Fax: (713) 834-6153
Glucose metabolism gene polymorphisms and clinical outcome in pancreatic cancer
Article first published online: 15 SEP 2010
Copyright © 2010 American Cancer Society
Volume 117, Issue 3, pages 480–491, 1 February 2011
How to Cite
Dong, X., Tang, H., Hess, K. R., Abbruzzese, J. L. and Li, D. (2011), Glucose metabolism gene polymorphisms and clinical outcome in pancreatic cancer. Cancer, 117: 480–491. doi: 10.1002/cncr.25612
- Issue published online: 20 JAN 2011
- Article first published online: 15 SEP 2010
- Manuscript Accepted: 29 JUL 2010
- Manuscript Received: 28 MAY 2010
- glucose metabolism;
- pancreatic adenocarcinoma;
- single nucleotide polymorphism;
- overall survival;
Altered glucose metabolism is the most common metabolic hallmark of malignancies. The authors tested the hypothesis that glucose metabolism gene variations affect clinical outcome in pancreatic cancer.
The authors retrospectively genotyped 26 single nucleotide polymorphisms from 5 glucose metabolism genes in 154 patients with localized disease and validated the findings in 552 patients with different stages of pancreatic adenocarcinoma. Association between genotypes and overall survival (OS) was evaluated using multivariate Cox proportional hazard regression models with adjustment for clinical predictors.
Glucokinase (GCK) IVS1 + 9652C > T and hexokinase 2 (HK2) N692N homozygous variants were significantly associated with reduced OS in the training set of 154 patients (P < .001). These associations were confirmed in the validation set of 552 patients and in the combined dataset of all 706 patients (P ≤ .001). In addition, HK2 R844K variant K allele was associated with a better survival in the validation set and the combined dataset (P ≤ .001). When data were further analyzed by disease stage, glutamine-fructose-6-phosphate transaminase (GFPT1) IVS14-3094T>C, HK2 N692N and R844K in patients with localized disease and GCK IVS1 + 9652C>T in patients with advanced disease were significant independent predictors for OS (P ≤ .001). Haplotype CGG of GPI and GCTATGG of HK2 were associated with better OS, respectively, with P values of .004 and .007.
The authors demonstrated that glucose metabolism gene polymorphisms affect clinical outcome in pancreatic cancer. These observations support a role of abnormal glucose metabolism in pancreatic carcinogenesis. Cancer 2011. © 2010 American Cancer Society.
A common property of malignant tumors is altered glucose metabolism. The Warburg effect (aerobic glycolysis, a persistently high rate of glucose conversion into lactate even under normoxic condition) is a distinctive metabolic characteristic of malignancies that distinguishes them from normal cells.1 Possibly this effect is an adaptation to intermittent hypoxia in premalignant lesions. Enhanced glycolysis at the expense of mitochondrial energy production causing microenvironment acidosis triggers evolution to phenotypes resistant to acid toxicity, provides precursors for macromolecule biosynthesis, and protects cells from excessive toxic reactive oxygen species.2 Subsequent cell populations with intensified glycolysis and acid resistance have a strong growth advantage, which promotes malignant proliferation, unrestrained growth, and invasion.3 On the basis of this prominent phenotype, positron emission tomography (PET) imaging has become a major method for cancer detection and surveillance. The worldwide clinical application of PET has resulted in a resurgence of interest in tumor metabolism.4 PET using the glucose analogue tracer 2-(18F)-2-deoxy-D-glucose (FDG) has shown that most cancers profoundly strengthen glucose uptake, which depends on glycolysis rate. FDG uptake/trapping results from up-regulation of glucose transporters and hexokinases (HK1/2) in pancreatic cancer.5, 6 This is a marker that can be used to monitor cancer progression; the augmented glucose uptake correlates with enhanced tumor aggression, advanced clinical stage, and poorer prognosis.7, 8
Pancreatic cancer is the fourth leading cause of cancer mortality in the United States, with an estimated 42,470 new cases and 35,240 deaths in 2009.9 Pancreatic cancer is 1 of the most difficult malignancies to treat, with a 5-year survival rate <5%.9 Glucose intolerance and diabetes are common manifestations of pancreatic cancer. Whether and how genetic variations in glucose metabolism affect the clinical outcome of this disease are unknown.
