INTRODUCTION
Gait is a fundamental physiologic process that requires coordinated functioning of the motor, sensory, and central nervous systems.1 Moreover, gait is a critical human behavior that facilitates the performance of daily activities and the ability to live independently. Walking speed is a robust indicator of an individual's health state and is considered by some to be the sixth vital sign.2,3 Reduction in walking speed, as occurs with aging, has been shown to be a powerful prognosticator of adverse outcomes including falls and mortality.2,3
The visual, proprioceptive, and vestibular sensory systems contribute critically to maintaining static balance and posture, and studies have shown that they also play important roles in regulating gait function including walking speed. All 3 sensory systems are known to experience declining functions with healthy aging.4 However, whether age-related imbalance is associated with the well-established age-associated decline in walking speed has not been fully established. We hypothesized that balance limitation would be associated with slower walking speed in a cohort of older adults.
In this study, we used previously collected data from the National Health and Nutrition Examination Survey (NHANES) to evaluate the association between age-related imbalance and walking speed in the US population. The NHANES is a population-based study conducted by the US Centers for Disease Control and Prevention that performed balance testing and the 20-ft timed walk on more than 4000 individuals ages 50 to 85 years between 1999 and 2002.5 We performed a cross-sectional analysis to investigate whether imbalance in healthy aging is associated with slower walking speed. We also examined whether imbalance may mediate the association between age and slower walking speed using structural equation modeling.
METHODS
Study Population
The NHANES is an ongoing cross-sectional survey of the civilian, noninstitutionalized population. Every 2 years, the NHANES selects participants on the basis of age, gender, race, and place of residence. It uses a stratified, multistage probability sampling design and selectively oversamples lower income individuals, racial minorities, and older adults.5 Sample weights account for this complex design and yields results that are generalizable to the US population. The NHANES protocol was approved by the National Center for Health Statistics (NCHS) Institutional Review Board, and informed consent was obtained from participants whose rights were protected. We used previously collected, deidentified NHANES data in the public domain including sample weights and performed a cross-sectional analysis.
The 1999 to 2000 and 2001 to 2002 NHANES performed balance testing on a subset of adults 40 years and older (half of the adults 40-69 years in the 1999-2000 cycle and all adults 40-85 years in the 2001-2002 cycle). Gait testing on a 20-ft walk was performed in adults aged 50 to 85 years in both the 1999-2000 and 2001-2002 cycles. We combined these cycles to analyze 4 years of data per NCHS recommendations.5 A total of 21 004 people participated in the NHANES from 1999 to 2002; 4983 individuals were aged 50 to 85 years and eligible for both balance and gait testing. No subsequent cycles concurrently assessed balance and walking speed.
Balance Testing
The NHANES documented balance function using the modified Romberg Test of Standing Balance on Firm and Compliant Support Surfaces. This test examined the participant's ability to stand unassisted with feet together under 4 progressively challenging conditions. In the fourth condition of this balance test, participants had to maintain balance on a foam-padded surface with their eyes closed, thereby reducing visual and proprioceptive inputs and increasing their reliance on vestibular inputs. This was a more informative postural task as most adults in the NHANES were able to perform the tasks with visual and/or proprioceptive inputs available.
Balance testing was scored on a pass/fail basis. Test failure was defined as a subject opening his/her eyes, moving the arms or feet to achieve stability, or beginning to fall or requiring operator intervention to maintain balance within 30 seconds. Because each successive condition was more difficult than the prior, balance testing was ended whenever an individual failed to pass a test either during the initial trial or a retest if the participant opted for one. A maximum of 2 trials per condition was allowed. The NHANES balance testing procedure had been previously validated.6-8 Details of the NHANES balance testing protocols are available at http://www.cdc.gov/nchs/data/nhanes/ba.pdf.
Participants were excluded if they were unable to stand on their own, were having dizziness sufficient to cause unsteadiness, weighed more than 275 lb, had a waist circumference that could not accommodate proper fitting of the safety gait belt, needed a leg brace to stand unassisted, or had a foot or leg amputation. Of the 4195 adults 40 years and older who were eligible for balance testing, 850 participants (20.3%) were excluded because they did not complete a Mobile Exam Center examination for reasons including "safety exclusion," "participant refusal," or "illness/emergency," etc, resulting in 3345 subjects who completed balance testing. Of these, 3270 subjects (97.8%) passed conditions 1 to 3 and participated in condition 4. Thirty-three participants had missing data for condition 4 and were excluded, resulting in 3237 participants (96.8%). Of these, 2183 participants were 50 years and older and eligible for gait testing.
