Correlations of clinical and laboratory measures of balance in older men and women

Authors


Abstract

Objective

It is known that impaired balance is associated with falls in older adults; however, there is no accepted gold standard on how balance should be measured. Few studies have examined measures of postural sway and clinical balance concurrently in large samples of community-dwelling older adults. We examined the associations among 4 types of measures of laboratory- and clinic-based balance in a large population-based cohort of older adults.

Methods

We evaluated balance measures in the Maintenance of Balance, Independent Living, Intellect and Zest in the Elderly Boston Study (276 men and 489 women ages 64–97 years). The measures included laboratory-based anteroposterior (AP) path length and mean sway speed, mediolateral (ML) mean sway and root mean square, and area of ellipse postural sway; the Short Physical Performance Battery (SPPB); the Berg Balance Scale; and the one-leg stand test. Spearman's rank correlation coefficients were assessed among the balance measures.

Results

The area of ellipse sway was highly correlated with the ML sway measures (r = >0.91, P < 0.0001) and sway speed was highly correlated with AP sway (r = 0.97, P < 0.0001). The Berg Balance Scale was highly correlated with the SPPB (r = 0.74, P < 0.001) and the one-leg stand test (r = 0.82, P < 0.001). Correlations between the laboratory- and clinic-based balance measures were low but statistically significant (−0.29 ≤ r ≤ −0.16, P < 0.0001).

Conclusion

Clinic-based balance measures, and laboratory-based measures comparing area of ellipse with ML sways or sway speed with AP sway, are highly correlated. There is less correlation between the clinic- and laboratory-based measures. Since both laboratory- and clinic-based measures inform balance in older adults, but are not highly correlated with each other, future work should investigate the differences.

INTROUCTION

One-third of US adults ages 65 years or older experience one or more falls each year (1). Nearly 40% of these falls result in hospitalization due to a fall-related injury (2). Impaired balance in older adults has been shown to increase the risk of falls (3–5), yet there is no consensus on the best way to measure balance (i.e., laboratory- or clinic-based measures) in this population. Laboratory-based balance tests (e.g., force plate measurements) are conducted in a research setting using specialized equipment and typically include multiple series of data collected from platform measurements of sway from center of pressure (COP) in different directions and require extensive postmeasurement data processing. Clinical performance-based measures (e.g, the Short Physical Performance Battery [SPPB], Berg Balance Scale, and one-leg stand test, among others) may be performed in the field and typically would require completion of multiple but common tasks, often with a lesser time commitment compared with laboratory measures.

One study has suggested that the laboratory-based tests of balance should serve as the gold standard because they are more sensitive to slight changes in postural sway (6). However, the clinical tests of balance are often more accessible to clinicians and researchers since they do not require specialized equipment. The Berg Balance Scale has gained widespread support for use as a clinical measure of balance, given its documented validity and reliability (7). Some studies have used both the laboratory and the clinical measures of balance (8, 9), but this approach may not be feasible in a study of older adults who may tire easily. The issue of respondent burden in elderly participants is also possible with the Berg Balance Scale or the SPPB, since each comprises multiple time-consuming tasks.

Although the Berg Balance Scale and other clinical balance measures have been compared with platform (laboratory-based) balance measures in small studies of elderly participants (10, 11), individuals who have experienced a stroke (12, 13), and elderly nursing home residents (14), there has been no comprehensive evaluation of the SPPB, Berg Balance Scale, and laboratory-based measures of balance in a large sample of community-dwelling older adults. In addition, there are conflicting results regarding possible sex differences in some of the balance measures (15). The purpose of this study was to compare several laboratory-based and clinical measures of balance in a population-based cohort of older adults to determine if the clinical tests are equivalent to the laboratory-based measures, whether simpler clinical measures such as the SPPB or the one-leg stand test are comparable to the Berg Balance Scale in measuring balance, and whether measures of balance differ by sex. We would expect high correlation within clinic-based measures and among laboratory-based measures, but correlation may differ between clinic- and laboratory-based measures.

