The Relationship Between Leg Power and Physical Performance in Mobility-Limited Older People

Authors

  • Jonathan F. Bean Md MS,

    1. Department of Physical Medicine and Rehabilitation, Harvard Medical School, Spaulding Rehabilitation Hospital, Boston, Massachusetts;
    2. Research and Training Institute, Hebrew Rehabilitation Center for the Aged, Boston, Massachusetts; and
    3. Department of Health Sciences, Sargent College of Health and Rehabilitation Sciences, Boston University, Boston, Massachusetts.
    Search for more papers by this author
  • Dan K. Kiely MPH, MA,

    1. Research and Training Institute, Hebrew Rehabilitation Center for the Aged, Boston, Massachusetts; and
    Search for more papers by this author
  • Seth Herman BA,

    1. Research and Training Institute, Hebrew Rehabilitation Center for the Aged, Boston, Massachusetts; and
    Search for more papers by this author
  • Suzanne G. Leveille PhD,

    1. Research and Training Institute, Hebrew Rehabilitation Center for the Aged, Boston, Massachusetts; and
    Search for more papers by this author
  • Kelly Mizer BS,

    1. Department of Health Sciences, Sargent College of Health and Rehabilitation Sciences, Boston University, Boston, Massachusetts.
    Search for more papers by this author
  • Walter R. Frontera MD, PhD,

    1. Department of Physical Medicine and Rehabilitation, Harvard Medical School, Spaulding Rehabilitation Hospital, Boston, Massachusetts;
    Search for more papers by this author
  • Roger A. Fielding PhD

    1. Department of Health Sciences, Sargent College of Health and Rehabilitation Sciences, Boston University, Boston, Massachusetts.
    Search for more papers by this author

Address correspondence to Jonathan F. Bean, MD, MS, Hebrew Rehabilitation Center for Aged, 1200 Centre St., Boston, MA 02131. E-mail: bean@mail.hrca.harvard.edu

Abstract

OBJECTIVES:

The purpose of this study was to assess the influence of leg power and leg strength on the physical performance of community-dwelling mobility-limited older people.

DESIGN:

Cross-sectional analysis of baseline data from a 12-week randomized controlled exercise-intervention study.

SETTING:

Exercise laboratory within the Department of Health Science of an urban university.

PARTICIPANTS:

Forty-five community-dwelling mobility-limited people (34 women, 11 men), aged 65 to 83.

MEASUREMENTS:

Health status, depression, cognition, physical activity, and falls efficacy; physiological measures of lower extremity strength and power; and measures of physical performance.

RESULTS:

Through bivariate analyses, leg power was significantly associated with physical performance as measured by stair-climb time, chair-stand time, tandem gait, habitual gait, maximal gait, and the short physical performance battery describing between 12% and 45% of the variance (R2). Although leg power and leg strength were greatly correlated (r = .89) in a comparison of bivariate analyses of strength or power with physical performance, leg power modeled up to 8% more of the variance for five of six physical performance measures. Despite limitations in sample size, it appeared that, through quadratic modeling, the influence of leg power on physical performance was curvilinear. Using separate multivariate analyses, partial R2 values for leg power and leg strength were compared, demonstrating that leg power accounted for 2% to 8% more of the variance with all measures of physical performance.

CONCLUSION:

Leg power is an important factor influencing the physical performance of mobility-limited older people. Although related to strength, it is a separate attribute that may exert a greater influence on physical performance. These findings have important implications for clinicians practicing geriatric rehabilitation. J Am Geriatr Soc 50:461–467, 2002.

