The gait speed advantage of taller stature is lost with age

Taller individuals walk faster but it is unknown whether this advantage persists at older ages. We examined the cross-sectional/longitudinal associations of height with gait speed (GS) in participants from the Dijon-Three-City cohort study (France) over 11 years. In 4011 participants (65–85 y), we measured usual/fast GS (6 m) up to five times. We examined whether the baseline height-GS association varied with age using linear regression, and whether height influenced GS change using linear mixed models. Taller participants 65 y at baseline walked faster than shorter ones (fast GS difference between top/bottom height quartiles, 0.100 m/s, P < 0.001); this association weakened with age (P-interaction = 0.02), with a 0.012 m/s (P = 0.57) difference at 80 y. Ten-year fast GS decline was 51% greater (P < 0.001) in younger participants in the top height quartile (−0.183 m/s) compared to those in the bottom quartile (−0.121 m/s), leading the GS difference between the two groups to be attenuated by 50% over the follow-up. The height-related difference in fast GS decline was not explained by time-dependent comorbidities or height shrinkage. Analyses for usual GS yielded consistent findings. The height-GS relation is more complex than previously thought, as the height related advantage in GS disappears as persons grow older due to faster decline in taller compared to shorter persons.


Supplementary methods
Cross-sectional association of height with fast gait speed at baseline (1999)(2000)(2001) (table 2) We used linear regression to model gait speed at baseline as a function of height at baseline while adjusting for covariates (model 1): = 0 + 2 2 ℎ + 3 3 ℎ + 4 4 ℎ + + 2× ( 2 ℎ × ) + 3× ( 3 ℎ × ) + 4× ( 4 ℎ × ) + + ∑ 2 1 + ∑ 2 1 + where is fast gait speed at baseline of the ℎ participant (m/s), -0 (intercept) is the average baseline fast gait speed in subjects with reference values for the covariates at baseline (women, 65 years, normal BMI, low education, Q1 of height), -The other coefficients are the regression coefficients associated with the covariates, -ℎ is the baseline height of the ℎ participant categorized in sex-specific quartiles ( 1 − 4 ℎ , 1 ℎ is the reference), is age at baseline of the ℎ participant, centered at 65 years and divided by 10, -2 − 4 ℎ × are the three interactions between age and quartiles 2 − 4 of height, -is coded 0 for women (reference) and 1 for men, -is the education level of the ℎ participant in 3 categories (low, reference), -is the baseline body mass index of the ℎ participant in 3 categories (reference, <25 kg/m²), -is the residual of the ℎ participant.
Model 2 is further adjusted for baseline MMSE, history of cardiovascular disease (stroke, coronary heart disease, lower-limb arteritis), hypertension, and hypercholesterolemia.
Association between height at baseline and change in fast gait speed over the follow-up (table 3) We used a linear mixed model to model change in gait speed over time as a function of height at baseline while adjusting for covariates (model 1): -0 is the random intercept and 1 is the random slope of the ℎ participant, -is the residual of the ℎ participant at the ℎ visit.
Model 2 is further adjusted for time-dependent MMSE, history of cardiovascular disease (stroke, coronary heart disease, lower-limb arteritis), hypertension, hypercholesterolemia.
In supplementary table 2-A, height is modelled as a continuous variable, rather than in quartiles: where ℎ is continuous height at baseline, centered at 162 cm and divided by 10.
In supplementary table 2-B, height is modelled as a time-dependent variable, rather than a time-invariant variable fixed at baseline: where ℎ is the time-dependent height of the ℎ participant at the ℎ visit.     Beta, regression coefficients; CI, confidence interval; Q1 (shortest)-Q4 (tallest), sex-specific quartiles of height; BMI, body mass index. a Analyses based on 3,779 participants in whom usual gait speed (m/s) was measured at study baseline. Adjusted for sex (reference, women), education (reference, low education), baseline BMI (reference, normal). The intercept corresponds to the average usual gait speed for women aged 65 years at baseline, with normal BMI, low education, and in the lower quartile of height. b Analyses based on 4,069 participants in whom usual gait speed (m/s) was measured at least once. Adjusted for sex (reference, women), education (reference, low education), time-dependent BMI (reference, normal), timedependent BMI  sex, baseline height (reference, first quartile), baseline age (years, centered at 65 years), baseline height  baseline age. Beta, regression coefficients; CI, confidence interval; Q1 (shortest)-Q4 (tallest), sex-specific quartiles of height; BMI, body mass index; MMSE, mini-mental state examination. a Adjusted for sex (reference, women), education (reference, low education), time-dependent BMI (reference, <25 kg/m 2 ), baseline height (reference, Q1), baseline age (centered at 65 years), baseline heightbaseline age. b We included an indicator to identify home measures of gait speed at wave 6.

Supplementary Figure 1. Individual trajectories of fast gait speed over the follow-up
Individual trajectories were studied in a proportional random sample of 1200 subjects after stratification by sex and median age. Panel A shows individual trajectories of gait speed in grey and a fairly linear decline. The red line corresponds to a smoothed summary curve and also shows a fairly linear decline. Panel B shows OLS (ordinary least square regression) individual trajectories in grey that assume a linear decline and the corresponding pooled summary curve that is very similar to the pooled curve from Panel A. Men ≤73y, n=238; men >73y, 222; women ≤73y, 363, women >73y, 377.