Gait Testing
Participants were asked to walk 20 ft at their usual pace in a Mobile Exam Center corridor with a "start" and "finish" tape strip. Participants underwent the timed walk without room for acceleration or deceleration. A certified examiner started timing when the subject's first foot crossed the "start" line and stopped timing when his/her foot crossed the "finish" line. Walking speed (in meter/second, or m/s) was calculated as walking distance (20 ft = 6.10 m) divided by time (in seconds). The NHANES gait assessment methodology had been previously validated.9-11 Details of the NHANES gait testing protocols are available at http://www.cdc.gov/nchs/data/nhanes/ms.pdf.
General exclusion criteria for gait testing included a history of myocardial infarction within 6 weeks of testing, chest or abdominal surgery within 3 weeks of testing, any knee surgery, severe back pain, or a history of brain aneurysm and stroke. Participants were also excluded if they could not walk without holding onto someone else. From 1999 to 2002, 4449 participants 50 years and older were qualified. Of these, 489 individuals (11.0%) were excluded specifically from the 20-ft walk, resulting in 3960 participants who completed the walk. In this study, we examined a subset of 2116 participants aged 50 to 85 years who underwent both condition 4 and gait testing.
Demographic and Health-Related Covariates
Participant age was categorized into the decades 50 to 59, 60 to 69, 70 to 79, and 80 or more. Race/ethnicity was grouped as non-Hispanic white, non-Hispanic black, Mexican American, or others. In 2116 subjects who completed both condition 4 and gait testing, race/ethnicity information was missing in 49 participants (2.3%). Education was grouped as less than high school, high school diploma, and beyond high school. Education was missing in 3 subjects (0.14%). Body mass index (BMI) in kg/m2 was grouped as normal (BMI <25), overweight (BMI 25-30), and obese (BMI >30). BMI was missing in 2 subjects (0.09%).
Smoking history consisted of the number of years smoked and current number of cigarettes per day. Pack-years of smoking were calculated, and participants were grouped as never smoked, fewer than 20 pack-years, and 20 pack-years or more. Many participants (n = 1003, or 47.4%) had unknown smoking history or missing information required to calculate pack-years. So we created a separate category, "unknown or missing smoking status" to retain adequate power in the analysis. Hypertension was defined on the basis of a physician diagnosis, use of antihypertensive medication, an average systolic blood pressure of more than 140 mm Hg, or an average diastolic blood pressure of more than 90 mm Hg at the time of examination. Average blood pressure consisted of up to 4 readings on 2 occasions. A total of 7 subjects (0.33%) had missing blood pressure data. Diabetes was defined on the basis of a physician diagnosis, use of antihyperglycemic medication, an 8-hour fasting glucose of 126 mg/dL or more, or a nonfasting glucose of 200 mg/dL or more. No participant had missing diabetes information. Demographic and health information was included in the analysis as potential confounders for the association between age-related imbalance and slow gait speed.12,13
We also considered visual acuity in our regression models to evaluate whether the association between balance function and walking speed could in part be confounded by vision loss. Left and right visual acuity was measured using a Snellen chart and averaged. Participants were grouped into "normal and better than 20/20 vision" (>=20/20) or "worse than 20/20 vision" (<20/20). There were 539 participants (25.4%) with missing vision information; thus, a category of "missing vision measurement" was created. Finally, peripheral neuropathy was assessed in the NHANES using the Semmes-Weinstein 10-g monofilament.14 Peripheral sensory neuropathy was defined as one or more insensate sites of 3 sites tested per foot. In previous studies, severe sensory neuropathy had been shown to be associated with poor ankle proprioception, and was thus included in our analysis.15,16 In this cohort, 82 participants (3.9%) had missing peripheral neuropathy data.