The “disablement pathway,” as originally described by Nagi and adapted by Jette, progresses from impairments to functional limitations to disability.1,2 Functional limitations, as manifested by measures of lower extremity physical performance, have been shown to be predictive of disability, the most distal disablement outcome.3,4 As a result, many investigators have theorized that impairments in strength serve as a more proximal determinant of physical performance.5–7 It has been demonstrated that strength impairments are related to function as measured by a variety of physical performance measures, in a curvilinear pattern.6,8,9

More recently, it has been theorized that impairments in muscle power are more influential than strength on function.10 Muscle power is a related but different attribute than strength and is defined as the ability to perform muscular work per unit of time (power = work/time). In simpler terms, if strength is defined as the ability to exert force, power is defined as the ability to exert force quickly (power = force × velocity). Muscle power has been reported to be associated with nursing home resident's physical performance. Bassey et al. found that leg power, measured by using the Nottingham Leg Rig, was correlated with gait speed, chair-rise time, and stair-climb time.11 More recently, in community-dwelling older women, leg power was shown to be predictive of self-reported disability.12 The influence of muscle power on the physical performance of community-dwelling older people has yet to be demonstrated. Furthermore, although power has been theorized as potentially more influential on function than strength, the relative contribution of these attributes has yet to be elucidated.

Thus, the purpose of this study is to understand the influence of impairments in muscle power on the function (physical performance) of mobility-limited, community-dwelling older people. We hypothesize that muscle power is a predictor of function, as defined by physical performance, and that its influence on function is greater than that of muscle strength.

METHODS

Study Design

This was a cross-sectional analysis performed on data collected as part of a randomized single-blind trial evaluating the effect of two 12-week exercise programs on strength, power, and physical performance in mobility-limited older people. After acceptance into the study and the completion of baseline testing, subjects were randomly assigned to a standardized walking program or a stair-climbing exercise program. This study analyzed baseline data.

Study Population

Recruitment of subjects (N = 45) was conducted in the greater Boston metropolitan area and facilitated through the Harvard Cooperative Program on Aging and through advertising in local newspapers.13 At the initial screening visit, consent was obtained, a screening physical performance test was conducted, and a comprehensive history and physical examination was performed. Inclusion criteria were: age 65 and older, mild to moderate mobility limitations as defined by a score of 11 or lower (out of 12) on the Short Physical Performance Battery,14 and the ability to climb a flight of stairs independently without the use of an assistive device, although the unilateral use of the handrail was allowed. Exclusion criteria included unstable acute or chronic disease, a score of less than 23 on the Folstein Mini-Mental State Examination,15 or a neuromusculoskeletal impairment interfering with independent stair climbing. Subjects who met these criteria were invited back for a second visit that included a screening submaximal exercise tolerance test performed and conducted according to American Heart Association guidelines. If eligible, subjects returned to the laboratory for a subsequent visit to complete baseline testing. All subjects lived independently in the community.

One hundred eighty-six inquiries were solicited via advertising in local papers and newsletters and direct mailings. After screening these individuals via telephone and eliminating subjects who were not eligible or were unable to commit to the study, 57 potential subjects were invited to participate in a screening assessment. Of these, four were excluded for medical reasons and eight chose not to commit to the study. Therefore, 45 subjects (34 women, 11 men) were eligible and randomized, representing 79% of the potential subjects.

Physical Performance Measures (Dependent Variables)

Tandem Gait

Forward and backward tandem walking was timed using a stopwatch to the nearest 0.01 second over a 20-foot course. Subjects were asked to walk heel to toe through the course's entirety as quickly and as safely as possible. The mean of two trials for each direction was recorded. Tandem gait was determined by summing the forward and backward tandem walk times.

Stair Time

A standard 10-stair flight, with handrails on each side, was used for this test. The subjects were instructed to ascend the stairs as quickly as possible, using the handrail if necessary. The stopwatch was stopped when both feet were planted at the top of the 10th step. Time was recorded to the nearest 0.01 second, and the average of two trials was taken.

Chair-stand Time

A chair with arms and a seat 0.43 meters from the ground was placed against a wall for support and safety purposes. The subject was instructed to stand up and sit back down as fast as they could for 10 repetitions with their arms crossed at their chest. One trial was conducted, with time recorded using a stopwatch to the nearest 0.01 second. Although the five-repetition chair stand is a component of the Short Physical Performance Battery (SPPB) described below, we chose to use the 10-repetition chair-stand test individually to better discriminate differences in performance between subjects and to eliminate potential ceiling effects.

