Telomere Length in White Blood Cell DNA and Lung Cancer: A Pooled Analysis of Three Prospective Cohorts

Telomeres shorten with each cell division, leading to loss of chromosomal and genetic integrity, and eventually either apoptosis, cellular senescence, or neoplasia (1). Shorter telomeres and telomerase inactivation are frequently observed in peripheral blood leukocytes of patients with cancer for many different malignancies, including lung cancer (2–6). However, most of these findings were derived from retrospective case–control studies, which may have shown telomere length shortening as a consequence of tumor progression. In contrast, there is growing evidence that longer telomere length is associated with increased risk for cancers (7–13). It has also been suggested that the direction of the association between telomere length and cancer may vary by cancer types (14). Positive associations between telomere length and lung cancer were recently reported in two prospective cohort studies of Caucasian male smokers and Asian female never-smokers (15, 16).

To investigate this association in more populations, we carried out a pooled analysis of the new prospective Prostate, Lung, Colorectal and Ovarian (PLCO) Cancer Screening Trial, which was conducted among men and women in the United States, and two previously published case–control studies nested in prospective cohorts: the Alpha-Tocopherol, Beta-Carotene Cancer Prevention (ATBC) trial conducted among male smokers in Finland and the Shanghai Women's Health Study (SWHS) cohort, which is comprised primarily of female never-smokers (15, 16). Here, we report the results from the pooled analysis as well as from each of the three studies individually.

Data from the following three case–control studies nested in prospective cohorts were pooled: The PLCO trial [403 lung cancer cases and 403 controls individually matched by age at baseline (±5 years), sex, race, PLCO study center, and date of baseline blood draw (±3 months); ref. 17], the ATBC trial [229 lung cancer cases and 229 controls individually matched on date of birth (±5 years)], and the SWHS [215 lung cancer cases and 215 controls individually matched on date of birth (±2 years) and date of blood sample collection (±3 months)], resulting in a total of 847 cases and 847 matched controls, individually matched by age, sex, and study. Briefly, the PLCO trial includes 77,500 men and 77,500 women ages 55 to 74 years who were recruited in the United States between September 1993 and July 2001. The average time between sample collection and diagnosis among cases was 7.41 years, and only the screening-arm participants provided DNA (18). The ATBC trial includes 29,133 male smokers 50 to 70 years of age, who were recruited from southwest Finland from 1985 to 1988, with an average of 5.23 years between sample collection and diagnosis among cases (19). The SWHS recruited 74,942 Chinese women aged 40 to 70 years who were primarily never-smokers between 1997 and 2000, with an average of 4.27 years between sample collection and diagnosis among cases (20).

All study participants provided written informed consent before participation, and the study protocols were approved by Institutional Review Boards of each study center and the NCI.

All blood samples were collected before diagnosis of lung cancer among cases. DNA was extracted using the phenol–chloroform method for the ATBC trial and the SWHS study, and using the ProMega ReliaPrep, a magnetic bead-based extraction method, for the PLCO trial. Telomere length for all three studies was measured in the same laboratory using the same monochrome multiplex qPCR method (15, 16, 21). Briefly, PCR (10 μL) was done using 10 mmol/L Tris–HCl pH 8.3, 50 mmol/L KCl, 3 mmol/L MgCl2, 0.2 mmol/L each dNTP, 1 mmol/L DTT, 1 mol/L betaine, 0.75× SYBR Green I, and AmpliTaq Gold DNA polymerase, 0.625 U. Four primers were used in each reaction: telg (at 100 nmol/L), telc (at 900 nmol/L), to amplify telomere sequences; and either hbgu (at 500 nmol/L) and hbgd (at 500 nmol/L) to amplify the β-globin gene, or albugcr2 (700 nmol/L) and albdgcr2 (500 nmol/L) to amplify the albumin gene. Relative telomere lengths were measured as T/S (telomere/single copy gene) values, which are proportional to the average telomere length per cell as described elsewhere (15, 16, 21). The main advantage of this method is that it eliminates pipetting steps that might introduce variation.

Cases were plated next to their matched controls to increase precision. Blind quality controls (5% of batch size) were assigned to all batches in random locations to evaluate assay reproducibility. All samples were assayed in triplicate, and the averaged data were used in the analysis. The intraclass correlation coefficient (ICC) of the assay was 87% for the PLCO trial, 80% for the ATBC trial, and 87% for the SWHS. The overall coefficient of variation (CV) was 7.3% for the PLCO trial, 7.0% for the ATBC trial, and 11% for the SWHS.