HK2, glucokinase (GCK), glutamine-fructose-6-phosphate transaminase (GFPT1), glucose phosphate isomerase (GPI), and O-linked N-acetylglucosamine transferase (OGT) are key enzymes involved in glucose metabolism. For its crucial role in determining the cell fate (survival or death),10 glucose metabolism pathway has become a therapeutic target for cancer treatment,11 with clinical trials on HK2 inhibitors being conducted.12, 13 We have previously shown that obesity and diabetes are associated with reduced overall survival in patients with pancreatic cancer.14, 15 Whether genetic variations in glucose metabolism contribute to the poor clinical outcome of pancreatic cancer has never been explored. To test the hypothesis that genetic variation in glucose metabolism genes is related to clinical outcome in pancreatic cancer, we evaluated 26 single nucleotide polymorphisms (SNPs) of GCK, GFPT1, GPI, HK2, and OGT genes (Fig. 1) in reference to the overall survival (OS) and response to chemoradiotherapy in 706 patients with pancreatic cancer.
MATERIALS AND METHODS
Patient Recruitment and Data Collection
The 706 patients included 154 patients with resectable tumor who were enrolled in clinical trials of preoperative gemcitabine-based chemoradiation16 and 552 patients who were recruited in a case-control study conducted at The University of Texas M. D. Anderson Cancer Center from February 1999 to May 2007, with follow-up to August 2009.17 Patients were eligible for the current study if they had a diagnosis of pathologically confirmed pancreatic ductal adenocarcinoma and had an available DNA sample. All patients signed an informed consent for medical record review and DNA sample collection. The study was approved by the institutional review board of The University of Texas M. D. Anderson Cancer Center and conducted in accordance with all current ethical guidelines.
We reviewed patients' medical records to collect demographic (age, sex, and self-reported race) and clinical information on date of diagnosis; date of death or last follow-up; clinical tumor stage; tumor resection, site, size, and differentiation; performance status; serum markers for liver, kidney, and pancreas functions; and serum carbohydrate antigen 19-9 (CA19-9) level at diagnosis. Clinical tumor staging followed the objective computed tomography (CT) criteria: a localized or potentially resectable tumor is defined as a tumor with no evidence of extrapancreatic disease (extensive peripancreatic lymph node involvement), no involvement of the celiac axis, superior mesenteric artery, inferior vena cava, or aorta, or encasement or occlusion of the superior mesenteric vein-portal vein confluence. Tumor abutment and encasement of the superior mesenteric vein, in the absence of vessel occlusion or extension to the superior mesenteric artery, was considered resectable. Locally advanced tumors are those unresectable but without distant metastasis. Tumor response to preoperative therapy was evaluated by CT at restaging in patients who had localized tumor and received preoperative chemoradiotherapy. Tumor margin and lymph node status were evaluated in patients with resected tumors only. Dates of death were obtained and cross-checked using the following sources: The University of Texas M. D. Anderson Cancer Center tumor registry, inpatient medical records, or the US Social Security Death Index (www.deathindexes.com/ssdi.html). OS time was calculated from the date of diagnosis to the date of death or last follow-up.
DNA Extraction, SNP Selection, and Genotyping
DNA was extracted from peripheral lymphocytes using Qiagen (Valencia, Calif) DNA isolation kits. Seventeen tagging SNPs were selected using SNPbrowser software (Applied Biosystems, www.allsnps.com/snpbrowser) with a cutoff of r2 = 0.8 and a minor allele frequency ≥10% in Caucasians from the HapMap Project database (www.hapmap.org). We also included 9 coding SNPs (nonsynonymous or synonymous) or untranslated region (UTR) SNPs that have a minor allele frequency ≥5% in Caucasians. The genes, nucleotide substitutions, functions, reference SNP identification numbers, and minor allele frequencies of the 26 SNPs are described in Table 1. The protein sequences, structures, homology models, mRNA transcripts, and predicted functions for the SNPs were evaluated by F-SNP (Queen's University, Kingston, Ontario, Canada).18 Genotyping used the mass spectroscopy-based MassArray method (Sequenom, San Diego, Calif). We randomly genotyped 20% of total samples in duplicate, showing 99.8% concordance. The inconsistent data were excluded from final analysis.
