• Open Access

Telomere length in white blood cells is not associated with morbidity or mortality in the oldest old: a population-based study


Prof. Thomas von Zglinicki, Henry Wellcome Laboratory for Biogerontology, University of Newcastle, Newcastle General Hospital, Newcastle upon Tyne, NE 46BE, UK. Tel.: +44 191 256 3310; fax: +44 191 256 3445; e-mail: t.vonzglinicki@ncl.ac.uk


Cross-sectional studies have repeatedly suggested peripheral blood monocyte telomere length as a biomarker of aging. To test this suggestion in a large population-based follow-up study of the oldest old, we measured telomere length at baseline in 598 participants of the Leiden 85-plus Study (mean age at baseline 89.8 years). We also obtained second telomere measurements from 81 participants after an average time span of between 3.9 and 12.9 years. Telomere length at baseline was not predictive for mortality (P > 0.40 for all-cause, cardiovascular causes, cancer or infectious diseases, Cox regression for gender-adjusted tertiles of telomere length) or for the incidence of dementia (P = 0.78). Longitudinally, telomere length was highly unstable in a large fraction of participants. We conclude that blood monocyte telomere length is not a predictive indicator for age-related morbidity and mortality at ages over 85 years, possibly because of a high degree of telomere length instability in this group.


In human somatic cells, telomeres shorten with cell proliferation due to numerous factors including chromosome end-related inability of the DNA replication machinery and insufficient repair of oxidative damage (von Zglinicki, 2002). Telomerase can preserve telomere length by adding tandem repeats de novo at chromosome ends; however, its activity in somatic cells and haematopoetic progenitor cells is too low to enable full maintenance of peripheral blood mononuclear cell (PBMC) telomere length. Accordingly, average PBMC telomere length shortens with donor age, although there is high variation between individuals of the same age. Telomere length is under strict genetic control, and the intraindividual correlation between telomere lengths in different tissues is high (von Zglinicki et al., 2000). Recently, correlations between PBMC telomere length and the incidence of major age-related diseases, i.e. vascular dementia (von Zglinicki et al., 2000), Alzheimer's dementia (Panossian et al., 2003), myocardial infarction (Brouilette et al., 2003), atherosclerosis (Benetos et al., 2004) or solid tissue tumours (Wu et al., 2003), have been found. One retrospective study reported a significantly higher mortality risk, specifically from cardiovascular or infectious diseases, for individuals with shorter telomeres (Cawthon et al., 2003).

In vitro, short telomeres trigger cellular replicative senescence (Bodnar et al., 1998; d’Adda di Fagagna et al., 2003). Therefore, short telomeres in PBMC might be indicative of immunosenescence (Effros & Pawelec, 1997), while failure of immunosurveillance may contribute to the development of age-related diseases (Franceschi et al., 2000; Pawelec et al., 2004). Moreover, oxidative damage to telomeric DNA is a major cause of telomere shortening (von Zglinicki, 2002), and PBMC telomere length in vivo is stress-dependent (Epel et al., 2004). PBMC telomere length could thus be a biomarker of the individual's cumulative exposure to oxidative stress and his or her ability to mount an effective antioxidant response, both regarded as prognostic indicators of age-related morbidity and mortality.

While the available data point to PBMC telomere length as a potential biomarker of aging, it must be stressed that there are few prospective studies and no longitudinal data on telomere dynamics in humans. Moreover, most published studies have been performed in relatively small and/or relatively young collectives. Thus, we tested the hypothesis that telomere length in white blood cells is predictive for morbidity and mortality in a large population-based sample of the oldest old including, for the first time, longitudinal data on telomere dynamics in individual adults.


In the cross-sectional sample at baseline, average telomere length of all participants was 4310 ± 906 bp. This is lower than typical telomere restriction fragment (TRF) lengths measured by Southern blotting (Benetos et al., 2004; Valdes et al., 2005), but higher than reported telomere lengths measured by real-time PCR (Cawthon et al., 2003; Epel et al., 2004), probably because our data are standardized against TRF length in every single batch of measurements (see Experimental procedures). PBMC telomere length was not significantly associated with age (85–101 years; telomere change rate 5 ± 13 bp year−1). This is in accord with published data that show a continuous decrease of the rate of telomere loss with increasing age (Frenck et al., 1998; Mondello et al., 1999; Rufer et al., 1999; Unryn et al., 2005). The oldest populations examined so far (median ages between 78 and 80 years) had telomere loss rates of only 14–15 bp year−1 (Mondello et al., 1999; Cawthon et al., 2003). Baseline PBMC telomere length was not associated with a history of cardiovascular disease or hypertension (Table 1). However, a history of myocardial infarction was more than twice as frequent in participants in the lowest tertile of telomere length (P = 0.038, Table 1) in accord with published data (Brouilette et al., 2003). There were no significant associations between telomere length and prevalence of diabetes, dementia and cognitive impairment (Table 1).

