Relationships between mobility of axial structures and physical performance.

[Schenkman M, Shipp KM, Chandler J, et al. Relationships between mobility of axial structures and physical performance. Phys Ther. 1996;76:276-285.

Key words: Geriatrics, Physical performance, Spine

Physical activities often require mobility of the neck, trunk, and pelvic complex. Bed mobility, for example, may be most easily accomplished when the thoracic spine rotates relative to the pelvis. Bending over to tie one's shoelaces requires the ability to flex and extend the spine. The easiest way to look over one's shoulder while driving is to rotate the neck and thoracic spine. Each of these motions requires the ability to bend or twist the spine in order to move the upper body with respect to the pelvis. Because the axial structures (neck, trunk, and pelvis) form the supportive base from which movement of the limbs occurs, the ability to move the neck and trunk also should contribute to the ability to move the extremities. Loss of mobility of the axial structures and altered postural alignment are observed among elderly persons[1-4] and occur commonly with a variety of disorders.[5-9] The most direct causes are injuries affecting the spine (eg, sports and automobile-related injuries), chronic low back pain,[5,6] and degenerative diseases.[7-9] Neurologic disorders such as Parkinson's disease also result in decreased axial mobility and abnormal postural alignment.[10,11]

The association between structural characteristics of the axial segments and specific physical activities is not known. Whether losses of axial mobility are associated with compromised physical performance also is unknown Mobility can often be improved through exercise.[5,12-15] Losses of axial mobility, therefore, also might be reversible with appropriate intervention, and, if physical performance is associated with losses in axial mobility, function might improve.

We investigated relationships between sets of variables related to regional measures of the spine, measures of combined spinal motion, and measures of physical performance that incorporate axial mobility (Fig. 1). Three sets of variables are related to measures of the spine, including measures of spinal sagittal-plane configuration (ie, lumbar lordosis, thoracic kyphosis) and range of motion (ROM) (ie, cervical, lumbar). One set of variables is related to combined spinal motion. We defined combined spinal motion as triplanar motion that is not restricted to a single segment of the spine. We developed a measure of combined spinal motion referred to as "functional axial rotation" (FAR).[16] (The conceptual basis for this functional measure of spinal mobility has been presented elsewhere.16) We believe that this measure can be used to infer rotation that is available for functional activities. We anticipate that data from the current investigation will support this supposition. One set of variables is related to physical performance. The activities chosen (moving from a supine position to a sitting position, turning 360' while standing, and reaching while standing) are all representative of movements that incorporate some degree of axial mobility.

The specific aims of our investigation were (1) to determine the effect of age on axial mobility and on selected physical performance measures and (2) to determine the strength of association between five sets of variables (cervical ROM, lumbar ROM, sagittal configuration [ie, lumbar lordosis, thoracic kyphosis], combined spinal motion, and physical performance). We hypothesized that the following sets of variables would be correlated: (1) the cervical and sagittal configuration sets with the combined spinal motion set, (2) the combined spinal motion set with the physical performance set, and (3) the physical performance set with the three axial sets (cervical, lumbar, sagittal configuration). These relationships, which form the conceptual basis for this investigation, are outlined in Figure 1.

Method

Subjects

Male and female volunteers were recruited through the Duke University Medical Center Aging Center registry, the Department of Veterans Affairs Medical Center in Durham (NC), the Duke University Aging Center, and the Duke University Graduate Program in Physical Therapy. Because we wanted to investigate the effect of age on axial mobility and physical performance independent of specific diseases and injuries, subjects were excluded if they had primary pathology or surgery of the axial skeleton (eg, fusion of the spine or compression fractures), if they reported being on bed rest during the past year because of back pain, or if they had back pain on the day of testing. Subjects were also excluded if they reported having neurological conditions (eg, Parkinson's disease, stroke) that might contribute to functional limitations. Subjects who were not excluded by self-report were further screened by the examiner to rule out any neurological, rheumatological, or orthopedic disorders. The subjects provided written informed consent prior to their participation.