|Gene||Chromosome||SNP||Function||RS No.||Allele Frequency|
The distribution of genotypes was tested for Hardy-Weinberg equilibrium with the goodness-of-fit chi-square test. Genotype and allele frequency of the SNP were determined by direct gene counting. Haplotype diversity and linkage disequilibrium index (Lewontin's D′ and r2) were calculated using SNPAlyze (DYNACOM Co., Mobara, Japan). The median follow-up time was computed using censored observations only. The association between genotype/haplotype and OS was evaluated by Cox proportional hazard regression models. Hazard ratios and 95% confidence interval (CIs) were calculated with adjustment of sex, race, and any clinical factors that are significant predictors for OS in multivariate Cox regression models. The association of genotype with categorical variables such as sex, race, and tumor response to therapy was examined using the chi-square test and a logistic regression model with adjustment for clinical factors. Statistical analysis used SPSS (SPSS Inc, Chicago, Ill). The false discovery rate was calculated using the Beta-Uniform Mixture method.19 For 77 comparisons in a total of 26 SNPs (38 SNPs in dominant and 39 in recessive inheritance modes) for OS in all patients, we found that a P value of.002 corresponded to an false discovery rate of 5%. Thus, P ≤ .002 in the genotype analysis was considered statistically significant.
The patients' demographics and clinical predictors for OS are summarized in Table 2. There were 333 patients with localized disease, 211 with locally advanced disease, and 162 with metastatic disease. Of the 333 patients with localized tumor, 275 (83%) had tumor resection. Of the 706 patients, 138 (19.5%) were alive at the end of the study, with a median follow-up time of 46.0 months. The median survival time for the entire patient population was 17.2 months (95% CI, 15.8-18.5). Advanced tumor stage, unresected tumor, an elevated CA19-9 serum level or biochemical index, or poor performance status remained as significant predictors for worse OS in multivariate Cox regression models (data not shown).
|Variable||No. of Patients||No. of Deaths||MST, mo||Plog-rank|
|Tumor size, cm||<.001|
|Well to moderate||355||261||26.2|
|Tumor response to therapyb||<.001|
Genotype Distribution and Allele Frequencies
The observed allele frequencies in this study population were comparable to the previously reported allele frequencies in the general population (Table 1). The distribution of 26 SNPs followed Hardy-Weinberg equilibrium (P > .05) except for OGT IVS18-424A>G (P = .001). Linkage disequilibrium data of the 26 SNPs are described in Table 3. There were significant sex and racial differences in the genotype distributions; for example, the HK2 N692N CC genotype frequency was 22.4% for men but 10.3% for women (P < .001), and the HK2 R844K GG genotype frequency was 63.9%, 53.5%, and 25.9% for whites, Hispanics, and blacks, respectively (P < .001) (data for other SNPs are not shown). Therefore, sex and race were included in all Cox regression models.
|GCK||−515G>A||IVS1- 11823G>A||IVS1+ 6037T>C||IVS1+ 9652C>T||IVS1+ 11382G>A||IVS3- 1489C>T||IVS6+87A>C|
|Ex1+318A>G||−0.0719||−4.62 × 10−3||−0.0611|
|Ex4+51A>T||0.0272||0.0772||0.0309||−3.07 × 10−3|
|Ex7+62T>C||0.1307||4.49 × 10−3||−0.0983||0.0262||0.9261|
|Ex15+41C>T||−2.03 × 10−3||0.0337||−0.0651||0.0172||0.2625||0.1285|
|Ex17-79G>A||0.0564||5.33 × 10−3||−0.2623||4.72 × 10−3||0.1457||0.2392||−0.5486||0.0218|
Associations of Genotype with Overall Survival
The association of each genotype with OS was first analyzed in a relatively homogenous population of 154 patients who had resectable tumor and were treated on protocol for preoperative chemoradiotherapy. SNPs with a P value <.05 in the multivariate Cox regression models are listed in Table 4. Of the 26 SNPs evaluated, GCK IVS1 + 9652 C > T and HK2 N692N homozygous mutants were significantly associated with OS at the level of 5% false discovery rate (P < .002). The significant associations of both SNPs with OS were confirmed in the validation set of 552 patients (Table 4). In addition, the homozygous K variant of the nonsynonymous SNP HK2 R844K was significantly associated with a better OS in the validation set (P = .001). When data of the training set and the validation set were pooled to increase power, the significant associations of GCK IVS1 + 9652C>T, HK2 N662N, and R844K genotype with OS all remained highly significant.
