Constitutive telomere length and gastric cancer risk: Case‐control analysis in Chinese Han population

Abstract

The shortening of telomeres may result in chromosome instability and thus promote tumorigenesis. Previous studies have demonstrated clear involvement of telomere shortening in the carcinogenesis of several malignancies. However, the association between constitutive telomere shortening and gastric cancer development has yet to be established. Therefore, in the present study, we measured average telomere length using quantitative real‐time PCR in peripheral blood lymphocytes from a gastric cancer (GC) case‐control study consisting of 396 cases and 378 controls. The results showed that GC patients had significantly shorter average telomere length than matched controls (mean ± SD 0.89 ± 0.19 vs 1.06 ± 0.25, P < 0.001). We further categorized telomere length using the 50% value in the controls as a cut‐off point and assessed the association between telomere length and GC risk using multivariate logistic regression analysis. We found that short telomere length was associated with a significantly increased GC risk (adjusted odds ratio = 2.14, 95% confidence interval = 1.52–2.93). Quartile stratification revealed a dose–response relationship between telomere shortening and GC risk (P for trend < 0.001). Stratified analysis showed that sex, age, and alcohol drinking, but not smoking and Helicobacter pylori infection, seem to have a modulating effect on the average telomere length in both cases and controls. We also found that telomere shortening and smoking had a significant joint effect on GC risk. Collectively, our findings provide the first evidence linking the short telomere length in peripheral blood lymphocytes to elevated GC risk, which warrants further investigation in other populations. (Cancer Sci 2009; 100: 1300–1305)

Abbreviations:

CI

confidence interval

GC

gastric cancer

OR

odds ratio

PBL

peripheral blood lymphocyte

S

number of single copy genes

T

number of telomere repeat products

The telomere is a specialized structure located at the very end of eukaryotic chromosomes. It is composed of tandem DNA sequence repeats (TTAGGG in human), which are surrounded by many specific telomere‐binding proteins.^{( 1} , ^{2 )} The main function of telomeres is to stabilize the chromosome ends and protect them from degradation, atypical recombination, and end to end fusion.^{( 3 )} In normal somatic cells, telomeres are progressively shortened with successive cell divisions, primarily as a result of incomplete replication of linear DNA molecules by conventional DNA polymerases (known as the end replication problem). Thus, age has been recognized as a factor inversely associated with telomere length. However, there is considerable variation in telomere length among normal individuals of the same age.^{( 4 )} In addition, cigarette smoking, oxidative stress, chronic inflammation, and epigenetic changes may also cause telomere shortening.^{( 5} , ⁶ , ^{7 )} The progressive shortening of the telomere limits the replicative capacity of human somatic cells and serves as a “mitotic clock” to define the lifespan of somatic cells.^{( 8} , ^{9 )} When the telomeres reach a critical length, cells undergo either irreversible growth arrest or apoptosis.^{( 10} , ^{11 )}

Critically short telomeres are significantly related to baseline and mutagen‐induced genetic instability possibly caused by a process of so‐called breakage‐fusion‐bridge.^{( 12} , ^{13 )} In fact, a number of previous studies have reported that short telomere length, as a measure of telomere dysfunction, is associated with the initiation and progression of malignant tumors, such as cancers of the breast,^{( 14 )} prostate,^{( 15 )} lung, bladder,^{( 12 )} and esophagus.^{( 16 )} Further evidence was also obtained from telomerase‐knockout mouse models, indicating that animals with critically short telomeres exhibit an increased cancer incidence. For example, Artandi and DePinho have reported that when telomere dysfunction is induced experimentally by deficiency in the telomerase RNA component (mTerc) in a p53 mutant mouse background, high levels of breast adenocarcinomas and other epithelial cancers are observed.^{( 17 )} These tumors do not normally occur in these mouse strains. They have also reported that increased telomere dysfunction is correlated with increased incidence of colon carcinomas using the same Terc‐null, p53‐mutant mice.^{( 18 )}

