The Effect of Exercise on Gait Patterns in Older Women: A Randomized Controlled Trial

Journal of Gerontology: MEDICAL SCIENCES 1996. Vol. 51A, No. 2, M64-M70 Copyright 1996 by The Geromological Society of America The Effect of Exercise on Gait Patterns in Older Women: A Randomized Controlled Trial Stephen R. Lord, 1 David G. Lloyd, 2 Meredith Nirui, 2 Jacqueline Raymond, 2 Philippa Williams, 1 and Rachel A. Stewart 2 'Prince of Wales Medical Research Institute, New South Wales, Australia, department of Safety Science, University of New South Wales. Background. This study was undertaken to determine (a) whether a program of regular exercise can improve gait patterns in older women, and (b) whether any such improvement in gait is mediated by increased lower limb muscle strength. Method. A 22-week randomized controlled trial of exercise was conducted as part of the Randwick Falls and Fractures Study in Sydney, Australia. Subjects were 160 women aged 60-83 years (Mean age 71.1, SD = 5.2) who were randomly recruited from the community. Exercise and control subjects were tested prior to and at the end of the trial. At initial testing, exercisers and controls performed similarly in the strength and gait parameters. They were well matched in terms of age and a number of health and life-style characteristics. Results. At the end of the trial, the exercise subjects showed improved strength in five lower limb muscle groups, increased walking speed, cadence, stride length, and shorter stride times as indicated by both reduced swing and stance duration. There were no significant improvements in any of the strength or gait parameters in the controls. Within the exercise group, increased cadence was associated with improved ankle dorsiflexion strength, and increased stride length was associated with improved hip extension strength. Exercise subjects with initial slow walking speed showed greater changes in velocity, stride length, cadence, and stance duration than those with initial fast walking speed. Conclusion. These findings show that exercise can increase gait velocity and related parameters in older persons, and that part of this increase may be mediated by improved lower limb muscle strength. WHILE the role of exercise in the maintenance of good health and mobility in older adults is generally accepted, the specific benefits of different types of exercise on various physiological systems are still largely in question. This is particularly the case for balance and gait (1). It has been found that older persons actively engaged in exercise perform better in gait parameters such as selfselected walking speed compared with matched groups of older nonexercisers (2,3)- The implications of such findings are unclear, however, as it is not possible to conclusively determine any causal directions of associations found, that is, whether high exercise levels maintain an individual's fitness and subsequent walking patterns, or conversely, whether poor health or impaired stability and gait preclude exercise (4). Only a few studies have assessed the effects of exercise and physiotherapy trials on the gait of elderly persons, and most of these trials have been of a pilot nature only (5-10). Meier (5) had six nursing home residents participate in intensive daily physiotherapy for 5 weeks and found a high degree of improvement in the gait of five of these subjects compared to six controls. Sauvage et al. (6) also studied nursing home residents, this time undertaking a moderate- to high-intensity strengthening and aerobic exercise program for 12 weeks. They found that significant, though limited, improvements could be achieved in clinical mobility scores, strength, muscular endurance, and certain gait parameters including velocity. Fiatarone et al. (7) also found that highintensity strength training can improve tandem gait speed in nonagenarians in institutional care. Studies of the effects of exercise on gait patterns of healthy older people have produced uncertain results. Brown and Holloszy (8) investigated the effects of low-intensity exercise for one hour/day, 5 days/week for 3 months on 62 older exercisers and compared their results with 13 controls. They found significant improvements in strength, range of motion of hip and trunk, and standing balance, but none in endurance or gait. In contrast, Judge et al. (9) found clinically significant improvement in strength and gait velocity in 31 subjects following