Analysis
To account for the NHANES survey design, sample weights were incorporated into analyses per NCHS instructions.5 The STATA 13 statistical software was used to conduct all analyses (StataCorp, College Station, Texas). The study cohort was tested for normality using the quantile-quantile plot per NCHS instructions, and the plot was found to be consistent with a normal distribution. Furthermore, the Breusch-Pagan/Cook-Weisberg test and residual plot were used to test for heteroscedasticity, and variance of the residuals was found to be homoscedastic (P = .59). We first compared mean walking speeds among demographic and health covariate categories using the t test and the analysis of variance (ANOVA) F test. Then, we performed simple linear regression using robust standard errors per NCHS recommendation and multivariable linear regression to estimate the association between performance on condition 4 and walking speed adjusting for these covariates as potential confounders. Furthermore, we constructed structural equation models (SEMs) to test the extent to which age-related imbalance mediated the association between age and slower walking speed. SEMs utilized a series of multivariate regressions and allowed for the modeling of multiple outcomes simultaneously. By testing imbalance as an intermediate outcome, the SEM allowed us to estimate the fraction of the association between age and slow walking speed that was mediated by imbalance. A P value less than .05 was considered statistically significant for our analyses.
RESULTS
From 1999 to 2002, 2116 participants aged 50 to 85 years underwent concurrent assessment of balance function and walking speed (Table 1). Mean walking speed for all participants was 1.03 (standard deviation: 0.23) m/s, which was comparable to previously published studies in older adults.3,6-11 A walking speed change of 0.1 m/s has been shown to be a clinically useful predictor of well-being.17 In our cohort, walking speed decreased with age from 1.12 m/s in participants aged 50 to 59 years to 0.84 m/s in those 80 years and older (P < .01). In the ANOVA F test, significant walking speed differences were observed with respect to other demographic and health characteristics (Table 1). Post-hoc tests indicated significant differences for all pairwise comparisons (P < .01, data not shown), except for walking speed differences between subjects with normal BMI (1.07 m/s) and overweight BMI (1.04 m/s, P = .06). Furthermore, we observed clinically significant changes (0.11 m/s) in walking speed among participants who failed condition 4 compared with those who did not (P < .01).
The association between age-related imbalance and walking speed was next evaluated using simple and multivariable linear regression incorporating sample weights (Table 2). In the unadjusted regression model with robust standard errors, inability to perform the balance task was associated with 0.10 m/s slower walking speed (95% confidence interval [CI]: -0.13 to -0.07; P < .01). In the multivariable regression model, inability to perform the balance task was associated with 0.06 m/s slower walking speed after adjusting for potential confounders (95% CI: -0.09 to -0.03; P < .01).
To assess the magnitude of the reduction in gait speed associated with age-related imbalance, we estimated the difference in chronological age that would be equivalent to the association between imbalance and gait speed decline. In a fully adjusted multivariate model accounting for demographic and health risk factors, a 1-year difference in age was associated with a gait speed difference of 0.005 m/s (95% CI: -0.007 to -0.003; P < .01), whereas imbalance was associated with a gait speed difference of -0.062 m/s (95% CI: -0.091 to -0.032; P < .01). Therefore, the difference in age equivalent to the gait speed difference associated with age-related imbalance is approximately 12 years. We performed the multivariable regression with and without visual acuity and peripheral neuropathy in the models. Neither visual acuity nor peripheral neuropathy was significantly associated with walking speed, and adding both variables to the model did not appreciably change the linear regression coefficient (data not shown).
We also evaluated whether age-related changes in walking speed differed as a function of balance function. In an age-stratified analysis, we found that walking speed was slower among individuals with age-related imbalance across all age categories (Figure 1). Moreover, the difference in walking speed between individuals with and without age-related imbalance widened with increasing age. The differences in walking speed between individuals with and without imbalance were 0.07 m/s among 50- to 59-year-olds, 0.06 m/s among 60- to 69-year-olds, 0.07 m/s among 70- to 79-year-olds, and 0.11 m/s among individuals 80 years and older. Of note, an imbalance-age interaction term was not significant in the regression model (P = .80).
Finally, we used structural equation modeling to estimate the extent that imbalance mediated the relationship between age and slower walking speed (Figure 2). We observed that imbalance that occurred with normal aging mediated 12.2% of the association between age and slower walking speed in this national cohort.
