Fall and Balance Outcomes After an Intervention to Promote Leg Strength, Balance, and Walking in People With Diabetic Peripheral Neuropathy: “Feet First” Randomized Controlled Trial

Background

Weight-bearing exercise has been discouraged for people with diabetes mellitus and peripheral neuropathy (DM+PN). However, people with diabetes mellitus and insensate feet have an increased risk of falling. Lower-extremity exercise and balance training reduce fall risk in some older adults. It is unknown whether those with neuropathy experience similar benefits.

Objective

As part of a study of the effects of weight-bearing exercise on foot ulceration in people with DM+PN, the effects of a lower-extremity exercise and walking intervention on balance, lower-extremity strength (force-generating capacity), and fall incidence were determined.

Design

The study was an observer-masked, 12-month randomized controlled trial.

Setting

Part 1 of the intervention took place in physical therapy offices, and part 2 took place in the community.

Patients

The participants were 79 people who were mostly sedentary, who had DM+PN, and who were randomly assigned to either a control group (n=38) or an intervention group (n=41).

Intervention

Part 1 included leg strengthening and balance exercises and a graduated, self-monitored walking program; part 2 included motivational telephone calls. Both groups received regular foot care, foot care education, and 8 sessions with a physical therapist.

Measurements

The measurements collected were strength, balance, and participant-reported falls for the year after enrollment.

Results

There were no statistically significant differences between the groups for falls during follow-up. At 12 months, there was a small increase in the amount of time that participants in the intervention group could stand on 1 leg with their eyes closed. No other strength or balance measurements differed between the groups.

Limitations

The study was designed to detect differences in physical activity, not falls. The intensity of the intervention was insufficient to improve strength and balance in this population.

Conclusions

The training program had a minimal effect on participants' balance and lower-extremity strength. Increasing weight-bearing activity did not alter the rate of falling for participants in the intervention group relative to that for participants in the control group. People who are sedentary and who have DM+PN appear to be able to increase activity without increasing their rate of falling.

In patients with diabetes, regular participation in physical activity can improve glycemic control and reduce cardiovascular mortality.^1,2 Exercise that improves lower-extremity balance and strength (force-generating capacity) has been shown to be effective in reducing falls in older adults.^3–7 However, people who are sedentary and people with insensate feet⁸ have an increased risk of falling. Although 1 nonrandomized study of a small sample of older adults with peripheral neuropathy showed that lower-leg strengthening and balance exercises improved tandem and one-leg stance time and functional reach (clinical balance measures),⁹ we are aware of no prior randomized controlled trial of activity in people with peripheral neuropathy and in which fall incidence was an outcome.

Although there have been concerns that weight-bearing exercise could place people with diabetes mellitus and peripheral neuropathy (DM+PN) at higher risk of foot ulcers and amputation,^10,11 there also is evidence suggesting that daily weight-bearing activity may decrease the risk of foot ulceration.^12,13 In our recently completed randomized trial to increase weight-bearing activity in people with DM+PN, compared with participants in a control group, participants in the intervention group experienced an increase in weight-bearing, bout-related physical activity without an increase in foot ulcer rates.¹⁴ The change in total daily steps, as measured with an accelerometer, did not differ significantly between the study groups at either 6 or 12 months. In the control group, however, total steps decreased by 13% over 12 months (P<.01). Steps taken during 30-minute exercise bouts increased 14% between the baseline and 6 months in the intervention group but decreased 6% in the control group (P<.05); however, the difference did not remain statistically significant at 12 months. The number of days of participating in an exercise program (ie, walking) increased from 0 days per week in both groups at the baseline to 3 days per week in the intervention group and 1.5 days per week in the control group by 6 months (P<.05) but no longer differed between the groups at 12 months. Foot ulcer rates did not differ significantly between the groups.

Patients with diabetes can achieve better glycemic control and reduced cardiovascular mortality through regular physical activity. Increasing lower-extremity strength and balance also reduces falls in older adults. People with DM+PN, however, have an increased risk of falling,⁸ and exercise has been discouraged in this population because of concerns about foot ulcers and amputation. People most at risk of falling have been excluded from most exercise studies.¹⁵ The paucity of research in this population makes it difficult to predict the effects of exercise on falls. Thus, our intervention aimed to increase participants' lower-extremity strength and balance without increasing falls. We hypothesized that the physical activity intervention would not increase falls in participating adults, who were all at least 50 years of age and had DM+PN. Here we report on the effects of this lower-extremity exercise and walking intervention on balance, lower-extremity strength, and fall incidence.

