INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

INCREASED GAIT INSTABILITY, marked unsteadiness and inconsistency from one stride to the next, is common in many older adults, even in the absence of overt disease (36, 57, 82). Gait instability predisposes individuals to falls and likely contributes to changes in neuropsychological and functional status, including a fear of falling, decreased confidence, and self-imposed mobility restrictions (Fig. 1) (57, 80, 81). Marked gait instability may also detract from quality of life and lead to functional dependence and institutionalization.

View larger version (34K): [in this window] [in a new window]   Fig. 1.   Simplified block diagram illustrating some of the physiological and neuropsychological factors that may be associated with gait instability. Locomotor system and certain age-associated changes (shaded boxes) in physiological capacity that may mediate gait instability are shown. Goals of the present study were 1) to quantify the association between gait instability and these potential contributing factors and 2) to determine whether modification of proximate mediators of gait instability would reduce gait instability. Note the highly interdependent nature of many of these relationships and their putative effect on gait instability. CNS, central nervous system; PNS, peripheral nervous system; ROM, range of motion.

In persons with neurodegenerative (e.g., Parkinson's) disease, deficits in the central nervous system's ability to regulate and coordinate motor outputs are largely responsible for locomotor instability (7, 35, 36, 39, 58, 61). In contrast, in older adults without apparent neurological disease, the reasons for the increased gait instability are less clear. It is likely that they are multifactorial. Physiological deficits secondary to aging and disuse may perturb the locomotor system of the older adult to bring about gait instability (Fig. 1) (2, 36, 57, 82). Many age-associated changes may contribute to this process, including reduced range of motion, decreased aerobic capacity, decreased muscle function, and impaired balance (1, 2, 20, 70, 77). Other factors, such as impaired neuropsychological function, may also heighten instability, further exacerbating the effects of decreased physiological capacity (Fig. 1). However, the interrelationships between gait instability and physiological capacity, neuropsychological status, and health-related quality of life have not been well studied. Few studies have quantitatively examined the effects of these age-associated changes on gait instability, the plasticity of gait instability, or the ability to modify physiological and neuropsychological parameters and improve gait stability (51, 57).

In the present study of gait instability in the elderly, we focus on one of the outputs of the locomotor system: the temporal fluctuations in the gait cycle duration (8, 29, 57, 89). The gait cycle duration, also known as the stride time or the stride interval, reflects the walking rhythm. Because it relies on central and peripheral inputs and feedback, it can be viewed as a final output of the locomotor system. Moreover, because a variety of physiological and neuropsychological systems may impact on stride time dynamics, instability measures may be a sensitive marker of dynamic system integration (8, 29, 35, 38, 57, 61).

Stride-to-stride changes in the gait cycle duration can be described with respect to the fluctuation magnitude (e.g., the variance, the size of the fluctuations) and the fluctuation dynamics (how the stride time changes from one stride to the next, independent of the variance) (29, 35, 38, 57, 64, 87). Although both aspects may be referred to as gait instability, a priori, different factors may participate in the regulation of the fluctuation magnitude and the fluctuation dynamics, and the clinical relevance of these two aspects of gait instability may be different as well. For example, in healthy older adults, the fluctuation magnitude is similar to that observed in healthy young adults, although there are subtle age-related changes in the fluctuation dynamics (29, 39).

The goals of the present study, therefore, are to quantitatively investigate gait instability, the factors that contribute to it, its relationship to neuropsychological and functional status, and its potential reversibility in functionally impaired older persons. Functional impairment is an important problem in elderly individuals, leading to great personal and societal burden, including clinical sequelae, such as falls and institutionalization (2, 33, 48, 57, 80). A more complete understanding of its complex multifactorial etiology is necessary to design preventive and rehabilitative approaches to this syndrome (2, 60). Functionally impaired elders often present with mobility problems, physiological deficits (balance, strength, endurance, flexibility), and psychosocial problems (2, 10, 80), and therefore all these domains should be assessed when approaching this area of investigation. A common thread among these domains of function and capacity is abnormality of gait characteristics in frail elders. However, gait instability is a characteristic of gait that has been minimally studied in this regard compared with gait velocity and other simpler constructs. There is reason to believe that gait instability may be a more powerful predictor of functional impairment than characteristics that focus on the mean or "typical" value of gait, e.g., average walking velocity (40, 57).