Significance & Innovations

  • Few studies have examined measures of postural sway and clinical balance concurrently in large samples of community-dwelling older adults. We examined the associations among 4 types of measures of laboratory- and clinic-based balance in this population.

  • We found that clinic-based balance measures, and laboratory-based measures comparing area of ellipse with mediolateral sways or sway speed with anteroposterior sway, are highly correlated. There is less correlation between the clinic- and laboratory-based measures.

SUBJECTS AND METHODS

Study population.

Participants in the current analysis were members of the Maintenance of Balance, Independent Living, Intellect and Zest in the Elderly (MOBILIZE) Boston Study, a longitudinal study designed to examine novel risk factors for falls in a population-based sample of older adults living near Boston, Massachusetts. Details of the MOBILIZE Boston Study cohort have been previously reported (16, 17). In brief, between 2005 and 2008, the study enrolled 765 participants 70 years of age or older (which includes 16 spouses ages 64–69 years) who were able to communicate in English, lived within the Boston area, could walk for 20 feet unassisted, and were deemed cognitively intact (i.e., Mini-Mental State Examination score of 18 or higher).

As previously described (16–18), enrollment included door-to-door recruitment and telephone screenings. A group of 4,303 potential participants 70 years of age or older were identified from the 5,655 households sampled. Among these 4,303 participants, 1,581 were ineligible and 1,973 could not be located or refused to participate. The standard Council of American Survey Research Organization response rate was 52% after screening for eligibility criteria, and 30% for the door-to-door phase. In comparison with US Census data for the population ages 65 years and older in the Boston area, the study sample was representative of Boston area elders in terms of age, sex, race, and ethnicity. Participants were examined by research nurses at the MOBILIZE Boston Study clinic. The Institutional Review Board at the Hebrew Rehabilitation Center approved the study, and all participants provided informed consent.

Measures of balance.

During study visits, trained clinical staff assessed balance among the participants. The measures of balance included laboratory-based COP data using a Kistler force plate (Model 9286AA, Kistler Instrument Corporation) and clinical balance tests, including the timed balance component from the SPPB, the score from the Berg Balance Scale, and the ability to stand on 1 leg for up to 20 seconds.

Prior to measuring balance using the laboratory-based force plate tests, we performed a daily calibration trial where we placed a 50-pound weight on the force plate and collected data for 30 seconds to ensure accuracy. We then asked participants to stand comfortably with bare feet approximately hip width apart, with their arms at their sides, eyes open, and looking straight ahead. For each participant, we conducted 5 trials of quiet standing, during which COP trajectory data were collected using a single force plate, with each trial lasting 30 seconds in duration. Traditional sway parameters were used to describe the direction, distance, and velocity of a participant's trajectory, with greater sway indicating poorer balance. Five traditional sway measures have been shown to predict falls in prospective studies (19–24). Therefore, the current study examined the following measures: mediolateral (ML) mean sway (the length in mm of ML sway from geometric center averaged across the COP data), ML root mean square (RMS; the ML RMS or the SD of all ML measurements, in mm), area of ellipse (the area that fits 95% of the sway data points, in mm2), anteroposterior (AP) path length (the overall AP movement over 30 seconds, in mm), and sway speed (the combination of the AP and ML path lengths divided by time, in mm/second). These measures have been shown to have high interrater reliability and test–retest reliability, with intraclass correlation coefficients (ICCs) ranging from 0.70–0.89 (25). For the majority of the study participants, all measures were averaged across the 5 trials; in <10 participants, the measures were averaged across 4 trials to obtain acceptable reliability (26).