a 12-week resistance and balance training exercise program. In another study of 55 community-dwelling older people, Topp et al. (10) reported an unexpected finding, in that following a 12-week dynamic resistance training program, exercisers actually demonstrated reduced gait velocity. While the results of the above studies are encouraging, the methodological problems inherent in many of these trials, including small sample sizes, rudimentary gait analysis techniques, nonrandomization of intervention and control groups, short duration of the intervention, and differing outcome measures limit our understanding of the beneficial effects of exercise on gait. In this study, we attempted to address these problems by undertaking a long-term (22-week) randomized controlled trial of exercise in a large sample of older women who were recruited from the community, with the aim of assessing M64

EFFECT OF EXERCISE ON GAIT PATTERNS M65 whether exercise has beneficial effects on gait patterns. Further, we examined whether any such improvements in gait are related to concomitant increases in lower limb muscle strength. METHODS Subjects The sample comprised women aged 60 years and over who had taken part in the Rand wick Falls and Fractures Study (11). The women, who were living independently in the community, were initially recruited from 64 randomly selected census collector's districts in the Randwick local government area in Sydney, Australia. All women aged 60 years and over living within these districts (who were identified using extracted information from the electoral roll) were invited to take part in the study. The only exclusion criteria were not living at the dwelling at the time of the study or speaking/understanding no or very little English. Seven hundred and four women (60% of those eligible) participated in the initial phase of the study by completing a structured interview containing questions about falls frequency and related health and life-style factors. No significant differences were found between the initial study sample and the reference population with regard to age structure, marital status, and employment status (whether or not still in the work force). In phase 2 of the study, the coded identification numbers of 374 subjects were randomly assigned to either the exercise or control recruitment pools so as to have equal numbers in each pool. Randomization prior to recruitment was conducted so as to avoid possible inducement of increased exercise in the control subjects. At this stage, it was found that 9 women had died since taking part in the initial survey and 21 were not living at the same address. An additional 35 women could not be contacted despite six attempts (three letters and three home visits). Women were excluded from taking part if they were ill and/or immobile (n = 30), were in hospital (n = 4), had a medical condition involving the neuromuscular, skeletal, or cardiovascular system that precluded taking part in an exercise program (as determined by a physician at pretest) (n = 2), spoke little English (n = 2), or if they were already attending exercise classes of equivalent intensity to the study intervention (n = 5). The participation rates of eligible subjects were similar in the exercise and control groups: 70.9% and 77.6%, respectively (x 2 = 1.54, df = \,p =.21). Of the 197 subjects who took part in the exercise trial, 37 did not complete the baseline gait assessments. Fifteen women were not assessed because the gait apparatus was unavailable, and 22 women were unable to undertake the test (usually because their stride was too short for the apparatus, so that two consecutive foot strikes were recorded by the force plate, instead of the strike from a single foot). The mean age of the 80 women in the exercise group was 71.1 years (SD = 5.2), which was almost identical with the mean age of the 80 controls 71.2 years (SD = 5.2), (t m =. 