- Email: [email protected]
Inertial sensor based reference gait data for healthy subjects
Gait & Posture 33 (2011) 673–678
Contents lists available at ScienceDirect
Gait & Posture journal homepage: www.elsevier.com/locate/gaitpost
Inertial sensor based reference gait data for healthy subjects Rene´ Schwesig a,*, Siegfried Leuchte a, David Fischer a, Regina Ullmann b, Alexander Kluttig c a
Department of Sport-Science, Martin-Luther-Universita¨t Halle-Wittenberg, Selke Str. 9F, 06099 Halle, Saale, Germany Motion Analysis Lab, Children‘s Hospital, Rorschacherstr. 168, 9006 St. Gallen, Switzerland c Institute of Medical Epidemiology, Biostatistics and Informatics, Martin-Luther-Universita¨t Halle-Wittenberg, Magdeburger Str. 8, 06097 Halle, Saale, Germany b
A R T I C L E I N F O
A B S T R A C T
Article history: Received 13 July 2010 Received in revised form 22 February 2011 Accepted 25 February 2011
Inertial sensor gait analysis systems (ISGAS) facilitate gait analysis in an unobstructed environment outdoors or outside of a conventional gait laboratory. However, their use in clinical settings and in large scale studies necessitates thorough evaluation of their performance in different settings and populations and reference data on healthy subjects. The purposes of this study were to obtain spatio-temporal gait parameters using a large cohort of subjects of all ages and to identify relationships between gait parameters and subject characteristics. An inertial sensor based system (RehaWatch1, HASOMED1) was used to collect gait data for 1860 healthy subjects (919 men; aged 5–100 years). Following two practice trials, data of one trial were collected for each subject while walking on a 20 m walkway. Spatiotemporal gait parameters were calculated and normalized to body height. Demographic and morphological data including age, gender, body height and body mass were recorded. Multifactorial regression models were used to evaluate determinants of different gait parameters. Strong non-linear relationships between predictors (age) and gait parameters were identified. Overall, the predictors explained the largest portion of variance for stride length (R2 = 0.46). Normalized cadence showed one peak across all ages. Normalized walking speed and normalized stride length showed two peaks across all ages. The largest and smallest variations across the ages were observed for normalized walking speed (98%) and for normalized stride length (89%), respectively. This reference database is the foundation for future evaluations of gait disorders in patients of all ages and has been integrated in the RehaWatch1 system. ß 2011 Elsevier B.V. All rights reserved.
Keywords: Gait analysis Life span Reference data Inertial sensor
1. Introduction Gait analysis is a useful tool for surgical planning and for evaluating factors associated with pathological gait, rehabilitation and treatment interventions. While conventional gait analysis provides detailed information on kinematic and kinetic parameters, these systems are limited to laboratory use. In contrast, inertial sensor gait analysis systems (ISGAS) facilitate gait analysis in an unobstructed environment outdoors and outside of a conventional gait laboratory. However, their use in clinical settings and in large scale studies necessitates thorough evaluation of their performance. In addition, reference data on healthy subjects of all ages are needed to allow for the detection of, sometimes subtle, gait impairments in patients. For instance, Auvinet et al. [1] showed that a custom-made portable biaxial-accelerometer setup provided accurate reference data for common gait parameters during walking. Most studies reported gait data for small groups of healthy subjects [1–10]. While reference gait data for children and
* Corresponding author. Tel.: +49 0345 552 4450; fax: +49 0345 552 7054. E-mail address: [email protected] (R. Schwesig). 0966-6362/$ – see front matter ß 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.gaitpost.2011.02.023
adolescents are available [11–17], the primary focus of most of these studies was the speed dependency of gait parameters rather than on the establishment of a reference database. One of the largest cohorts reporting data across the ‘‘life span’’ (age range: 20– 85 years) consisted of gait data for 293 subjects [18,19] and showed clear changes in gait patterns in subjects aged 70 years and older. In most studies, complex gait analysis systems have been used for the measurement of spatio-temporal gait parameters [3,10,15,16]. In comparison, direct inertial sensor based systems have been used more scarcely [1,8]. In general, populations in previous studies providing reference gait data were too small to generalize gait patterns for all age groups. The purposes of this study were to obtain spatio-temporal gait parameters in a large cohort of subjects of all ages and to identify relationships between gait parameters and subject characteristics. 2. Methods 2.1. Subjects 1860 healthy subjects (919 men, 941 women) were recruited from local schools, sports clubs, sports groups, companies and institutions. All tests were conducted at these respective locations and on a hard flat surface (hardwood floor, asphalt, sport surface) between 7.30 am and 9.00 pm from March to December. Prior to data
674
R. Schwesig et al. / Gait & Posture 33 (2011) 673–678