Method

Design Overview

The “Feet First” trial was an observer-masked, randomized controlled trial of an individually adapted, behavior change physical activity intervention. Detailed trial methods, including CONSORT (Consolidated Standards of Reporting Trials) requirements, were previously described.¹⁴ In brief, 79 people with DM+PN were recruited over 18 months and randomly allocated to intervention (n=41) and control (n=38) groups. Physical activity and balance measurements were collected at baseline (before enrollment) and after 3, 6, and 12 months of participation. Falls were reported by participants throughout the study. In an intention-to-treat analysis, we compared the changes from the baseline to 6 and 12 months for participants in the 2 groups.

Setting and Participants

People in central Missouri who were at least 50 years old were invited to join the study. Eligible participants had a diagnosis of type 1 or 2 diabetes mellitus; did not take part in weight-bearing sessions of activity for 20 minutes or longer more than twice per week; had absent sensation on at least 1 point of 10 possible sites on each foot; and had loss of vibratory sensation, as measured with a biothesiometer (≥25 V=insensate).^16–18 We excluded people who lacked telephone access or had medical conditions that might contraindicate exercise.¹⁴ Of 260 potential participants, 105 declined participation; 76 of the remaining 155 potential participants did not meet inclusion criteria, leaving 79 participants for randomization.

Randomization and Interventions

After eligible participants provided written informed consent, baseline measurements were collected, and the participants were randomly allocated to intervention and control groups. Using randomization blocks of various sizes, we stratified participants by type of clinical site (eg, facilities affiliated with the Department of Veterans Affairs, academic family medicine or specialty clinics, nonacademic primary care facilities) to ensure balanced treatment of participants in the intervention and control groups.¹⁹ Allocation was concealed by use of opaque envelopes, which were opened by the study nurse at the time of the randomization visit. Research personnel who collected measurements or outcome data were unaware of participants' study group identities.

The intervention was delivered in 2 parts. During months 1 to 3 (part 1), participants in the intervention group attended 8 individual sessions with a physical therapist that focused on exercises to progressively strengthen legs and promote balance. These exercises ( Appendix) have been successful in improving balance in people with peripheral neuropathy⁹ and in reducing falls in older adults.^4,20–24 Participants were asked to perform progressively more difficult exercises during 3 additional weekly 1-hour sessions at home. The physical therapist and study nurse helped each participant develop a walking program tailored to his or her physical abilities, readiness to undertake exercise, activity preferences, and social or environmental constraints. Participants were encouraged to increase activity slowly, at minimum adding 100 steps to their daily activities every 2 weeks.^10,25 Participants used a waist-worn pedometer (Accusplit Eagle 170)* to monitor walking.

Participants in the control group also attended 8 visits with the therapist, at which they received diabetes self-care instruction. They were not given activity education, were not taught leg strengthening or balance exercises, and were not guided to undertake a walking program or use a pedometer.

During months 4 to 12 (part 2), we used motivational techniques to enhance exercise and fall-related self-efficacy via regular telephone calls from the study nurse.^26–28 The nurse called participants every other week (with calls lasting at least 10 minutes) to prompt them to continue their leg strengthening and balance exercises and follow their walking plan and to assist them in solving problems related to participation. Participants in the control group received calls at a similar frequency, during which they reported their recent activities but did not receive motivational assistance from the study nurse.

All participants received usual medical care from their own providers. Because poorly fitting shoes may predispose older people to falling,²⁹ we referred all participants to local orthotists or podiatrists to obtain reasonably priced therapeutic footwear at enrollment. The study did not pay for the footwear. Participants were asked to wear the footwear at the sessions with the physical therapist or when exercising or walking inside and outside their homes.

Outcomes and Follow-up

At the baseline examination, we collected demographic and health characteristics, including age, sex, marital status, education, ethnicity, smoking status, type and duration of diabetes mellitus, comorbid illnesses, and availability of health insurance. Body mass index was measured as (weight in kilograms)/(squared height in meters). Participants reported all falls that occurred in the year before enrollment. Variables related to those falls included the date on which each fall occurred, the participant's activity and footwear at the time of the fall, and whether medical care or hospitalization was required as a result of the fall. Adherence to study procedures was measured by defining “protocol completers” as those who attended at least half of the required study protocol elements (physical therapy sessions during months 1 to 3; timely attendance at the 3-, 6-, and 12-month visits; and weekly telephone calls for reporting foot lesion outcomes).

At the baseline assessment and at 3, 6, and 12 months, each participant used a water-resistant, computerized accelerometer (StepWatch)† to measure physical activity. Using the validated Summary of Diabetes Self-Care Activities Scale,³⁰ participants also reported the number of days per week on which they participated in any exercise program.