The participants were also asked to perform several clinical balance tests while wearing their own typical shoe wear (no participants wore high heels during this test). The balance component of the SPPB (27, 28) included 10-second timed measurements of unsupported standing with feet in a side-by-side parallel stand (a score of 1 means the participant was able to hold for 10 seconds and 0 means unable), a semitandem foot stand with the heel of one foot placed to the side of the big toe of the other foot (a score of 1 means the participant was able to hold for 10 seconds and 0 means unable), and a tandem foot stand with the heel of one foot placed directly in front of the toes of the other foot (a score of 2 means the participant was able to hold for 10 seconds, 1 means able to hold for 3–9 seconds, and 0 means unable), with a total score range of 0–4. The balance component of the SPPB has been shown to have good reliability, with ICCs ranging from 0.70–0.82 (29).

The Berg Balance Scale (30) has an overall range of 0–56 and comprises 14 items (or tasks), with each item score ranging from 0–4. The Berg Balance Scale includes the following tests: the ability to place the feet together independently and standing unsupported for up to 1 minute, standing unsupported with one foot in front of the other for up to 30 seconds, sitting to standing/single chair stand, transferring from one chair to another, turning 360°, reaching forward with an outstretched arm while standing, standing unsupported for up to 2 minutes without holding onto anything for support, sitting with the back unsupported and the feet supported on the floor or on a stool, standing to sitting, standing unsupported with eyes closed for 10 seconds, picking up an object from the floor and returning to a standing position, turning to look behind over the left and right shoulders while standing, placing the alternate foot on a step or stool while standing unsupported, and standing on 1 leg for up to 20 seconds. The following is an example of the scoring technique for item 1 (placing the feet together independently and standing unsupported for up to 1 minute): a code of 0 would be given if the person needed help to attain the position and was unable to hold for 15 seconds; 1 if the person needed help to attain the position, but was able to stand for 15 seconds with the feet together; 2 if the person was able to place the feet together independently, but was unable to hold for 30 seconds; 3 if the person was able to place the feet together independently and stand for 1 minute with supervision; and 4 if the person was able to place the feet together independently and stand for 1 minute safely. The Berg Balance Scale has been shown to have high interrater and intrarater reliability, with ICCs of 0.98 and 0.99, respectively (31).

In our study, we also examined the last component of the Berg Balance Scale (the one-leg stand test [time in seconds to stand on 1 leg without holding on for support]) separately to determine its correlation with the overall Berg Balance Scale as well as with the SPPB and the force plate platform COP balance measures, since the one-leg stand test component has also been associated with falls in other studies of older adults (32, 33). For all clinical balance measures examined in the current study, higher values indicate better balance.

Statistical analysis.

Using the balance data collected from the participants at the baseline MOBILIZE Boston Study examination (entered and verified from October 2008), we generated descriptive statistics, including medians, ranges, means, and SDs, of the balance measures for the overall group, and separately by sex. We compared differences in distributions between men and women using Wilcoxon's rank sum tests. Spearman's rank correlation coefficients were generated to determine the associations among the different measures of COP-based static balance, among the different clinic-based measures of balance, and between the static balance and the clinic-based balance measures for the entire group and by sex. We also examined possible sex differences in the correlation coefficients after normalizing the static balance measures by accounting for height differences between the men and women (34). More specifically, the COP measures were divided by body height to account for the different ranges of height for women and men. Moreover, we used generalized linear regression to examine the association between COP-based static measures of balance and those from clinic-based measures, and to formally test for interaction between sex and clinic-based balance measures in predicting COP-based measures adjusting for height, weight, and age.

We used a 2-sided P value level of 0.05 to indicate statistical difference from 0 or comparison to the reference group. All analyses were conducted using SAS, version 9.1.

RESULTS

The baseline characteristics of study participants and balance measures are shown in Table 1. The mean age of participants was 78 years (range 64–97 years). In general, men had a higher degree of sway than women in 3 of the 5 COP measures, indicating poorer static balance. Specifically, men had higher ML mean sway, ML RMS, and area of ellipse sway than women (P < 0.0001 for all). However, men had relatively better clinic-based balance measures, as indicated by the balance component of the SPPB, the Berg Balance Scale, and the one-leg stand test balance measures (P ≤ 0.002 for all).