We compared characteristics of cases and controls for each of the three studies and pooled data using the Wilcoxon signed-rank test for continuous variables and the Fisher exact test for categorical variables. Correlations for continuous variables were evaluated using Spearman correlation coefficients. Telomere length was log-transformed because telomere length was right-skewed, and analyzed as both continuous and categorical (based on quartiles among controls). We also further categorized the fourth quartile (longest) of telomere length based on the lower and upper half to evaluate the association with very long telomere lengths. Analyses by telomere length quartile that retained the initial quartile categorization used in each study (to minimize the influence of potential assay variation across studies), and using a new categorization based on pooling the original telomere length data, produced similar findings, and results are presented for the former. We found no evidence of heterogeneity across studies using the Cochran Q test (Q = 0.97, P = 0.62) and therefore we pooled data from the three studies. We used conditional logistic regression to calculate ORs and 95% confidence intervals (CI) for the association between telomere length and lung cancer risk. All models were adjusted for age, to account for residual confounding, and pack-years of smoking as continuous variables. Other potential confounders were evaluated and did not affect the findings significantly; therefore, they were excluded from the final models. Tests of trend were calculated by treating the level of telomere length as a continuous variable using the original telomere length values and adjusting for age and pack-years of smoking. We also stratified the analysis by lung cancer subtypes (adenocarcinoma or squamous cell carcinoma), time between blood draw and lung cancer diagnosis (≤ median of 6 years or >6 years), sex, and smoking status (ever or never). Among cases, the association between lung cancer subtypes, time to diagnosis, and sex was evaluated using the Pearson χ² test. Only adenocarcinoma and squamous cell carcinoma were considered in the analysis as these two are the major subtypes (both subtypes consist >50% of cases) in all three studies, and the other subtypes were too small in numbers for meaningful analyses (Supplementary Table S1).

All statistical analyses were conducted using R, version 3.0.1 (http://www.R-project.org/), and SAS, version 9.3 (SAS Institute Inc.). All statistical tests were conducted as two-sided, and a P value of <0.05 was considered significant.

The pooled study population consisted of 847 incident lung cancer cases (mean age, 61.4 years; SD = 6.5) and 847 matched controls (mean age, 61.1 years; SD = 6.4; Table 1). Cases and controls were significantly different in their BMI (25.6 kg/m² among cases and 25.9 kg/m² among controls, P = 0.004), smoking status (71.9% smokers among cases and 54.5% smokers among controls, P < 0.001), and smoking pack-years (mean of 35.0 years among cases and 7.0 years among controls, P < 0.001). Telomere length was inversely correlated with age (Spearman correlation r = −0.19 in cases and −0.17 in controls, both P < 0.0001) and pack-years (Spearman correlation r = −0.38 in cases and −0.34 in controls, both P < 0.0001). Among controls, telomere length in the SWHS was longer than in the other two studies [median (interquartile range)]: 1.50 (0.34) in SWHS, 1.08 (0.31) in ATBC, and 1.13 (0.31) in the PLCO trial (Table 1 and Supplementary Table S2). This is at least in part due to women in the SWHS being almost all nonsmokers and younger on average than subjects from one (PLCO trial) of the other two cohorts (Table 1). Telomere length was significantly longer in cases compared with controls in the pooled population (Table 1).

Individuals with longer telomere length had significantly higher risk of lung cancer compared with individuals with shorter telomere length [OR (95% CI) by increasing quartile: 1.00; 1.24 (0.90–1.71); 1.27 (0.91–1.78); and 1.86 (1.33–2.62); P trend = 0.000022; Table 2]. Findings were consistent across the three cohorts and particularly evident for subjects with very long telomere length, i.e., lung cancer risks for telomere length [OR (95% CI)] in the upper half of the fourth quartile were 2.41 (1.28–4.52), 2.16 (1.11–4.23), and 3.02 (1.39–6.58) for the PLCO trial, the ATBC trial, and the SWHS, respectively (Table 2).

We found in the pooled study population that longer telomere length was associated with increased risk of lung cancer among both smokers and nonsmokers, and among both males and females (Table 3). Further stratification by tobacco use and sex showed similar associations, although the number of nonsmoking males was too small to analyze.

The association between telomere length and lung cancer persisted for cases diagnosed more than 6 years from time of blood draw to lung cancer diagnosis (Table 4). Further, the association was most apparent for adenocarcinoma cases [OR (95% CI) by quartile: 1.00; 1.54 (0.84–2.83); 1.16 (0.64–2.09); and 2.52 (1.38–4.60), P trend = 0.0029] versus the null finding for squamous cell cancer cases (P = 0.026 for difference in the association for adenocarcinoma vs. squamous cell carcinoma for highest vs. lowest telomere length quartile; Table 4).

A significant association between telomere length and lung cancer was also found among patients with adenocarcinoma diagnosed more than 6 years from time of blood draw to lung cancer diagnosis [OR (95% CI) by quartile: 1.00; 3.12 (1.17–8.30); 2.08 (0.84–5.12); and 4.61 (1.81–11.8), P trend = 0.0019]. Further, the association seemed to be stronger among female patients with adenocarcinoma [OR (95% CI) by quartile: 1.00; 1.46 (0.65–3.31); 1.01 (0.46–2.24); and 3.08 (1.39–6.84), P trend = 0.0090] compared with male cases (P = 0.0081 for difference, highest vs. lowest telomere length quartile; Table 5).