Balance measurements were collected at the baseline assessment and after 6 and 12 months of participation. The measures included the Berg Balance Scale, the Timed “Up & Go” Test, and the one-leg static stance test, each of which is associated with fall risk.^31,32 The Berg Balance Scale consists of 14 items, including standing from a seated position, reaching forward with arms extended while bending from the waist, picking up an object from the ground, turning to look over one's shoulder, turning 360 degrees, alternately tapping each toe on a low stool, and performing progressive balance challenges (standing with feet together, standing with feet in tandem, and then standing on one leg).³³ Lower scores indicate a higher risk of falling; scores above 45 are often categorized as indicating a low risk of falling.^31,34 In the Timed “Up & Go” Test, participants rise from a seated position, walk forward approximately 3 m (10 ft), turn around, walk back to their seat, turn around, and sit down, all without using their hands for assistance. The inability to complete the test within 30 seconds is associated with an increased risk of falling.³⁵ In the one-leg stance test, participants stand on 1 leg (with the other leg totally clearing the floor) without assistance. Richardson and colleagues^36,37 found that this test was highly associated with peripheral neuropathy and predicted fall risk in older adults. Vellas et al³⁸ found that people who could not maintain one-leg balance for at least 5 seconds were at risk for injurious falls.

We also collected measurements of motor strength at the baseline and after 6 and 12 months of participation. A single observer tested all participants. We used a handheld dynamometer (microFET2)‡ to test ankle dorsiflexion by using the “break” method.³⁹ This method measures the minimum effort required by an individual who is performing an isometric contraction to overcome a gradual buildup of resistance by the observer. Among older adults with a history of falls, the ankle dorsiflexor muscle shows greater decrements in strength than several other muscles.⁴⁰ The right ankle was always measured first. Although the test was conducted while participants were wearing shoes, it was mildly aversive, usually causing participants to offer less resistance with the second (left) foot. We present only values for the right ankle in the results.

Each participant also completed the Falls Efficacy Scale^41,42 and the Foot Function Index Disability Scale⁴³ at baseline, at 6 months, and at 12 months. The Falls Efficacy Scale measures respondents' confidence in accomplishing 10 activities without falling, with lower scores being associated with a greater fear of falling and lower self-efficacy or confidence. Our data correspond to the earliest version of the instrument,⁴¹ in which lower scores indicate higher confidence. More recent versions reverse the coding,^42,44 with scores of 75 or lower indicating low self-efficacy, scores of 76 to 99 indicating intermediate self-efficacy, and a score of 100 indicating high self-efficacy (no concerns about falling).^44,45 In our data, these ranges correspond to scores of 25 or higher, scores of 1 to 24, and a score of 0, respectively. The Foot Function Index Disability Scale is a foot-related disability scale that measures the effect of foot pathology on function in terms of disability and activity restriction. Higher scores indicate greater dysfunction.

Falls were reported continuously throughout the study. Participants were instructed to call a dedicated telephone “hot line” to report all falls or to call weekly to report no falls. We monitored all participants' use of the hot line and conducted follow-up calls within 1 week for those who did not call in a timely fashion. A fall was defined as an unintentional, uncontrolled descent to a lower level, except onto a bed or soft furniture, while standing or walking. Within 24 hours of a reported fall, study staff contacted the participants to obtain details about the fall (ie, their activities, footwear, and use of any assistive devices at the time of the fall). Participants also indicated whether they received medical care as a result of the fall and how the fall affected their activities of daily living for 3 days after the fall, specifically, whether they had difficulty walking or conducting instrumental activities of daily living (eg, cooking and cleaning) or were unable to perform these tasks.

Data Analysis

We used SAS for Windows (version 9.1)§ for all analyses. We conducted an intention-to-treat analysis to compare participants in the intervention group with those in the control group, regardless of their level of protocol adherence and study participation. Comparisons were 2-sided, and a P value of <.05 was considered significant. On the basis of a simulation study, the present study had 80% power to detect a difference between groups of approximately 1 fall in 18 months of participation, assuming a mean fall rate of 2.25 falls in 18 months for the control group.

We compared the baseline characteristics of the participants in the intervention and control groups by using chi-square analysis for categorical variables. For continuous variables with normal distributions, we used the t test, and for skewed distributions, we used the Wilcoxon rank sum test. We used the Mantel-Haenszel chi-square test to compare falls in the control and intervention groups in the year before enrollment by dividing them into 3 categories (0, 1, and 2 or more), so that no category contained fewer than 10% of observations. The proportions of protocol completers in the groups were compared by using chi-square analysis.