This study was undertaken with two major purposes in mind: 1) The physical and psychological factors that are associated with gait instability in functionally impaired elders were examined to better understand this element of overall gait and mobility. Understanding its etiology will advance the understanding of the neurophysiological changes associated with aging and the attempts to intervene in this domain and, ultimately, improve function. 2) The efficacy of a multifactorial intervention designed to improve multiple physical capacities simultaneously, i.e., muscle strength, balance, aerobic capacity, and flexibility, was tested. The elements of this intervention were chosen to specifically target those physical capacities that are known to be impaired with age and also believed to be modifiable with appropriate exercise modalities (12, 26, 62, 92). We hypothesized that modification of the proximate physiological mediators of gait instability would affect the distal outcome, gait instability itself. To this end, we measured gait instability and potential contributing factors in older men and women with functional impairment before and after they participated in a randomized placebo-controlled, 6-mo multimodal exercise trial.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subjects

Assessments

Gait instability. The protocol and analytic methods were similar to those employed previously to study gait instability in neurological disorders (35, 39). Subjects were instructed to walk on level ground at their normal pace around a 91.4-m-long indoor carpeted pathway for 2 min. To measure the gait rhythm and the timing of the gait cycle, force-sensitive insoles (37) were placed in the subject's shoe. These inserts produce a measure of the force applied to the ground during ambulation. A small, light-weight (5.5 × 2 × 9 cm, 0.1 kg) recorder was worn on each ankle. Subsequently, the digitized data were transferred to a workstation for analysis using software that extracts the initial contact time of each stride (37). With this information, the stride time or duration of the gait cycle (time from initial contact of one foot to subsequent contact of same foot) was determined for each stride during the 2-min walk, and a time series of stride times was generated for each subject. As we now describe, several measures were used to quantitatively assess different aspects of the stride-to-stride fluctuations in each subject's stride time (35, 39, 41).

Fluctuation magnitude. To study the intrinsic dynamics of the gait rhythm, the time series of the stride time was analyzed as follows (35, 39, 41). The first 5 s of the recorded data were excluded to minimize any start-up effects, and then a median filter was applied to remove data points that were 3 SDs greater than or less than the median value. The average stride time was determined, and a measure of stride-to-stride variability (fluctuation magnitude) was calculated. The stride time variability is defined here as the stride time coefficient of variation (CV), where the variability is normalized to each subject's mean stride time (CV = 100 × SD/mean).

Fluctuation dynamics. We quantified the temporal "structure" or ordering of each time series (independent of the overall variance) by calculating two measures of the stride time fluctuation dynamics: the nonstationary index and the inconsistency of the variance (38). Two time series can be very similar with respect to the mean and variance, i.e., the fluctuation magnitude, but may change in time in different ways and have different fluctuation dynamics (see http://reylab.bidmc.harvard.edu/DynaDx/index.html). To minimize the effects of any differences in mean or variance, each time series was first normalized with respect to its mean and SD, yielding new time series each with mean = 0 and SD = 1, but with different dynamic properties depending on how the local values change with time (38). This normalized time series was then divided into blocks of five strides each, and in each segment the (local) average and (local) SD were computed. The nonstationary index, defined as the SD of the local averages, was then calculated to estimate the dispersion of these normalized, local means. The nonstationary index provides a measure of how the local average values change during the walk, independent of the overall variance (the fluctuation magnitude) of the original time series. A higher nonstationary index indicates greater range among the local averages. We also calculated the inconsistency of the variance, the SD of the local SDs, to evaluate how the local SD changes with time. Higher inconsistency of the variance values indicates greater dispersion and more inconsistent local SDs; the normalized variance is "unstable" and fluctuating with time. Compared with healthy controls (38), the nonstationary index is decreased in subjects with advanced Huntington's disease, while the inconsistency of the variance is increased. The nonstationary index and the inconsistency of the variance are unitless. A composite instability index was next constructed by combining the stride variability and the inconsistency of the variance after normalizing each to a 0-1 scale, with 0 corresponding to low instability. If gait instability changes in the present study subject are parallel to those seen in Huntington's disease, one would anticipate that the nonstationary index would be relatively decreased in persons with greater dysfunction, while the stride time variability, the inconsistency of the variance, and the composite instability index would be relatively increased in persons with greater levels of impairment.