collection, sensors were warmed up for approximately 5 min to facilitate a stable operation temperature of 31 8C during data collection. The study was approved by the local ethics committee. All participants or parents of young participants, respectively, gave their informed consent. They were required to have no history of musculoskeletal or neurological diseases, gait disorders or any painful condition that could affect their gait. Subjects were excluded if they presented with apparent marked pelvic asymmetry or scoliosis, or if they took medications that could have affected their gait. Body height and body mass were measured using a personal scale (WPT100/200 OW; RADWAG, Hilden, Germany). All tests were conducted by the same investigator. 2.2. Gait analysis Gait data were collected using a mobile inertial sensor based system (RehaWatch1 system; HASOMED1 GmbH, Magdeburg, Germany). The sensors were attached to the lateral ankle using a special device (Fig. 1) [20]. Each sensor contains three accelerometers and three gyroscopes (Analog Devices, Norwood, MA, USA) measuring foot motion in six degrees of freedom. The measurement range of the accelerometers and gyroscopes is 5 g and 6008/s (sampling rate: 512 Hz), respectively. Gait events were determined using offline evaluation of the gait data [21]. The primary gait event was heel-strike. All other gait events (including full contact, heel-off, and toe-off) were identified relative to heel-strike. Based on these gait events, spatio-temporal gait parameters and gait phases were calculated automatically, including kinematic parameters (e.g. stride length, foot height, and walking speed), gait phases [22,23] and symmetry [24]. In addition to standard gait parameters including cadence and walking speed, RehaWatch1 measures parameters describing gait patterns (e.g. foot height, and stance phase symmetry) that are not readily measured using step monitors and, in contrast to conventional motion capture systems, it is portable. The measurement system consists of a mobile computer (DataLogger) that is attached to the patient using straps (Fig. 1). Data of the inertial sensors are received via cable-connection and temporarily stored in the DataLogger. The data transfer from the DataLogger to the PC is performed subsequently using an USB-stick. Measured data comprises accelerations and angular velocities used to calculate spatio-temporal parameters. Angular velocities were integrated to obtain information on the spatial orientation of the sensor. Sensor position is calculated
by twice integrating acceleration data while taking into account the spatial orientation of the sensor. To avoid drift, the integration process is reset following each step while factoring in the boundary conditions. Hence, the sensor velocity is forced to equal zero at the end of the step and final foot height must match initial foot height. High intra-observer reliability (ICC-range: 0.691–0.959) of the inertial sensor based system has been reported previously [20]. Briefly, data for 44 healthy subjects (age: 27.7 4.2 years) were collected in three sessions per subject (sessionto-session time: 48 h). Each subject performed three trials per session while walking for 20 m. General linear models did not reveal significant differences in gait parameters between sessions. In addition to variance and correlation analyses, Bland–AltmanPlots suggested a high reliability of the device. The validity of the RehaWatch1 system was tested in an unpublished study. Ten healthy subjects (six females and four males, mean age 27.2 9.2 years, range: 5.9– 36.8 years) were recruited for this study. Subjects performed level walking at their preferred self selected walking-speed along a 12 m-walkway (repeated trials, n = 10). All repeated measurements were recorded simultaneously with the inertial based and an electro-optical system (VICON, Oxford Metrics Ltd., Oxford, UK). The ‘‘Plug-In-Gait’’ [25] marker set was used consisting of 16 reflective markers attached to anatomical landmarks using a standardized protocol. High validity (ICC-range: 0.776–0.991) was found for cadence, stride length, stride time and walking speed. Lowest test–retest correlations (ICC < 0.75) were found for time in double support and time in single support. Hence, this study concentrates on the spatio-temporal parameters. Gait analysis depends on three main test conditions: test distance, walking speed and the shoes worn by the subjects [2]. Self-selected speed appears to be the optimal condition for gait analysis and for providing functionally relevant data [1]. In this study, subjects wore their personal walking shoes. Subjects were asked to walk straight at their self-selected speed toward a target at a 20 m-distance to ensure that a sufficient number of stable walking cycles were recorded between the start and the end of the test. Each subject performed two walking trials to adjust to the test conditions while wearing the ISGAS. Data of the third trial were used for further analysis. 2.3. Statistical analysis All statistical analyses were performed using SPSS version 18.0 for Windows (SPSS Inc., Chicago, IL, USA) and SAS version 9.2 (SASInstitute, Cary, NC, USA). Sex differences of gait parameters were tested using a general linear model. Differences of means were considered statistically significant if p-values were less than 0.05 and partial eta-squared (h2)-values were greater than 0.05. Because of the large number of cases, decisions of significance were made primarily based on h2-values. The SAS quantreg procedure was used for modeling the effects of age on the conditional quantiles (5th, 10th, 25th, 50th, 75th, 90th and 95th) of different gait parameters by means of quantile regression analysis. Quantile regression analysis [26] extends the regression model to conditional quantiles of the response variable, such as the median or the 10th and 90th percentiles. A priori defined third-order polynomials were used to calculate curve fitting regression lines. Walking speed, stride length and cadence are strongly dependent on leg length. However, because leg length was not recorded in this study, gait parameters were normalized to body height/mean body height separately for each gender (Table 1):
Table 1 Demographic and morphological characteristics of subjects (n = 1860) categorized by decades of age and gender. All values are means (one standard deviations). Age (years) All subjects 5–100
Fig. 1. Photograph of a subject wearing the inertial sensor gait analysis system (RehaWatch1, HASOMED1 GmbH): (1) data logger attached to a belt, (2) sensor cable, fixed below the knee with an elastic belt, and (3) inertial sensors (6DOF) attached to a device fixed to each foot device.