We used 3 methods to compare the fall incidence in the intervention group with that in the control group. First, we examined categories of fall counts by using the Mantel-Haenszel chi-square test. Falls were divided into 3 categories (0, 1, and 2 or more) so that no category contained fewer than 10% of observations. Second, to determine whether the number of falls in the intervention group differed from that in the control group, we performed a Poisson regression in PROC GENMOD, with the actual number of falls as the outcome variable and the natural logarithm of days in the study as an offset variable to standardize the counts over time. We used model coefficients and standard errors to calculate a mean fall rate and 95% confidence interval for each group. Finally, although we found no difference between study groups in the frequency of weekly reporting of falls (data not shown), we were concerned that participants who fell frequently might have become disinterested in reporting all of their falls throughout the study, resulting in the potential for reduced counts of later falls. We, therefore, performed a survival analysis by using time to first fall as the outcome, censoring participants who did not fall by their date of study exit.

We examined between- and within-group differences over time. Most continuous balance measures were highly skewed; therefore, we used nonparametric analysis for these outcomes. For dichotomous measures, we compared groups at both 6 and 12 months, with stratification on dichotomized baseline values, by using the Cochran-Mantel-Haenszel test of general association. For continuous measures, we compared row mean scores of groups at both 6 and 12 months, with stratification on grouped baseline values, by using the Cochran-Mantel-Haenszel method. Strata were defined as quintiles that were based on all baseline values (data were pooled to determine cutoff values for strata). We compared within-group changes from baseline to 6 and 12 months by using the signed rank test for the difference between values measured at 0 and 6 months and values measured at 0 and 12 months.

Role of the Funding Source

This study was funded by the Robert Wood Johnson Foundation, which approved the study design but did not direct the study team with regard to conduct, analysis, or interpretation of the study.

Results

There were 38 participants in the control group and 41 participants in the intervention group. We found no statistically significant differences between the groups at the baseline for demographic characteristics, health conditions, balance measures, or lower-extremity strength (Tab. 1). Follow-up time was variable because of difficulties in scheduling the final visits for participants. Of the 79 participants enrolled, 1 died of causes unrelated to the study and 2 withdrew early, 1 voluntarily and 1 because of cognitive decline. The remainder completed at least 11 months of participation. There was no statistically significant difference between the groups for time in the study (Tab. 2). Protocol adherence was fair, with 45% of participants in the intervention group and 35% of participants in the control group completing more than half of the required study protocol elements. Adherence did not differ between the groups (P=.38). The degree of protocol adherence did not affect the analytic results; therefore, we did not adjust for it.

Table 1

Baseline Participant Characteristics^a

Participant Characteristic	Control Group (n=38)	Intervention Group (n=41)	P
Demographic characteristics
Age, y, X̅ (SD)	64.8 (9.4)	66.3 (10.6)	.55
Women, no. (%)	20 (53)	20 (49)	.73
Health conditions
Body mass index, kg/m2, X̅ (SD)	37.3 (8.0)	36.0 (8.2)	.36
Joint pain in lower limbs, no. (%) of participants	27 (71)	30 (73)	.83
Type 2 diabetes, no. (%) of participants	35 (92)	39 (95)	.80
Fall-related measures
Falls Efficacy Scale score, X̅ (SD)a	8.3 (12.0)	10.4 (13.9)	.85
Falls in previous year, no. (%) of participants			.40
None	16 (42)	21 (51)
1	9 (24)	9 (22)
≥2	13 (34)	11 (27)
Strength and balance measures
Berg Balance Scale score, X̅ (SD)a	49.0 (5.3)	48.1 (6.8)	.80
Foot Function Index Disability Scale score, X̅ (SD)a	25.7 (18.9)	25.3 (20.0)	.85
Right-ankle dynamometry, kg, X̅ (SD)	27.4 (6.9)	26.1 (10.2)	.49
One-leg stance test with eyes open, s, X̅ (SD)	9.54 (15.9)	10.1 (13.0)	.29
≥5 s, no. (%) of participants	14 (36.8)	20 (48.8)	.28
One-leg stance test with eyes closed, s, X̅ (SD)	1.0 (1.6)	1.5 (2.6)	.63
≥2 s, no. (%) of participants	10 (26.3)	10 (24.4)	.84
Timed “Up & Go” Test, s, X̅ (SD)	12.3 (3.5)	12.8 (4.7)	.83

^a

Scores on the Falls Efficacy Scale range from 0 to 100, with 0 representing the highest self-efficacy. Scores on the Berg Balance Scale range from 0 to 56, with 56 representing the highest functioning. Scores on the Foot Function Index Disability Scale range from 0 to 100, with 100 representing the greatest difficulty.

Table 2

Time in Study and Fall Incidence by Group^a

Measure	Control Group (n=38)	Intervention Group (n=41)	P
Days in study			.53b
Median	394	408
Range	168–575	11–511
Falls during follow-up, no. (%) of participants			.97c
None	22 (58)	25 (61)
1	9 (24)	9 (22)
≥2	7 (18)	7 (17)
Falls/1,000 person-days of follow-up, X̅ (95% CI)	2.02 (1.42–2.88)	2.06 (1.46–2.89)	.95d
Days to first fall during follow-up, X̅ (95% CI)	226 (192–260)	251 (208–295)	.98e

^a

CI=confidence interval.