|Genotype||Training Set||Validation Set||Combined Dataset|
|No. of Patients/No. of Deaths||MST||HR (95% CI)a||P||No. of Patients/No. of Deaths||MST||HR (95% CI)a||P||MST||HR (95% CI)a||P|
|CT||39/32||19.8||1.56 (1.03-2.38)||.03||149/117||16.0||1.08 (0.87-1.34)||.48||16.9||1.10 (0.90-1.33)||.35|
|TT||2/2||5.5||18.0 (3.63-89.5)||<.001||20/19||7.1||2.61 (1.61-4.21)||<.001||7.1||2.69 (1.71-4.22)||<.001|
|CC vs CT/TT||1.61 (1.06-2.45)||.026b|
|TC||39/31||21.5||0.84 (0.54-1.32)||.46||152/129||13.9||1.26 (1.02-1.55)||.03||16.0||1.25 (1.03-1.50)||.022|
|CC||2/2||4.9||7.64 (1.69-34.6)||.008||23/20||11.9||1.46 (0.91-2.33)||.12||11.9||1.57 (1.01-2.45)||.045|
|TT vs TC/CC||0.96 (0.61-1.49)||.84b|
|CG||43/29||27.5||0.61 (0.39-0.94)||.027||147/120||16.4||0.88 (0.71-1.09)||.25||17.8||0.85 (0.70-1.03)||.09|
|GG||6/2||—c||0.11 (0.03-0.48)||.003||23/18||22.9||0.60 (0.36-0.99)||.049||27.8||0.51 (0.32-0.83)||.006|
|CC vs CG/GG||0.53 (0.34-0.84)||.007b|
|GA||26/19||28.7||0.56 (0.34-0.93)||.02||59/45||20.5||0.80 (0.58-1.09)||.16||22.0||0.77 (0.59-1.02)||.06|
|GG||3/1||—c||0.21 (0.03-1.61)||.13||4/2||33.7||0.32 (0.08-1.33)||.12||33.7||0.31 (0.10-0.99)||.048|
|AA vs GA/GG||0.52 (0.31-0.86)||.01b||0.75 (0.55-1.03)||.07b||0.73 (0.56-0.96)||.02b|
|TC||59/53||17.6||1.94 (1.24-3.03)||.004||255/218||15.1||1.26 (1.01-1.57)||.037||15.3||1.39 (1.15-1.69)||.001|
|CC||16/16||8.3||5.62 (2.67-11.8)||<.001||108/89||12.0||1.80 (1.36-2.38)||<.001||11.6||2.01 (1.55-2.61)||<.001|
|GA||60/50||16.9||1.66 (1.10-2.48)||.01||291/242||15.0||1.29 (1.05-1.58)||.01||15.7||1.29 (1.08-1.55)||.006|
|AA||22/14||38.5||0.71 (0.38-1.30)||.27||66/53||16.9||1.10 (0.80-1.50)||.57||18.7||0.97 (0.73-1.29)||.85|
|GA||61/45||27.8||0.69 (0.46-1.04)||.07||182/144||18.9||0.77 (0.61-0.96)||.02||20.5||0.76 (0.62-0.93)||.008|
|AA||13/6||—c||0.38 (0.15-0.93)||.03||14/9||48.5||0.31 (0.16-0.63)||.001||84.3||0.37 (0.21-0.63)||<.001|
Next we analyzed the association of each genotype and OS by disease stage. In a total of 333 patients with localized disease, GFPT1-3094T>C, HK2 N692N and R844K were significantly associations with OS (Table 5). The GCK IVS1 + 9652C>T and HK2 N692N genotype showed some associations with OS among the 211 patients with locally advanced disease, but neither reached the significance level (P = .027 and .013). Among the 162 metastatic patients, GCK IVS1 + 9652C>T was the only SNP that had significant association with OS (P < .001). When data were pooled from patients with locally advanced and metastatic disease, GCK IVS1 + 9652C>T remained as the sole significant genetic predictor for OS (P ≤ .001).
|Stage||Genotype||No. of Patients/ No. of Deaths||MST||HR (95% CI)a||P|
|Locally advanced||GCK IVS1+9652C>T|
|CC/CT vs TT||1.01 (0.71-1.43)||.958|
|Locally advanced + metastatic||GCK IVS1+9652C>T|
Associations of Haplotype Diversity With OS
Haplotype frequencies and their associations with OS are described in Table 6. The GPI IVS6-378T>C, IVS9 + 2363C>G and G163G CGG haplotype was associated with a better OS (P = .004) and the CCG haplotype with a worse OS (P = .01). The different associations with OS of these 2 haplotypes were obviously determined by the IVS9 + 2363C>G genotype. Two haplotypes of HK2 gene, GCTATGG and ATTACAT, were associated with a better or worse OS, with a P value of .007 and .03, respectively, in multivariate Cox regression (Table 6). Two other haplotypes, GCCGCAT and ATTGCAT, showed nonsignificant associations with OS (P = .055 and .06). Apparently, haplotypes containing CAT of N692N, L766L, and Ex18 + 407T>G (3′UTR SNP) all conferred a worse OS.