Men/women
M (n = 919) W (n = 941) Subjects by age group 5.0–10.0 M (n = 112) W (n = 110) 10.1–15.0 M (n = 115) W (n = 115) 15.1–20.0 M (n = 111) W (n = 113) 20.1–30.0 M (n = 166) W (n = 166) 30.1–40.0 M (n = 93) W (n = 94) 40.1–50.0 M (n = 100) W (n = 102) 50.1–60.0 M (n = 79) W (n = 83) 60.1–70.0 M (n = 83) W (n = 86) 70.1–100.0 M (n = 60) W (n = 72)
Body mass (kg)
Body height (m)
BMI (kg/m2)
70.6 (22.7) 58.3 (17.0)
1.71 (0.17) 1.61 (0.13)
23.5 (4.98) 22.2 (4.81)
31.1 28.9 44.9 43.5 72.8 60.0 78.8 62.7 86.2 66.8 86.0 68.8 86.6 69.0 82.3 67.9 79.4 64.7
1.36 1.34 1.55 1.54 1.80 1.68 1.82 1.68 1.82 1.68 1.80 1.67 1.77 1.64 1.74 1.63 1.72 1.58
16.6 16.1 18.5 18.0 22.4 21.3 23.9 22.1 26.1 23.6 26.5 24.8 27.6 25.6 27.1 25.4 26.8 25.8
(7.17) (5.77) (12.4) (12.0) (14.0) (9.77) (10.6) (8.98) (12.9) (11.9) (13.0) (13.8) (12.5) (12.7) (13.0) (11.0) (12.3) (10.6)
(0.08) (0.04) (0.12) (0.11) (0.07) (0.06) (0.07) (0.06) (0.06) (0.06) (0.08) (0.06) (0.07) (0.07) (0.07) (0.05) (0.06) (0.06)
(2.46) (2.26) (3.21) (3.39) (3.57) (3.06) (2.59) (2.85) (3.69) (4.01) (3.35) (4.79) (3.84) (4.18) (3.72) (3.75) (3.50) (4.12)
R. Schwesig et al. / Gait & Posture 33 (2011) 673–678 normalized cadence = cadence sqrt (body height/mean body height), normalized stride length = stride length/(body height/mean body height), and normalized speed = speed/sqrt (body height/mean body height). A sex-stratified multifactorial regression model was used for each gait parameter to identify predictive variables for each gait parameter. Linear, quadratic and cubic terms were used for age, body height and body mass, respectively. A backward selection method with a significance level of 0.10 was used to find the best prediction model. Bivariate two-sided Pearson correlation was used to detect relationships between spatio-temporal parameters.
3. Results 3.1. Subjects Demographic and morphological data for all subjects are given in Table 1. The 1860 subjects included 941 women and were aged between 5 and 100 years (Table 1). The number of steps used for gait analysis ranged from 19 to 65. 3.2. Age, body height, body mass, gender and gait parameters A strong non-linear relationship between predictors (age, body height, and body mass) and gait parameters was found (Table 2). In combination, the predictors explain the greatest portion of variance for stride length (men: R2 = 0.47; women: R2 = 0.45). Men had longer stride times and walked at a lower cadence than women (mean standard deviation of stride time: men: 1.07 0.08 s; women: 1.01 0.08 s; h2 = 0.111; cadence: men: 113 9 steps/min; women: 119 9 steps/min; h2 = 0.136). This gender difference was observed in subjects aged 15 years or older. 3.3. Reference data for gait parameters Curves fitted to normalized cadence (Fig. 2a and b) data over age showed a clear difference between men and women. Male subjects in the 50th percentile showed a smaller variability than female subjects (men: 50th percentile = 109–115 steps/min; women: 50th percentile = 99–126 steps/min). Until the age of 65 years, normalized cadence slightly increased from 109 steps/
675
min to about 115 steps/min (interdecile range: 97–124 steps/ min) for men and from 113 steps/min to about 125 steps/min (interdecile range: 99–134 steps/min) for women. Between 65 and 90 years cadence declined from 115 to 109 steps/min (interdecile range: 94–123 steps/min) for men and from 125 to 99 steps/min (interdecile range: 95–134 steps/min) for women. Normalized stride length showed a two peak pattern (Fig. 2c and d), and non-normalized (Fig. 3c) stride length showed a single peak pattern. The age of occurrence of minimum and maximum normalized stride length for male subjects was 90 and 6 years, respectively (50th percentile: minimum = 0.95 m; maximum = 1.60 m) and for female subjects 100 and 5 years, respectively (50th percentile: minimum = 0.88 m; maximum = 1.48 m). Stride time data over age showed a local maximum and a local minimum (Fig. 3b). Stride time increased from the age of 5 years (50th percentile: 0.96 s), reached a local maximum at the age of 28 years (50th percentile: 1.07 s), decreased to a local minimum at the age of 64 years (50th percentile: 1.00 s) and increased beyond the local minimum until the age of 100 years (50th percentile: 1.31 s) (Fig. 3b). While non-normalized walking speed showed a single peak pattern (Fig. 3a), walking speed normalized for body height showed two peaks for male and female subjects (Fig. 2e and f). The age of occurrence of the maxima during childhood was the same for male and female subjects (male: 1.37 m/s at 6 years; female: 1.32 m/s at 5 years). Walking speed decreased during adolescence (male: 1.26 m/s at 22 years; female: 1.30 m/s at 13 years). The second maximum was reached during adulthood (male: 1.38 m/s at 56 years; female: 1.43 m/s at 55 years). Especially above the age of 70 years, the decrease in walking speed occurs more rapidly: male subjects reach their minimum of 0.79 m/s at 90 years, while female subjects reach their minimum of 0.64 m/s at 100 years. The largest and smallest variations (percent difference between minimum and maximum of the 5th and 95th percentiles) across the ages were found for normalized walking speed (98%) and normalized stride length (89%), respectively.
