1
Causal Associations of Self-Reported Walking Pace with Telomere
Length in 405,981 middle-aged adults: a UK Biobank study
Article Type: Original article
Authors and Affiliations: Paddy C. Dempsey
1,2,3
, Crispin Musicha
2,6
,
Alex V. Rowlands
1,2
,
Melanie Davies
1,2,4
, Kamlesh Khunti
7,8
, Cameron Razieh
1,2
, Iain Timmins
4
, Francesco
Zaccardi
7,8
, Veryan Codd
2,6
, Christopher P. Nelson
2,6
, *Tom Yates
1,2
, *Nilesh J Samani
2,6
1. Diabetes Research Centre, University of Leicester, Leicester General Hospital, Leicester, UK.
2. NIHR Leicester Biomedical Research Centre, Leicester General and Glenfield Hospitals,
Leicester, UK
3. Baker Heart and Diabetes Institute, Melbourne, Australia.
4. Leicester Diabetes Centre, Leicester General Hospital, University Hospitals of Leicester NHS
Trust
5. Department of Health Sciences, University of Leicester, Leicester, United Kingdom
6. Department of Cardiovascular Sciences, University of Leicester, Leicester, United Kingdom
7. NIHR Collaboration for Leadership in Applied Health Research and Care—East Midlands,
University as Leicester, Leicester, United Kingdom
8. Leicester Real World Evidence Unit, University of Leicester, Leicester General Hospital,
Gwendolen Rd, Leicester, United Kingdom
*Contributed equally
Author and address for correspondence: Paddy C. Dempsey (pcd5@leicester.ac.uk
)
Main text word count = 3045; Number of Tables = 2; Number of Figures = 1; Number of
Supplementary Tables = 1; Number of Supplementary Figures = 3; Number of references = 52
. CC-BY-ND 4.0 International licenseIt is made available under a
perpetuity.
is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint
The copyright holder for thisthis version posted September 12, 2021. ; https://doi.org/10.1101/2021.09.06.21263163doi: medRxiv preprint
NOTE: This preprint reports new research that has not been certified by peer review and should not be used to guide clinical practice.
2
ABSTRACT
Objectives: Walking pace is a strong marker of functional and health status. We investigated
whether walking pace is also associated with leucocyte telomere length (LTL), which is causally
associated with several diseases and has been proposed as a marker of biological age.
Methods: We used baseline data from UK Biobank participants recruited from March-2006 to
July-2010. Walking pace was self-reported as slow, steady/average, or brisk. Accelerometer-
assessed measures of total physical activity and intensity were included to support interpretation
of walking pace data. LTL was measured by qPCR assay. Bi-directional Mendelian randomization
(MR) analyses were undertaken to inform likely causal directions.
Results: The analysed cohort comprised 405,981 adults (54% women) with mean age of 56.5
years (SD, 8.1) and body mass index 27.2 kg/m
2
(SD, 4.7). Steady/average and brisk walkers
had significantly longer LTL compared with slow walkers, with a Z-standardised LTL difference of
0.066 (0.053-0.078) and 0.101 (0.088-0.113), respectively. Associations remained but were
attenuated following full covariate adjustment: 0.038 (0.025-0.051) and 0.058 (0.045-0.072),
respectively. Accelerometer data (n=86,002) demonstrated a non-linear association between LTL
and habitual movement intensity, but not total activity. MR analysis supported a causal
association of walking pace on LTL, with an increase in Z-standardised LTL of 0.192 (0.077,
0.306) for each difference in walking pace category. No evidence of a causal association was
observed for LTL on walking pace.
Conclusion: Faster walking pace may be causally associated with longer LTL, which could
explain some of the beneficial effects of brisk walking on health status.
