A statistical approach to distinguish telomere elongation from error in longitudinal datasets


Telomere length and the rate of telomere attrition vary between individuals and have been interpreted as the rate at which individuals have aged. The biology of telomeres dictates shortening with age, although telomere elongation with age has repeatedly been observed within a minority of individuals in several populations. These findings have been attributed to error, rather than actual telomere elongation, restricting our understanding of its possible biological significance. Here we present a method to distinguish between error and telomere elongation in longitudinal datasets, which is easy to apply and has few assumptions. Using simulations, we show that the method has considerable statistical power (>80 %) to detect even a small proportion (6.7 %) of TL increases in the population, within a relatively small sample (N = 200), while maintaining the standard level of Type I error rate (α ≤ 0.05).

This is a preview of subscription content, log in to check access.

Fig. 1


  1. Aviv A, Chen W, Gardner JP, Kimura M, Brimacombe M, Cao X, Srinivasan SR, Berenson GS (2009) Leukocyte telomere dynamics: longitudinal findings among young adults in the Bogalusa Heart Study. Am J Epidemiol 169:323–329

    PubMed  Article  Google Scholar 

  2. Barrett ELB, Burke TA, Hammers M, Komdeur J, Richardson DS (2013) Telomere length and dynamics predict mortality in a wild longitudinal study. Mol Ecol 22:249–259

    PubMed  Article  Google Scholar 

  3. Beaulieu M, Reichert S, Le Maho Y, Ancel A, Criscuolo F (2011) Oxidative status and telomere length in a long-lived bird facing a costly reproductive event. Funct Ecol 25:577–585

    Article  Google Scholar 

  4. Bize P, Criscuolo F, Metcalfe NB, Nasir L, Monaghan P (2009) Telomere dynamics rather than age predict life expectancy in the wild. Proc R Soc B 276:1679–1683

    CAS  PubMed  Article  Google Scholar 

  5. Boonekamp JJ, Simons MJP, Hemerik L, Verhulst S (2013) Telomere length behaves as biomarker of somatic redundancy rather than biological age. Aging Cell 12:330–332

    CAS  PubMed  Article  Google Scholar 

  6. Chen W, Kimura M, Kim S, Cao X, Srinivasan SR, Berenson GS, Kark JD, Aviv A (2011) Longitudinal versus cross-sectional evaluations of leukocyte telomere length dynamics: age-dependent telomere shortening is the rule. J Gerontol A Biol Sci Med Sci 66A:312–319

    Article  Google Scholar 

  7. Crawley MJ (2005) Statistics: an introduction using R. Wiley, Chichester

    Google Scholar 

  8. Crawley MJ (2007) The R book. Wiley, Chichester

    Google Scholar 

  9. Ehrlenbach S, Willeit P, Kiechl S, Willeit J, Reindl M, Schanda K, Kronenberg F, Brandstatter A (2009) Influences on the reduction of relative telomere length over 10 years in the population-based Bruneck Study: introduction of a well-controlled high-throughput assay. Int J Epidemiol 38:1725–1734

    PubMed  Article  Google Scholar 

  10. Epel ES, Merkin SS, Cawthon R, Blackburn EH, Adler NE, Pletcher MJ, Seeman TE (2009) The rate of leukocyte telomere shortening predicts mortality from cardiovascular disease in elderly men. Aging (Albany NY) 1:81–88

    CAS  Google Scholar 

  11. Foote CG, Gault EA, Nasir L, Monaghan P (2011) Telomere dynamics in relation to early growth conditions in the wild in the lesser black-backed gull. J Zool 283:203–209

    Article  Google Scholar 

  12. Gorbunova V, Seluanov A (2009) Coevolution of telomerase activity and body mass in mammals: from mice to beavers. Mech Ageing Dev 130:3–9

