Figure 2. Telomere attrition vulnerability increases with cancer severity.
A. Telomere length (TL) decreases with participants’ age. TL is reported as an adjusted telomere : single-copy gene (T:S) ratio, derived from partial residuals of age in a generalized linear model (GLM) adjusted for sex, smoking status, cancer stage, lymphocyte ratio, neutrophil ratio, and batch. N = 100. In panels A-E, each dot represents one individual patient. B. No significant difference in measured TL between cancer cases (mauve) and matched healthy controls (purple). The paired comparison was performed using a linear mixed-effects model (LMM), with each linked line representing matched pairs. N = 24 for patients, 24 for controls. C. Age-associated telomere shortening was found only in cancer patients (mauve). Solid lines depict smoothed trends fitted by generalized additive models (GAM) using penalized regression splines, representing the optimal non-linear relationship between age and TL. Shaded areas around each line denote the 95% confidence intervals (CIs). D. TL decreases with increasing cancer stage severity. Each central bold line shows the mean TL of each stage group, with error bars representing 95% CIs. Statistical significance from pairwise comparisons between stage groups is denoted as *p < 0.05, **p < 0.01. A trend test assessing TL across cancer stages was performed by treating cancer stage as a continuous variable in a GLM. N = 18 for early stage, 45 for locally advanced stage, 29 for oligometastatic stage, 8 for widely metastatic stage. E. Age-associated telomere shortening was found in patients with metastatic cancer. Smoothed trend lines represent generalized additive model (GAM) fits with penalized regression splines, depicting the relationship between TL and age in patients with local (blue; includes early-stage and locally advanced, n = 63) and metastatic (green; includes oligometastatic and widely metastatic, n = 37) disease. Shaded areas around each line denote the 95% CIs. F. Association between patient characteristics and TL at baseline. Each panel displays the estimated relative TL changes (ratios and 95% CIs represented by dots and error bars) for each variable based on crude (univariate), adjusted (multivariate), and pooled (multiple imputation) analyses. The vertical dashed line at 1 indicates no change; values <1 indicate relatively shorter TL, and values >1 indicate relatively longer TL corresponding to changes in the variable indicated at left. Statistical significance is indicated by a 95% CI that does not cross 1. For categorical variables, effect estimates represent the ratio of TL relative to the indicated reference category. For continuous variables, estimates represent the relative TL change per one standard deviation (SD) increase. N = 100.