Table 2 Results of multifactorial regression model (method: backward selection) for spatio-temporal gait parameters by sex. Dependent variable
Walking speed (m/s)
Stride time (s)
Stride length (m)
Cadence (steps/min)
Men
Women
Predictor variables
R2
Predictor variables
R2
Body height squ Age lin Age squ Age cb Body mass squ Body height cb Age lin Age squ Age cb Body mass squ Body height squ Age lin Age squ Age cb Body mass squ Body mass cb Body height lin Age lin Age squ Age cb Body mass squ
0.171
Body height lin Age lin Age squ Age cb Body mass squ Body height lin Age lin Age squ Age cb Body mass squ Body height lin Age squ Age cb Body mass squ Body mass cb
0.291
Body height lin Age lin Age squ Age cb Body mass squ
0.172
List of abbreviations: lin, linear; squ, squared; cb, cubed. Units of predictor variables: m (body height); years (age); kg (body mass).
0.223
0.472
0.243
0.156
0.445
R. Schwesig et al. / Gait & Posture 33 (2011) 673–678
676
Normalized cadence = cadence * sqrt(body height/mean body height) a) men
b) women 170
150
..
Normalized cadence [steps/min]
Normalized cadence [steps/min]
170
.
. . . . . . . . .... .. . ...... .. .. . . . ... .. . .. .. ... .. .. . . . ... . . . . . .. . . .. ............... .................................... .. . ...................... ......... .. .. ............ ..................... ... . ....... ...... ......... ........... ........ ....... ... ....... ... .. .. ... ......... ... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ................... ......................... .... . .............. ....... .. ............. . .... .. ..... .... . . .................... ............ ............ . .. .. .... .. .. ... . ... . .. . .. .... . . . . . ............ .. ...... ..... ..... ..... .. . . . . .... . . . .... . .. ............. .. .. .. . . . . . .... . . . ...... .. . .. .
130
110
90
.
70 0
10
150
. .. . .... . . . .. . . . .. .. . . . .. .. .. .. . ... .... .... ....... ....... ..... ............... ...... .. . . .. . . ............. . . . ... .. . .. .. .... ..... . ..... ...... . .. . . .. .. ... .... . ....... .... .. . ....... ....... ..... . ... .... . ...... .. ..... . . . ............. ............... .......... ........ . ..... ..... .. . .. ... . .... ... ... . ..... . . . .................................................................. ......... .......... ....... ................ ............... ... . .... ... ... . . . . . .. . ... ...... . ...................... ........................... .. ... . . . . . . ......... . ....... ..... . .. . . . .. . . . .. . . . ................ .. .... .. . .. . . . . . . . ...... .. . . . . . . . .. .. . . . .. . . .
130
110
90
70 20
30
40
50
60
70
80
90
100
0
10
20
30
40
Age [years]
50
60
70
80
90
100
Age [years]
Normalized stride length = stride length / (body height/mean body height) c) men 2.1
2.1
. . . . . .. . . . .. . . .......... . ... . . . . ... . . ... . . . ... .. . . . ............... ........... ..... . .. .. .. ......... ... . ... .... .. .. . ....... . .......................................................... ......................... .......................... .... ............................ .. . . ................ .......................................... ....... ................. .. . ...... .... ...... ............. . . ... ... ... .. ... .. .. .. . .. . . .. .... .. . ...... .... . . . .. ............... ........................ .. .. . . .. ..... ... .... . . ... . ... ..... ... ... ..... . . ... .. . . . ............ ............. ...... ...... . ... . . . . . . .. . . .. . .. . . . . .
1.7 1.5 1.3 1.1 0.9 0.7
. . . . . .. . .. . . . . ... . . . . . .. . .. . .. .. . ...................... ................... ....... ... ...... ........ ... .................. ....... ... ... . . . . . .................................................... ........... . ................ ....... .. .... ..... ............... ... ... .. .................................................................................................................................. ............................................ .. .. .. ..... . . . . ............ ................... .. . .. ...... ... .............. . ....... .. . .. ...... .. . . .. ... .. . ........ . . . . .. . .. . .. . . . . . . . .. .. ... .. . . . . . . . . ... . .