Keywords: walking pace; physical activity; telomere; ageing; mendelian randomisation;
accelerometer; lifestyle
. CC-BY-ND 4.0 International licenseIt is made available under a
perpetuity.
is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint
The copyright holder for thisthis version posted September 12, 2021. ; https://doi.org/10.1101/2021.09.06.21263163doi: medRxiv preprint
3
INTRODUCTION
Wal
king is a simple and accessible form of physical activity (PA) for all ages, conferring many
physical, mental, and social health benefits with minimal adverse effects [1-4]. It therefore holds
strong potential as a pragmatic target for intervention [5]. Strong associations with health status
have been seen for habitual or self-rated walking pace, which has been associated with better
physical fitness and reduced risk of cardiovascular disease and all-cause mortality [6-10], with
brisk walkers having up to 20 years greater life expectancy compared to slow walkers [11].
Indeed, walking pace has been shown to have a stronger association with survival and be a
substantially better prognostic marker for all-cause or cardiovascular mortality than other
measures of PA volume, diet, or physical function [12, 13]. Similarly, accelerometer-assessed
measures of PA in UK Biobank suggest that as little as 10 min of brisk walking a day is associated
with longer life expectancy [14]. A genome-wide association study (GWAS) on self-reported
walking pace within UK Biobank identified 70 independent SNPs at genome-wide significance
[15], close to an order of magnitude greater than the number reported for other self-reported
or
ac
celerometer-assessed measures of PA traits within the same cohort [16].
The importance of walking pace as a marker and potential promoter of health is likely to reflect it
being a complex functional activity influenced by many factors, such as motor control,
musculoskeletal health, cardiorespiratory fitness and lung capacity, habitual activity levels,
cognition, motivation, and mental health [3, 13, 15]. These factors also align to the concept of
biological age, which relates to an individual’s ability to maintain a robust homeostasis when
subject to stressors [17]. Therefore, it is possible that walking pace acts as both a marker and
modulator of biological age. However, whether walking pace is causally associated with potential
indicators of biological age remains unknown.
Although the relationship between leukocyte telomere length (LTL) and disease is complex [18],
LTL has been proposed as a marker of biological age and is associated with higher risk of several
age-related diseases; including coronary artery disease and several cancers [18-23]. Telomeres
are DNA–protein complexes that protect the ends of chromosomes from degradation, end-to-end
fusion, and abnormal recombination of DNA strands (genomic instability). The DNA component
progressively shortens with each cell cycle, decreasing in most cell types as humans age,
ultimately contributing to replicative senescence [19, 24]. Along with reflecting cellular replicative
history, telomere shortening is also moderated by factors such as oxidative stress and
inflammation [24]. Telomere length is usually measured in leukocytes (LTL), which is reflective of
telomere length in other tissues, along with reflecting the senescent status of circulating cells
related to the immune system [25].
Previous research suggests an association of higher levels of PA and cardiorespiratory fitness
with longer LTL [26, 27], supporting the hypothesis that higher levels of PA and cardiorespiratory
fitness may act to slow markers of biological ageing. However, most studies in humans to date
have been small and/or observational in nature, with some studies showing weak or null
associations [26-28]. Therefore, the current literature is not definitive and does not support
inferences around causal direction. Moreover, there remains insufficient research investigating
the association between simple functional habitual movements, such as walking pace, and LTL.
The aim of this study was, therefore, to investigate the association between self-reported walking
pace and LTL in middle aged adults. This included harnessing previously defined genetic
instruments for both walking pace and LTL to undertake bi-directional Mendelian randomisation
(MR) analyses, to help clarify the causal nature and relative importance of any observed
associations. We support observational analyses for self-reported walking pace using
accelerometer-assessed total PA and intensity, to aid broader interpretation. Our observational
hypothesis was that a brisker walking pace would be causally associated with longer LTL.
METHODS
D
ata source and study population
This analysis used data from participants within UK Biobank, a large prospective cohort of middle-
aged adults designed to support biomedical analysis focused on improving the prevention,
. CC-BY-ND 4.0 International licenseIt is made available under a
perpetuity.
is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint
The copyright holder for thisthis version posted September 12, 2021. ; https://doi.org/10.1101/2021.09.06.21263163doi: medRxiv preprint
4
diagnosis, and treatment of chronic disease through phenotyping and genomics data [29].