18, p =.85). The exercise and control groups were also similar across a number of health and life-style measures, including medical conditions, falls, instability, drug use, and inactivity (12). The Exercise Program The exercisers participated in approximately one-hour exercise sessions twice weekly for two 10-week sessions (with an inter-term two-week break). The classes were divided into four sections: a 5-minute warm-up period, a 35- minute conditioning period, a 15-minute stretching period, and a 5-10-minute relaxation (cool down) period (13). The exercises were undertaken as group activities, with a major emphasis on social interaction and enjoyment. Most of the activities were accompanied by music. Warm-up period. The warm-up period commenced with moderate-paced walking. After 2-3 minutes, arm movements were included to increase heart rate. Conditioning period. The conditioning period contained aerobic exercises, activities for balance, hand-eye and foot-eye coordination, and strengthening exercises. The aerobic exercises involved continuous movement of the legs and trunk and intermittent movement of the arms. The leg movements were designed to use the full range of movement of the hip, knee, and ankle joints, and to condition and strengthen all major muscle groups. These included movements that extend, flex, abduct, adduct, and rotate the leg and foot such as side-stepping; fast walking; forward and backward stepping; leg lifts; placing foot to the front, side and behind; kpee bends; forward and side lunging; and heel rises. The trunk movements were designed to maintain flexibility of the spine and to condition and strengthen the back, chest, abdominal, and pelvic floor muscle groups. These included movements that rotate, flex, and extend the neck, back, and pelvis such as twisting the upper body, body bends, neck side flexion and rotation, knee lifts, opposite elbow to raised knee, pelvic rocking, pelvic floor contractions, and belly dancing techniques. The arm movements were designed to use the full range of movement of the shoulder, elbow, and wrist joints, and to strengthen all major muscle groups. These included movements that extend, flex, abduct, adduct, and rotate the arm and hand, such as circling the arms, biceps curls, bench press, upright row, short and long arm shoulder lever, mock boxing, shoulder rolls, and shrugs. The activities for balance, hand-eye, and foot-eye coordination included standing on one leg with the other raised; ball games requiring catching with one hand while standing or moving; kicking a moving ball; throwing to a moving target; running under a skipping rope; and team ball games. Strengthening exercises included lifting one's own body weight (e.g., modified push-ups) and opposing-musclegroups resistance exercises (e.g., while seated, lifting leg off floor while resisting movement with hand pressing down on knee). Stretching. Participants undertook the stretching exercises while sitting on a chair or on the floor. All muscle groups were stretched. The muscles were slowly elongated and held for at least 20 seconds. Participants were encouraged to breath and relax throughout each stretch. Relaxation. In this period, participants sat on a chair or the floor, or lay on the floor. A variety of techniques were

M66 LORDETAL. used including muscle relaxation, concentration on specific body areas, controlled breathing, and guided imagery. Outcome Measures Gait. The gait assessment was carried out in the Department of Safety Science Biomechanics Laboratory at the University of New South Wales. The apparatus consisted of a heavy, rigid wooden decked walkway (11.2 meters in length) containing a KISTLER 928IB 11 load platform at its center (14). This platform, which measured ground reaction forces of a single foot strike, was mounted level with the walkway on a concrete base. The signals from the load platform were passed through KISTLER charge amplifiers to the data collection computer. The subject's heel strikes were detected by two sensitive accelerometers attached to the walkway. The timing of the electronically processed signal from the accelerometers was automatically recorded by the data collection computer which permitted step, stride duration, and cadence to be measured. This method has been shown to be more accurate than heel switches and affords no encumbrance to the subject (15). Walking speed was measured using two proximity sensors at a set distance apart with the load table centered in between. Time to traverse this distance was recorded automatically by the data collection computer. In each walk, the foot which struck the load platform (right or left) was recorded so as to calculate the left-to-right and right-to-left step duration. From the cadence value, stride duration was calculated. Stance duration was determined from the vertical ground reaction force signal, which when subtracted from stride duration, gave swing duration. All subjects walked barefoot, so as to