M. The codon 72 polymorphic variants of p53 have markedly different apoptotic potential. Nat Genet. 2003;33:357–365. MEDLINE |
CrossRef
[35]. [35]Pim D, Banks L. p53 polymorphic variants at codon 72 exert different effects on cell cycle progression. Int J Cancer. 2004;108:196–199. MEDLINE |
CrossRef
[36]. [36]Siddique M, Sabapathy K. Trp53-dependent DNA-repair is affected by the codon 72 polymorphism. Oncogene. 2006;25:3489–3500. MEDLINE |
CrossRef
[37]. [37]Hu Y, McDermott MP, Ahrendt SA. The p53 codon 72 proline allele is associated with p53 gene mutations in non-small cell lung cancer. Clin Cancer Res. 2005;11:2502–2509. MEDLINE |
CrossRef
[38]. [38]Mechanic LE, Marrogi AJ, Welsh JA, et al. Polymorphisms in XPD and TP53 and mutation in human lung cancer. Carcinogenesis. 2005;26:597–604. MEDLINE |
CrossRef
[39]. [39]Szymanowska A, Jassem E, Dziadziuszko R, et al. Increased risk of non-small cell lung cancer and frequency of somatic TP53 gene mutations in Pro72 carriers of TP53 Arg72Pro polymorphism. Lung Cancer. 2006;52:9–14. Abstract | Full Text |
Full-Text PDF (117 KB)
|
CrossRef
[40]. [40]Dotto GP. p21(WAF1/Cip1): more than a break to the cell cycle?. Biochim Biophys Acta. 2000;1471:M43–M56. MEDLINE [41]. [41]Su L, Sai Y, Fan R, et al. P53 (codon 72) and P21 (codon 31) polymorphisms alter in vivo mRNA expression of p21. Lung Cancer. 2003;40:259–266. Abstract | Full Text |
Full-Text PDF (206 KB)
[42]. [42]Sakiyama T, Kohno T, Mimaki S, et al. Association of amino acid substitution polymorphisms in DNA repair genes TP53, POLI, REV1 and LIG4 with lung cancer risk. Int J Cancer. 2005;114:730–737. MEDLINE |
CrossRef
[43]. [43]Barbone F, Bovenzi M, Cavallieri F, Stanta G. Cigarette smoking and histologic type of lung cancer in men. Chest. 1997;112:1474–1479. MEDLINE |
CrossRef
[44]. [44]Anderson GP, Bozinovski S. Acquired somatic mutations in the molecular pathogenesis of COPD. Trends Pharmacol Sci. 2003;24:71–76. MEDLINE |
CrossRef
[45]. [45]Mayne ST, Buenconsejo J, Janerich DT. Previous lung disease and risk of lung cancer among men and women nonsmokers. Am J Epidemiol. 1999;149:13–20. MEDLINE [46]. [46]Williams MD, Sandler AB. The epidemiology of lung cancer. Cancer Treat Res. 2001;105:31–52. MEDLINE [47]. [47]Goode EL, Ulrich CM, Potter JD. Polymorphisms in DNA repair genes and associations with cancer risk. Cancer Epidemiol Biomarkers Prev. 2002;11:1513–1530. MEDLINE [48]. [48]Garte S, Gaspari L, Alexandrie AK, et al. Metabolic gene polymorphism frequencies in control populations. Cancer Epidemiol Biomarkers Prev. 2001;10:1239–1248. MEDLINE a Division of Toxicology and Cancer Risk Factors, German Cancer Research Center (DKFZ), Im Neuenheimer Feld 280, D-69120 Heidelberg, Germany b Division of Biostatistics, German Cancer Research Center (DKFZ), Heidelberg, Germany c Department of Tumor Biology, Center of Oncology, M. Sklodowska-Curie Memorial Institute, Gliwice, Poland d Thoraxklinik am Universitätsklinikum Heidelberg, Germany  Corresponding author. Fax: +49 6221 42 3359.
PII: S0169-5002(06)00485-5 doi:10.1016/j.lungcan.2006.09.006 © 2006 Elsevier Ireland Ltd. All rights reserved. | |
|
FULL TEXT