    Google Scholar 

  13. Heidinger BJ, Blount JD, Boner W, Griffiths K, Metcalfe NB, Monaghan P (2012) Telomere length in early life predicts lifespan. Proc Natl Acad Sci USA 109:1743–1748

    CAS  PubMed  Article  Google Scholar 

  14. Mather KA, Jorm AF, Parslow RA, Christensen H (2011) Is telomere length a biomarker of aging? A review. J Gerontol A Biol Sci Med Sci 66:202–213

    PubMed  Article  Google Scholar 

  15. Nordfjäll K, Svenson U, Norrback K-F, Adolfsson R, Lenner P, Roos G (2009) The individual blood cell telomere attrition rate is telomere length dependent. PLoS Genet 5:e1000375

    PubMed Central  PubMed  Article  Google Scholar 

  16. R Development Core Team (2011) A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna

    Google Scholar 

  17. Riethman H (2008) Human telomere structure and biology. Annu Rev Genomics Hum Genet 9:1–19

    CAS  PubMed  Article  Google Scholar 

  18. Salomons HM, Mulder GA, van de Zande L, Haussmann MF, Linskens MHK (2009) Telomere shortening and survival in free-living corvids. Proc R Soc B 276:3157–3165

    CAS  PubMed  Article  Google Scholar 

  19. Shalev I, Moffitt TE, Sugden K, Williams B, Houts RM, Danese A, Mill J, Arseneault L, Caspi A (2012) Exposure to violence during childhood is associated with telomere erosion from 5 to 10 years of age: a longitudinal study. Mol Psychiatry 18:576–581

    PubMed Central  PubMed  Article  Google Scholar 

  20. Steenstrup T, Hjelmborg JB, Kark JD et al (2013) The telomere lengthening conundrum—artifact or biology? Nucleic Acids Res 14:e131

    Article  Google Scholar 

  21. Wichura MJ (1988) Algorithm AS 241: the percentage points of the normal distribution. J R Stat Soc Ser C Appl Stat 37:477–484

    Google Scholar 

Download references


MJPS is supported by the Natural Environment Research Council (J024597/1) (United Kingdom). SN is supported by the Rutherford Discovery Fellowship (New Zealand). GS is supported by a grant by The Netherlands Organisation for Scientific Research (452-10-012), granted to M. Mills.

Author information



Corresponding author

Correspondence to Mirre J. P. Simons.

Electronic supplementary material



The derivation of Equation 5

A two-level regression which model telomere length (TL) can be expressed as:

$$y_{ij} = \beta_{0} + \gamma_{j} + (\beta_{1} + \varphi_{j} )t_{ij} + \varepsilon_{ij} ,$$
$$\left( {\begin{array}{*{20}c} {\gamma_{j} } \\ {\varphi_{j} } \\ \end{array} } \right)\sim N\left( {\left( {\begin{array}{*{20}c} 0 \\ 0 \\ \end{array} } \right),\left( {\begin{array}{*{20}c} {\sigma_{\gamma }^{2} } & {\rho \sigma_{\gamma } \sigma_{\varphi } } \\ {\rho \sigma_{\gamma } \sigma_{\varphi } } & {\sigma_{\varphi }^{2} } \\ \end{array} } \right)} \right) ,$$
$$\varepsilon_{ij} \sim N(0, \, \sigma_{\varepsilon }^{2} )$$

where t ij is the ith time point at which TL, y ij is measured for the jth individual (i = 1, 2,…, n; n is the number of TL measurements and n > 2; j = 1, 2,…, N; N is the number of individuals in a study), β 0 is the grand intercept (TL at t = 0), β 0 is the grand slope (regression coefficient for t), γ j is the deviation from β 0 for the jth individual, φ j is the deviation from β 0 for the jth individual, γ j and φ j has a multivariate normal distribution with the variance–covariance structure specified in A2, and ε ij is the ith residual value and residuals are normally distributed with a variance of \(\sigma_{\varepsilon }^{2}\).