Our findings demonstrate a consistent, positive, and statistically significant association between longer leukocyte telomere length and risk of lung cancer—overall and among both smokers and nonsmokers. This paper adds new analyses from two previously published cohorts (ATBC trial and the SWHS; refs.15, 16) and new data from a third cohort (the PLCO trial). The novel pooled analyses of the combined data show strong and highly consistent effects across the three prospective cohorts and demonstrate for the first time that risks are particularly high for individuals who have extremely long telomere length. Further, we make the novel observation that the effect of long telomere length occurs well in advance (i.e., more than 6 years) of diagnosis so that it is unlikely that disease bias could influence the association. Finally, our findings demonstrate that the effect of long telomere length and lung cancer is particularly evident for adenocarcinoma, and especially among females.

Several small case–control studies have reported shorter telomeres to be associated with lung cancer (2, 22, 23), whereas results from our three prospective cohort studies have found the opposite effect (15, 16). At the same time, our findings contrast with a Danish prospective study that found a null association with lung cancer (14). It is possible that findings from case–control studies were driven by reverse causation bias (24). In our study, we observed a similar association among cases diagnosed more than 6 years after blood draw, and it is highly unlikely that telomere length in these subjects was altered by disease.

Our studies are ethnically diverse, including both Caucasians of European descent and Asians, and the positive association between longer telomere length and lung cancer was consistent in all three cohorts. Telomere length in the three studies was measured in the same laboratory, which developed the original assay and has extensive experience conducting analyses for epidemiologic studies. Assays were done in triplicate, cases were plated next to their matched control, and the measurements were precise with high ICC values and low CVs determined through the use of blinded quality control samples.

Both shorter and longer telomere lengths are biologically plausible in driving carcinogenesis (25). Longer telomere length may increase cancer risk by promoting immortality of the cells, leading to aberrant cell proliferation and tumor formation (26, 27). Consistent with recently published mechanistic evidence showing that longer telomeres may be associated with genomic instability (28), we found very long telomere length measured in peripheral white blood cell DNA to be associated with increased risk of lung cancer. Telomere length in blood leukocytes is a relatively good surrogate for telomere length in tissues (29) and may be used as a biomarker for lung cancer risk and potentially a screening tool for early cancer detection (30). Telomere shortening signifies replicative senescence, whereas telomere elongation signifies cell immortalization and unlimited cellular proliferation (31).

The effect of telomere length may differ according to cancer types and subtypes (14, 32). In our study, we also found the association between telomere length and lung cancer to vary by lung cancer subtype. Studies have reported differential gene expression, suggesting different tumorigenesis pathways between adenocarcinoma and squamous cell carcinoma (33, 34). Adenocarcinoma is more commonly found in women than men and is the most common lung cancer subtype among lifelong nonsmokers (35, 36). Consistent with this, our findings showed that the association within the patients with adenocarcinoma appeared to be stronger among females than males. We also observed women to have longer telomere lengths than men, especially nonsmoking women. Moreover, genetic studies have demonstrated relative telomerase activity and polymorphisms of the telomerase reverse transcriptase gene TERT, which is instrumental for telomere replication, to be associated with higher risk of adenocarcinoma and longer telomere length, suggesting that patients with adenocarcinoma may be affected by telomerase activity and therefore, telomere length (37–39). Limitations of this study include the relatively small sample size in the lung cancer subtype analyses, which necessitates replication of our results by a larger study.

In summary, we found a statistically significant positive association between longer telomere length and lung cancer risk and evidence that this association persisted among cases diagnosed more than 6 years after blood collection and differs by lung cancer subtype. Telomere length in white blood cell DNA may be an important biomarker of future increased risk of lung cancer.

R.M. Cawthon has ownership interest in a patent on method of measuring telomere length and is a consultant/advisory board member for Telomere Diagnostics, Inc.

The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the article.

Conception and design: X.-O. Shu, S. Min, S. Chanock, N. Rothman, Q. Lan

Development of methodology: R.M. Cawthon

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): R.M. Cawthon, M.P. Purdue, Y.-T. Gao, W.-Y. Huang, S.J. Weinstein, B.-T. Ji, J. Virtamo, X.-O. Shu, Y.-B. Xiang, S. Min, W.-H. Chow, D. Albanes, W. Zheng, N. Rothman

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): W.J. Seow, W. Hu, W.-Y. Huang, H.D. Hosgood III, N.E. Caporaso, S. Chanock, N. Rothman, Q. Lan

Writing, review, and/or revision of the manuscript: W.J. Seow, R.M. Cawthon, M.P. Purdue, Y.-T. Gao, W.-Y. Huang, S.J. Weinstein, B.-T. Ji, J. Virtamo, H.D. Hosgood III, B.A. Bassig, X.-O. Shu, Q. Cai, Y.-B. Xiang, W.-H. Chow, S.I. Berndt, C. Kim, U. Lim, D. Albanes, N.E. Caporaso, S. Chanock, W. Zheng, N. Rothman, Q. Lan

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): W.J. Seow, W. Hu, Y.-T. Gao, S.J. Weinstein, B.-T. Ji, Y.-B. Xiang, W. Zheng

Study supervision: W. Zheng, N. Rothman, Q. Lan

This work was supported by intramural funds from the NCI and extramural funds, including U.S. Public Health Service contracts N01-CN-45165, N01-RC-45035, N01-RC-37004, and HHSN261201000006C from the NCI, NIH, and Department of Health and Human Services.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.