Demographics, health, and neuropsychological status. Standardized tests were used to evaluate baseline status and to control for any group differences. As described below, all the tests were previously validated and have been shown to be reliable and sensitive to change in elderly populations. The number of chronic diseases was ascertained by the study physician through the physical examination and a review of the medical history. On the day of screening, subjects brought in all prescription and over-the-counter medications taken on a consistent basis (at least once a week) to establish the total number of medications used regularly. The Mini-Mental State Exam (27) was used to estimate cognitive impairment. This test has been used extensively to assess mental function in various geriatric populations (28, 31, 43). The Geriatric Depression Scale (GDS), a 30-item yes/no questionnaire with scores ranging from 0 to 30, was used to quantify depressive symptoms (93). The GDS is a reliable and valid self-rating depression screening scale for elderly populations (93). Physical activity levels were estimated from the Physical Activity Scale for the Elderly (PASE) (85). Higher scores correspond to greater levels of activity. PASE scores have been shown to be positively associated with grip strength, static balance, and leg strength and negatively correlated with resting heart rate, age, and perceived health status, with test-retest reliability, assessed over a 3- to 7-wk interval, of 0.75 (95% confidence interval = 0.69-0.80) (85).

Functional status and health-related quality of life. Self-report of the ability to perform activities of daily living (ADLs) and instrumental ADLs (IADLs) were assessed using the Katz ADL and Lawton IADL questionnaires (50, 53). Higher scores on the Lawton IADL correspond to greater functional ability, and lower scores on the Katz ADL correspond to greater functional ability. These instruments are often included in geriatric assessments and have been shown to correlate with depressive symptoms, confidence in avoiding falls while performing usual activities, and physical measures such as gait velocity, balance function, and grip strength (45, 67, 80, 86).

Physiological capacity. muscle function. The one repetition maximum (1 RM) was used to measure dynamic concentric muscle strength of the legs, arms, and trunk using the knee extension, double leg press, and lateral pull-down machines (Keiser Sports Health Equipment, Fresno, CA). The 1-RM test has been widely used to evaluate muscle function and the ability of the muscle to generate force in various populations (25, 47, 59, 62). In older adults, reproducibility of the 1 RM is high, with a high correlation between repeat measures 1 wk apart (62). To determine the 1 RM, incremental loads were added until failure occurred despite verbal encouragement to exert maximal effort (26, 62). Knee extension endurance was measured by determining the total number of repetitions achieved with proper form and no pauses at 90% of the 1 RM. Maximal leg extensor power was also determined using a double-leg-press machine (Keiser Sports Health Equipment). Maximal hand grip strength was determined using a hand grip dynamometer (Takei Scientific Instruments). Reduced grip strength, a validated measure of upper extremity strength, is a strong predictor of disability in older people and is also associated with increased risk of mortality in acute illness (19, 30, 68, 71). The highest of three measurements was used for maximal grip strength.

Gait speed. Maximal and habitual gait speed were assessed over the middle 2 m of a 4-m course with an ultrasonic timer (Ultratimer, DCPB Electronics, Glasgow, Scotland). The average of two measurements made to the nearest 0.01 s was used to determine the maximal and habitual gait speed. Gait speed is a commonly used, reliable measure of physical function in geriatric patients and is a valid indicator of ADL function and lower extremity strength (11, 14, 69, 78).

Flexibility. As performed as part of standard physical therapy practice, passive ankle dorsiflexion and plantarflexion ranges of motion were measured using a goniometer, standardized bony landmarks, and positions.