Gastric cancer is one of the most frequent malignant tumors in China. Although more and more environmental risk factors have been identified,^{( 19 )} such as cigarette smoking, alcohol consumption, Helicobacter pylori infection, and excessive salt intake, the genetic factors associated with sporadic GC remain mostly unclear. Maruyama et al. analyzed telomere length in biopsy samples from gastric mucosa and found that telomere length in the gastric mucosa became reduced as the mucosa underwent metaplasia and developed into adenoma.^{( 20 )} Similar results have also been reported by Yang's research group.^{( 21 )} Therefore, dysfunctional telomeres are considered an early initiating event in GC development by inducing chromosomal instability.

Previous data has suggested that the distribution of telomere lengths among chromosomes is genetically determined.^{( 22 )} In addition, a few case‐control studies have observed that individuals with shorter telomeres in PBL are at an increased risk for the development of human cancers, such as bladder,^{( 23 )} head and neck,^{( 12 )} lung,^{( 24 )} breast,^{( 14 )} and renal cell cancer.^{( 25 )} However, to date, the relationship between telomere length and GC susceptibility has not been reported. To further verify the role of telomere length in the risk of GC, we investigated the association between telomere length and GC risk in a case‐control study consisting of 396 GC cases and 378 matched healthy controls by measuring the average telomere length of all chromosomes in PBL.

Materials and Methods

Study population.  In an ongoing case‐control study, a total of 396 incident cases who were newly diagnosed with histologically confirmed primary gastric adenocarcinoma were recruited consecutively from the Department of Gastroenterological Surgery in two hospitals (Xijing and Tangdu) affiliated with The Fourth Military Medical University, Xi’an, Shaanxi, China, between October 2007 and July 2008, which represented 90% of all new cases diagnosed at the same study period in both recruitment places. All cases had no prior chemotherapy or radiotherapy. No age, sex, or disease stage restrictions for case recruitment were used. The 378 healthy controls without previous cancer history (except non‐melanoma skin cancer) were recruited during the same time period as cases recruited from individuals who visited the same hospital for physical examination with a response rate of approximately 75%. The controls were frequency matched to the cases on the basis of age, sex, and residential area. When matching, age was classified into five 10‐year groups and the residential area was divided into two categories: Xi’an area and others. The ethnicity of all participants was Chinese Han.

Epidemiological data.  After signed informed consent was obtained from each individual, all participants were interviewed by trained staff interviewers to collect information regarding demographics, smoking history, alcohol consumption, dietary habits, and family history of cancer by using a standardized epidemiological questionnaire. After interview, a venous blood sample from each subject was drawn into coded tubes (3 mL into a heparinized tube and 2 mL into a regular tube) and delivered to the laboratory for analysis. Blood samples were obtained for 92.0% cases and 99.5% controls. This study was approved by the institutional review board of The Fourth Military Medical University.

Measurement of serum antibody IgG to H. pylori.  The 2 mL of coagulated blood was centrifuged for 10 min at 400 g to collect the serum. The serum was then divided into three aliquots for storage in –80°C. H. pylori infection in all subjects was determined by a pylori DTect test using a commercial IgG enzyme‐linked immunosorbent assay kit (Diagnostic Technology, Pymble, Australia) according to the manufacturer's instructions. The test has been validated in Chinese populations with high sensitivity and specificity for detection of H. pylori infection.^{( 26 )}

Extraction and quantification of genomic DNA.  Genomic DNA was extracted from the 3 mL of whole blood (anticogulated) using the EZNA blood DNA midi kit (Omega Bio‐Tek, Norcross, GA, USA) according to the manufacturer's protocol. The concentration of DNA was measured by absorbance at 260 nm. DNA purity is determined by calculating the ratio of absorbance at 260 nm to the absorbance at 280 nm. Pure DNA has an A₂₆₀/A₂₈₀ ratio of 1.7–1.9. The DNA samples were then diluted with sterile water to 1 ng/µL and stored at –30°C until analysis.