Habitual and Maximal Gait Velocities

Gait velocities were measured to the nearest 0.01 second as the mean of two trials using an ultrasonic gait speed monitor (Ultratimer, DCPB Electronics, Glasgow, Scotland). Participants were instructed to walk at their normal, comfortable pace to determine their habitual gait velocity and then told to walk as fast as they could to determine their maximal gait velocity. Velocity was recorded after subjects walked an initial 2 meters to control for individual differences in acceleration.

Short Physical Performance Battery

The SPPB is a well-established, reliable, and valid measure of lower extremity performance.3,4,14,16 Testing was performed as previously described and involves an assessment of standing balance, a timed 2.4-meter walk, and a timed test of 5 repetitions of rising from a chair and sitting down.14 All timing was measured to the nearest 0.1 seconds using a stopwatch. Each of the three tests is scored based on performance between 0 and 4, with a maximum score of 12. Lower scores on the SPPB have been found to predict disability over 1 to 6 years in several older populations.4,16

Physiological Measures: Leg Power and Strength Measurements (Independent Variables)

Maximal dynamic strength of the lower extremities was assessed by one repetition maximum (1RM) measures of bilateral leg press (Newtons) and individual left and right knee extensors (KE) (Newton-meters). The 1RM measurements were conducted using seated KE and recumbent leg press (LP) pneumatic resistance machines customized with software and digital displays (Keiser Sports Health Equipment Inc., Fresno, CA). An ultrasonic system measuring position, and therefore relative motion, aided examiners in establishing a subject's full range of motion (ROM) by observing the excursion of a lighted bar on the output screen during performance of the measure with minimal resistance. Subjects performed the concentric phase, maintained full extension, and performed the eccentric phase of each repetition over 2, 1, and 2 seconds, respectively. The examiner progressively increased the resistance for each repetition until the subject could no longer move the lever arm one time through the full ROM. An adapted version of the Borg Scale was used to evaluate the subject's perceived effort for each repetition and to assist the examiner in tailoring increments in resistance to achieve the 1RM in approximately 8 to 10 repetitions.

After measurement of the 1RM, assessment of bilateral LP and individual left and right KE peak muscle power was performed using the same pneumatic resistance machines used for 1RM testing. Performance of power tests using the pneumatic resistance machines as summarized below has previously been described and validated in our laboratory.12,17 Peak power (watts) was computed by the performance of one repetition at 8 relative intensities (40%, 50%, 60%, 70%, 75%, 80%, 85%, and 90%) of the 1RM. For a given muscle action, maximal power varies in a curvilinear fashion by percentage of the 1RM. Generally, lower maximal power values are found at the lowest and highest percentages of the 1RM, with peak power occurring between 60% and 85% of the 1RM.12,17 Beginning with 40%, subjects performed the lift at each established percentage of their 1RM as fast as possible through the full ROM. At each force setting, subjects performed one maximal effort with a 30- to 45-second rest between repetitions. The software engineered for the testing equipment calculated work and power during the concentric phase of each repetition by sampling the system pressure (force) and position 400 times per second. Data collected between the start and stop positions of the concentric phase of the repetition were used to compute work (joules). To eliminate noisy data generated at the start and end of the motion, data collected between 5% and 95% the concentric phase were used to calculate power (watts). After each repetition, work and power data were stored and then displayed on the output screen. The highest power achieved, regardless of percentage of the 1RM at which it was achieved, was recorded as the peak power. For the majority of subjects, this occurred at 70% of the 1RM.

Composite Peak Power and Strength Measure

To create a composite measure for peak leg power and leg strength, we summed the values of the LP and right and left KE for peak power and strength, respectively. This composite value was created to better represent the diverse lower extremity muscle actions that constitute the physical performance tests utilized in the study.

Covariates (Adjustment Variables)

Body Mass Index

Body mass index (BMI) was calculated from measurements of height and weight obtained during the screening physical examination.

Depression

The Iowa Short Form of the Center for Epidemiological Studies—Depression scale was conducted through an interview and used as an index of depressive symptomology.18 Scores of 16 or higher have been associated with increased risk of clinical depression.