Interventions

Exercise group. Subjects in the exercise group were given a 6-mo, home-based exercise program that focused on progressive strength and balance training with encouragement to increase overall aerobic and physical activity (63). The exercise program was designed 1) to follow the standards and target levels recommended by the American Council for Sports Medicine; 2) to modify three major components of physical capacity that have been related to function and mobility: strength, balance, and exercise capacity; 3) to mimic, to the degree possible, the laboratory-based interventions that successfully improved strength, balance, and function in similar populations; 4) to be simple to learn and performed by functionally impaired older adults at home; and 5) to provide opportunities for increasing the intensity of the exercise and a range of levels (12, 26, 46, 62, 63). An exercise trainer came to the home to instruct the subject on how to safely and properly perform the exercises. During the first home visit, each subject was given a detailed booklet outlining the program, several sets of dumbbells, and a pair of 20-pound adjustable ankle weights (AllPro Equipment, Jericho, NY). The ankle weights were adjustable from 1 to 20 pounds, in 1-pound increments (we note that none of the subjects "maxed" out). Six home visits were conducted over the 1st mo, with one home visit each month thereafter. The exercises were as follows: chair stands (3 levels from using hands to no hands), circle turns (3 levels), tandem walk (3 levels), plantarflexion (3 levels), ankle dorsiflexion, knee extension (with ankle weights), standing hip extension (with ankle weights), standing hip abduction (with ankle weights), overhead raise (dumbbells), biceps curl (dumbbells), and triceps extension (dumbbells). The upper extremity exercises were included because flexibility and upper extremity strength may impact on function and gait. Subjects were asked to perform the balance and strength exercises three times per week. To tailor the intervention "dose" to each individual, two sets of eight repetitions for each exercise were performed at a target-relative, perceived exertion of 7-8 on the 10-point Borg scale (9). This target level is parallel to that used in laboratory, machine-based studies of progressive resistance training, where the exercise level is at 80% of the 1 RM and is based on American College of Sports Medicine recommended practice and our previous successful experiences in work with older adults (25, 26, 59, 62). Every 2 wk, subjects were encouraged to increase the level of weight at which they were working to maintain the proper intensity and to ensure adequate dose. Subjects were also asked to accumulate 30 min of aerobic activities such as walking and stair climbing each day. All physical activities were documented in exercise logs that were returned to the investigators each month.

Attention control group. Subjects in the attention control group received 6 mo of nutrition education. A registered dietitian made four home visits to each attention control subject in the 1st mo of study and one home visit per month thereafter. As with the exercise group, attention control subjects were given logs to be filled out daily to keep track of the number of servings of fruits, vegetables, and calcium-rich foods consumed.

Statistical Analysis

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Subject Characteristics

Before the intervention, there was a wide range of gait instability among the study subjects (Table 1). Stride time variability ranged from 1.6 to 10.4%. The average stride time variability was 3.6%, almost twice that seen in healthy young and older adults (29, 39). The two measures of the stride time dynamics, the nonstationary index and the inconsistency of the variance, ranged from 0.31 to 0.93 and from 0.11 to 0.45, respectively. Measures of the fluctuation dynamics, the nonstationary index and the inconsistency of the variance, were not correlated with the fluctuation magnitude (stride time variability) but were moderately correlated with one another (Pearson's r = 0.59, P < 0.0001).

Baseline Predictors of Gait Instability

The baseline univariate predictors of gait instability are shown in Table 2. Increased stride time variability generally indicated dysfunction in other domains. For example, high variability was associated with decreased functional status (self-report and performance based), decreased health-related quality of life (i.e., lower physical function and vitality), reduced neuropsychological status (i.e., increased depressive symptoms), decreased physiological capacity (e.g., reduced exercise capacity, dynamic balance, and functional reach), lower physical activity levels, and lower health status [i.e., increased body mass index (BMI)]. Similar relationships were observed for the composite instability index.

Both measures of the stride time fluctuation dynamics were associated with health-related quality of life and neuropsychological status (Table 2). The nonstationary index was higher with self-report of better health status, vitality, and physical function, fewer chronic conditions, increased physiological capacity (e.g., gait speed), and fewer depressive symptoms. A lower inconsistency of the variance was associated with higher health-related quality-of-life measures and fewer depressive symptoms and tended to be correlated with increased upper extremity strength. None of the measures of gait instability were significantly correlated with the number of medications (despite a wide range) or mental status (which was confined to a relatively small range).

Multivariate analysis was performed to identify the independent predictors of the different gait instability measures. As noted in Table 2, the independent predictors of stride time variability were average stride time, flexibility (shelf reach and back reach), dynamic balance, health status (number of chronic conditions), functional status (ADLs), and BMI. These factors explained 77% of the variance in stride time variability.