A sample of convenience of 31 men and 26 women, aged 20 to 91 years (X=58.4, SD=24.4), participated in this study. Subjects were recruited to fit into three age groups: 20 to 40 years (group 1), 60 to 74 years (group 2), and 75 years and older (group 3) (Tab. 1). Mean heights ranged from 170.4 to 172.7 cm. Mean weights ranged from 67.0 to 75.9 kg.

Measures

Measurements were obtained for each subject during a single session. Unless otherwise noted, each variable was measured three times. The first trial was for practice. The second and third trials were averaged to compute the score used for analysis.

Measures of regional ROMs and sagittal configuration included the following variables: standing sagittal-plane spinal configuration for the thoracic and lumbar regions; lumbar ROM in the sagittal plane (flexion and extension) and the frontal plane (lateral flexion); and cervical ROM in the sagittal plane (flexion and extension), the frontal plane (lateral flexion), and the transverse plane (rotation). The Debrunner kyphometer[TM](*) was used to measure sagittal-plane configuration (ie, thoracic kyphosis, lumbar lordosis)(17) (Fig. 2). Midpoints between T2-3, T11-12, and S1-2 were identified by palpation and marked. Subjects stood in their normal erect posture with feet hips-width apart and arms resting by their sides. The degrees were read directly from the scale with the blocks of the kyphometer spanning T2-3 and T11-12 for thoracic kyphosis and T11-12 and S1-2 for lumbar lordosis. A single measurement was recorded for each. Excellent reliability has been reported for lordosis and kyphosis of asymptomatic persons by others[17] and for lordosis and kyphosis of symptomatic elderly persons at our center[9] (intraclass correlation coefficients [ICCs]=.90 and above).

Lumbar mobility was determined using the Back Range of Motion (BROM[TM]) instrument[7]([dagger]) (Fig. 3). The spinous processes of T-12 and S-1 were palpated and marked. Measurements were made with the subjects in a standing position. Resting position was recorded. Subjects were instructed to bend as far as possible forward, backward, and to the side. For flexion and extension, the upper contact point of the base of the BROM[TM] instrument was placed at S-1 and the sliding arm of the device was placed at T-12. The degrees of maximum flexion and extension were read directly from the outer scale of the unit. For lateral flexion, the positioning frame OF-THE-BROM[TM] was placed at T-12. The degrees of maximum right and left lateral flexion were read from the inclinometer. Good reliability in elderly subjects with osteoarthritis has been reported (ICCs=.72-.94).[8]

The Cervical Range of Motion (CROM[TM]) instrument[18,19]([dagger]) was used to measure cervical mobility (Fig. 4). Subjects were seated with the back supported. With the CROM[TM] instrument on the head and the magnetic yoke on the chest, resting position was recorded from the appropriate inclinometer. Subjects were then instructed to move as far as possible into each plane of motion flexion, extension, lateral flexion, and rotation). The maximum degrees of motion for each trial were read from the appropriate inclinometer. Reliability has been established for measurements in all planes of motion for asymptomatic individuals (ICCs=.76 or above).[18,19] Measures of combined spinal motion included functional axial rotation, physical and visual (FAR-p and FAR-v). Functional axial rotation[16] was assessed with each subject seated and the pelvis stabilized by Velcro[R]([double dagger]) straps (Fig. 4). A hoop with symbols (numbers and letters) was suspended at eye level by two tripods, one in front of the subject and the other behind the subject. The CROM[TM] headpiece was donned by the subject, and the forward head arm of the unit was used as a pointer oriented toward the hoop. The subject was instructed to turn as far as possible in one direction (right or left), letting the arms swing by the sides, and to report the farthest symbol that could be seen. The symbol with which the pointer aligned determined FAR-p. The symbol the subject reported determined FAR-v. The symbols corresponded to 5-degree increments, with 0 degrees aligned with the midline of the subject's face. The degrees of rotation for each trial were computed from the symbols. Interrater reliability for asymptomatic subjects (ICC=.97, right and left sides) and test-retest reliability for asymptomatic subjects (ICC=.95, right side; ICC=.90, left side) were determined in our laboratory.(16)