Between March 2006 and July 2010, individuals living within 25 miles of one of the 22 study
assessment centres located throughout England, Scotland, and Wales were recruited and
provided comprehensive data on a broad range of demographic, clinical, lifestyle, and social
outcomes. All participants provided written informed consent and the study was approved by the
NHS National Research Ethics Service (Ref: 11/NW/0382).
Self-reported walking pace
Self-reported walking pace was ascertained using a touchscreen question: “How would you
describe your usual walking pace?” with response options of “slow”, “steady/average” or “brisk”.
Participants could access further information which defined a slow pace as less than 3 miles per
hour, a steady/average pace as between 3-4 miles per hour, and a brisk pace as more than 4
miles per hour.” We excluded participants whose answers were “None of the above or “Prefer not
to answer” (n = 3,956).
Covariate measurement
We utilized demographic and lifestyle related characteristics of age, sex, ethnicity (white/non-
white), Townsend Index of deprivation, highest educational level achieved (degree or above/any
other qualification/no qualification), employment status (unemployed/in paid or self-employment),
alcohol drinking status (never/previous/current), salt added to food (never/sometimes), oily fish
intake (never/sometimes), fruit and vegetable intake (a score from 0-4 taking into account
questions on cooked and raw vegetables, fresh and dried fruit consumption), processed and red
meat intake (average weekly frequency in days per week), and sleep duration (<7, 7-8, >8 hours),
and a diagnosis of cardiovascular disease or cancer prior to baseline. The latter two prevalent
disease variables were derived from the self-reported history of heart attack, angina, stroke, or
cancer variables, and from linked hospital episode data (corresponding ICD 10 codes I20-25, I60-
69, or C00-99). Health-related covariates of blood pressure and cholesterol medication, doctor
diagnosed diabetes or prescribed insulin medication and mobility limitations (self-reported
longstanding illness or disability or chest pain at rest), white blood cell (leukocyte) count, and
body mass index (BMI) in three categories (<25, 25-30, ≥30 kg/m
2
) were included in models. Total
MET-minutes/week of PA was derived from weekly frequency and duration of walking, moderate,
or vigorous intensity PA using the short-form International Physical Activity Questionnaire (IPAQ)
[30]. Further details for each variable are available on the UK Biobank Website
https://www.ukbiobank.ac.uk/
Accelerometer PA measurements
PA was assessed using accelerometry in a subset of participants (n=~100,000; see
Supplemental Figure S1) between 2013-2015 who were invited to wear an Axivity AX3 triaxial
accelerometer (Axivity Ltd., Newcastle, UK) continuously on their dominant wrist for seven
consecutive days [31]. Accelerometer measures were average acceleration over the 24 h day
(proxy for total PA, mg) and intensity gradient over 24 h (a measure of the intensity distribution of
PA over the day) [31-33]. A higher average acceleration indicates more PA is accumulated across
the day, irrespective of the intensity. A higher intensity gradient indicates more time is habitually
spent in higher intensity activities, such as brisk walking (see Supplemental Accelerometer
Methods
for the accelerometer data processing methodology and PA variable interpretation).
Leucocyte telomere length measurements
LTL was measured using an established multiplex qPCR assay from 488,415 available DNA
samples of participants in UK Biobank, which are detailed elsewhere [34]. After extensive quality
checks and adjustment for technical factors valid LTL measurements were available for 472,577
individuals [34]. For analyses of data available for the full cohort we used log-transformed and z-
standardised LTL values (UK Biobank data field code 22192), of which log-transformed data were
re-standardised for analyses performed on the sub-set of participants with accelerometer data.
Statistical analyses
For analyses presented in this paper, we included all participants with LTL measured from the
UK Biobank baseline sample, where there was no mismatch in self-reported and genetic sex
. CC-BY-ND 4.0 International licenseIt is made available under a
perpetuity.
is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint
The copyright holder for thisthis version posted September 12, 2021. ; https://doi.org/10.1101/2021.09.06.21263163doi: medRxiv preprint
5
(n=472,248). Exclusions were also made for missing walking pace or covariate data, or for
missing accelerometer data in the subset analysis (Supplemental Figure S1).