Table 1. Total and sex-specific baseline characteristics and balance measures of the MOBILIZE Boston Study*
 All (n = 765)Men (n = 276)Women (n = 489)
Mean ± SDMedian (25th, 75th percentiles)Mean ± SDMedian (25th, 75th percentiles)Mean ± SDMedian (25th, 75th percentiles)
  • *

    MOBILIZE = Maintenance of Balance, Independent Living, Intellect and Zest in the Elderly; RMS = root mean square; SPPB = Short Physical Performance Battery.

  • P < 0.0001 for the difference between men and women.

  • P ≤ 0.002 for the difference between men and women.

Characteristics      
 Age, years78.1 ± 5.474.0 (74.0, 82.0)78.3 ± 5.277.0 (74.0, 82.0)78.0 ± 5.677.0 (74.0, 82.0)
 Weight, pounds162.4 ± 34.2160.0 (139.0, 181.3)179.5 ± 32.2175.5 (156.8, 195.0)152.8 ± 31.5149.0 (130.0, 171.0)
 Height, inches64.4 ± 3.964.0 (62.0, 67.0)68.0 ± 2.868.0 (66.0, 70.0)62.4 ± 2.862.0 (61.0, 64.0)
Static balance measures      
 Mediolateral mean sway, mm2.5 ± 1.12.3 (1.8, 3.1)2.8 ± 1.22.7 (2.0, 3.4)2.4 ± 1.02.2 (1.7, 2.8)
 Mediolateral RMS, mm3.2 ± 1.32.9 (2.2, 3.8)3.5 ± 1.43.3 (2.5, 4.3)3.0 ± 1.22.8 (2.2, 3.6)
 Area of ellipse sway, mm2183.3 ± 140.7146.3 (97.1, 225.4)222.1 ± 165.2193.1 (118.1, 271.6)161.1 ± 119.1132.6 (88.1, 189.7)
 Anteroposterior path, mm446.9 ± 124.5424.7 (367.3, 492.4)454.7 ± 139.1427.8 (368.4, 506.6)442.5 ± 115.2422.4 (367.3, 487.0)
 Mean sway speed, mm/second19.1 ± 4.918.3 (16.1, 21.3)19.5 ± 5.418.6 (16.0, 21.8)19.0 ± 4.518.2 (16.1, 21.1)
Clinical balance measures      
 SPPB tandem balance score3.3 ± 1.04.0 (3.0, 4.0)3.4 ± 0.94.0 (3.0, 4.0)3.1 ± 1.14.0 (2.0, 4.0)
 One-leg stand test, seconds8.1 ± 7.35.5 (2.1, 14.5)9.2 ± 7.47.4 (2.2, 16.8)7.4 ± 7.14.5 (1.9, 12.1)
 Berg Balance Scale score49.7 ± 6.751.0 (48.0, 55.0)50.5 ± 6.352.0 (48.0, 56.0)49.3 ± 6.950.0 (48.0, 54.0)

The unadjusted overall and sex-specific Spearman's rank correlation coefficients among the 5 COP static balance measures are shown in Table 2. All correlation coefficients among the different measures of balance were statistically significant (P < 0.0001). Nonetheless, area of ellipse sway was highly correlated with ML mean sway (r = 0.91) and ML RMS sway (r = 0.92), but was less correlated with AP sway (r = 0.31). Sway speed was more highly correlated with AP sway (r = 0.97) than with sways in the ML directions (0.34 ≤ r ≤ 0.35) and area of ellipse sways (r = 0.36). Comparisons of the AP sway with the ML mean sway (r = 0.27) and ML RMS (r = 0.28) balance measures were less correlated. This general pattern was similar for both men and women; however, men had higher correlation coefficients than women when comparing sway speed with ML mean (r = 0.44 versus r = 0.27), ML RMS (r = 0.46 versus r = 0.28), and area of ellipse sways (r = 0.48 versus r = 0.29). Similar sex differences were observed when comparing AP sway with sway in the ML directions and with area of ellipse sway. These sex-specific patterns remained after normalizing the static balance measures by scaling them to account for height (data not shown).