Measures of physical performance included the 360-degree turn, functional reach, and timed performance of going from a supine to a sitting position. The 360-degree turn measures the number of steps and the time required to turn around in place in a standing position. A digital stopwatch was used. One trial to the right side and one trial to the left side were recorded. Reliability of this measure has been established in our laboratory for persons with Parkinson's disease (ICC=.77 for number of steps and .80 for time).[20]

Functional reach,[21] a clinical measure of balance control, was taken with each subject standing next to a wall with a yardstick at the height of the acromion and the dominant arm in 90 degrees of shoulder flexion. The distance the subject could reach forward was determined relative to this starting position. There were two practices and three trials. Excellent reliability (ICCs=.90 or above) has been reported for asymptomatic subjects.[21]

Movement from a supine to a sitting position was measured with each subject moving at a self-selected pace and as fast as possible. The subject began this test lying supine on a low treatment table with arms by the side and lower extremities extended. A digital stopwatch was used to determine the time from the signal to start to the completion of the activity. The subject was instructed to move as he or she normally would (self-selected pace). After two self-selected pace trials were complete, the subject was instructed to perform the task as fast as possible for the remainder of the trials.

Data Analysis

Subjects were assigned to one of three age groups (20-40 years, 60-74 years, and 75 years or over). Data from the test trials to each side were averaged, and the means of averaged data were used in all further analyses. We performed a preliminary analysis of data using a paired t test to compare the right and left sides; no differences were found. Data from the right side only were used in further analyses. The following sets of variables were used to analyze the data: cervical (including measures in four planes of motion), lumbar (including measures in three planes of motion), sagittal configuration (including thoracic and lumbar measures), combined spinal motion (FAR-v and FAR-p), and physical performance (including functional reach, supine to sit, and 360[degrees] turn in a standing position [time and steps]).

To determine whether there were differences by age group, we performed a multivariate analysis of variance (MANOVA), controlling for gender, for each of the five sets of variables. For sets of variables in which differences were found, we then used an analysis of variance to identify those variables within the set for which the effect of age was significant. Finally, when individual variables were different, we used Tukey's post hoc analysis.

We used canonical correlations to determine how the sets of variables correlated with one another. Canonical correlation[22] is the correlation between a linear combination of the variables in one set and those in another set. The idea is to determine the linear combination of the variables that have maximum correlation.([sections])

Results

The sample represents a wide range of ages and movement capabilities, with a wide range of values for all variables (Tab. 2). There was a 55-degree difference between the lowest and highest measurements of transverse-plane cervical rotation and a 100-degree difference between the lowest and highest FAR measurements. The number of steps to accomplish the 360-degree turn in a standing position ranged from 4 to 11, and the measurements of functional reach ranged from 19.8 to 48.3 cm.

Our first question was whether there was a difference by age group in the regional and combined measures of the spine or in the measures of physical performance. All of the cervical and lumbar measures changed in the expected direction across the three age groups. With increasing age, thoracic kyphosis increased and lumbar lordosis decreased. Similarly, measurements of ROM decreased. For example, there was a decline of cervical extension of more than 20 degrees and a decline of cervical rotation of more than 30 degrees between the youngest and oldest age groups. The oldest age group had less than half as much lumbar extension than did the younger age groups. Similarly, all four measures of physical performance showed declines across the age groups, with a 60% decrease for the number of steps required in a 360-degree turn (7.9 steps for the oldest age group compared with 5.2 steps for the youngest age group).

We first tested for the effects of age within sets of variables. The MANOVA was significant at P=.0001 for all sets of variables except sagittal configuration (P=.007) (Tab. 2). The analysis of variance, therefore, was performed for the individual variables in each of the sets and revealed that there was an effect of age for each of the variables (P [greater than or equal to] .0005 for 12 of the 15 variables). We therefore performed Tukey's post hoc analysis.