^b

Independent t test.

^c

Mantel-Haenszel chi-square test.

^d

Poisson regression.

^e

Kaplan-Meier survival analysis.

For most strength and balance measures, there were no statistically significant differences between the groups at either 6 or 12 months, after controlling for baseline values (Tab. 3). For the most difficult test, the one-leg stance test with eyes closed, there was a difference between groups. At baseline, participants in the control group could perform this test for a mean duration of 1.0 second; the mean time for participants in the intervention group was 1.5 seconds (P=.63). At 6 months, the time increased for both groups, but the difference was not significant. At 12 months, the mean time for participants in the control group returned to the baseline (1.0 second), but the mean time for participants in the intervention group was 1.9 seconds (P=.05). The proportion of participants who could perform this test for at least 2 seconds showed a similar pattern, with 20.6% of participants in the control group and 41.2% of participants in the intervention group able to perform this task at 12 months (P=.03). Lower-extremity strength declined in both groups over time (P<.05 for both 6 and 12 months compared with the baseline in both groups).

Table 3

Comparison of Strength, Balance, Efficacy, and Functional Measures Over Time^a

Measureb	Control Group (n=38)	Intervention Group (n=41)	Pc
Right-ankle dynamometry, kg
Baseline (0, 0)	27.4 (25.2–29.7)	26.1 (22.9–29.3)
6 mo (2, 4)	23.8 (21.2–26.4)d	24.3 (21.2–27.5)d	.11
12 mo (4, 7)	20.4 (16.4–24.5)d	22.0 (17.9–26.0)d	.22
Berg Balance Scale score
Baseline (0, 0)	49.1 (47.3–50.8)	48.1 (46.0–50.3)
6 mo (2, 4)	49.9 (48.0–51.8)	48.1 (45.1–51.1)	.91
12 mo (4, 6)	47.9 (45.5–50.4)	47.1 (43.4–50.8)	.80
Falls Efficacy Scale score
Baseline (0, 0)	8.3 (4.4–12.2)	10.4 (6.0–14.8)
6 mo (3, 3)	15.5 (8.3–22.7)d	10.2 (5.2–15.3)	.07
12 mo (4, 5)	10.9 (6.8–15.1)	13.0 (7.1–18.9)	.73
Foot Function Index Disability Scale score
Baseline (0, 0)	25.7 (19.5–31.9)	25.3 (19.0–31.6)
6 mo (3, 3)	24.4 (17.8–31.1)d	25.3 (18.3–32.3)	.68
12 mo (4, 5)	24.4 (17.8–31.0)	25.5 (18.1–32.9)	.91
One-leg stance test with eyes open, s
Baseline (0, 0)	9.5 (4.3–14.7)	10.1 (6.0–14.2)
6 mo (2, 4)	11.6 (6.0–17.1)	15.7 (8.7–22.7)	.89
12 mo (4, 6)	10.8 (5.4–16.2)	14.6 (8.4–20.7)	.97
One-leg stance test with eyes open, ≥5 s, no. (%) of participants
Baseline (0, 0)	14 (36.8)	20 (48.8)
6 mo (2, 4)	18 (50.0)	18 (48.6)	.35
12 mo (4, 6)	13 (38.2)	21 (60.0)	.12
One-leg stance test with eyes closed, s
Baseline (0, 0)	1.0 (0.5–1.5)	1.5 (0.6–2.3)
6 mo (2, 4)	1.7 (1.1–2.3)d	2.8 (1.1–4.4)d	.67
12 mo (4, 7)	1.0 (0.4–1.5)	1.9 (1.1–2.8)	.046
One-leg stance test with eyes closed, ≥2 s, no. (%) of participants
Baseline (0, 0)	10 (26.3)	10 (24.4)
6 mo (2, 4)	13 (36.1)	16 (43.2)	.43
12 mo (4, 7)	7 (20.6)	14 (41.2)	.03
Timed “Up & Go” Test, s
Baseline (0, 0)	12.3 (11.2–13.5)	12.8 (11.4–14.3)
6 mo (2, 4)	12.2 (11.3–13.2)	12.8 (11.4–14.3)	.71
12 mo (4, 7)	13.2 (10.1–16.3)	13.8 (10.6–17.1)	.54

^a

Values are reported as the mean (95% confidence interval of the mean) unless otherwise indicated.

^b

Numbers of participants (control group, intervention group) shown in parentheses.