control for the effects of different shoes (between trials and among subjects). In addition, subjects wore a standard set of test clothing which consisted of a pair of close-fitting stretch bicycle shorts and a sleeveless shirt. The trials were undertaken at a self-selected comfortable walking speed, and data collection commenced only after subjects were accustomed to walking in the laboratory environment. Data were collected for up to 20 walks 10 when the left foot hit the force platform and 10 when the right foot hit the force platform. The gait facility also permits sagittal plane motion to be recorded using a NAC HSV400 camera with retro-reflective markers on major body landmarks. These data, with the synchronized ground reaction forces, permit both a kinematic and joint kinetic analysis of gait; however, such analyses are not presented here. Muscle strength. The strength of five muscle groups in the dominant leg was measured. Testing of the hip and knee muscle groups was performed using a strap assembly, incorporating a strain gauge load cell, which was connected to an amplifier with the outgoing signal recorded on a chart recorder. The hip flexors and extensors were measured while the subject was standing, by placing a strap around the subject's leg just proximal to the knee joint. The subject's torso was supported by a padded rest during the tests to minimize recruitment of other muscle groups. The knee flexors and extensors were measured while the subject was sitting on a tall chair by placing a strap around the leg just proximal to the ankle joint. In three experimental trials per muscle group, the subject pulled against the strap assembly with maximal force; the greatest force for each muscle group was recorded. The testing of the ankle dorsiflexion was performed using a specially designed ankle strength testing device which used a pivoted platform attached to a strain gauge load cell and connected to the same instruments as above. While the subject was sitting on a tall chair, the foot was secured to the pivoted platform. In three experimental trials, the subject attempted to maximally dorsiflex in the device, and the greatest force was recorded. Statistical Analysis Associations among the gait parameters and between the gait parameters and age were assessed using Pearson correlations. Student's /-tests were used to compare the means of the gait and strength measures of the exercise and control groups at initial assessment. Two separate repeated-measure multiple analyses of variance (ANOVAs) were then used to compare changes in test performance at the end of the program in the gait and strength measures between the exercise and control groups. In these analyses, the gait and strength variables were treated as the within-subjects factors, and group allocation was treated as the betweensubjects factor. Univariate analyses were then performed to assess which of the individual variables demonstrated significant differences. Within the exercise group, Pearson correlation coefficients were calculated to assess relationships between degree of improvement during the trial for each test and (a) class attendance and (b) initial test performance. Finally, Mests and ANOVA procedures were used to determine whether improvements in the gait measures in the exercisers at retest were related to concomitant increases in lower limb muscle strength. The data were analyzed using the SPSS computer package (16). RESULTS Attendance and compliance. Of the 80 subjects who were recruited into the study as exercise subjects, 68 were still attending and available for retest at 22 weeks (although in two cases, gait retests were not possible because of equipment failure). Of the 12 subjects who failed to complete the program, one died, one suffered a stroke, and one suffered a laceration to the leg that prevented participation. The remaining nine dropped out after attending no more than six sessions. The mean number of classes attended for those who completed the program was 29.0 (72.5%). The range was 10-40 classes (25%-100%) with 51 subjects (77%) attending 25 or more classes. Sixty-four control subjects were also retested. Six control subjects were not reassessed because of equipment unavailability, and 10 declined a second test. Baseline-post-intervention comparisons. Table 1 shows that all of the gait parameters, with the exception of stride length and swing duration, were significantly intercor-