Average telomere length assessment by quantitative PCR.  Average telomere length was measured by a quantitative PCR‐based method described by Cawthon,^{( 27 )} using an Applied Biosystems 7500 PCR System (Foster City, CA, USA). The same method in principle was also used to measure relative telomere length in other previous studies.^{( 23} , ^{24 )} This is currently the most economical and high‐throughput method for the assessment of telomere length, especially in epidemiological studies with a large number of samples. The traditional method for measuring telomere length in human genomic DNA is to determine terminal restriction fragment length by use of Southern blotting.^{( 28 )} This method can directly evaluate the absolute telomere length; however, it requires a large amount of DNA (0.5–5 µg/individual) and time (3–5 days). Furthermore, the relative mean terminal restriction fragment lengths of individuals can vary by as much as 5% depending on the particular restriction enzymes used. In addition, the quantitative fluorescence in situ hybridization laser scanning cytometry technique has also been used to measure telomere length, such as in the study by Wu et al.^{( 12 )} This method is not as simple or as amenable to rapid high‐throughput processing of large numbers of samples. Also it has a higher intra‐ and interassay variation compared with other methods. In the present study, the relative telomere length was determined by real‐time PCR through two steps of relative quantification. First, the ratio of T to S (36B4, encoding acidic ribosomal phosphoprotein PO) was determined for each sample using standard curves. The T/S ratio of one individual relative to another should correspond to the relative telomere lengths of their DNA. Then, the T/S ratio for each sample was normalized to a calibrator DNA in order to standardize between different runs. All samples were assayed in duplicate on a 96‐well plate. The telomere and 36B4 PCR reactions were always carried out on separate 96‐well plates with the same samples in the same well positions. The primers used were Tel.1 (GGTTTTTGAGGGTGAGGGTGAGGGTGAGGGTGAGGGT) and Tel.2 (TCCCGACTATCCCTATCCCTATCCCTATCCCTATCCCTA) for amplification of telomeric repeats, and 36B4‐1 (CAGCAAGTGGGAAGGTGTAATCC) and 36B4‐2 (CCCATTCTATCATCAACGGGTACAA) for the single‐copy gene 36B4 reaction. The PCR reaction (total volume 25 µL for each reaction) for the telomere and 36B4 amplification consisted of 1 × SYBR Green mastermix (Takara, Dalian, China), 300 nM Tel.1 (or 300 nM 36B4‐1), 800 nM Tel.2 (or 300 nM 36B4‐2), and 10 ng of genomic DNA. The thermal cycling profile for the telomere amplification was 95°C for 10 min, followed by 35 cycles of 95°C for 15 s and 54°C for 2 min, and for the 36B4 amplification it was 95°C for 10 min, followed by 35 cycles of 95°C for 15 s and 60°C for 1 min.

A standard curve of a diluted reference DNA (the same DNA sample for all runs), two negative controls, two positive controls from two different tumor cell lines, as well as the calibrator DNA were included in each run. For each standard curve, one reference DNA sample was diluted serially by 4‐fold per dilution to produce a 5‐point standard curve between 40 and 0.625 ng DNA in each reaction. The data were analyzed with ABI SDS software (version 1.2, ABI Company, Foster City, CA, USA). R² for each standard curve was ≥0.98. Standard deviations (for Cycle of threshold (Ct) values) were accepted at 0.2. Otherwise, the sample was repeated. For testing the interassay variation, two samples with relatively low and high telomere lengths were tested using three different runs. For testing the intra‐assay variation, these two samples were also tested with eight replicates for each sample in the same run. Researchers performing the laboratory assays were blinded to the case‐control status of the samples.