Health Status

All medical diagnoses were obtained via a questionnaire and subsequent interview of each subject during the history and physical examination that the principal investigator conducted. Background medical information that the subject's primary care physicians provided to the investigators confirmed this information. The principal investigator was solely responsible for tabulating and coding all medical diagnosis and standing medications.

Physical Activity

Leisure, household, and occupational activity levels were estimated using the Physical Activity Scale for the Elderly questionnaire, a brief, reliable, and valid instrument assessing physical activity in older adults.19

Falls Efficacy

The Falls Efficacy Scale is a reliable and valid 10-question assessment of fear of falling.20

Statistical Analysis

We initially calculated descriptive statistics for subject characteristics. The relationships between the dependent (physical performance) and independent variables (power and strength) were evaluated using regression methods, beginning with unadjusted models and proceeding to multivariate adjusted models. Both sets of bivariate analyses were repeated, using weight as a covariate to control for any influence body size might have had on the results. The simple correlation between leg power and leg strength was determined.

Based on clinical experience and a literature review, we included established covariates in our model to adjust for the potentially confounding effects of these factors on the relationship between muscle power (or strength) and physical performance. Using a stepwise regression, all variables that were significantly associated with performance outcomes (alpha = .15) were retained in an additional multiple regression model analysis for each physical performance measure. In an additional step, factors from the additional multivariate model that were not significant at the alpha = .05 level were dropped from the final analysis, leaving a common pool of covariates to be used in the final multiple regression analyses. These same procedures were repeated for leg strength. For any of the dependent variables in which a normal distribution was questionable, log transformation was performed in an effort to normalize its distribution. To compare the influence of leg power and strength on their associations with physical performance, we derived partial variance (R2) for each predictor (power and strength) in the fully adjusted models.

Lastly, to determine the best fitting model of the relationship between impairment (power or strength) and function (physical performance), we included a quadratic term in the unadjusted models. Consistent with prior studies, comparisons were made between the linear and quadratic (curvilinear) models for each dependent variable.6,9

RESULTS

Baseline Characteristics

Summary baseline characteristics and physiological and physical performance measures are presented in Table 1. The mean age ± standard deviation (SD) of the subjects was 72.7 ± 4.6 (range 65–83, with 76% being women and 89% Caucasian. Their health status was characterized by a mean of 2.1 ± 1.5 chronic conditions (range 0–5) and 2.6 ± 2.0 daily medications (range 0–8). On average, participants were college educated and had minimal depressive symptoms. Their level of physical activity was quite variable but consistent with established means for the age groups examined.19 In general, they had a strong sense of self-efficacy with regard to falls (mean score ± SD on falls efficacy scale 11.4 ± 2.5, range of potential values 10–40).20 Baseline mean values for power and strength are compatible with other community-based studies in similar populations.12 A mean habitual gait speed (meters/second) of 1.18 ± 0.21 (range 0.47–1.57) suggests a varied group with mixed levels of gait limitations.4 The mean SPPB score of 9.4 ± 1.36 (range 6–11) demonstrates that there was a relatively even mix between those with mild and moderate functional limitations as defined by the SPPB.4,14

Table 1.  Baseline Characteristics for 45 Subjects (34 Female, 11 Male)
CharacteristicsMean ± SDRange
  1. SD = standard deviation; CES-D = Center for Epidemiological Studies Depression Scale; PASE = Physical Activity Scale for the Elderly; 1RM = One Repetition Maximum, SPPB = Short Physical Performance Battery.