Independent predictors of the nonstationary index were vitality, maximal gait speed, and the number of chronic conditions, explaining 19% of the variance. Independent predictors of the inconsistency of the variance were upper extremity strength, body pain, balance (i.e., tandem stance time), and self-report of health status (SF-36), accounting for 27% of the variance. The independent predictors of the composite instability index were stride time, upper extremity strength, self-report of vitality, and functional status (ADLs), explaining 61% of the variance.

Effects of Exercise

Functional status and physiological capacity. The effects of the exercise on functional status and physiological capacity are described in detail elsewhere (63). Briefly, compared with the control group, the exercisers significantly improved dynamic balance and performance-based measures of functional status after participating in the 6-mo, multimodal exercise program. Scores on the PPT increased by 6 ± 14% in the exercisers and decreased by 3 ± 14% in the control group (P < 0.01). Scores on the SPPB improved by 28 ± 38% in the exercise group and decreased by 2 ± 22% in the control group (P < 0.001). Dynamic balance improved by 32 ± 16% in the exercisers compared with 9 ± 25% in the control group (P < 0.001). However, although all measures of physiological capacity tended to improve in the exercisers, there was no significant group × time effect on any measure of strength, muscle endurance, exercise capacity, or habitual or maximal gait speed. For example, knee extension strength was 170 ± 62 and 165 ± 64 N at baseline in the exercise and control groups, respectively, and increased to 200 ± 75 and 183 ± 71 N, respectively, after the intervention. There were also no differences between the groups over time for the self-reported quality-of-life measures (SF-36) or depressive symptoms.

Gait instability outcomes. As shown in Table 3, there was a small, but significant, exercise effect on the inconsistency of the variance and the composite instability index; both measures decreased in the exercise group and increased in the control group. Changes in stride time variability and the nonstationary index in response to the exercise intervention were more variable and were not statistically significant.

View larger version (16K): [in this window] [in a new window]   Fig. 2.   Effect of exercise on gait instability. In this 75-yr-old woman with relatively large gait instability at baseline, stride time variability (coefficient of variation) decreased by ~50%, from 7.6 to 4.0%, and the other gait instability measures also improved (see text) as knee extension strength improved by 42% after the intervention.

Correlates of change in gait instability in response to exercise. Compared with baseline values, a decreased inconsistency of the variance was significantly associated with increased upper extremity flexibility (decreased back reach distance) among the exercisers (Table 4) but not in the control group. In turn, increased upper extremity flexibility (decreased back reach) was associated with decreased depressive symptoms (Spearman's r = 0.42, P = 0.021) and increased functional performance (Spearman's r = 0.34, P = 0.06 with respect to ADL score) in this group. Among the exercisers who showed an improvement in upper extremity flexibility after the intervention (n = 15), the inconsistency of the variance was reduced by 22%, while it increased by 13% in the exercisers who showed no improvement in back reach (P = 0.004).

View larger version (16K): [in this window] [in a new window]   Fig. 3.   Among the exercisers, subjects who showed any improvement in knee extension strength (strength gainers) significantly improved knee extension strength and stride variability after the intervention compared with those exercisers who showed no improvement in knee extension strength (other exercisers). Strength and all other measures were not different in these two subgroups of exercisers at baseline, except stride time variability was increased in those subjects who subsequently improved knee extension strength (P = 0.032). The improvement in stride time variability was associated with the improvement in strength (Pearson's r = 0.45, P = 0.014). For stride time variability, improvement here is defined as the negative of the percent change from baseline. Error bars, SE.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

This quantitative study of gait instability in older persons reveals several findings of interest: 1) Many factors contribute to gait instability. 2) Stride time variability is strongly associated with self-report of functional status and performance-based measures of function that have been shown to predict important clinical outcomes [e.g., morbidity and nursing home admission (33)]. 3) Neuropsychological status and health-related quality of life play important, independent roles in gait instability. 4) Improvement in physiological capacity is associated with reduced gait instability. Taken together, these findings suggest that although the etiology of gait instability in older persons with mild-moderate functional impairment is multifactorial, exercise interventions designed to reduce gait instability may be effective in bringing about a more consistent and more stable walking pattern. To some degree, the age-associated increase in gait instability is modifiable.