The results of Tukey's post hoc analysis revealed differences among all three age groups for many variables (cervical rotation, lumbar extension, lumbar lateral flexion, both measures of FAR, and functional reach). There were differences between the youngest age group and the oldest age group for all other variables except cervical flexion and kyphosis. We also found differences between groups 1 and 2 in cervical lateral flexion, lumbar flexion, and kyphosis. Finally, there was a difference between groups 2 and 3 for the number of steps in the 360-degree turn.

Table 3 shows relationships among the five sets of variables. Of the three axial sets, the cervical set correlated with the combined spinal motion set (r=.61, P=.0006). The first canonical variable for the cervical set of variables was the weighted sum of three variables, with the largest weight on cervical rotation (rotation [0.07], lateral flexion [0.05], and extension [0.02], respectively). The first canonical variable of the combined spinal motion set of variables was the weighted sum of the two variables, with the larger weight on FAR-v (FAR-v [0.04] and FAR-p [0.03]).

Table 3.
Correlations Between Sets of Variables

Sets Correlated                         r       p

Axial and combined spinal
    motion (CSM) sets
  Cervical and CSM sets                .61     .0006
  Lumbar and CSM sets                  .23     .78
  Sagittal spinal configuration
    and CSM sets                       .36     .19
CSM and physical performance
    sets                               .60     .005
Physical performance and axial
    sets
  Physical performance and
    cervical sets                      .56     .21
  Physical performance and
    lumbar sets                        .51     .24
  Physical performance and
    sagittal spinal configuration
    sets                               .37     .44

The combined spinal motion set of variables correlated with the physical performance set (r=.60, P=.005). The first canonical variable for the combined spinal motion set of variables was the weighted sum of FAR-p and FAR-v, with the greater weight on FAR-p (FAR-p [0.93], FAR-v [0.51]). The first canonical variable of the physical performance set was the weighted sum of three variables, with the most weight on functional reach (functional reach [0.96], 360[degrees] turn-steps [0.31], and 360[degrees] turn-time [-0.46]). The physical performance set of variables did not correlate with any of the axial sets.

Discussion

Results of our cross-sectional study illustrate that as age increases, axial ROM decreases. We did not control for possible subtle, preclinical underlying disease and therefore do not know whether these declines represent aging alone or reflect other factors. Furthermore, the three age groups were chosen as a sample of convenience intended to represent relatively young adults and two groups of older adults. The results indicate that the measures of combined spinal motion correlate strongly with measures of physical performance, suggesting relationships between the ability to move the spine and the ability to perform selected physical activities.

Declines of Mobility Across the Three Age Groups

As available motion of spinal segments of the neck and trunk decline, individuals have increasingly fewer options for functional movement. Decreases of FAR illustrate the magnitude and potential importance of losses of axial mobility with age. The oldest group of subjects had a mean of 109 degrees for FAR-p. In contrast, the youngest group of subjects had a mean of 152 degrees for FAR-p. Accompanying the loss of FAR-p was a decreased ability to visualize objects to the posterior measured by FAR-v. Limitations of FAR might contribute to the difficulty that older individuals typically report when performing simple daily activities such as looking over their shoulders while driving and should be investigated in future studies.

Declines in regional measures of spine mobility parallel the declines with age of FAR. Most regional measures of axial mobility declined by at least 25% across the three age groups investigated, with lower ranges of cervical and lumbar mobility for the two older age groups compared with the youngest age group. For lumbar extension, this decline was greater than 50%. These findings are consistent with those of other investigators.[1-4] Regional measures should be closely related to the structural limitations for spinal motion and thus are an indicator of mechanical constraints of movement.

Relationship Between Axial Motion and Physical Performance

The results of the canonical correlations demonstrate relationships between axial measures, combined spinal motion, and physical performance. The cervical set of variables, but not the lumbar set of variables, correlated with the combined spinal motion set. This finding is consistent with the fact that the cervical region contributes much more transverse- and coronal-plane motion than does the lumbar region.[23] The sagittal configuration set of variables did not correlate with the combined spinal motion set, although we had hypothesized this relationship.