Statistical analysis.  All statistical analyses were done using the Statistical Analysis System (Version 9.1.3; SAS Institute, Cary, NC, USA). Smoking and drinking status were categorized as dichotomized variables. Individuals who had smoked less than 100 cigarettes in his or her lifetime were defined as never smokers, and those that consumed three and more standard cups of alcohol a week for over 6 months were considered as ever drinkers. Pearson χ² test was used to examine differences in the distribution of cases and controls in terms of sex, age, smoking, and drinking status. Student's t‐test was used to analyze normally distributed continuous variables, such as pack‐years and average telomere length. The average telomere length was also analyzed as a categorical variable by grouping it based on the median or quartile values in the controls. The association between GC risk and average telomere length was estimated using OR along with corresponding 95% CI. To account for the potentially confounding effects of age, sex, H. pylori infection, smoking, and drinking status, unconditional logistic regression analysis with multiple covariates was carried out. In addition, subjects were dichotomized by age according to the median age of the controls. Then, stratified analyses were carried out to compare average telomere length among different subgroups of cases or controls, to assess the GC risk associated with telomere length shortening in those subgroups, and to evaluate the joint effect of average telomere length and smoking status on GC risk. All P‐values were based on two‐sided tests. A probability level of 0.05 was used as the criterion for statistical significance.

Results

The characteristics of the GC cases and controls are summarized in Table 1. The cases and controls were well matched in terms of sex (P = 0.48) and age (P = 0.97). The mean age of cases and controls was 52.7 and 53.1 years respectively. However, there were statistically significant differences between the cases and the controls with respect to smoking status, pack‐years, drinking status, and H. pylori infection. More patients with GC than controls reported a history of smoking (54 vs 35%; P < 0.001) and drinking (44 vs 29%; P < 0.001). GC patients had a significantly higher percentage of H. pylori infection than controls (68 vs 52%; P < 0.001), indicating that H. pylori infection might be involved in GC development as a risk factor. This finding is in agreement with previous reports in a Chinese population. In our study, cases and controls were recruited around the Xi’an area where there is a higher incidence of GC and higher rate of H. pylori infection compared with other places in China. Therefore, H. pylori positivity seems not to be overrepresented in the GC cases. This result also suggests that other factors, such as somatic genetic variants and smoking, might cooperate with H. pylori infection to promote gastric carcinogenesis. In addition, GC patients were self‐reported heavier smokers than control subjects (45.9 ± 26.1 vs 34.3 ± 22.1, P < 0.001). The real‐time PCR was used to measure the average telomere length in PBL from all subjects with 7.4 and 4.8% of mean interassay coefficient variation and intra‐assay coefficient variation respectively. We found that the average telomere length was significantly shorter in cases than in controls (mean ± SD, 0.89 ± 0.19 vs 1.06 ± 0.25; P < 0.001).

Table 1.

Distribution of selected characteristics of gastric cancer cases and healthy controls

Variable	Case (n = 396) n (%)	Control (n = 378) n (%)	P‐value
Sex
Male	258 (65)	237 (63)
Female	138 (35)	141 (37)	0.48
Smoking status
Never	182 (46)	246 (65)
Ever	214 (54)	132 (35)	<0.001
Alcohol drinking
Never	222 (56)	268 (71)
Ever	174 (44)	110 (29)	<0.001
HP infection
Yes	269 (68)	197 (52)
No	127 (32)	181 (48)	<0.001
Age (years)
<40	39 (10)	33 (9)
40–50	86 (22)	80 (21)
50–60	148 (37)	144 (38)
60–70	90 (23)	91 (24)
>70	33 (8)	30 (8)	0.97
Pack‐years † , mean (SD)	45.9 (26.1)	34.3 (22.1)	<0.001
Average telomere length	0.89 (0.19)	1.06 (0.25)	<0.001

^† Ever smokers only. HP, Helicobacter pylori.