Table 2. Total and sex-specific Spearman's rank correlation coefficients for the COP-based static balance measures of the MOBILIZE Boston Study*
 Mediolateral mean sway, mmMediolateral RMS, mmArea of ellipse sway, mm2Anteroposterior path, mmMean sway speed, mm/second
  • *

    P < 0.0001 for all correlation coefficients. COP = center of pressure; MOBILIZE = Maintenance of Balance, Independent Living, Intellect and Zest in the Elderly; RMS = root mean square.

All     
 Mediolateral mean sway, mm1.0†0.9970.9100.2680.340
 Mediolateral RMS, mm 1.0†0.9170.2800.354
 Area of ellipse sway, mm2  1.0†0.3060.362
 Anteroposterior path length, mm   1.0†0.970
 Mean sway speed, mm/second    1.0†
Men     
 Mediolateral mean sway, mm1.00.9960.9130.3380.444
 Mediolateral RMS, mm 1.00.9200.3520.460
 Area of ellipse sway, mm2  1.00.3930.481
 Anteroposterior path length, mm   1.00.967
 Mean sway speed, mm/second    1.0†
Women     
 Mediolateral mean sway, mm1.00.9970.8970.2110.266
 Mediolateral RMS, mm 1.00.9030.2230.280
 Area of ellipse sway, mm2  1.00.2510.288
 Anteroposterior path length, mm   1.00.972
 Mean sway speed, mm/second    1.0†

The unadjusted overall and sex-specific Spearman's rank correlation coefficients of the COP-based static balance measures with the clinical balance measures, and the correlation coefficients among the clinic-based measures, are shown in Table 3. Although all correlations were statistically significant, those between the static measures and clinic-based measures of balance were generally low, with coefficients ranging from −0.29 to −0.16 overall. Among the clinic-based balance measures, the balance component of the SPPB was highly correlated with the Berg Balance Scale (r = 0.74) and the one-leg stand test (r = 0.62), and there was a high correlation between the Berg Balance Scale and the one-leg stand test (r = 0.82). These general patterns of associations were similar for both men and women. Again, coefficients were higher for men than women when comparing sway speed or AP sway with each of the clinic-based measures of balance. These sex-specific differences were observed even after the COP-based measures were normalized to account for height (data not shown).

Table 3. Total and sex-specific Spearman's rank correlation coefficients for the COP-based static balance measures with clinical balance measures of the MOBILIZE Boston Study*
 AllMenWomen
SPPBBerg Balance ScaleOne-leg stand test, secondsSPPBBerg Balance ScaleOne-leg stand test, secondsSPPBBerg Balance ScaleOne-leg stand test, seconds
  • *

    Higher values in the static balance measures indicate poorer balance, while higher SPPB tandem balance, Berg Balance Scale, and one-leg stand test values indicate better balance. COP = center of pressure; MOBILIZE = Maintenance of Balance, Independent Living, Intellect and Zest in the Elderly; SPPB = Short Physical Performance Battery; RMS = root mean square.

  • P < 0.0001 for all correlation coefficients.