1.9 1.7 Normalized stride length [m]
1.9
Normalized stride length [m]
d) women
1.5 1.3 1.1 0.9 0.7
0.5
0.5
0.3
0.3
0.1
.
0.1 0
10
20
30
40
50
60
70
80
90
100
0
10
20
30
40
Age [years]
50
60
70
80
90
100
Age [years]
Normalized speed = speed / sqrt(body height/mean body height) e) men . . . .. . . . . . . . .... . . . . . . . .......... ... . .. . .. . .. . .. . ... .. .. ... . . . . . . . . . . ...... ........ . ... .. . ... . ...... . . ... .. .... . . . . . ..... . . . .. . . . . ................... .. ................ ....... ......... ......................... ...... .... ...................... . .. .............. .. ............... ......... ...... .......................... . ... ................. ......... .. ......................... ....... ...... ... . ..... . ... .... . . ... .. . . . . ............. .. ........................................... . .. .... .. ...... ..... .... ...... ... . .... . . .... . .. ................ . . . .... .. . . . ..... .. ... ...... .... . ............ ...... . ........ .... . . . . . . .. . . . . . . . . . . .. . . ... . .. . .. . . . . . .. . .. ... . . . . . . . . . . .
1.8 1.6 1.4 1.2 1.0 0.8 0.6
2.0
. . . . . . .. . . . .. . .. . . . .. . .. . .... . .. . . . ... ... ... .. .. ... .. .... .. . . ..... .. .... .. ..... ... .. .... .. . ............ . ....................... . ..... ................. ........ . . ... ........ ............ .......... . ........... ... ..................... ..... ........ . .. ..... ... .. ..... . ..... ...... .. . . .......................................................................... ........... ......... .. ......... .............. ... .... .... . . . . . . . . . . . . . . . . . ............................................................ .. ..... ... .. ...... ...... ........... . ... .. .... ..... . . ............... ............ .. . . .... ....... ... . ... . . ... .. . .. .. ... . .. . . . .. . ... .. . . .. .. . . . ... . .. . . . ... . .. ... . . . . . . .. .. .. . . .. . . .. .
1.8 1.6 Normalized speed [m/s]
2.0
Normalized speed [m/s]
f) women
1.4 1.2 1.0 0.8 0.6
0.4
0.4 0
10
20
30
40
50
60
70
80
90
100
Age [years]
0
10
20
30
40
50
60
70
80
90
100
Age [years]
Fig. 2. (a–f) Age and sex dependent presentation of the body height normalized cadence (a and b), normalized stride length (c and d) and normalized speed (e and f) based on quantile regression analysis. Various lines describe the 5th, 10th, 25th, 50th, 75th, 90th and 95th quantiles (bottom-up).
4. Discussion In this study, reference data for gait parameters across all ages were presented. Results of the quantile regression analysis (Figs. 2a–f, 3a–c) showed that analyzing age and gender aspects
using age groups (e.g. decades) is not sufficient. Because of the large variation within each age group especially during childhood and adolescence and at advanced age, an ordinal scale for the predicting parameter age should not be used. Hence, comparison of this data set with results of previous studies [1,8] prove
R. Schwesig et al. / Gait & Posture 33 (2011) 673–678
677
Fig. 3. (a–c) Age dependent presentation of (a) walking speed, (b) stride time and (c) stride length across the entire life span using quantile regression analysis. In addition, bivariate two-sided Pearson correlations are given. Various lines describe the 5th, 10th, 25th, 50th, 75th, 90th and 95th quantiles (bottom-up).
challenging because those studies typically reported mean values for age groups (decades) or gender specific mean values. In general and in agreement with many previous studies [2,6,8,10], men took larger steps (stride length: women: 1.42 m; men: 1.47 m; h2 = 0.021), showed longer stride time (women: 1.01 s; men: 1.07 s; h2 = 0.111) and had lower cadence than women (women: 119 steps/min; men: 113 steps/min; h2 = 0.109). In particular, middle-aged women tend to walk at a higher cadence compared to men. To date, the largest studies on reference gait data [18,19] with respect to size of study population and age range reported a mean self-selected walking speed of 1.37 m/s (women: 1.28 m/s; men: 1.43 m/s) across all ages (20–85 years). In our study, nonnormalized mean walking speed across all ages was 1.31 m/s (women: 1.33 m/s; men: 1.30 m/s) and was within the range reported by Murray et al. [18,19]. In general, women walked slightly faster than men. However, in contrast to previous studies, the gender difference in this study was not significant (h2non-normalized ¼ 0:004; h2normalized ¼ 0:004). According to Kirtley et al. [27], a self-selected walking speed of less than 1 m/s indicates a disability or pathology. In our study, non-normalized walking speed was lower than 1 m/s only in subjects aged 93 years or older (50th percentile: 0.94 m/s). In addition, walking speed lower than 1 m/s was only observed in the 5th percentile (age 16 and 69 years), 10th percentile (age 7 and 73 years) and 25th percentile (age 80 years). Hence, the results of our study support the use of walking speeds of 1.0 m/s and lower as an indicator for disability or pathology. Normal mean (non-normalized) stride length of 1.41 m as reported by