We carried out unconditional logistic regression analysis to assess the association between average telomere length and GC risk (Table 2). We first dichotomized the average telomere length into long and short groups by using the median value in the controls as the cut‐off value. After adjustment for the confounding effects of age, sex, smoking status, drinking status, and H. pylori infection, we found that individuals with short average telomere length had a significantly increased risk of GC, with an OR of 2.14 (95% CI 1.52–2.93). Next, we categorized the subjects into four groups based on the quartile values of average telomere length in the controls. We observed a dose–response relationship between GC risk and telomere length shortening. The χ² tests for trend were significant (P < 0.001). Using the subjects in the first quartile as a reference, the adjusted OR (95% CI) of average telomere length for individuals in the second, third, and fourth quartiles were 1.45 (95% CI 0.97–2.23), 2.11 (95% CI 1.36–3.24), and 3.12 (95% CI 2.01–4.79) respectively.

Table 2.

Gastric cancer risk as estimated by average telomere length

Average telomere length	Case, n (%)	Control, n (%)	Adjusted OR † (95% CI)
By median
Long	127 (32)	189 (50)	1 (Reference)
Short	269 (68)	189 (50)	2.14 (1.52–2.93)
By quartile
First	51 (13)	94 (25)	1 (Reference)
Second	76 (19)	95 (25)	1.45 (0.97–2.23)
Third	110 (28)	95 (25)	2.11 (1.36–3.24)
Fourth	159 (40)	94 (25)	3.12 (2.01–4.79)
P for trend			<0.001

^† Adjusted by age, sex, Helicobacter pylori infection, smoking, and drinking status.

CI, confidence interval; OR, odds ratio.

We also assessed average telomere length according to host characteristics (Table 3). A modulating effect of sex, age, and alcohol drinking on average telomere length was found in both cases and controls. Female persons had a significant longer average telomere length than male persons for both GC cases (0.93 ± 0.21 vs 0.87 ± 0.15; P = 0.003) and controls (1.14 ± 0.27 vs 1.01 ± 0.22; P < 0.001). Individuals at least 53 years old had a significantly shorter average telomere length than did those younger than 53 years among the GC cases (0.85 ± 0.17 vs 0.93 ± 0.20; P < 0.001) and healthy controls (0.99 ± 0.23 vs 1.13 ± 0.26; P < 0.001). A similar trend for age‐related telomere shortening was also obtained in both cases and controls when age was stratified into four groups (data not shown). In addition, there was a significantly longer telomere length in ever drinkers than that in never drinkers for both cases (0.92 ± 0.21 vs 0.87 ± 0.16; P = 0.010) and controls (1.11 ± 0.28 vs 1.04 ± 0.19; P = 0.016). However, there were no significant associations between average telomere length and smoking status or H. pylori infection in either GC cases or controls. We next carried out stratified analysis to examine GC risk associated with average telomere length shortening by selected host characteristics (Table 4). There was no significant difference for GC risk associated with shorter telomere length among different subgroups. In addition, average telomere length and smoking status were shown to have a joint effect on GC risk by comparing the risks for ever smokers with short telomere length, never smokers with short telomere length, and ever smokers with long telomere length against the risk for never smokers with long telomere length (i.e. the reference group) (Table 5). In brief, the GC risk for ever smokers with short telomere length was higher than that for never smokers with short telomere length and ever smokers with long telomere length (OR [95%CI], 4.87 [3.09–6.97] vs 2.31 [1.54–3.44] vs 2.52 [1.59–3.98] respectively).

Table 3.

Comparison of average telomere length among different subgroups in GC cases or controls

Subgroup	Cases			Controls
	n	Mean (SD)	P‐value*	n	Mean (SD)	P‐value*
Sex
Male	258	0.87 (0.15)	0.	237	1.01 (0.22)
Female	138	0.93 (0.21)	0.003	141	1.14 (0.27)	<0.001
Age (years)
<53	194	0.93 (0.20)		188	1.13 (0.26)
≥53	202	0.85 (0.17)	<0.001	190	0.99 (0.23)	<0.001
Smoking status
Never	182	0.90 (0.20)		246	1.05 (0.22)
Ever	214	0.88 (0.18)	0.298	132	1.08 (0.27)	0.270
Alcohol drinking
Never	222	0.87 (0.16)		268	1.04 (0.19)
Ever	174	0.92 (0.21)	0.010	110	1.11 (0.28)	0.016
HP infection
Yes	269	0.88 (0.18)		197	1.07 (0.20)
No	127	0.91 (0.22)	0.180	181	1.05 (0.26)	0.410

^* P‐values were determined by Student's t‐test to assess the difference of average telomere length between two different subgroups in cases or controls.