^c

For continuous measures, between-group comparison over time at 6 or 12 months, controlling for grouped baseline value; for proportions, Cochran- Mantel-Haenszel test for between-group comparison, 6 or 12 months compared with baseline value.

^d

Signed rank test for within-group comparisons of the differences between 6- or 12-month values and baseline values (P<.05).

More than half (51%) of the participants in the intervention group reported they had not fallen in the year before the study; 9 participants (22%) had fallen once, and 11 participants (27%) had fallen 2 or more times (Tab. 1). Compared with participants in the intervention group, fewer participants in the control group reported no falls in the previous year (n=16, 42%); 9 participants (24%) had fallen once, and 13 participants (34%) had fallen 2 or more times. The difference between the groups was not statistically significant (P=.40).

During follow-up, 33 falls occurred in participants in the intervention group and 31 falls occurred in participants in the control group (Tab. 2). More than half of the participants in each group reported no falls during follow-up; 9 participants in each group fell once, and 7 participants in each group fell 2 or more times (P=.97). Using Poisson regression to account for all falls and time in the study, we calculated that participants in the control group experienced 2.02 falls per 1,000 person-days; for participants in the intervention group, this value was 2.06 (P=.95). In the survival analysis, the mean times to the first fall were 251 days for participants in the intervention group and 226 days for participants in the control group (P=.98).

Although participants in the intervention and control groups did not differ with respect to falls either before or after study inception, participants in both groups reported fewer falls during the study than in the year before the study. When we omitted one extreme prestudy value that we believed to be inaccurate (34 falls), the mean rate for participants in the control group decreased from 3.89 to 2.02 falls per 1,000 person-days and the value for participants in the intervention group decreased from 3.08 to 2.06 (P<.001). The interaction of time period with group was not significant (P=.82), indicating that the effect did not differ by group. There was no difference between groups at any time point for either the Falls Efficacy Scale or the Foot Function Index Disability Scale, suggesting no intervention effect on either fall-related self-efficacy or foot-related disability.

Discussion

In this randomized controlled trial, we found that our exercise intervention did not increase fall rates despite increasing bout-related physical activity. At the same time, the intervention was not helpful because no reduction in falls or improvement in balance or strength was apparent. Consistent with our hypothesis, we found that fall rates did not differ significantly between the groups at any point in the study. This finding suggests that a community-based intervention promoting walking at an intensity similar to that achieved by participants in the “Feet First” trial is not harmful to people with DM+PN, even when an increase in balance is only modest.

With the exception of increased time on the one-leg stance test with eyes closed, the intervention did not improve lower-extremity strength or balance. Vellas et al³⁸ found that one-leg stance times (eyes open) of less than 5 seconds were associated with injurious falls in older adults. Because almost no participants were able to stand on one leg with their eyes closed for 5 seconds, we used a lower cutoff of 2 seconds for this measure. Twice as many participants in the intervention group as in the control group were able to complete this task for 2 seconds or more at 12 months. Although the improvement in the mean time differed by only 1 second at 12 months, with a standard deviation of 2.0, the effect size appears to be moderate (0.5). To our knowledge, the one-leg stance test with eyes closed is a new measure without established norms. This and other measures merit further investigation. Traditional measures of strength and balance may not be sufficiently sensitive to detect changes in people with insensate feet.

The amount of walking increased most notably between baseline and 6 months, after participants had completed part 1 of the intervention (physical therapist sessions with leg strengthening and balance exercises).¹⁴ Generally, improvements in physical activity regressed toward the baseline during the second 6 months of study participation. Participants in both groups exhibited decreases in strength, as measured by ankle dynamometry. This finding could reflect a return to more sedentary activity patterns at the end of the study and the fact that ankle dynamometry was mildly aversive, causing participants to offer less resistance in later measurements. Two studies with a similar intervention in people with DM+PN reported improvements in lower-extremity strength and balance.^9,46 It is possible that our intervention was not intense enough to achieve more extensive improvements in balance and lower-extremity strength, particularly in our sedentary population.

Deficits in balance and strength have been identified as risk factors for falls in older adults, suggesting that exercise interventions that aim to reduce these deficits also could help reduce falls.⁴⁷ Guidelines for preventing falls in older adults recommend exercise programs, particularly exercises to improve balance.^3,6,7 Some randomized trials of exercise and balance retraining in older adults achieved reductions in fall rates,^5,23,48–50 whereas others had no effect on falls.^15,51 No randomized trials of this nature have been carried out in people with DM+PN.

