Hearing and gait: an association that did not survive adjustment

Design

Cross-sectional study with prespecified composite outcomes, HC3-robust linear regression and Holm multiplicity adjustment

Population

172 cognitively unimpaired adults aged 50-80 from the DataCOG study (119 women, mean age 63.9 ± 8.1 years)

Primary outcome

Three composite instrumented Timed Up and Go gait measures against better-ear pure-tone average

Effect

Unadjusted Holm p ≤ 0.015 on all three; age- and sex-adjusted global duration Holm p = 0.050, turning 0.064, walking 0.131

The idea that hearing loss contributes to mobility decline is plausible and widely repeated. This cross-sectional study of 172 cognitively unimpaired adults aged 50 to 80 from the DataCOG cohort (119 women, mean age 63.9 ± 8.1 years) tested it properly, and the discipline of the analysis is what makes it worth reading.

Gait was measured with an instrumented Timed Up and Go, hearing with better-ear pure-tone average. Three composite gait outcomes — global duration, turning, walking duration — were prespecified and modelled unadjusted, adjusted for age and sex, and stratified by sex, with a Holm-adjusted threshold of p<0.05 set in advance to handle the multiple comparisons.

Unadjusted, worse hearing was associated with poorer performance on all three composites (Holm p ≤ 0.015). Adjusted for age and sex, global duration became borderline (Holm p = 0.050) and turning and walking duration fell out entirely (p = 0.064 and 0.131). Sex-stratified estimates were larger in men (n=53) than women (n=119) but neither reached the threshold (men p = 0.054; women p ≥ 0.459).

What this shows is mostly age. Hearing declines with age and so does gait, and a study that does not adjust will find them associated. The authors report exactly that and call the result hypothesis-generating rather than claiming the sex difference their own stratification hinted at.

• Do not counsel patients that treating hearing loss will improve their mobility — that is not established
• The unadjusted association is real and uninformative; it is age acting on both
• Prespecified multiplicity correction is why this paper reads honestly — look for it in others
• Men showed a larger estimate but the study had 53 of them; that is a sample size problem, not a sex finding
• Falls assessment in an older patient with hearing loss is still worth doing on its own merits

Why it matters

A widely repeated link between hearing and mobility turns out to rest largely on both declining with age.

Don't overread it

Cross-sectional and modest in size; this cannot rule out a real association, only show that this study did not demonstrate one.

The statistics, in plain English

Holm adjustment raises the bar when several outcomes are tested, because testing three things gives three chances to find something by luck. Here the unadjusted results cleared that bar comfortably and the age-and-sex-adjusted ones did not — which is the signature of confounding by age rather than a real effect. A p of 0.050 is not a near miss to be talked past; with 172 participants the study simply cannot settle the question.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.