HP, Helicobacter pylori.

Table 4.

Estimates of gastric cancer risk associated with average telomere length stratified by selected variables

Average telomere length	Cases, n (%)	Controls, n (%)	OR (95% CI) †
Sex
Male
Long	88 (34)	117 (49)	1 (Reference)
Short	170 (66)	120 (51)	1.85 (1.37–2.69)
Female
Long	39 (28)	72 (51)	1 (Reference)
Short	99 (72)	69 (49)	2.66 (1.69–4.35)
Age (years)
<53
Long	68 (35)	96 (51)	1 (Reference)
Short	126 (65)	92 (49)	1.92 (1.24–2.97)
≥53
Long	59 (27)	93 (49)	1 (Reference)
Short	143 (73)	97 (51)	2.59 (1.68–4.12)
Drinking status
Never
Long	71 (32)	135 (50)	1 (Reference)
Short	151 (68)	133 (50)	2.14 (1.42–3.05)
Ever
Long	56 (32)	54 (49)	1 (Reference)
Short	118 (68)	56 (51)	2.01 (1.29–3.35)
HP infection
Yes
Long	91 (34)	97 (49)	1 (Reference)
Short	178 (66)	100 (51)	1.84 (1.25–2.65)
No
Long	36 (28)	92 (51)	1 (Reference)
Short	91 (72)	89 (49)	2.66 (1.60–4.36)

^† Adjusted for age, sex, Helicobacter pylori (HP) infection, smoking, and drinking status, where appropriate.

Table 5.

Joint effect of telomere length and smoking in gastric cancer risk

Telomere length	Smoking status	Cases, n (%)	Controls, n (%)	Adjusted OR † (95%CI)
Long	Never	57 (14)	127 (34)	1 (Reference)
Short	Never	125 (32)	119 (31)	2.31 (1.54–3.44)
Long	Ever	70 (18)	62 (16)	2.52 (1.59–3.98)
Short	Ever	144 (36)	70 (19)	4.87 (3.09–6.97)

^† Adjusted for age, sex, Helicobacter pylori infection and drinking status.

Discussion

In our study, we evaluated the average telomere length in PBL from GC patients and controls by using real‐time PCR. Our findings demonstrated that the GC cases exhibited significantly shorter telomere length than healthy controls. We also presented evidence of an increased risk for GC associated with progressively shorter telomere length. In addition, a significant joint effect was indicated to exist between smoking and telomere length in elevating GC risk. This is the first epidemiological study to link average telomere length with EC risk.

In a previous study, Wu et al. demonstrated that telomere shortening in PBL is significantly associated with an increased cancer risk of head and neck, bladder, lung, and kidney.^{( 12 )} Broberg et al.^{( 23 )} and McGrath et al.^{( 29 )} confirmed the association of telomere length with cancer risk by showing that constitutive short overall telomere length in PBL is a strong genetic susceptibility marker for bladder cancer. In addition, Jang et al. also reported that a short telomere is associated with a significantly increased risk of lung cancer.^{( 24 )} In the present study, for the first time, we obtained a similar result in GC, suggesting that telomere shortening in PBL may be useful as a predisposition marker to assess GC risk. A series of previous studies provided strong biological support for our findings by showing that telomere dysfunction is an early and common genetic alteration acquired in the multistep process of malignant transformation and telomere shortening leads to increased frequencies of chromosome instability.^{( 30} , ^{31 )} Animal studies have shown that mice with shorter telomeres have an increased incidence of tumors and enhanced risk of epithelial cancers due to the formation of complex non‐reciprocal translocations (a classical cytogenetic feature of human carcinoma).^{( 18} , ^{32 )} Clinical observations of tumor tissues showed that telomere length in patients with colorectal carcinoma or prostate cancer is shorter than in matched normal controls.^{( 33} , ^{34 )} In addition, cells with telomere dysfunction are also found to be associated with decreased DNA repair capacity and complex cytogenetic abnormalities.^{( 35 )} Taken together, our result of an association between short telomeres and GC risk is comparable with the telomere‐driven chromosomal instability theory, and suggests that individuals with constitutively short telomeres may be more prone to acquiring chromosome instability, and therefore be at an increased risk for GC development.