Murray et al. [18,19] was reached already at the age of 16 years (50th percentile) with an average stride length of 1.45 m across all ages (men: 1.47 m; women: 1.42 m). The same investigators reported a norm value for non-normalized cadence of 113 steps/ min. In our study, a mean non-normalized cadence across all ages of 116 steps/min was found (men: 113 steps/min; women: 119 steps/min). A cadence of 113 steps/min was reached at 22 years and at 81 years (50th percentile). Results of the multifactorial regression analysis stratified by gender (Table 2) and those of the quantile regression analysis (Figs. 2a–f and 3a–c) suggest a non-linear influence of the predictors on gait parameters. In addition, the results of our study (r = 0.076) did not support earlier reports [1,8] of age dependency of walking speed (r = 0.58 in women, r = 0.49 in men). Similar observations were made for stride length (r = 0.119) and cadence (r = 0.060) where relevant correlations were not found. It is possible that these discrepancies between results are caused by difference in age range between studies (20–86 years and 20–98 years vs. 5–100 years) and by the lack of information on children, adolescents and the elderly in previous studies. Ashton-Miller [28] observed a 1%-decrease in walking speed per year in persons above the age of 60 years. Those results support our observation. Between the age of 60 years (50th percentile: 1.40 m/s) and the age of 90 years (50th percentile: 1.01 m/s), walking speed decreased by 28% corresponding to a yearly change of 0.93%. However, when considering the decline in walking speed from the maximum walking speed which was found for subjects at 50.1 years (50th percentile: 1.42 m/s), the annual decline in walking speed would be 0.73%. It is important to note
678
R. Schwesig et al. / Gait & Posture 33 (2011) 673–678
that all of these values are mean values and that there is no linear relationship between walking speed and age. Moreover, above the age of 70 years, the decrease in cadence, stride length and walking speed and increase in stride time occurs more rapidly. Grieve and Gear [29] observed a very strong linear relationship between cadence and walking speed (r > 0.95) within one subject. In contrast, in our study we found a moderate relationship between cadence and the square root of walking speed (r = 0.544) for the entire population. A strong relationship between stride length and the square root of walking speed was found (r = 0.843). These relationships were independent of age and gender (rage = 0.118; rsex = 0.084). Possible reasons for the discrepancies between the results of these studies include different age ranges (1–35 years vs. 5–100 years) and different methods employed in these studies. The clinical importance of our study is the extensive data base that provides reference gait data that may be used in the diagnosis or evaluation of orthopaedic or neurological patients with conditions, such as osteoarthritis, and/or diseases, such as cerebral palsy or Parkinson’s disease. Frequently, information on spatiotemporal parameters is sufficient for these patient populations, especially when considering the trade-off between effort and time spent on data collection and processing and the limited additional information as it relates to the specific disorder or pathology. 5. Conclusion The presented reference data on gait parameters of men and women aged 5–100 years (self-selected walking speed, 20 m straight corridor, subjects wearing their own comfortable shoes) forms the foundation for future studies that aim to evaluate gait pathologies and disorders of older adults. In particular, specific gait parameters can now be more precisely evaluated not only in comparison to age groups but also at specific time points throughout a person’s life span. The inclusion of additional data especially on subjects above the age of 70 years is desirable to improve the precision of the information extracted from the database. The data collected in this study has been integrated as reference database into the RehaWatch1 gait analysis system. Since there are no data on the comparison of different gait analysis systems the assignability and comparability of the reference data is limited. Further studies including different gait analysis systems are necessary to provide information about reliability and comparability of gait parameters. Acknowledgements The authors thank Christoph Hintze for his help with data acquisition and Dr. Annegret Mu¨ndermann for her writing assistance. The authors would like to express their appreciation to all participating institutions (n = 62) and subjects (n = 1860) for their participation in this study. Funding for this study was provided by Investitionsbank Sachsen-Anhalt, Germany. Conflict of interest statement The authors do not have any conflicts of interest.