Static balance measures         
 Mediolateral mean sway, mm−0.18−0.28−0.29−0.21−0.33−0.37−0.22−0.31−0.29
 Mediolateral RMS, mm−0.18−0.28−0.28−0.21−0.33−0.36−0.22−0.30−0.28
 Area of ellipse sway, mm2−0.16−0.26−0.27−0.24−0.37−0.39−0.20−0.27−0.26
 Anteroposterior path length, mm−0.22−0.24−0.26−0.27−0.34−0.38−0.21−0.19−0.19
 Mean sway speed, mm/seconds−0.21−0.24−0.26−0.28−0.36−0.39−0.19−0.17−0.19
Clinical balance measures         
 SPPB tandem balance0.740.620.710.570.750.64
 Berg Balance Scale0.820.820.82

The results from the generalized linear regression indicate that there were sex differences in the associations between clinic- and COP-based measures after accounting for height, weight, and age (Table 4). The magnitude of the association between each of the clinic-based measures of balance and sway speed (as measured by the beta coefficients from the models) was seen to be nearly twice as great in men than women (P < 0.05 for all interactions). The beta coefficients were also higher in men than women between the clinic-based measures and area of ellipse sway, although beta coefficients were at borderline statistical significance for the SPPB and one-leg stand test.

Table 4. Sex-specific associations between the clinic-based and COP-based static balance measures of the MOBILIZE Boston Study*
 Mean sway speed, mm/secondArea of ellipse sway, mm2
Men, β (SE)Men, R2Women, β (SE)Women, R2Men, β (SE)Men, R2Women, β (SE)Women, R2
  • *

    COP = center of pressure; MOBILIZE = Maintenance of Balance, Independent Living, Intellect and Zest in the Elderly; SPPB = Short Physical Performance Battery.

  • Crude R2 is unadjusted, all others are adjusted R2; all crude and adjusted P < 0.01 for the association between clinic-based balance predicting COP-based static balance.

  • P < 0.001 for interaction between sex and SPPB predicting sway speed adjusting for height, weight, and age; P = 0.054 for interaction predicting area of ellipse sway.

  • §

    P < 0.001 for interaction between sex and Berg Balance Scale predicting sway speed adjusting for height, weight, and age; P = 0.0005 for interaction predicting area of ellipse sway.

  • P = 0.001 for interaction between sex and one-leg stand test predicting sway speed adjusting for height, weight, and age; P = 0.057 for interaction predicting area of ellipse sway.

Crude SPPB balance−2.4 (0.37)0.132−1.1 (0.20)0.057−50.0 (11.8)0.060−25.5 (5.3)0.046
 Adjusted height−2.6 (0.39)0.139−0.9 (0.20)0.079−54.0 (12.3)0.073−27.9 (5.4)0.058
 Adjusted height and weight−2.6 (0.39)0.142−1.1 (0.19)0.209−50.9 (12.2)0.095−25.8 (5.4)0.074
 Adjusted height, weight, and age−2.1 (0.41)0.176−1.0 (0.19)0.218−34.3 (13.0)0.130−24.4 (5.6)0.074
Crude Berg Balance Scale−0.44 (0.056)0.185−0.19 (0.037)0.052−12.0 (1.8)0.148−5.8 (0.95)0.074
 Adjusted height−0.48 (0.059)0.199−0.15 (0.037)0.068−13.3 (1.8)0.171−6.6 (0.98)0.094
 Adjusted height and weight−0.50 (0.059)0.210−0.25 (0.035)0.227−12.7 (1.8)0.183−6.2 (1.0)0.102
 Adjusted height, weight, and age§−0.43 (0.063)0.231−0.22 (0.039)0.230−10.8 (2.0)0.198−6.2 (1.1)0.100
Crude one-leg stand test, seconds−0.27 (0.042)0.133−0.12 (0.029)0.031−7.3 (1.3)0.107−4.4 (0.75)0.067
 Adjusted height−0.28 (0.042)0.137−0.09 (0.029)0.055−7.4 (1.3)0.113−4.9 (0.77)0.084
 Adjusted height and weight−0.29 (0.042)0.143−0.14 (0.028)0.192−7.0 (1.3)0.130−4.5 (0.77)0.095
 Adjusted height, weight, and age−0.23 (0.046)0.171−0.12 (0.030)0.200−5.3 (1.4)0.151−4.5 (0.83)0.093

DISCUSSION

In this cross-sectional analysis of balance measures in a population-based cohort of older adults, we found that within the measures of COP-based static balance, the area of ellipse sway was much more strongly correlated with the ML than with AP measures of balance, while sway speed was more strongly correlated with the AP than with ML measures. Moreover, these static balance measures were weakly correlated with the clinic-based balance measures, while the clinic-based balance tests were strongly correlated with each other, including the one-leg stand test. The correlations when comparing sway speed or AP sway with ML sway, ML RMS, area of ellipse sway, and clinic-based measures of balance were higher in men than in women, even after accounting for differences in height.