|Haplotypea||Frequency||MST||HR (95% CI)b||P|
Associations of Genotype With Other Clinical Parameters
The association between each genotype and tumor response to therapy was evaluated in 261 patients who had resectable tumors and received preoperative chemoradiotherapy. HK2 N692N and R844K genotype showed associations with tumor response (Table 7). Interestingly, the genotype distribution of these 2 SNPs was significantly different by disease stage and tumor resection status. For example, the HK2 N692N CC genotype was detected in 12.6%, 19.0%, and 25.9% of patients with localized, locally advanced, and metastatic disease (P < .001, chi-square test). The HK2 R844K GG genotype was present in 52.9% of the patients with localized disease and 69.7% of the patients with advanced disease (P < .001, chi-square test).
|Variable||Genotype||No. (%)||No. (%)||P (Chi-Square Test)||ORa (95% CI)||Pa|
|Response to therapy|
|TT||126 (93.3)||9 (6.7)||1.0|
|TC||90 (77.6)||26 (22.4)||4.31 (1.84-10.1)||.001|
|CC||22 (62.9)||13 (37.1)||5.96 (2.09-17.0)||.001|
|GG||114 (47.9)||35 (72.9)||1.0|
|GA||105 (44.1)||13 (27.1)||0.36 (0.16-0.82)||.015|
|TT||156 (46.8)||112 (30.0)||1.0|
|TC||135 (40.5)||179 (48.0)||1.57 (1.10-2.25)||.013|
|CC||42 (12.6)||82 (22.0)||2.38 (1.46-3.86)||<.001|
|GG||176 (52.9)||260 (69.7)||1.0|
|GA||137 (41.1)||106 (28.4)||0.43 (0.30-0.64)||<.001|
|AA||20 (6.0)||7 (1.9)||0.29 (0.11-0.78)||.014|
|TT||146 (53.1)||122 (28.3)||1.0|
|TC||101 (36.7)||213 (49.4)||2.29 (1.58-3.31)||<.001|
|CC||28 (10.2)||96 (22.3)||3.62 (2.14-6.14)||<.001|
|GG||136 (49.5)||300 (69.6)||1.0|
|GA||118 (42.9)||125 (29.0)||0.43 (0.29-0.64)||<.001|
|AA||21 (7.6)||6 (1.4)||0.15 (0.05-0.42)||<.001|
We identified glucose metabolism gene variations associated with clinical outcome in pancreatic cancer. GCK IVS1 + 9652C>T, HK2 N692N and R844K in all patients, GFPT1 IVS14-3094T>C, HK2 N692N and R844K in patients with localized tumor, and GCK IVS1 + 9652C>T in patients with advanced diseases were significant independent predictors for OS. We also found a significant association of HK2 N692N and R844K genotype with disease stage, tumor resection status, and response to preoperative chemoradiotherapy. These data support a role of glucose metabolism gene polymorphisms in modifying the clinical outcome in pancreatic cancer.
Hexokinases catalyze the phosphorylation of glucose to glucose-6-phosphate. This is the first and rate-limiting step in glucose metabolism. HK2 localizes to the outer membrane of the mitochondria and is the major hexokinase isoform expressed in cancer cells.20 HK2 expression is insulin responsive and responsible for the accelerated glycolysis in cancer cells.21 Overexpression of HK2 in tumor tissues has been correlated to poor prognosis in breast cancer and liver cancer but not in pancreatic cancer,7, 22, 23 although the negative finding in pancreatic cancer could be partially explained by the heterogeneity of the study population.23 We observed 2 HK2 SNPs, R844K and N692N, significantly associated with OS, tumor stage, tumor resection status, and tumor response to therapy. HK2 R844K, an evolutionary conservative SNP, K variant, is computationally predicted to deleteriously affect protein coding and RNA splicing, which change the solvent accessibility and hydrophobicity of the protein.18 The K variant thus may confer a dysfunctional low enzymatic activity of HK2, impose restraint on glycolysis rate, and dampen tumor progression because of lack of energy supply. Indeed, a better response to therapy, a higher tumor resection rate, and a longer OS were observed among patients carrying the K variant allele (GA/AA genotype). The functional significance of the synonymous SNP HK2 N692N is unknown. By computational prediction, such an SNP may result in altered conformation, substrate affinity, and mRNA splicing.18 Whether such changes result in a higher enzyme activity, which may explain the association of the variant allele with reduced OS, needs further investigation. We observed that the HK2 N692N CC frequency in men was higher than that in women, and concluded that CC represents higher enzyme activity; whether the genotype difference contributes to a previously reported higher HK enzyme activity in men than women needs further investigation.24 We also observed that haplotypes containing variant alleles of HK2 N692N, L766L, and Ex18 + 407T>G (3′-UTR SNP) were associated with worse OS. It is possible that these genotypes/haplotypes conferred a higher level/activity of HK2, contributing to a higher rate of glycolysis, accelerated tumor progression, and reduced OS.