EFFECT OF EXERCISE ON GAIT PATTERNS M67 Table 1. Correlations Among the Gait Parameters at Pretest (n = 160) Cadence Stride Stance Duration Stance % Swing Duration L-R Step R-L Step Velocity Cadence Stride length Stance duration Stance percentage Swing duration Left-right step.77**.83**.28** -.83** -.96** -.40** -.62** -.43** -.57**.62** -.43** -.78**.04.74** -.22* -.76** -.98** -.29**.96**.38**.81** -.73** -.97** -.25**.97**.36**.82**.97** *p<.01;**p<.001. related. Of particular note, velocity was strongly positively associated with stride length and strongly inversely associated with stance duration. Cadence was strongly inversely associated with stance duration and right-to-left and left-toright step duration. All of the gait parameters with the exception of swing duration were significantly associated with age (Table 2). Most of the baseline gait and strength scores for the exercisers and controls were similar, although exercisers had longer stance duration and right-to-left step times (Table 3). Pre- and post-intervention scores and mean percentage change in performance for the two groups are shown in Table 4. Paired r-tests revealed that the exercise subjects showed significant improvements in every gait and strength measure. In contrast, the control subjects recorded scores at retest that were very similar to or marginally inferior to the test scores at pre-test for the tests of velocity, stride length, stance duration as a percentage of total step time, right-toleft step time, and left-to-right step time. Further, cadence was significantly lower, and stance and swing duration longer at retest. In one strength measure, ankle dorsiflexion, control subjects showed a notable (though not significant) improvement which may indicate a practice effect. The multiple analysis of variance for the gait parameters revealed a significant Group x Time effect (F = 3.50, df = 8,121, p <.001), indicating an improvement in the test measures in the intervention group, but little or no change in the control group. Compared with baseline scores, the univariate analyses revealed that exercise subjects showed significantly improved performance in all tests (compared with controls) with the exception of stance duration as a percentage of total stride time. There was also a significant Group x Time effect (F = 9.78, df = 5,124,/? <.001) in the multiple analysis of variance for the strength measures, and the univariate analyses revealed significant differences between exercisers and controls in every strength test. Table 5 shows that within the exercise group, the degree of change in the tests of velocity, cadence, stance duration, left-to-right and right-to-left step duration was significantly associated with number of sessions attended. Table 5 also shows that within this group, initial test performance was significantly inversely associated with amount of change in all of the gait parameters with the exception of stance duration as a percentage of total stride time, indicating that those with initial slow walking speed (and related parameters including small step length and reduced cadence) improved more than those with initial high walking speed. Those with Gait Parameter Table 2. Correlations Between the Gait Parameters and Age at Pretest (n = 160) Left-right foot step duration (msecs) Right-left foot step duration (msecs) *p<.05;**p<.01. Table 3. Mean Values and (SDs) for the Gait and Strength Measures at Pretest Gait parameters Left-right foot step duration (msecs) Right-left foot step duration (msecs) Strength (kg force) Ankle dorsiflexion Knee extension Knee flexion Hip extension Hip flexion *p <.05. Exercisers («= 80) 1.11 (0.19) 115.4 (11.2) 1.15 (0.13) 675 (79)* 64.2 (1.8) 376 (33) 517 (51) 527 (54)* 8.0 (2.4) 21.9 (6.9) 12.6 (4.0) 22.6 (7.8) 23.8 (6.8) Correlation -.36** -.21** -.36** -.24**.17*.11.19*.20* Controls (n = 80) 1.14 (0.19) 118.9 (12.5) 1.15 (0.12) 649 (71) 64.0 (2.5) 366 (37) 504 (49) 510 (50) 7.7 (2.2) 22.8 (8.0) 12.4 (4.3) 21.1 (8.1) 23.8 (7.5) initial velocities less than 1 m/sec (approximately the lowest velocity quartile) improved 11.8% at retest, compared with only 3.9% in those in the second and third quartile groups and 1.8% in those in the highest quartile group. The number of exercisers with velocities less than 1 m/sec declined from 21 (31.8%) at initial test to 11 (16.7%) at retest, whereas the number of controls with velocities less than 1 m/sec remained virtually unchanged: 15 (23.4%) at initial test and 16 (25.0%) at retest (chi square = 5.29, df = \,p<.05).