Table 1.  Cross-sectional correlates of telomere length at baseline
 Tertiles of telomere lengthP-value*
Lowest(n = 225)Intermediate(n = 228)Highest(n = 226)
  • *

    Differences in proportions were tested by Chi-squared (linear by linear), and differences in continuous variables were tested by anova test for linearity.

Females164 (73%)166 (73%)165 (73%)
Age at baseline (year) 90.0 (SD 3.2) 89.7 (SD 3.1) 89.5 (SD 2.7)0.069
Institutionalized living124 (56%)110 (50%)122 (55%)0.8
History of cardiovascular disease 68 (33%) 60 (29%) 62 (31%)0.66
History of hypertension 51 (25%) 44 (21%) 51 (26%)0.9
History of myocardial infarction 23 (11%) 17 (8%) 11 (5%)0.038
History of diabetes 25 (12%) 30 (14%) 26 (13%)0.8
Diagnosis of dementia 46 (23%) 49 (24%) 51 (25%)0.7
MMSE score (points) 24.0 (SD 6.1) 23.5 (SD 7.0) 24.1 (SD 6.1)0.9

We obtained a second blood sample after an average follow-up of 3.7 years in 67 subjects and of 12.9 years in 14 subjects. These data confirmed the absence of a significant telomere shortening with time in this age group as a whole. However, they also showed that PBMC telomeres in individual subjects can either shorten or lengthen with time by amounts greatly exceeding the methodological variation (Fig. 1). Telomere shortening rates exceeding 200 bp year−1, i.e. more than tenfold the rate measured in cross-sectional studies for people over 50 years of age (Mondello et al., 1999; Cawthon et al., 2003; Unryn et al., 2005), were measured in 11 out of 81 participants (14%). In a further 8 participants (10%), telomeres became elongated between the first and second measurement by more than the 99% prediction interval as defined by repeated measurements in the young control group (Fig. 1). These telomere dynamics did not correlate which differences in white blood cell counts or frequencies of B cells, CD4+, CD8+ or CD16+ cells over time (data not shown).

Figure 1.

Instability of PBMC telomere length in the oldest old. Every data point compares the telomere length measurements taken at two different time points from the same individual. Filled dots: controls; age range, 25–54 years; time interval between blood samples, 10 days. The regression line (solid line) and the 99% prediction interval (dotted lines) for the controls are indicated. Open squares: study participants; mean time interval, 3.7 years. Open triangles: study participants; mean time interval, 12.9 years.

In the prospective analysis, telomere length at baseline was not predictive for all-cause mortality (Fig. 2, 598 events), neither for mortality from cardiovascular causes (242 events), from infectious diseases (67 events) or from malignancies (94 events, Table 2, all P > 0.40). Baseline telomere length did not predict the incidence of dementia during an average follow-up period of 3 years (P = 0.78) or a change in Mini-Mental State Examination (MMSE) scores during this period (P = 0.48). Moreover, there was no significant difference in terms of age-related mortality between subjects showing gain or loss in telomere length and between subjects showing fast or slow rates of telomere length change (data not shown).

Figure 2.

Survival of the participants of the Leiden 85-plus Study (n = 598) is not dependent on telomere length. Kaplan–Meier curves with left censoring for age at baseline, the tertiles are gender dependent.

Table 2.  Mortality risks depending on telomere length at baseline
 Tertiles of telomere length at baselineP trend
  • Tertiles of telomere length were determined for both sexes separately and then pooled.

  • *

    reference category.

  • Hazard ratios and corresponding 95% confidence intervals were estimated by Cox regression analyses with left censoring to adjust for differences in age at baseline.

Death from all causes11.07 (0.87–1.30)1.00 (0.82–1.21)0.98
Death from cardiovascular causes10.93 (0.67–1.27)1.02 (0.76–1.37)0.87
Death from infectious diseases11.45 (0.79–2.65)1.30 (0.72–2.39)0.41
Death from malignancies11.13 (0.70–1.86)0.81 (0.48–1.35)0.43