Age, years72.7 ± 4.665–83
Body mass index, kg/m227.0 ± 4.819.6–42.4
Medications2.6 ± 2.040–8
Chronic conditions2.1 ± 1.50–5
Education, years16.0 ± 2.88–20
Mini-Mental State Examination29.4 ± 1.224–30
Depressive symptoms (CES-D)14.1 ± 3.36–24.2
Physical activity (PASE)121.8 ± 73.610–314
Falls efficacy11.4 ± 2.510–24
Physiologic measures of motor impairment  
 Total leg peak power, watts573.9 ± 260.2142–1304
 Double leg press peak power375.2 ± 179.876–859
 Right leg extension peak power98.8 ± 47.820–253
 Left leg extension peak power99.9 ± 46.034–207
 Total leg strength, Newton-Meters2227.2 ± 876.2876–5380
 Double leg press, 1RM2117.4 ± 837.0842–5148
 Right leg extension, 1RM54.4 ± 23.316–126
 Left leg extension, 1RM55.4 ± 22.818–106
Physical performance measures  
 Stair climb time, seconds6.27 ± 1.564.60–12.52
 Chair stand (10) time, seconds29.7 ± 10.3217.84–71.57
 Tandem gait time, seconds77.2 ± 34.235.4–172.7
 Habitual gait velocity, m/second1.18 ± .210.47–1.57
 Maximal gait velocity, m/second1.61 ± .300.65–2.41
 Short Physical Performance Battery9.4 ± 1.366–11

Bivariate Regressions Between Leg Power or Leg Strength and Physical Performance

The bivariate relationships between leg power and physical performance are illustrated in Table 2 and included with the same analyses of leg strength for comparison. Leg power was significantly associated with all of the measures of physical performance and accounted for between 12% and 45% (R2) of the variance of these measures. In all the models, there was a significant association between leg power and the physical performance tests, with the strongest R2 values associated with stair-climb time and habitual and maximal gait speed.

Table 2.  Unadjusted Bivariate Linear Regression Models for Leg Power or Leg Strength Modeling Physical Performance
Physical
Performance
CoefficientStandard
Error
R 2P-value
  1. SPPB = Short Physical Performance Battery.

Stair climb time    
 Leg power−.003.0007.27<.001
 Leg strength−.008.0002.19.003
Chair stand (10) time    
 Leg power−.015.006.14.011
 Leg strength−.004.002.10.032
Tandem gait time    
 Leg power−.045.019.12.021
 Leg strength−.010.006.06.103
Habitual gait velocity    
 Leg power.0004.0001.26<.001
 Leg strength.0001.00003.21.002
Maximal gait velocity    
 Leg power.0007.0001.45<.001
 Leg strength.0002.00004.46<.001
SPPB    
 Leg power.002.0007.15.008
 Leg strength.0005.0002.12.017

In comparison, leg strength reached statistical significance with all of the measures except tandem gait, where P = .103. However, with only one exception, leg power accounted for more of the variance than leg strength. In the case of maximal gait, the influences of leg strength (R2 = .46) and leg power (R2 = .45) were essentially equivalent. Although demonstrating differing bivariate influences on the dependent variables, the correlation between leg strength and leg power was strong, r = .89. Adjustment for body weight did not materially alter the findings (data not shown).

Determination of Linearity

A quadratic term was modeled in each of the unadjusted models to ascertain whether the relationship between leg power and physical performance were curvilinear. As illustrated in Table 3, the quadratic term was significant (P = .05) in three of the models (stair-climb time, maximal gait, and SPPB). In the remaining models, the P-value was between .087 and .206, suggesting the possibility of Type II error; that is, given the relatively small sample size, we may not have the power to detect these associations, if in fact they exist. In every case, the quadratic model predicted more of the variance than the linear model. Taken together, these findings suggest that the relationship between leg power and physical performance may be curvilinear. The bivariate relationship between leg strength and physical performance was analyzed in the same way. Under none of the conditions was the quadratic term for strength statistically significant (data not shown).

Table 3.  Linear and Quadratic Models of Bivariate Relationships Between Leg Power and Physical Performance
Physical Performance
(Dependent Variable)
CoefficientStandard
Error
R 2P-value
  1. SPPB = Short Physical Performance Battery.