References [1] Auvinet B, Berrut G, Touzard C, Moutel L, Collet N, Chaleil D, et al. Reference data for normal subjects obtained with an accelerometric device. Gait Posture 2002;16:124–34. ¨ berg T, Karszina A, O ¨ berg K. Basic gait parameters: reference data for normal [2] O subjects, 10–79 years of age. J Rehabil Res Dev 1993;30:210–23. [3] Nigg BM, Fisher V, Ronsky JL. Gait characteristics as a function of age and gender. Gait Posture 1994;2:213–20. [4] Ostrosky KM, VanSwearingen JM, Burdett RG, Gee Z. A comparison of gait characteristics in young and old subjects. Phys Ther 1994;74:637–44. [5] Blanc Y, Balmer C, Landis T, Vingerhoets F. Temporal parameters and patterns of the foot roll over during walking: normative data for healthy adults. Gait Posture 1999;10:97–108. [6] Cho SH, Park JM, Kwon OY. Gender differences in three dimensional gait analysis data from 98 healthy Korean adults. Clin Biomech (Bristol Avon) 2004;19:145–52. [7] Kimura T, Kobayashi HNE, Hanaoka M. Effects of aging on gait patterns in the healthy elderly. Anthropol Sci 2007;115. [8] Senden R, Grimm B, Heyligers IC, Savelberg HH, Meijer K. Acceleration-based gait test for healthy subjects: reliability and reference data. Gait Posture 2009;30:192–6. [9] Menz HB, Latt MD, Tiedemann A, Mun San KM, Lord SR. Reliability of the GAITRite walkway system for the quantification of temporo-spatial parameters of gait in young and older people. Gait Posture 2004;20: 20–5. [10] Winter DA, Patla AE, Frank JS, Walt SE. Biomechanical walking pattern changes in the fit and healthy elderly. Phys Ther 1990;70:340–7. [11] Norlin R, Odenrick P, Sandland B. Development of gait in the normal child. J Pediatr Orthop 1981;1:261–6. [12] Todd FN, Lamoreux LW, Skinner SR, Johanson ME, St HR, Moran SA, et al. Variations in the gait of normal children. A graph applicable to the documentation of abnormalities. J Bone Joint Surg Am 1989;71:196–204. [13] Wheelwright EF, Minns RA, Law HT, Elton RA. Temporal and spatial parameters of gait in children. I: normal control data. Dev Med Child Neurol 1993;35:102–13. [14] Sutherland TH, Olshen R, Biden EN, Wyatt MP. The development of mature walking. London: MacKeith Press; 1988. [15] Schwartz MH, Rozumalski A, Trost JP. The effect of walking speed on the gait of typically developing children. J Biomech 2008;41:1639–50. [16] Samson W, Desroches G, Cheze L, Dumas R. 3D joint dynamics analysis of healthy children’s gait. J Biomech 2009;42:2447–53. [17] Beck RJ, Andriacchi TP, Kuo KN, Fermier RW, Galante JO. Changes in the gait patterns of growing children. J Bone Joint Surg Am 1981;63:1452–7. [18] Murray MP, Drought B, Kory RC. Walking patterns of normal men. J Bone Joint Surg Am 1964;46:335–60. [19] Murray MP, Kory RC, Sepic SB. Walking patterns of normal women. Arch Phys Med Rehabil 1970;51:637–50. [20] Schwesig R, Kauert R, Wust S, Becker S, Esperer HD, Leuchte S. Reliability of the novel gait analysis system RehaWatch. Biomed Tech 2010;55:115. [21] Negard NO, Kauert R, Andres S, Schauer T, Raisch J. Gait phase detection and step length estimation of gait by means of inertial sensors. In: 3rd European medical & biological engineering conference, IFMBE European conference on biomedical engineering. 2005. p. 11. [22] Perry J. Ganganalyse. Norm and Pathologie des Gehens. Mu¨nchen, Jena: Urban & Fischer; 2003. [23] Beckers D, Deckers J. Ganganalyse and Gangschulung. Therapeutische Strategien fu¨r die Praxis. Berlin u.a.: Springer; 1997. [24] Robinson RO, Herzog W, Nigg BM. Use of force platform variables to quantify the effects of chiropractic manipulation on gait symmetry. J Manipulative Physiol Ther 1987;10:172–6. [25] Kadaba MP, Ramakrishnan HK, Wootten ME. Measurement of lower extremity kinematics during level walking. J Orthop Res 1990;8:383–92. [26] Koenker R, Bassett G. Regression quantiles. Econometrica 1978;46:33–50. [27] Kirtley C, Whittle MW, Jefferson RJ. Influence of walking speed on gait parameters. J Biomed Eng 1985;7:282–8. [28] Ashton-Miller JA. Age-associated changes in the biomechanics of gait and gaitrelated falls in older adults. In: Hausdorff JM, Alexander NB, editors. Gait disorders: evaluation and management. Boca Raton: Taylor & Francis; 2005. p. 63–100. [29] Grieve DW, Gear RJ. The relationships between length of stride, step frequency, time of swing and speed of walking for children and adults. Ergonomics 1966;9:379–99.
Contents lists available at ScienceDirect
Gait & Posture journal homepage: www.elsevier.com/locate/gaitpost
Inertial sensor based reference gait data for healthy subjects Rene´ Schwesig a,*, Siegfried Leuchte a, David Fischer a, Regina Ullmann b, Alexander Kluttig c a
Department of Sport-Science, Martin-Luther-Universita¨t Halle-Wittenberg, Selke Str. 9F, 06099 Halle, Saale, Germany Motion Analysis Lab, Children‘s Hospital, Rorschacherstr. 168, 9006 St. Gallen, Switzerland c Institute of Medical Epidemiology, Biostatistics and Informatics, Martin-Luther-Universita¨t Halle-Wittenberg, Magdeburger Str. 8, 06097 Halle, Saale, Germany b
A R T I C L E I N F O
A B S T R A C T
Article history: Received 13 July 2010 Received in revised form 22 February 2011 Accepted 25 February 2011
Inertial sensor gait analysis systems (ISGAS) facilitate gait analysis in an unobstructed environment outdoors or outside of a conventional gait laboratory. However, their use in clinical settings and in large scale studies necessitates thorough evaluation of their performance in different settings and populations and reference data on healthy subjects. The purposes of this study were to obtain spatio-temporal gait parameters using a large cohort of subjects of all ages and to identify relationships between gait parameters and subject characteristics. An inertial sensor based system (RehaWatch1, HASOMED1) was used to collect gait data for 1860 healthy subjects (919 men; aged 5–100 years). Following two practice trials, data of one trial were collected for each subject while walking on a 20 m walkway. Spatiotemporal gait parameters were calculated and normalized to body height. Demographic and morphological data including age, gender, body height and body mass were recorded. Multifactorial regression models were used to evaluate determinants of different gait parameters. Strong non-linear relationships between predictors (age) and gait parameters were identified. Overall, the predictors explained the largest portion of variance for stride length (R2 = 0.46). Normalized cadence showed one peak across all ages. Normalized walking speed and normalized stride length showed two peaks across all ages. The largest and smallest variations across the ages were observed for normalized walking speed (98%) and for normalized stride length (89%), respectively. This reference database is the foundation for future evaluations of gait disorders in patients of all ages and has been integrated in the RehaWatch1 system. ß 2011 Elsevier B.V. All rights reserved.