The combined spinal motion set of variables correlated with the physical performance set, indicating a relationship between overall ability for combined spinal motion and ability to perform selected physical tasks. The axial set of variables showed a strong correlation with the physical performance set, although the association was not statistically significant. Lack of statistical significance may reflect the relatively small sample size. Significant associations between the axial and physical performance sets might be obtained with a larger data set. The axial spine also may function as a unit; thus, regional losses alone may not correlate with performance deficits as strongly as the whole unit. In addition, performance of the physical tasks investigated depends on many factors besides motion of the spine (eg, sensory status, ability to coordinate motor tasks, attentional ability). Thus, these measures are more distally related to regional spinal motion than is FAR. Larger data sets may help to clarify whether individual measures of axial motion and sagittal configuration relate to specific measures of physical performance.

Little other information is available examining the importance of axial mobility for physical performance. In one of the few available reports, Bergstrom and colleagues24 examined the relationship between spinal motion and several physical tasks for a cohort of 70-year-old subjects. That investigation was restricted to subjects whose functional ability was determined by self-report. Only a few questions relevant to axial mobility were examined, including the ability to reach the big toe, touch the earlobe, raise hands above the head, and enter public transportation. The results of their study indicated that loss of thoracic ROM was modestly associated with difficulty entering public transportation and with difficulty climbing stairs. Losses of both thoracic and lumbar ROM were associated with ability to touch the big toe, which is relevant to a person's ability for self-care of the feet.

One of the barriers to carrying out investigations of the relationship between axial mobility and physical performance has been the lack of available measures with established clinimetric properties that depend on combined axial motion. Most available measures of the axial spine are designed to quantify isolated regions (eg, cervical, lumbar) or single-plane motion (eg, frontal, sagittal).[1-7,17-19 Physical activities are performed by integrating mobility across many regions and planes of spinal motion. The "functional axial rotation" measure[16] was designed to quantify spinal motion in a combined manner. The substantial correlation between the combined spinal motion set of variables and the physical performance set of variables supports our contention that these measures are indeed relevant and functional measures of axial mobility. Additional measures of combined spinal motion in which other planes dominate would be useful (eg, a combined measure with predominantly sagittal-plane motion could be developed, such as reaching downward in a sitting position and tying shoelaces).

A second barrier to examining the relationship between measures of axial mobility and physical performance is the paucity of measures of physical performance that incorporate motion of the spine. Based on clinical observations, we chose three performance tasks that emphasize different requirements for spinal motion: reaching, moving from a supine to a sitting position, and turning in place while in a standing position. Reaching forward (functional reach)[21] includes extension of the spine and thoracic rotation. Moving from a supine to a sitting position often includes thoracic rotation and lumbar lateral flexion. Turning in place while in a standing position generally involves lumbar lateral flexion. Measures should be developed for other tasks that require spinal motion such as putting on a jacket and bending down to tie shoelaces.

We used a sample of convenience and assigned subjects to three age groups. Future studies investigating subjects of all ages will be necessary to further characterize the effects of age on axial mobility and physical performance. Furthermore, there are limitations to crosssectional data analyses. Although the associations between combined spinal mobility and physical performance are intriguing, this investigation cannot determine whether and which of those associations are causally related. Studies are needed to further investigate how regional axial mobility measures relate to specific types of physical performance and to determine whether changes in axial mobility are related to changes in performance.

Conclusion

We demonstrated an age effect for measures of axial mobility and sagittal configuration and for measures of physical activities that require axial motion. In addition, a relationship has been demonstrated between the combined spinal motion set of variables and the physical performance set of variables. Axial mobility may be modifiable. The extent to which changes in axial mobility will influence physical performance is unknown but worth exploring. Functional decline occurs with impairments in many systems (eg, sensory loss, motor control). Many of these impairments are not reversible. Loss of axial mobility is a potentially modifiable contributor to functional decline that has not been recognized previously.