In our study, stratified analysis was performed by host characteristics. Our findings showed that telomere length was significantly shorter in the male subjects than in the female subjects among both cases and controls, suggesting a modulating effect of age on average telomere length, which is possibly attributed to the different kinds and levels of sex hormone. This result is consistent with previous observations.^{( 36} , ^{37 )} Telomere length is a biomarker of biological age, and age‐dependent shortening of telomeres in most somatic cells impairs cellular function and viability.^{( 38 )} Therefore, we evaluated the correlation between age and average telomere length. Our results support earlier findings by Valdes et al.^{( 5 )} and O'Sullivan et al.^{( 39 )} that there is an inverse relationship between telomere length and age. Cigarette smoking is an important risk factor for GC. In this current study, we also examined the effect of smoking on telomere shortening. Although it has been reported that telomere length decreases in a dose‐dependent manner as the amount of cigarettes smoked increases,^{( 5} , ^{40 )} we did not observe an association between the average telomere length and smoking status in either GC cases or controls. However, consistent with a previous study,^{( 12 )} we found a significant joint effect between average telomere shortening and smoking in elevating cancer risk. For the first time, our findings indicated that ever drinkers had a significantly longer telomere length than never drinkers. Telomerase activation by alcohol drinking in target tissue might be a potential explanation for this result. If true, these observations in our study could point toward differences in the effects of some host characteristics on average telomere length. These observations support the notion of gene–environment interactions in cancer etiology. However, we could not rule out the possibility of chance findings because of the limited size in each subgroup. Further investigations are needed to confirm these findings.

In the present study, average telomere length was measured in a surrogate tissue, PBL. It may be questioned whether the telomere length in PBL is representative of those in target tissues. Friedrich et al. measured telomere length in three unrelated tissues and found a significant linear correlation in each pair of the two different tissues donated by the same participant.^{( 41 )} A study of monozygotic and dizygotic twins has shown that the majority of the interindividual variation in telomere length is genetically determined.^{( 42 )} These results support the idea that easily accessible tissues such as blood could serve as surrogates for target tissues when measuring the relative telomere length.

Our study had its own strengths and limitations. We used a case‐control study design with a large sample size, which can significantly improve the statistical power. However, we could not determine the cause‐and‐effect relationship between average telomere length and GC development in this retrospective analysis. Further confirmation of this relationship using a prospective study design is warranted.

In summary, our data show for the first time that shorter telomere length in PBL is associated with increased GC risk. This study is an initial step to demonstrate that telomere length can be used as a marker to assess GC risk. Further investigations are needed to validate our findings and determine the underlying mechanism of telomere shortening in GC development and progression. In addition, although many GC risk factors have been elucidated and a series of molecular and cytogenetic biomarkers have been identified for GC risk assessment over the last three decades, to date, no single biomarker can be used specifically for effective screening of the high‐risk GC subgroup. More and more evidence indicates that a single biomarker commonly only plays a weak role in cancer risk prediction. Therefore, a statistical risk prediction model incorporating multiple risk factors and biomarkers shows a promising future for identification of the high‐risk GC subgroup. Telomere length evaluation might be used in future establishment of such a model.