Stair climb time    
 Linear model−.003.0007.27<.001
 Quadratic model    
 Quadratic term.000006.000002.37.014
 Linear term.011.003 .001
Chair stand (10) time    
 Linear model−.015.006.14.011
 Quadratic model    
 Quadratic term.000029.000016.20.087
 Linear term−.053.022 .023
Tandem gait time    
 Linear model−.045.019.12.021
 Quadratic model    
 Quadratic term.000072.000056.15.206
 Linear term−.140.076 .073
Habitual gait velocity    
 Linear model.0004.0001.26<.001
 Quadratic model    
 Quadratic term−.0000004.0000003.29.198
 Linear term.001.0004 .032
Maximal gait velocity    
 Linear model.0007.0001.45<.001
 Quadratic model    
 Quadratic term−.0000008.0000004.50.052
 Linear term.002.0005 .001
SPPB    
 Linear model.0020.0007.15.008
 Quadratic model    
 Quadratic term−.000004.0000002.23.050
 Linear term.008.003 .011

Multivariate Regressions

Based on the analyses described in the methods section, a pool of common covariates (age, BMI, and chronic conditions) was included in adjusted analyses. Falls efficacy was added as an adjustment variable for habitual gait speed, given its close association with this physical performance measure as determined by our analysis and past publications.20 Additionally, contrary to previous reports, gender was never found to be a significant adjustment variable.21 To confirm this finding, all multivariate analysis were repeated, with the inclusion of gender, and under no conditions did this materially alter the multivariate results.

For three of the dependent variables (stair-climb time, chair-stand time, and SPPB) normality of the distribution was questioned. Analyses that included log transformation of all three of these variables did not alter the results concerning the associations between the variables.

In comparison to their adjustment variables, leg power and leg strength were the only variables that significantly predicted all physical performance measures at the alpha = .05 level, and in the case of chair-stand time were the only significant variables. Table 4 allows a comparison of the leg power and leg strength adjusted models of physical performance. In every case, statistical significance is achieved. However, in comparing the R2 values, it can be seen that the leg power models consistently account for more of the variance than the leg strength models.

Table 4.  Adjusted Multivariate Linear Regression Models for Leg Power or Leg Strength Modeling Physical Performance
Physical Performance
(Dependent Variable)
Physiological Impairment
(Independent Variable)
CoefficientStandard
Error
R 2P-value
  • Note: All models were adjusted for age, BMI, and chronic conditions.

  • *

    Habitual gait also adjusted for falls efficacy.

Stair climb time Leg power−.003.0007.50<.001
  Leg strength−.0007.0002.42.005
Chair stand (10) time Leg power−.016.006.24.010
  Leg strength−.004.002.20.025
Tandem gait time Leg power−.052.018.29.008
  Leg strength−.013.006.24.041
Habitual gait velocity* Leg power.0004.0001.61<.001
  Leg strength.0001.00003.57<.001
Maximal gait velocity Leg power.0007.0001.58<.001
  Leg strength.0002.00004.56<.001
Short Physical Performance Battery Leg power.002.0007.38.002
  Leg strength.0006.0002.35.006

This is further illustrated in Figure 1, in which partial R2 values for leg power and leg strength are plotted. Leg power R2 values are higher than the partial R2 values for strength for all physical performance measures, with the greatest differences being seen with stair-climb time, chair-stand time, and tandem gait.

Figure 1.

Partial R2 values for leg power and leg strength derived from the multivariate regression models. Of all of the measures of physical performance, leg power explains between 2% and 8% more of the variance than leg strength. SCT = stair-climb time; CST = chair-stand time; TG = tandem gait; HG = habitual gait; MG = maximal gait; SPPB = Short Physical Performance Battery.

DISCUSSION

The major finding of this study is that, in older people with mild to moderate mobility limitations, leg power is a significant predictor of physical performance. This is the first study of community-dwelling mobility-limited people to demonstrate that leg power impairments serve as a major determinant of function. The nature of this relationship appears to be curvilinear. Furthermore, though leg power and leg strength are related attributes being strongly correlated (r = .89), our results underscore differences as they relate to physical performance. Leg power, in comparison to strength, appears to exert a greater influence on function.