A series of linear regression models were used to quantify the associations of “steady/average”
and “brisk” self-reported walking pace with Z-standardised log-LTL (β-coefficient with 95% CI),
compared to “slow” walkers as the reference group. Model 1 adjusted for age, sex, ethnicity and
total white blood cell count, as these variables are known to be associated with LTL [34]. Model
2 additionally adjusted for other included confounders. Models 3 and 4 additionally adjusted for
total PA and then body mass index, which were considered last given their potential role as
confounders or mediators.
One-sample bidirectional MR
We conducted a bi-directional one-sample Mendelian randomisation (MR) analysis (see
Supplemental Figure S2) to evaluate a potential causal relationship between longer LTL and
self-rated walking pace, using the inverse-variance weighted (IVW) [35] method with sensitivity
analyses using both the weighted median [36] and robust adjusted profile score [37] methods.
We also used MR-Egger regression to assess robustness to horizontal pleiotropy [38]. Each of
these approaches makes a slightly different set of assumptions about the pleiotropic effects of
genetic instruments, hence if the effect estimates are consistent across methods this provides
stronger evidence of causality. To examine the causal association of LTL on walking pace (part
1) the LTL instrument utilised a set of 130 genome-wide significant (P<8.31x10
-9
), conditionally
independent, uncorrelated, and non-pleiotropic genetic variants we recently identified as genetic
instruments for LTL [18]. We matched variants to the publicly available walking pace GWAS data
that were unadjusted and adjusted for BMI, matching 121 variants. For walking pace on LTL (part
2) we considered 70 genome-wide significant (P<5.0x10
-8
) independent genetic variants as the
instrument from the unadjusted GWAS using weights from both the unadjusted GWAS and the
BMI-adjusted GWAS [15]. These were matched to the LTL GWAS [18], matching all variants. To
interpret the causal effect estimate of 1-SD increased LTL length on differences in walking pace
category, the coded values 0, 1 and 2 for self-reported slow, steady/average and brisk walking
pace can be thought of as threshold values for an underlying continuous trait, as has been
demonstrated previously [15].
Sensitivity analyses
To support the findings and interpretation for self-reported walking pace, we included sensitivity
analyses examining the sub-set of the UK Biobank cohort with accelerometer data and LTL
(n=86,002; see Supplemental Figure S1 and Supplemental Table S1 for sub-sample
descriptive characteristics), focussing on two key metrics summarizing total PA (average
acceleration) and the intensity distribution (intensity gradient) of PA over each 24-hour day (see
Supplemental Accelerometer Methods). Due to some evidence of non-linearity, associations
for these two accelerometer exposures with LTL were examined using
restricted cubic splines
(three evenly-spaced knots), with reference values set at the 10
th
percentile of the exposure.
Observational analyses were conducted using Stata v15.1 (StataCorp, TX, USA) and statistical
significance was set at p<0.05 (two-tailed). MR analyses were performed using the
MendelianRandomisation package implemented in R software [39].
RESULTS
Descriptive characteristics of the observational analytical sample
Descriptive characteristics for the analysis sample are shown in Table 1. Mean age was 56.5
years (SD, 8.1); mean BMI was 27.2 kg/m
2
(SD, 4.65); and 54% and 95% were female and white,
respectively. Approximately half the participants reported an average/steady walking pace
(n=212,303; 52.3%), with 6.6% (n=26,835) reporting a slow walking pace and 41.1% (n=166,843)
reporting a brisk pace. Compared to slow walkers, those who reported being average/steady and
brisk walkers were slightly younger, were more likely to have never smoked, and were less likely
to be taking cholesterol/blood pressure medications, have a chronic disease, or have mobility
limitations. Slow walkers reported engaging in less PA and had a higher deprivation index and
prevalence of obesity compared to average and brisk walkers. Differences observed between the
. CC-BY-ND 4.0 International licenseIt is made available under a
perpetuity.
is the author/funder, who has granted medRxiv a license to display the preprint in(which was not certified by peer review)preprint
The copyright holder for thisthis version posted September 12, 2021. ; https://doi.org/10.1101/2021.09.06.21263163doi: medRxiv preprint