Cross-sectional data suggest a continuous loss of PBMC telomeres over the whole human lifespan, even if the rate of loss might diminish with advancing age (Frenck et al., 1998; Mondello et al., 1999; Rufer et al., 1999; Unryn et al., 2005). Telomere loss, like aging itself, is related to oxidative stress, both in vitro (von Zglinicki, 2002) and in vivo (Epel et al., 2004). Short telomeres are believed to contribute to immunosenescence (Effros & Pawelec, 1997). This would strongly suggest a closer association between short telomeres and age-related morbidity and mortality as people grow older. However, our data show that in the oldest population examined so far, telomere length in peripheral blood is not associated with survival or age-related disease at all, but is associated with an unexpectedly low longitudinal stability of PBMC telomere length. Because we performed rigorous quality testing of our DNA and could confirm very low sample-to-sample and batch-to-batch variation (see Experimental procedures and Fig. 1), we believe that we can exclude methodological variation as the cause for these unexpected results. Furthermore, longitudinal telomere length instability was not related to major shifts in PBMC composition. However, we could not assess telomerase activity in the available samples, and we could not identify the causes of this instability. The possible role of survivor effects could not be tested, because the study design did not include a younger control group.

To our knowledge, this is the largest study on human telomere length reported so far. It is population-based and prospective, and it is focused towards the oldest old. The complete lack of predictive power of telomere length in this age group contrasts sharply with expectations raised by reported associations between telomere length and morbidity and mortality in younger age groups (Brouilette et al., 2003; Cawthon et al., 2003; Benetos et al., 2004). However, it confirms and reinforces data suggesting the loss of correlation between telomere length and survival in individuals above 74 years (Cawthon et al., 2003).

Our results indicate that telomere length cannot be used unconditionally as a biomarker of the human aging process, especially at very advanced age. They also show that PBMC telomere length in this age group can go either up or down by a considerable amount within a few years. We believe that this telomere instability might be causal for the breakdown of telomere length as an aging biomarker at high age. However, the causes for this instability and its age dependency still need to be established.

Experimental procedures

Telomere length was measured within the framework of the Leiden 85-plus Study, a population-based prospective study (Heijmans et al., 2002; Mooijaart et al., 2004). For the present analyses we made use of the first cohort enrolled between December 1986 and September 1989. All inhabitants of Leiden, the Netherlands, born before 1902 (n = 1258) were contacted. A total of 977 people agreed to participate, 60 refused, and 221 had died before they could be enrolled. The Medical Ethical Committee of the Leiden University Medical Center approved the study, and informed consent was obtained from all subjects or their guardian in case of cognitively impaired subjects.

During two home visits, an interview and physical examination were performed to obtain an extensive range of clinical characteristics, use of medication and daily physical functioning. Participants were followed for mortality from the date of the baseline visit until 1 May 2001. The date and causes of death were obtained from the civic registries and the Dutch Central Bureau of Statistics. Causes of death were adjudicated according to the tenth version of the International Classification of Diseases (ICD-10; WHO, 1994). Cognitive function was assessed by MMSE and psychiatric interview at baseline and after a follow-up of 3 years.

High molecular weight DNA was isolated from PBMC samples stored at −196 °C. DNA concentration and quality were monitored by agarose gel electrophoresis. Samples were discarded if DNA degradation (as smear below 20 kb) was visible. We obtained high-quality DNA samples from 598 participants. From 81 of these participants, a second DNA sample was obtained after a time delay of up to 14 years. Telomere length was measured as abundance of telomeric template versus a single gene by quantitative real-time PCR (Cawthon et al., 2003) with modifications as described (Martin-Ruiz et al., 2004). Measurements were performed in quadruplicates. Three DNA samples with known telomere lengths (3.0, 5.5 and 9.5 kbp) were run as internal standards together with each batch of 16 study samples. To assess the methodological variation, two blood samples were taken from 14 volunteers 10 days apart. With the exception of storage time under liquid nitrogen, preparation of PBMCs, freezing, thawing, preparation of DNA and measurement were done as for the study samples. Average absolute variation between these repeated blood samples, independently drawn and processed and measured in independent batches, was 90 ± 58 bp. Similar values were obtained for repeated measurements of the same DNAs in separate batches (data not shown). The 99% prediction interval was calculated from the control repeats and used as a measure of methodological variation (see Fig. 1). Longitudinal measurements from study participants were repeated in separate batches, if changes were significantly above the methodological variation. These repeats confirmed the first results in all cases.

Tertiles of telomere length at baseline were determined for both sexes separately and then pooled. Mortality risks depending on these sex-adjusted tertiles of telomere length at baseline were estimated by Cox regression analyses. Because the age of entry varied between 85 and 101 years (mean 89.8 ± 3.0 years), we used left censoring (Kurtzke, 1989) to correct for the delayed entry into the risk set. stata was used for the statistical analyses.


The work was supported by a Medical Research Council Component Grant to TvZ.