The multifactorial nature of gait instability is not surprising. Gait dynamics are controlled by an array of physiological and neuropsychological systems that rely on neural, motor, sensory, and volitional inputs. On the other hand, the strong association between stride time variability and functional status, even with measures that are based on the ability to independently perform activities such as bathing and dressing, is less intuitive. Perhaps this can be better understood by examining the relative importance of balance and muscle function in gait instability. Although gait instability is intuitively linked to balance and balance capabilities are required for walking-related tasks (3), static balance was not a particularly strong predictor of any of the gait instability measures. Similarly, although lower extremity muscle function is certainly important for locomotion (46, 90, 92), in the present study, depressive symptoms were more closely associated with stride time variability than muscle function, and vitality was the most significant predictor of the metrics of gait instability that were based on fluctuation dynamics. Upper extremity function and flexibility also played a relatively critical role in predicting gait instability, consistent with other findings of the importance of these properties on gait. Thus many diverse factors apparently can contribute to gait instability. Perhaps the strong association with functional status and the ability to perform ADLs reflects a common reliance on a multiplicity of neuropsychological and physiological components. Similarly, in studies of community-living older adults (10, 24, 51), it was observed that lower extremity strength explains only a relatively modest amount of the variance of functional performance, highlighting the importance of other factors, in addition to strength, on function and, apparently, gait instability.

Two other factors may partially explain the significance of neuropsychological function and perceived health-related quality of life compared with strength and balance in gait instability in the present study. In general, many physiological systems that participate in walking are not fully taxed but are used only at submaximal levels. For example, only ~40% of maximal strength of the quadriceps is used during habitual ambulation. Similarly, in a study of relatively disabled older women, Ferrucci et al. (24) observed that strength is associated with lower extremity function, mainly in the lower portion of the range of strength, resulting in a curvilinear relationship between physiology and function (14). Thus mild weakness, such as that seen in our subjects, may not show a dramatic effect on gait instability. In addition, it is possible that, with only mild physical impairment, the effects of neuropsychological function become relatively exaggerated. To investigate this question further, it would be helpful to study these relationships among frail older adults with more reduced physiological capacity and to examine the role of conscious effort in the presence of reduced automaticity of function. Nonetheless, it is important to keep in mind that improvement in strength was indeed associated with reduced gait instability.

A number of studies have examined the effects of exercise on gait in older adults (12, 46, 55, 83, 91, 92). Different results have been obtained, in part, because of differences in the exercise intervention, in the study population, in the duration of the exercise program, in the study design, and in the assessment of gait. In a 12-wk randomized trial of dynamic resistance training using elastic tubing (83), significant exercise-related improvement in balance or gait was not achieved. In that study, baseline measures of habitual gait speed were >1 m/s. In another study (91), 12 wk of balance and strength training did not improve gait speed, presumably because baseline strength and balance were already adequate for gait and walking speed was already fairly high (>1 m/s) at baseline, while measures of balance did improve. In contrast, Judge et al. (46) and Lord et al. (55) found significant exercise-related increases in habitual gait speed after 12 wk of balance and strength training and after 22 wk of balance, strength, aerobic, and flexibility training, respectively. In both of these studies, subjects who walked more slowly initially showed the greatest gains in gait speed. In addition, Lord et al. found that, after 12 mo of participation in a multimodal, group-based exercise program, measures of postural stability (e.g., test of maximal standing balance range) improved compared with a control group (56). More recently, Krebs et al. (51) demonstrated that 6 mo of moderate exercise improves mediolateral stability.

The results of the present study extend our knowledge about the potential reversibility of gait instability to measures on the basis of the fluctuations of walking rhythm and features of gait that have previously been associated with fall risk (36, 40, 57). Age-associated changes in walking rhythm dynamics are apparently modifiable, at least in certain situations, and not necessarily an inevitable result of aging. A small, but significant, intervention effect was observed in the exercisers, and it appears that older adults with decreased physiological capacity were more responsive to the intervention. The randomized, placebo-controlled design suggests that the observed effects were due to the intervention and not merely to the "attention" received by the exercisers. We hypothesized that modification of the physiological mediators of gait instability would alter gait instability, and the results are consistent with this hypothesis. Nonetheless, the multimodal nature of the intervention and our finding that gait instability is significantly associated with multiple physiological and neuropsychological factors limit our ability to identify the specific mechanisms of action of the intervention. The ability of exercise to partially restore muscle strength and function in older adults has been well documented (26, 62). Although much remains unknown about the plasticity of the aged motor control system (23), previous studies also indicate that age-associated changes in human motor control and performance may be modifiable in response to exercise and training, independent of the effects on subspinal, peripheral function (4, 5, 15, 49, 74, 76). The present results are consistent with these findings; however, additional studies, perhaps combined with functional neuroimaging, are required to more clearly define the response of gait instability to exercise.