DISCUSSION
We observed in these national-level data that the loss of intact balance in healthy aging was associated with slower walking speed, with an effect equivalent to 12 years of age. Moreover, we observed that age-related imbalance partially mediated the well-established association between age and slower walking speed by 12.2%. There are several potential explanations for the observed association in our analysis. It is possible that the limitations in static balance cause a decline in walking speed, as we hypothesized in the structural equation models. Gait has been characterized as a task involving repetitive sequences of postural stability and controlled falls.18 With reduced postural stability, individuals may slow their gait, possibly because of less efficient transitions between static and dynamic phases of the gait cycle. Indeed, one study observed that standing balance training improved walking speed among patients with ischemic heart disease.19
It is also possible that the factors, which influence static balance, contribute to slowing of walking speed associated with healthy aging. The main sensory systems that contribute to static postural control are the visual, proprioceptive, and vestibular systems. Prior studies have shown that older adults rely more on visual feedback to maintain balance during walking than younger adults, and those with significant visual impairment have slower walking speeds compared with age-matched controls.20,21 Moreover, one study found that proprioceptive feedback exercises improved walking speed in individuals with Parkinson's disease, suggesting the importance of proprioceptive inputs to gait function.22 We adjusted for both visual acuity and monofilament testing in our regression models and did not observe that these factors explained the association between static balance and walking speed. However, vision loss and proprioceptive impairment are multifaceted, and also involve contrast sensitivity, visual fields, and depth perception in the case of vision and joint and muscle proprioception in the case of proprioception. These dimensions of sensory function were not measured in NHANES; therefore, our ability to fully evaluate the influence of vision and proprioception function was limited.
In this study, test condition 4 was designed to have the participant increase reliance on vestibular information, suggesting a specific association between age-related vestibular loss and slower walking speed. Indeed, we observed in prior work that older adults with impaired semicircular canal function had slower gait relative to adults with intact function.23 We also previously reported an association between saccular dysfunction and slower walking speed among older women, but not among older men.24 The peripheral vestibular system senses head movements and responds to high-frequency head oscillations that occur during walking.25,26 Vestibular information is also used to stabilize the head and trunk and serves both a postural and gaze stability purpose.27,28 Previous studies showed that individuals with bilateral vestibular loss were less able to coordinate head rotation and head translation during walking, possibly because of impairment of the vestibulocollic reflex.28 Therefore, older individuals with a partial loss of vestibular function may slow their gait to reduce high-frequency head oscillations and allow time for other sensory systems (eg, visual and proprioceptive systems) to compensate for gaze maintenance and head stabilization. The current study of imbalance in a nationally representative sample supports these smaller studies by demonstrating that the more widespread phenomenon of age-related imbalance is associated with slower walking speed in the US population.
Finally, age-related limitations in the musculoskeletal system may mediate the association between imbalance and slow gait. Muscle strength was not measured in the 1999 to 2002 NHANES; therefore, we could not evaluate whether it explained part of the association between postural imbalance and walking speed. Nevertheless, a prior study showed that higher walking speed was associated with increased electromyographic muscle activity in several muscle groups in older individuals.29 This is further supported by a meta-analysis that found that walking speed was fundamentally a function of both dynamic steady-state balance and lower extremity muscle strength in older adults.30
We note several limitations in this study. The NHANES data have good generalizability to the US population, but the data are cross-sectional and thus cannot support causal inferences. Other unmeasured confounders may account for the association between age-related imbalance and walking speed. One potential confounder we were unable to include is depression, which is prevalent in older adults and may be associated with motor slowing.31 In the 1999 to 2002 NHANES, data on depression were collected in younger adults aged 20 to 39 years and thus were unavailable in this study cohort (aged 50-85 years). Moreover, although causality is implied in constructing the SEMs, these models do not demonstrate causation and as such are more hypothesis-generating rather than hypothesis-proving. Finally, the only gait variable we considered was speed. Walking speed is a critical but single dimension of an individual's walking performance and may not sufficiently capture the complex association between balance limitation and gait. Walking speed is easily measured and is a strong predictor of health and physical function, which is why it was included in the NHANES. Future studies will be required to understand how age-related imbalance affects other dimensions of gait, including gait variability, center-of-mass sway, and energy expenditure. Along the same line, elements of dynamic postural control were not quantified in the NHANES. However, limitations in dynamic postural control may be equally important contributors to age-related imbalance and should be considered in future studies.
CONCLUSIONS
In a nationally representative sample, we observed an association between imbalance that occurs with normal, healthy aging, and slower walking speed that was equivalent to the effect of 12 years of age. Walking speed is a robust indicator of health status and a strong predictor of morbidity and mortality in older adults. Further research is needed to determine whether this association is causal. This study emphasizes the importance of considering age-related imbalance in screening for and evaluating slow gait among older adults.
REFERENCES