M68 LORDETAL. Table 4. Mean Values (SDs) for the Gait and Strength Measures at Pre- and Posttests Pretest Posttest % Change! Exercisers (/» = 66) Gait parameters Left-right foot duration (msecs) Right-left foot duration (msecs) 1.12 116.2 1.16 668 64.0 375 514 524 (0.19) (11.4) (0.13) (79) (1.7) (33) (51) (55) 1.18 118.8 1.19 648 63.7 369 499 511 (0.18) (9.6) (0.13) (66) (1.9) (29) (40) (44) 5.9** 2.6** 3.1** -2.7** -0.5-1.3* -2.7** -2.1** Strength (kg force) Ankle dorsiflexion Knee extension Knee flexion Hip extension Hip flexion 8.0 22.6 12.7 22.7 24.2 (2.4) (6.8) (3.9) (7.9) (6.6) 9.2 26.7 14.2 24.1 27.1 (2.0) (8.2) (4.0) (7.8) (7.1) 21.2** 20.6** 15.5** 9.1* 14.4** Controls (n = 64) Gait parameters Left-right foot duration (msecs) Right-left foot duration (msecs) 1.15 119.2 1.15 647 64.0 366 503 508 (0.19) (13.1) (0.12) (74) (2.6) (37) (51) (51) 1.12 117.1 1.15 662 64.0 373 507 514 (0.18) (11.8) (0.12) (70) (1.9) (40) (49) (50) -1.8-1.4-0.5 2.6 0.1 2.1 1.1 1.3 Strength (kg force) Ankle dorsiflexion Knee extension Knee flexion Hip extension Hip flexion 7.8 23.6 12.8 21.6 24.7 (2.2) (7.9) (4.4) (8.0) (7.4) 8.1 23.5 11.9 20.8 24.1 (2.1) (7.4) (4.4) (7.2) (7.0) 6.5 2.3-5.3 0.4-1.0 Notes. Increases in the tests of velocity, cadence, stride length, and strength, and decreases in the tests of stance duration, stance percentage, swing duration, left-right foot step duration, and right-left foot step duration indicate improvements. tchange expressed as: ((retest score - baseline score)/baseline score) x 100. */; <.01; **p <.001; MANOVA univariate comparisons. Table 5. Correlations Between Percentage Change in the Gait Parameters at Retest and Initial Performance and Attendance! Gait Parameter Left-right foot step duration (msecs) Right-left foot step duration (msecs) texercisers, n = 66. *p<.05;**p<.01. Initial Performance -0.44** -0.56** -0.25* -0.52** -0.21-0.48** -0.61** -0.57** Attendance 0.30* 0.32* 0.17-0.36** -0.23-0.15-0.34** -0.34** Changes in the gait parameters were compared between the exercisers who did and did not show improvements in the lower limb muscle strength tests (Table 6). Subjects with notable improvements in ankle dorsiflexion strength (percentage improvements greater than 10%) demonstrated significantly greater increases in cadence and corresponding reductions in the related measures of stance duration, swing duration, and left-to-right step duration. Subjects with notable improvements in hip extension strength (percentage improvements greater than 10%) demonstrated significantly greater stride length and significantly reduced stance duration as a percentage of stride duration. No significant differences were evident between improvement in the other three muscle groups and improvement in the gait measures. Subjects were then classified into groups according to whether they demonstrated notable improvements in neither, either, or both ankle dorsiflexion and hip extension strength. Those who showed no notable improvement in either strength measure (n = 15) showed a.20% (SD = 4.28) reduction in velocity, which was much less than the 6.69% (SD = 9.23) increase in the 39 subjects who showed a notable improvement in one strength measure and the 7.14% (SD = 7.13) increase in the 12 subjects who showed notable improvements in both measures (F = 4.11, df = 2,63, p <.05).

EFFECT OF EXERCISE ON GAIT PATTERNS M69 Table 6. Mean Percentage Change (plus SDs) in the Gait Parameters for Those Who Did and Did Not Show Notable Increases in Strength at Retest (n = 66) Gait Parameter <10% increase (n = 34) Ankle Dorsiflexion >10% increase (n = 32) <IO% increase (n = 36) Hip Extension >10% increase (n = 30) Velocity (msec) Left-right foot step duration (msecs) Right-left foot step duration (msecs) 3.38 (7.62) 0.71 (4.12) 2.67 (5.83) -1.01 (4.69) -0.49 (2.05) 0.44 (5.90) 0.70 (4.01) -0.79 (4.21) 7.27 (8.84) 3.65 (4.96)* 3.35 (5.04) -3.68 (5.46)* -0.42 (1.78) -2.58 (4.55)* -4.18 (4.79)** -2.89 (4.93) 4.32 (8.79) 2.29 (5.25) 1.84 (4.51) -1.93 (5.92) 0.02 (1.92) -2.11 (4.16) -2.58 (5.37) -1.84 (5.09) 6.40 (8.07) 1.81 (4.11) 4.52 (6.18)* -2.72 (4.37) -1.12 (1.68)* 1.59 (5.84) -2.10 (3.92) -1.54 (4.00) *p<.05;**/><.01. DISCUSSION The findings of this large-scale, randomized, controlled trial of exercise in older people revealed significant improvements in walking speed, cadence, and stride length in the