References

[1] Einkauf DK, Ghodes ML, Jensen GM, Jewell MJ. Changes in spinal mobility with increasing age in women. Phys Ther. 1987;57:370-375. [2] Fitzgerald GK, Wynveen KJ, Rheault W, Rothschild B. Objective assessment with establishment of normal values for lumbar spinal range of motion. Phys Ther. 1983;63:1776-1781. [3] Youdas JW, Garrett TR, Suman VJ, et al. Normal range of motion of the cervical spine: an initial goniometric study. Phys Ther. 1992;72:770-780. [4] Sullivan MS, Dickenson CE, Troup JDG. The influence of age and gender on lumbar spine sagittal plane range of motion: a study of 1,126 healthy subjects. Spine. 1994;19:682-686. [5] Marras WS, Wongsum PE. Flexibility and velocity of the normal and impaired lumbar spine. Arch Phys Med Rehabil. 1986;67:213-217. [6] Mayer T, Tencer A, Kristoferson S, ct al. Use of noninvasive techniques for quantification of spinal range-of-motion in normal subjects and chronic low-back dysfunction patients. Spine. 1984;9:588-595. [7] Weiner DK, Distell B, Studenski SA, et al. Does radiographic osteoarthritis correlate with flexibility of the lumbar spine? J Am Geriatr Soc. 1994;42:257-263. [8] Husby G, Gram JT. Ankylosing spondylitis: clinical aspects, comparisons, men vs women, hospitalized vs epidemiological patients. In: Calabro JJ, Dick WD, eds. Ankylosing Spondylitis. Boston, Mass: MTP Press Ltd; 1987:86. [9] Lyles KW, Gold DT, Shipp KM, et al. Association of osteoporotic vertebral compression fractures with impaired functional status. Am J Med. 1993;94:595-601. [10] Schenkman M. Physical therapy intervention for the ambulatory patient. In: Turnbull G, ed. Physical Therapy Management of Parkinson's Disease. New York, NY: Churchill Livingstone Inc; 1992:137-192. [11] Webster DA. Critical analysis of the disability in Parkinson's disease. Modem Treatment. 1968;5:257. [12] Mayer T, Tabor J, Borasso E, et al. Physical progress and residual impairment quantification after functional restoration, part I: lumbar mobility. Spine. 1994;19:389-394. [13] Manniche C, Hesselsoe G, Bentzen L, et al. Clinical trial of intensive muscle training for chronic low-back pain. Lancet. December 24-31, 1988:1473-1476. [14] Kohles S, Barnes D, Gatchel T, et al. Improved physical performance outcomes after functional restoration in patients with chronic low-back pain: early vs recent training results. Spine. 1990;15:1321-1324. [15] Elnaggar IM, Nordin M, Sheikhzadeh A, et al. Effects of spinal flexion and extension exercises on low-back pain and spinal mobility in chronic mechanical low-back pain patients. Spine. 1991;16:967-972. [16] Schenkman M, Hughes MA, Bowden MG, Studenski SA. A new device for measuring functional axial rotation. Phys Ther. 1995;75:151-156. [17] Ohlen G, Spangefort E, Tingvall C. Measurement of spinal sagittal configuration and mobility with Debrunner's kyphometer. Spine. 1989; 14:580-583. [18] Rheault W, Albright B, Byers C, et al. Intertester reliability of the cervical range of motion device. J Orthop Sports Phys Ther. 1992;15:147-150. [19] Capuano-Pucci D, Rheault W, Aukai J, et al. Intratester and intertester reliability of the cervical range of motion device. Arch Phys Med Rehabil. 1991;72:338-340. [20] Schenkman M, Cutson T, Chandler J, et al. Reliability of measures for Parkinson's disease. In: Proceedings of the Annual Meeting of the American Geriatrics Society; Los Angeles, Calif, May 21, 1994. 1994:SA-77. [21] Duncan PW, Weiner DK, Chandler J, Studenski SA. Functional reach: a new clinical measure of balance. J Gerontol 1990;45:Ml92-M197. [22] Anderson TW. An Introduction to Multivariate Statistical Analysis. New York, NY: John Wiley & Sons Inc; 1971. [23] White AA, Panjabi MM. Clinical Biomechanics of the Spine. Philadelphia, Pa: JB Lippincott Co; 1978. [24] Bergstrom G, Aniansson A, Bjelle A, et al. Functional consequences of joint impairment at age 79. Scand J Rehabil Med. 1985;17:183-190.