As previously stated, in mobility-limited older people, a relationship between function, as measured by physical performance, and disability has been demonstrated. Guralnik et al. have demonstrated that, in mobility-limited older people, the risk of disability, institutionalization, and mortality can be stratified based on SPPB score or habitual gait speed.4,16 Strength has been hypothesized as a primary physiological mediator on the disablement pathway.8 In contrast, our findings demonstrate that, within a cohort of older people with mild to moderate mobility limitations, leg power is different and exerts more influence than strength. In fact, using a curvilinear model, leg power generally predicted 15% to 50% of the variance in physical performance. Specifically in the case of habitual gait speed, in the unadjusted linear model, over one-quarter of the variance was predicted by leg power. This is a very significant finding, given that habitual gait speed is strongly predictive of disability, highlighting the importance of leg power as a proximal outcome on the disablement pathway. In reviewing all of the respective performance measures, it appears that, in comparison to strength, power may exert the greatest influence on stair-climb time, chair-stand time, tandem gait, and habitual gait. In contrast, the influence of power and strength on maximal gait appears to be more consistent.

The relationship between leg power and physical performance may be best described as curvilinear. The quadratic models explained more of the variance for all six dependent variables, although it only reached significance in three (stair-climb time, maximal gait, and SPPB). This finding is consistent with previous studies evaluating the relationship between motor impairments in strength and physical performance.6,8,9 The fact that the leg power quadratic models did not reach statistical significance for all of the functional outcomes can possibly be explained by an insufficient sample size. Other studies demonstrating a significant curvilinear relationship between strength and physical performance had sample sizes that were 4.5 to 20 times larger.6,8,9 In contrast, we did not find strong evidence for a curvilinear relationship between leg strength and the functional measures. It is possible that the point of change in slope for the regressions occurs at a different point for leg strength than it does for leg power. Along with larger sample sizes, the previous studies demonstrating curvilinearity included subjects who had little to no demonstrable functional limitations. By design, such subjects were excluded from our study. Taken together, these facts suggest that the changes in slope of the relationship between leg strength and physical performance occur at a higher functional level than for leg power. These results and hypotheses will need further confirmation through future research. Regardless, our findings underscore differences between leg power and leg strength as they relate to function.

Age, BMI, and number of chronic conditions (a measure of health status) were found to be significant adjustment variables for some of the physical performance measures. This is consistent with previous reports in similar populations.21 Additionally, in contrast to previous reports evaluating strength, adjusting for weight and gender did not influence the results.6,9,22 This finding, within a relatively small sample size, may reflect further differences between leg power and leg strength.

Despite the importance of its results, our study has limitations worthy of discussion. The relative homogeneity of the population places limits on the generalizability of the results. The majority of participants were college-educated Caucasians and represented a subset of mobility-limited older people. As stated previously, the failure to detect significant associations were likely due to our small sample size. Our methods should be applied to larger, more demographically and functionally diverse populations. As opposed to many previous studies that evaluated motor impairments, we did not examine specific measures of strength and power such as double leg press or right/left KE in our analysis.6,7 Instead we created composite scores of leg strength and power by summing all three measures. This represents a strength rather than a limitation of our methodology. Biomechanical analyses of gait and chair rise demonstrate that the performance of such tasks involve an interplay between muscle actions at both the hips and knees.23,24 Thus, our composite score better represents the diverse muscle actions composing our performance measures. Furthermore, this approach has been used by other investigators and is recognized as an enhanced means of statistically representing the relationship between impairment and function.9,22 We focused solely on lower extremity power and strength. The inclusion of upper extremity measures should be included in future investigations. Lastly, peak muscle power is not a measure easily obtainable on standard exercise equipment, thus limiting the translation of our findings to common clinical settings. However, using the specialized equipment available in this study, power measurements were no more challenging or time consuming than strength measurements. Recognizing the potential importance of measuring muscle power, in the future, greater availability of appropriate exercise equipment will be needed.

In conclusion, these findings from a cohort of community-dwelling mobility-limited older people identify impairments in leg power as important determinants of physical performance measures predictive of disability. Although related to leg strength, leg power is recognized as a separate attribute that exerts a greater influence on physical performance. These findings identify leg power as an important proximal determinant on the disablement pathway. They raise important questions for future geriatric and rehabilitative research.

Funded by Harvard/Hartford Foundation Center of Excellence; Harvard Older Americans Independence Center; and Department of PM&R, Harvard Medical School. American Academy of Physical Medicine and Rehabilitation Educational Research Foundation.

Ancillary