When we consider A1 at the time points 1 and n (i.e. i = 1 and i = n), TL can be written as:

$$y_{1j} = \beta_{0} + \gamma_{j} + (\beta_{1} + \varphi_{j} )t_{1j} + \varepsilon_{1j} ,$$
$$y_{nj} = \beta_{0} + \gamma_{j} + (\beta_{1} + \varphi_{j} )t_{nj} + \varepsilon_{nj} .$$

When we have two measurements in time, 1 and m (the final time point) of telomeres the difference in telomere length is described by:

$$y_{nj} - y_{1j} = \beta_{1} (t_{nj} - t_{1j} ) + \varphi_{j} (t_{nj} - t_{1j} ) + \varepsilon_{nj} - \varepsilon_{1j} .$$

By setting d j  = y nj  − y 1j , the variance of d j can be expressed as:

$${\text{Var}}(d_{j} ) = (t_{nj} - t_{1j} )^{2} \sigma_{\varphi }^{2} + 2\sigma_{\varepsilon }^{2} .$$

Note that the constant β 1(t nj  − t 1j ) disappears. Using the definition of variance and further rearranging;

$$\frac{1}{(N - 1)}\sum\limits_{j = 1}^{N} {(d_{j} - \bar{d})^{2} } = (t_{nj} - t_{1j} )^{2} \sigma_{\varphi }^{2} + 2\sigma_{\varepsilon }^{2} ,$$
$$\frac{1}{2(N - 1)}\sum\limits_{j = 1}^{N} {d_{j}^{2} } = \sigma_{\varepsilon }^{2} + \frac{{(t_{nj} - t_{1j} )^{2} \sigma_{\varphi }^{2} }}{2} + \frac{{\bar{d}^{2} }}{2},$$

where \(\bar{d}\) is the mean value of dj. As \(\bar{d} = \beta_{1} (t_{nj} - t_{1j} )\) and setting (t nj  − t 1j ) = u;

$$\frac{1}{2(N - 1)}\sum\limits_{j = 1}^{N} {d_{j}^{2} } = \sigma_{\varepsilon }^{2} + \frac{{u^{2} }}{2}(\sigma_{\varphi }^{2} + \beta_{1}^{2} ).$$

When we assume that TL does not increase or decrease, i.e. \(\left( {\sigma_{\varphi }^{ 2} + \beta_{1}^{ 2} } \right) = 0\), A10 reduces to:

$$\sigma_{\varepsilon }^{2} = \frac{1}{2(N - 1)}\sum\limits_{j = 1}^{N} {d_{j}^{2} } .$$

If we estimate \(\sigma_{\varepsilon }^{2}\) in A11 only from individuals that show an increase of TL, or d j  > 0 (set such d j as D j ), we have Eq. 5 from the main text;

$$\sigma_{\varepsilon }^{\prime 2} = \frac{1}{2(m - 1)}\sum\limits_{k = 1}^{m} {D_{k}^{2} } ,$$

where \(D_{k}^{2}\) is the difference in TL between the initial and last measurements in the kth individuals that showed an increase in TL (k = 1, 2,…, m; m is the number of individuals whose TL elongated). Note that we assume \(\sigma_{\varepsilon }^{\prime 2}\) is also normally distributed as with \(\sigma_{\varepsilon }^{2}\) (A3). Due the symmetric nature of the normal distribution, \(\sigma_{\varepsilon }^{\prime 2}\) can be correctly estimated from restricted data, D j under our assumption.

Rights and permissions

Reprints and Permissions

About this article

Cite this article

Simons, M.J.P., Stulp, G. & Nakagawa, S. A statistical approach to distinguish telomere elongation from error in longitudinal datasets. Biogerontology 15, 99–103 (2014). https://doi.org/10.1007/s10522-013-9471-2

Download citation


  • Telomere length
  • Statistics
  • Telomere shortening
  • Within individual
  • Aging
  • Human