exercise group, with no significant changes evident in the control group. A number of these gait parameters including decreased walking speed and stride length have been found to be associated with age (3,17-21) and also falling (22). In the present study we found that all gait parameters, with the exception of swing duration, were significantly correlated with age, and that on retest, only 22 weeks following initial testing, control subjects had significantly lower cadence, and longer stance and swing times. In marked contrast, the improved gait patterns demonstrated by the exercisers at the completion of the trial indicate that exercise programs can ameliorate age-related declines in functional gait in older people and produce more confident walking patterns. The concomitant improvements in strength and the strong associations between degree of improvement and class attendance indicate a clear training effect. Thus, it seems that the exercise program was of an appropriate nature, the stimulus sufficiently intensive, and the program of adequate duration to produce considerable improvements in walking patterns in the intervention group. The findings showed that the increased walking speed evident in the exercisers at retest resulted from increases in both stride length and cadence. The significant associations uncovered between strength in specific muscle groups and certain gait parameters suggest that (a) cadence was mediated, at least in part, by improved ankle dorsiflexion strength, and (b) increased stride length was mediated by improved hip flexion strength. Further, those subjects with improvements in either or both muscle groups demonstrated increased walking speed. These interesting associations add insight into the mechanisms by which exercise improves gait patterns. However, it is acknowledged that such associations will require confirmation in further studies. It is also acknowledged that a limitation of studies of this type is that subjects cannot be blinded to their "treatment" condition. Thus, the exercise subjects were aware that they were receiving the intervention, and it is possible that part of their improved gait performance may have been due to increased motivation and effort expended at retest. A second study limitation is the possibility of experimenter bias, as a result of the investigators who assessed the subjects not being blind to treatment status. Most previous studies that have examined the effects of exercise on gait have reported little or no measurable benefit, although these trials have had a number of limitations including small sample sizes and crude assessments of gait parameters (5-8,10). In contrast, the current study, which supports the findings of Judge et al. (9), measured gait patterns using sophisticated equipment in a large randomized sample, and may have revealed more of the potential role that exercise can play in improving functional mobility. Also in accord with Judge et al. (9), the effects of the intervention were heterogeneous in that those with initial slow walking speeds showed substantially improved walking speed at retest, while those with walking speeds that were normal to begin with showed little change. The finding that older women with gait speeds of less than approximately 1 m/sec (indicating the most impaired gait) benefited most from structured group exercise, has implications for targeting exercise interventions in future research trials and public health initiatives; it has been shown that those with walking speed less than 1 m/sec have difficulties undertaking functional activities such as crossing roads in urban environments (23). Our finding of higher gait velocity on retest in the exercisers is in conflict with those of Topp et al. (10), who ascribed their finding of slower velocity following exercise to a better control of the center of gravity. We feel that this explanation is unlikely, as longer strides and increased cadence which are more typical of younger persons are more likely indicators of confident walking patterns and control of the center of gravity. While shorter steps reduce friction demanded by the person to walk safely, shorter steps result in slower walking speeds which may increase the risk of a fall occurring once a slip has started (24). Further, the finding that there was no change in the stance duration as a percentage of stride time in the exercise subjects following the trial, indicates that walking speeds were increased, whereas stable gait patterns were maintained (25). Finally, in addition to demonstrating efficacy, exercise programs also have to maintain adequate compliance to be effective public health interventions. It is considered that the