A collection of clinical trials (FICSIT) to evaluate the efficacy and feasibility of a variety of interventions for reducing falls, frailty, or both in older participants included an exercise component in 7 of 8 trials.⁴⁷ The interventions, which were provided to residents in both the community and nursing homes, were 10 to 36 weeks in duration, with follow-up at 2 to 4 years. Two of the trials produced statistically significant decreases in falls, 2 had borderline significance, and 3 produced increases in falls, although none of these were statistically significant. Pooled incidence ratio estimates indicated that balance training and overall exercise significantly reduced falls (incidence ratios of 0.90 and 0.83, respectively). The exercise components were variable, and for some of the interventions, the effects of exercise could not be separated from the effects of other intervention components.

Campbell et al²³ randomized women who dwelled in the community and who were 80 years of age and older to an exercise and balance regimen very similar to that used in the present study, also with telephone calls to maintain motivation. At 1 year, participants in the intervention group showed improvements on balance and chair stand measures and had significantly fewer falls (hazard ratio=0.68, 95% confidence interval=0.52–0.90) than participants in the control group; the reduction in falls was sustained through 2 years in participants who chose to continue beyond year 1.⁴ Although the physical activity scores were quite low (a mean physical activity score of 51.5/400), this population differed from participants in our sample, all of whom had DM+PN.

In a study of 20 patients with DM+PN, Richardson et al⁹ found that a 3-week intervention to increase strength and balance resulted in improvements in time on the one-leg stance test with eyes open, tandem stance time, and functional reach. Mean times on the one-leg stance test were similar to our results. Richardson et al⁹ did not monitor falls. In the present study, we randomized participants, monitored them serially over 1 year, and monitored both balance and falls. Our intervention produced modest increases in physical activity without increasing either foot ulcers or falls. It was not beneficial, however, in terms of better glycemic control, reduced falls, or improved strength and balance.

The present study had a number of potential limitations. First, the study was primarily designed to detect differences in physical activity between groups rather than differences in fall rates. We designed the study this way realizing that any inferences regarding the effects of physical activity on falls would be dependent on the change in weight-bearing physical activity. The fall rates that we observed in the 2 groups for the full 12-month study were essentially identical, indicating no important differences between groups. On the basis of the Berg Balance Scale³³ and Timed “Up & Go” Test results,³⁵ our participants had a low risk of falling at baseline, perhaps making it difficult for the intervention to have a dramatic effect. However, approximately half of our participants were unable to stand on 1 leg with their eyes open for 5 seconds, suggesting an elevated risk of injurious falls.³⁸ Additional studies of patients with DM+PN are needed to determine reasonable cutoff values for elevated risks of falls and injuries. Subsequent studies should be designed to detect differences in important but rarer outcomes, such as injurious falls.

Second, the accelerometer measured only minutes of activity during stepping. Time spent standing immobile was not recorded; however, we did not observe any significant differences in time spent inactive between the groups. Third, participants in the control group did not receive motivational calls from the study nurse and may not have been as engaged in the study as participants in the intervention group. This situation could have led to reduced reporting of falls by participants in the control group; however, we did not find any significant differences in protocol adherence between the groups. Fourth, the method that we used to detect falls (a hot line) may have actually made it more difficult for participants to report falls, especially if they experienced multiple falls each day or week. Other studies have used “fall calendars,” on which participants record their total number of falls each day²¹; this method is preferable for future intervention studies monitoring falls. We found no differences in patterns of reporting falls between the intervention and control groups; therefore, any reductions in reporting that occurred because of the use of the hot line to report falls were similar in the study groups. Finally, we have no demographic information on people who declined to participate; therefore, we cannot assess generalizability.

In summary, participants in the intervention group achieved a modest increase in activity with no increase in falls compared with those in the control group. Lower-extremity strength and balance were unimproved. Although interventions that included leg strengthening and balance exercises to promote ambulatory physical activity (ie, walking) in people with DM+PN did not decrease fall rates, neither did they increase them. This finding suggests that physical activity interventions that increase activity at the level achieved by this cohort probably do not increase the risk of falling in people with this diabetic complication. To further increase physical activity and protocol adherence, we recommend a supervised, center-based exercise program rather than a self-administered, home-based program and more in-person contact between participants and study staff. Allet et al⁴⁶ achieved a median session attendance of 87% for participants with DM+PN, and in the study of Rubenstein et al,¹⁵ older adults attended 84% of group exercise sessions; these values are far better than what we achieved with our home-based program. Additional research is under way to explore a more intensive intervention for people with DM+PN and to further test current exercise guidelines (NIH R21HD058938, Dr Michael J. Mueller, principal investigator). We recommend close supervision of people with DM+PN as they attempt to increase their weight-bearing activity and make changes in this important self-management behavior.

The Bottom Line

What do we already know about this topic?

People with diabetes and insensate feet are at higher risk for falls. Lower-extremity exercise and balance training can reduce fall risk; however, weight-bearing exercise has been discouraged in this population due to concerns about foot ulcers and amputation.

What new information does this study offer?