Some of the correlation coefficients among the balance measures were modest (range 0.221–0.997 among the COP-based measures), although all were statistically significant, possibly due to the large sample size of the study population. Such modest correlation coefficients between the COP-based static balance measures and clinical balance measures suggest that static posture and clinic-based balance instruments may capture different aspects of balance. For example, postural sway may be more sensitive to sensorimotor function or impairment than clinical measures of balance (35). Furthermore, Frykberg et al (12) found poor correlations between total Berg Balance Scale scores with quiet stand COP-based force plate measures in a group of 20 subjects (mean age 50 years) who had experienced a stroke >6 months prior to study participation. Once Frykberg et al separated the Berg Balance Scale into components of maintaining a position (including standing or sitting unsupported or standing with eyes closed) and dynamic balance (including transfers or picking up objects), the correlation coefficient between the maintaining a position component and mean AP sway speed increased to 0.5. The authors reported that the maintaining a position component of the Berg Balance Scale would better mirror the static balance measures of the force plate. However, in our study, the SPPB, which should be similar to the maintaining a position component of the Berg Balance Scale, did not show a higher correlation with the COP-based static balance measures. These results imply that the laboratory- and clinic-based measures may possibly measure different aspects of balance, may complement each other, or one may be a poorer measure of balance than the other. It is also possible that one or the other might not truly be measuring balance. Further investigation of this matter appears to be warranted.

However, it is also possible that standing barefoot for static balance measures versus wearing shoes for the clinical balance measures may explain some of the differences between these measures. Although bare feet might allow greater sensory feedback, most researchers believe that wearing shoes improves balance (36) compared to bare feet, while others believe it is the type of shoes that may improve balance (37, 38). We are uncertain as to whether having bare feet or wearing shoes can entirely explain the lower correlation between COP-based static and clinical balance measures.

The high correlations among the various clinic-based balance tests were expected. First, 2 of the 3 SPPB balance items (side-by-side and tandem balance) were also part of the 14-item Berg Balance Scale (items 1 and 2). Also, the Berg Balance Scale was highly correlated with the timed one-leg stand test, possibly because the one-leg stand test was one of the most challenging tasks among all of the Berg Balance Scale items (9). It is entirely plausible that a participant's ability to successfully complete the one-leg stand test would enable a better overall Berg Balance Scale score.

However, in our study, it is unclear why men had higher associations than women when comparing AP sway or sway speed with ML and area of ellipse sways, as well as when comparing each of the clinical measures of balance with sway speed or area of ellipse sway, even after accounting for differences in height. While Bryant and colleagues (15) found no statistically significant differences between men and women in their mean COP-based ML and AP sways when the participants' eyes were opened (whether or not the data were normalized by height), we found statistically significant differences between men and women in their COP-based AP and ML sways with eyes opened, even after taking height into account (data not shown). There are several differences between our study and that from Bryant and colleagues. For each of our COP-based measures of balance, we used the mean of 5 trials, while Bryant and colleagues used the mean of 3 trials. Moreover, their study included 44 men and 53 women, while our study included 276 men and 489 women. Therefore, our study may have more stable estimates and better power to detect significant differences between men and women for some of these measures. Nevertheless, there may not be any biologic mechanism underlying the observed differences between men and women in the correlations between AP sway or sway speed with ML and area of ellipse sways, or between each of the clinical measures of balance with sway speed or ellipse sway.