The present study also lends support to the idea that balance, specifically, "standing still" performance, may be less germane to the functional balance needed during gait (51). Our findings of an exercise effect on gait instability without a parallel effect on gait speed also support the idea that stability at a self-selected gait speed "may be more important to elders' well-being than simply walking faster" (51). Perhaps, we should shift the emphasis from simply improving gait speed to restoring gait stability.

Similar to previous studies that found more profound exercise effects on gait in persons with more impaired gait (46, 55), we observed that exercisers were more likely to successfully reduce their gait instability if they started with relatively increased locomotor impairment and greater instability. This difference may reflect a general ceiling effect and limits on the modifiability of the involved underlying physiological systems. Alternatively, this effect may simply be due to the relatively moderate nature of the intervention. As is common with home-based exercise trials, the home-based exercise program was apparently not fully successful at eliciting a sufficient physiological response to the intervention in all subjects. Perhaps the ceiling effect would have been removed with an exercise program that produced more robust changes on muscle function and physiological capacity. Additional studies are needed to address this question. Nonetheless, the present findings suggest that these gait instability measures may be useful in identifying older adults who may be among those with the greatest need to reduce gait instability and also are likely capable of reducing gait instability, even with only relatively modest effort.

This study has a number of limitations. 1) Although all subjects underwent a physical examination by the study physician, a comprehensive neurological examination was not performed. It is possible that undetected neuropathology may have contributed to the observed gait instability. Perhaps future studies should use other diagnostic tools to more fully evaluate and account for each subject's neurological status and the response to exercise. 2) Ankle muscle weakness has been associated with postural instability and fall risk (56, 88) and may contribute to gait instability; however, we did not measure muscle strength of the plantar- or dorsiflexors. 3) Because the exercise intervention did not achieve marked improvement in muscle strength, we can only speculate about the effects of a more robust exercise intervention on gait instability. An exercise study in the laboratory setting where the strength of the intervention is more readily tailored to each individual and the "dose" is controlled more precisely may also enable further study of the potential reversibility and plasticity of gait instability and allow us to determine if and how gait instability may be modified, even in stronger individuals. 4) The relatively small sample size may have affected the multivariate analyses and limited our ability to determine all the factors that are independently associated with gait instability. 5) The intervention was designed to improve function. Perhaps an intervention that was targeted to specifically reduce gait instability would have had a greater impact.

Future work will also be helpful in determining the most appropriate exercises and the optimal levels for reducing gait instability in older adults. It will also be helpful to directly compare measures of gait instability on the basis of the fluctuations in the timing of the gait rhythm with those based on kinematics, balance platform-based measures of postural control, and dynamic balance (17, 51, 56, 91), to study further the factors that contribute to gait instability and the complex interactions between instability, physiological capacity, neuropsychological function, and other mediators of these properties, and to further clarify the mechanisms whereby neuropsychological function modifies gait instability. The findings of the present study highlight the multifactorial nature of gait instability and its potential reversibility, at least in certain populations. Gait instability may not be an irreversible by-product of aging. Rather, improvements in gait stability apparently may be achievable by way of physiological adaptation during targeted, multimodal exercise regimens. Apparently, enhancement of the proximate physiological mediators of gait instability may improve gait instability itself. In addition, it appears that when attempting to restore gait stability, we should focus on improving neuropsychological function as well as physiological capacity. Despite intensive prevention efforts, falls in the elderly remain a major public health problem (48). Perhaps, an approach that focuses on physiological and neuropsychological factors may be useful for gaining additional insight into the etiology of gait instability as well as for augmenting our ability to reduce gait stability and fall risk and their deleterious sequelae via targeted interventions.