We found 3 intronic SNPs that were associated with OS, GCK IVS1 + 9652 TT genotype in patients with advanced diseases, GFPT1 IVS14-3094T>C in all patients and in patients with localized tumors, and a GPI haplotype containing the IVS9 + 2363C>G G allele in all patients. GCK is another member of the hexokinase family, catalyzing the adenosine triphosphate-dependent phosphorylation of glucose. Unlike HK2, GCK activity is not inhibited by its product glucose-6-phosphate, but remains active while glucose is abundant. GCK plays a role in maintaining glucose homeostasis as the glucose-sensor and glycolysis pacemaker involved in regulating insulin secretion.25 We speculate that there is an increased demand for glucose phosphorylation in advanced tumors because of the rapid cell growth, so GCK is required to maintain a constantly active glucose metabolism. The GFPT1 gene encoding glutamine-fructose-6-phosphate transaminase, the first and rate-limiting enzyme of the hexosamine biosynthesis pathway, controls the glucose flux into the hexosamine biosynthesis pathway. The hexosamine biosynthesis pathway is responsible for shuttling glucose to cellular glycosylation events, for example, promoting N-linked glycosylation of Wnt-related proteins.26 Glucose flux into the hexosamine biosynthesis pathway initiates post-translational modifications of cytoplasmic and nuclear proteins that regulate signal transduction, transcription, and protein degradation.27 GPI catalyzes the reversible isomerization of glucose-6-phosphate and fructose-6-phosphate, and plays a central role in glycolysis and gluconeogenesis. GPI can guide the glucose flow to the pentose phosphate pathway to produce nicotinamide adenine dinucleotide phosphate and pentose.28 GPI also functions as an autocrine motility factor, secreted from the tumor cells to promote cell migration, progression, and metastasis and to help the cells survive and proliferate under hypoxic and nutrient-deprived conditions.29 Although the functional significance of these intron SNPs is unknown, the variant alleles may affect the binding of transcriptional factors to the gene, thus up-regulating the mRNA and protein expression.18 The possibility that these SNPs are in linkage with unidentified functional SNPs could not be excluded.
OGT catalyzes the addition of a single N-acetylglucosamine in O-glycosidic linkage to serine or threonine residues. O-linked glycosylation plays a role in controlling gene expression, fuel metabolism, cell growth, differentiation, and cytoskeleton organization.30 We did not observe any significant association of the OGT genotype/haplotype with OS, partly because only 2 SNPs were examined in this study. Further study of this gene is warranted when additional SNPs are revealed by DNA sequencing.
Strengths of our study includes detailed clinical information, a large sample size, a 2-step design, and a hypothesis-driven gene selection. Limitations of the study include the limited number of genes and SNPs evaluated and the potentially false-positive findings owing to multiple comparisons. To keep the false discovery rate <5%, we applied a P value of .002 as the significance level in the genotype analysis. However, the frequencies of most homozygotes with major effects on clinical outcome are relatively low, so the possibility that these observations are by chance alone cannot be excluded. Additional studies with larger samples in different patient populations are required to confirm these findings. Furthermore, demonstrating the functional significances of these gene traits is pivotal in understanding their role in pancreatic cancer. Nevertheless, our findings provided supporting evidence for the importance of the glucose metabolism pathway in pancreatic cancer. Whether these genetic markers have a potential value in predicting response to glucose metabolism-targeted therapy in pancreatic cancer is under current investigation.
CONFLICT OF INTEREST DISCLOSURES
Supported by National Institutes of Health (NIH) RO1 grant CA098380 (to D.L.), SPORE P20 grant CA101936 (to J.L.A.), NIH Cancer Center Core grant CA16672, and Lockton Research Funds (to D.L.).