(*) Protek AG, Bern, Switzerland. ([dagger]) Performance Attainment Associates, 958 Lydia Dr, Roseville, MN 55113. ([double dagger]) Velcro USA Inc, 406 Brown Ave, PO Box 5218, Manchester, NH 03108. ([sections]) For example, canonical correlation can find the association between the combined spinal motion measures and physical performance measures, where the combined spinal motion measures consist of the two variables FAR-P ([x.sub.1]) and F.A.R-v ([x.sub.2]) and the physical performance measures consist of the variables functional reach ([y.sub.1]), supine to sit ([y.sub.2], 360-degree turn-time ([y.sub.3]), and 360-degree turn-steps ([y.sub.4]). The linear combination (X) of combined measures of the spine is the weighted sum of the variables [x.sub.1] and [x.sub.2] and bas the form

X = [a.sub.1][x.sub.1] + [a.sub.2][x.sub.2]

where [a.sub.1] and [a.sub.2] are constants. The linear combination (Y) of physical performance measures is the weighted sum of [y.sub.1], [y.sub.2], [y.sub.3]. and [y.sub.4] and has the form

Y = [b.sub.1][y.sub.1] + [b.sub.2][y.sub.2] + [b.sub.3][y.sub.3] + [b.sub.4][y.sub.4]

where [b.sub.1], [b.sub.2], [b.sub.3], and b, are the constants in that linear combination. The correlation between X and Y, subject to certain constraints, is the canonical correlation. Canonical correlation analysis was carried out for the hypothesized pairs of sets of variables (Fig. 1). When presenting the results of a canonical correlation, the overall correlation coefficient (r) and probability value (P) are reported. In addition, the relative weights of the variables in the linear combination (ie, [a.sub.1] and [a.sub.2]; [b.sub.1], [b,sub.2], [b.sub.3], and [b.sub.4]) are given for the first canonical correlate. Note that only relative weights (and not the absolute values) are meaningful.

M Schenkman, PhD, PT, is Associate Professor, Graduate Program in Physical Therapy, Senior Fellow, Center for the Study of Aging and Human Development, and Co-Director, Claude D Pepper Older Americans Independence Center, Duke University, Durham, NC 27710. Address all correspondence to Dr Schenkman at the Center for the Study of Aging and Human Development, Duke University Medical Center, PO Box 3003, Durham, NC 27710 (USA) ([e-mail]).

KM Shipp, PT, is Associate in Research, Department of Physical Therapy, and Senior Fellow, Center for the Study of Aging and Human Development, Duke University.

J Chandler, PT, was Assistant Clinical Professor, Department of Physical Therapy, and Senior Fellow, Center for the Study of Aging and Human Development, Duke University, at the time this study was conducted.

SA Studenski, MD, is Director, Center for the Study of Aging and Human Development, University of Kansas, Lawrence, KS 66045.

M Kuchibhatla, Phd, is Statistician, Claude D Pepper Older Americans Independence Center, Center for the Study of Aging and Human Development, Duke University.

The study protocol was approved by the institutional review boards for clinical investigations at Duke University Medical Center and the Veterans Administration Medical Center in Durham.

This work was presented in part at the 1992 meeting the Gerontologic Society of America, Washington, DC.

This work was supported by the National Institutes of Health, National Institute on Aging, Claude D Pepper Older Americans Independence Center Grant 5 P60 AG 11268, and by a Duke University Medical Center Small Research Grant.