This randomized trial of leg-strengthening exercises, balance exercises, and walking in 79 sedentary adults with diabetes and peripheral neuropathy resulted in modest gains in weight-bearing activity. Falls and lower-extremity strength were not affected by the intervention, and there was a small improvement in one balance measure.

If you're a patient, what might these findings mean for you?

Sedentary people with diabetes and insensate feet can increase their activity without increasing their rate of falling. Studies examining the effect of more intensive activity on the incidence of foot breakdown and fall risk in people with diabetes and peripheral neuropathy are needed.

This study was approved by the institutional review boards at the University of Missouri Health Sciences Center Institutional Review Board, the University of Washington Institutional Review Board, and the Harry S Truman Research and Development Committee at the Harry S Truman Memorial Veterans' Hospital.

This study was funded by the Robert Wood Johnson Foundation.

Data from this study were presented at the 37th North American Primary Care Research Group Annual Meeting; November 14-18, 2009; West Montreal, Quebec, Canada.

ClinicalTrials.gov registration number: NCT00286598.

*

Accusplit, 3900 Independence Dr, Suite 150, Livermore, CA 94551.

^†

OrthoCare Innovations, 700 12th St NW, Suite 700, Washington, DC 20005.

^‡

Hoggan Health Industries Inc, 8020 South 1300 West, West Jordan, UT 84088.

^§

SAS Institute Inc, 100 SAS Campus Dr, Cary, NC 27513-2414.

Appendix

Intervention Program Part 1: Leg Strengthening and Balance Exercises^a

Level 1	Level 2	Level 3	Level 4
1. One-leg stand, with both hands21	1. One-leg stand, with one hand21	1. One-leg stand, with no hands21	1. Standing arm/leg march21
2. Hip circle (30° from body)21	2. Walk on exercise mat with arms extended21	2. Walk on exercise mat with arms folded21	2. Crossover walk21
3. Arm circles (30° from body)21	3. Step sideways, with both hands (10 steps, 4 times)21,22	3. Step sideways, with one hand (10 steps, 4 times)	3. Tandem walk, with no hands (10 steps, 4 times)
4. Knee lifts while seated, arms to side21	4. March while seated21	4. Standing, cross legs at ankles21	4. Heel-toe walk, with no hands (10 steps, 4 times)22
5. While sitting on an inflatable exercise ball, catch a beach ball20	5. Knee lifts while seated, arms across chest21	5. Standing leg lift (bend leg, lift to horizontal)21	5. Step sideways, with no hands (10 steps, 4 times)22
6. Toe raises, both feet (set of 10 raises, 3 sets), with one hand, if necessary9	6. Heel stand (lift the toe to balance on the heel), with one hand, on both feet22	6. Heel stand9,22	6. Walk backward, with no hands22
	7. Walk in a figure-of-8, twice, with one hand22	7. Tandem walk, with one hand (10 steps, 4 times)22	7. Toe-tap (as per Berg Balance Scale, item #12), with no hands
	8. Walk backward, with one hand	8. Walk in a figure 8, twice, with no hands22	8. Ankle inversion and eversion, with one hand always, on one foot (1 set of 10)9
	9. Ankle inversion and eversion (shift center of mass laterally), with both hands, both feet (set of 10, 2 sets)9	9. Step over irregular objects	9. Balance on one foot (10 s, 3 tries per session)9
		10. Toe raise, on one foot, with one hand, if necessary (set of 10, 2 sets)

^a

The number of repetitions represents the goal for each exercise at the indicated level. Participants started with 1 set and progressed no faster than 1 level per week for any given exercise. “Both hands” means holding a stable object with both hands (eg, a wall or an assistant’s hand) while performing the maneuver; “one hand” means holding on with only one hand. “Both feet” means performing the maneuver while standing on both feet; “one foot” means performing the maneuver while standing on one foot. All participants practiced the “both-hands” version of each exercise before attempting the maneuver without holding a steady object. Participants practiced all maneuvers initially under the supervision of the physical therapist before practicing the maneuvers at home. The maneuvers were drawn from the referenced intervention studies. Reprinted with permission of the American Physical Therapy Association from: LeMaster JW, Mueller MJ, Reiber GE, et al. Effect of weight-bearing activity on foot ulcer incidence in people with diabetic peripheral neuropathy: Feet First randomized controlled trial. Phys Ther. 2008;88:1385–1398. This material is copyrighted, and any further reproduction or distribution requires written permission from APTA.

Author notes

All authors provided concept/idea/research design and data analysis. Dr Kruse and Dr LeMaster provided writing. Dr LeMaster provided data collection, project management, fund procurement, patients, facilities/equipment, and institutional liaisons. Dr Madsen provided consultation (including review of manuscript before submission).