The MOBILIZE Boston population was a group with good balance on average compared to a hospital- or nursing home–based group of older adults. It is therefore possible that differences in variability or strength of associations between COP-based static and clinical balance measures could be observed in populations with clinical conditions. It is unclear what other aspects of balance could explain the significant but moderate correlations between the static and clinical balance measures, and possible sex differences in the correlations between sway speed or AP sway with many of the other measures of balance, since normalizing by height did not change the results. Future research should explore explanations for these differences.

Our study has several limitations. First, we chose the 5 COP-based sway measures because they have been reported to be highly correlated with falls in older adults (19–24). It is possible that our results would have been different had we included other more comprehensive static balance measures. Nevertheless, as indicated in a systematic review by Ruhe et al (26), no single COP measure is more reliable than the others, but any sway balance measure should include both a parameter of distance (e.g., area of ellipse) and time-distance (e.g., velocity or sway speed), as we did in our study. In addition, our measures of COP-based sway did not account for differences in base support or width of feet since these data were not collected; however, all participants were instructed to stand with legs approximately hip width apart so there were no extremes of the base of support. It is possible that not standing with feet together may account for the higher correlation between ML laboratory measures, since the base of support affects ML sway differently than AP sway (39, 40). Finally, our study did not formally address measures of possible reduction in the burden of participants regarding time and effort in using the SPPB or the one-leg stand test as a sole clinical balance measure instead of the Berg Balance Scale or the COP-based measures. The SPPB or the one-leg stand test may be preferable to many clinicians, given that the one-leg stand test involves a single task and the balance component of the SPPB involves 3 items to complete, as compared to the 14 items in the Berg Balance Scale. However, it is unclear whether the one-leg stand test and the balance component of the SPPB can reliably measure balance in relation to an outcome. Reducing patient burden by limiting the number of tasks in a balance study is appealing in terms of time, cost, and retention of study participants. Clinical measures are more complex and integrated more measures of balance than laboratory-based measures (which are often used to investigate mechanisms). For our purposes, we view our results as being in agreement with choosing clinic-based measures of balance in future epidemiologic studies. Nonetheless, it is up to researchers to choose the appropriate clinic-based balance measure that will be sufficient for their study needs. Future studies could compare the validity and reliability of using as few measures of balance as possible to reduce the burden on elderly participants, in particular examining whether the one-leg stand test can be the sole clinical measure of balance in relation to a specific outcome.

Our study concurrently examined several comprehensive classes of balance measures using validated instruments, providing a unique opportunity to examine the associations between clinic- and laboratory-based balance measures. Our results suggest that there may be sex differences when comparing sway speed or AP sway with measures of sway in the ML direction and area of ellipse, and with each of the clinic-based measures of balance. Moreover, our study participants were from a population-based sample of community-dwelling older men and women. Results that are generalizable to healthy older adults aging in one's community may be more useful in the creation of a gold standard than findings from studies conducted in institutional settings.

The study results show strong agreements among clinic-based balance measures (Berg Balance Scale, SPPB, and the one-leg stand test) and among laboratory-based balance measures (AP with sway speed and ML with area of ellipse sway); however, agreements between clinic- and laboratory-based measures were modest, suggesting that these 2 types of measures may capture different aspects of balance and likely complement each other. Since neither consensus nor any guidance on how to choose these measures for research exists, further investigation on the relationships among these measures seems warranted.

AUTHOR CONTRIBUTIONS

All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Hannan had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Nguyen, Kiel, Hannan.

Acquisition of data. Galica, Hannan.

Analysis and interpretation of data. Nguyen, Kiel, Li, Galica, Kang, Casey, Hannan.

Acknowledgements

The authors acknowledge the MOBILIZE Boston research team and study participants for the contribution of their time, effort, and dedication. We sincerely thank Dr. Jennifer Kelsey for her critical review of the manuscript drafts.

Ancillary