M70 LORDETAL. high adherence (85% of subjects completing the program with an average attendance rate of 73%) was due, in large part, to the fact that the subjects enjoyed the group activities that provided the major element of the program. Thus, it appears that exercise interventions of this nature may offer an effective health promotion strategy for improving mobility in older persons. ACKNOWLEDGMENT Address correspondence and requests for reprints to Dr. Stephen R. Lord, Prince of Wales Medical Research Institute, High Street, Randwick, N.S.W. 2031, Australia. REFERENCES 1. Buchner DM, Beresford SA, Larson EB, La Croix AZ, Wagner EH. Effects of physical activity on health status in older adults II: Intervention studies. Annu Rev Publ Health 1992; 13:469-88. 2. Cunningham DA, Rechnitzer PA, Pearce ME, Donner AP. Determinants of self selected walking pace across ages 19 to 66. J Gerontol 1982:37:560-4. 3. lnnis FJ, Edholm OG. Studies of gait and mobility in the elderly. Age Ageing 1981; 10:147-56. 4. Larsson B, Renstrom P, Svardsudd K, et al. Health and ageing characteristics of highly physically active 65-year-old men. Eur Heart J 1984,5:31-5. 5. Meier A. Rehabilitation following falls of undetermined etiology: results of an intervention study. Schweiz Rundsch Med Prax 1992;81:I4O5-1O. 6. Sauvage LR, Myklebust BM, Crow Pan J, et al. A clinical trial of strengthening and aerobic exercise to improve gait and balance in elderly male nursing home residents. Am J Phys Med Rehabil 1992,71:333-42. 7. Fiatarone MA, Marks EC, Ryan ND, Meredith CN, Lipsitz LA, Evans WJ. High intensity strength training in nonagenarians: effects on skeletal muscle. JAMA 1990,263:3029-34. 8. Brown M, Holloszy JO. Effects of low intensity exercise program on selected physical performance characteristics of 60- to 71-year olds. Aging 1991 ;3:129-39. 9. Judge JO, Underwood M, Gennosa T. Exercise to improve gait velocity in older persons. Arch Phys Med Rehabil 1993;74:400-6. 10. Topp R, Mikesky A, Wigglesworth J, Holt W, Edwards JE. The effect of a 12-week dynamic resistance strength training program on gait velocity and balance in older adults. Gerontologist 1993;3:!:501-6. 11. Lord SR, Ward JA, Williams P, Anstey K. Physiological factors associated with falls in older community-dwelling women. J Am GeriatrSoc 1994;42:1110-17. 12. Lord SR, Ward JA, Williams P, Strudwick M. The effect of exercise on balance and related factors in older women: a randomized control trial. J Am Geriatr Soc 1995;43:1198-1206. 13. American College of Sports Medicine. Guidelines for graded exercise testing and exercise prescription, 2nd ed. Philadelphia: Lea and Febiger; 1980. 14. KISTLER Instruments AG, Winterthur, Switzerland. 15. Lloyd DG. Development and application of a gait analysis system. PhD thesis, University of New South Wales, 1992. 16. SPSS Inc. SPSS reference guide. Chicago: SPSS Inc; 1990. 17. Finley FR, Cody KA, Finizie RV. Locomotion patterns in elderly women. Arch Phys Med Rehabil 1969;50:140-6. 18. Dobbs RJ, Charlett A, Bowes SG, O'Neill CJA, Weller C, Hughes J, et al. Is this walk normal? Age Ageing 1993;22:27-30. 19. Murray MP, Kory RC, Clarkson BH. Walking patterns in healthy old men. J Gerontol 1969;24:169-78. 20. Winter DA, Patla AE, Frank JS, Walt SE. Biomechanical walking patterns changes in the fit and healthy elderly. Phys Ther 1990;70: 340-7. 21. Ferrandez AM, Pailhous J, Durup M. Slowness in elderly gait. Exp Aging Res 1990; 16:79-89. 22. Wolfson L, Whipple R, Amerman P, Tobin JN. Gait assessment in the elderly: a gait abnormality rating scale and its relation to falls. J Gerontol Med Sci 1990;45:M12-9. 23. Robinett CS, Vondran MA. Functional ambulation velocity and distance requirements in rural and urban communities. Phys Ther 1988:68;1371-3. 24. Lloyd DG. Environmental requirements for elderly people to have stability underfoot while walking. Proceedings of the National Forum on Prevention of Falls and Injuries amongst Older People. Sydney, Australia, 1990. 25. Winter DA. The biomechanics and motor control of human gait: normal, elderly and pathological. 2nd ed. Waterloo, Ontario: University of Waterloo Press; 1991. Received May 23, 1994 Accepted July 14, 1995