Keywords: Gait analysis Life span Reference data Inertial sensor
1. Introduction Gait analysis is a useful tool for surgical planning and for evaluating factors associated with pathological gait, rehabilitation and treatment interventions. While conventional gait analysis provides detailed information on kinematic and kinetic parameters, these systems are limited to laboratory use. In contrast, inertial sensor gait analysis systems (ISGAS) facilitate gait analysis in an unobstructed environment outdoors and outside of a conventional gait laboratory. However, their use in clinical settings and in large scale studies necessitates thorough evaluation of their performance. In addition, reference data on healthy subjects of all ages are needed to allow for the detection of, sometimes subtle, gait impairments in patients. For instance, Auvinet et al. [1] showed that a custom-made portable biaxial-accelerometer setup provided accurate reference data for common gait parameters during walking. Most studies reported gait data for small groups of healthy subjects [1–10]. While reference gait data for children and
* Corresponding author. Tel.: +49 0345 552 4450; fax: +49 0345 552 7054. E-mail address: [email protected] (R. Schwesig). 0966-6362/$ – see front matter ß 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.gaitpost.2011.02.023
adolescents are available [11–17], the primary focus of most of these studies was the speed dependency of gait parameters rather than on the establishment of a reference database. One of the largest cohorts reporting data across the ‘‘life span’’ (age range: 20– 85 years) consisted of gait data for 293 subjects [18,19] and showed clear changes in gait patterns in subjects aged 70 years and older. In most studies, complex gait analysis systems have been used for the measurement of spatio-temporal gait parameters [3,10,15,16]. In comparison, direct inertial sensor based systems have been used more scarcely [1,8]. In general, populations in previous studies providing reference gait data were too small to generalize gait patterns for all age groups. The purposes of this study were to obtain spatio-temporal gait parameters in a large cohort of subjects of all ages and to identify relationships between gait parameters and subject characteristics. 2. Methods 2.1. Subjects 1860 healthy subjects (919 men, 941 women) were recruited from local schools, sports clubs, sports groups, companies and institutions. All tests were conducted at these respective locations and on a hard flat surface (hardwood floor, asphalt, sport surface) between 7.30 am and 9.00 pm from March to December. Prior to data
674
R. Schwesig et al. / Gait & Posture 33 (2011) 673–678
collection, sensors were warmed up for approximately 5 min to facilitate a stable operation temperature of 31 8C during data collection. The study was approved by the local ethics committee. All participants or parents of young participants, respectively, gave their informed consent. They were required to have no history of musculoskeletal or neurological diseases, gait disorders or any painful condition that could affect their gait. Subjects were excluded if they presented with apparent marked pelvic asymmetry or scoliosis, or if they took medications that could have affected their gait. Body height and body mass were measured using a personal scale (WPT100/200 OW; RADWAG, Hilden, Germany). All tests were conducted by the same investigator. 2.2. Gait analysis Gait data were collected using a mobile inertial sensor based system (RehaWatch1 system; HASOMED1 GmbH, Magdeburg, Germany). The sensors were attached to the lateral ankle using a special device (Fig. 1) [20]. Each sensor contains three accelerometers and three gyroscopes (Analog Devices, Norwood, MA, USA) measuring foot motion in six degrees of freedom. The measurement range of the accelerometers and gyroscopes is 5 g and 6008/s (sampling rate: 512 Hz), respectively. Gait events were determined using offline evaluation of the gait data [21]. The primary gait event was heel-strike. All other gait events (including full contact, heel-off, and toe-off) were identified relative to heel-strike. Based on these gait events, spatio-temporal gait parameters and gait phases were calculated automatically, including kinematic parameters (e.g. stride length, foot height, and walking speed), gait phases [22,23] and symmetry [24]. In addition to standard gait parameters including cadence and walking speed, RehaWatch1 measures parameters describing gait patterns (e.g. foot height, and stance phase symmetry) that are not readily measured using step monitors and, in contrast to conventional motion capture systems, it is portable. The measurement system consists of a mobile computer (DataLogger) that is attached to the patient using straps (Fig. 1). Data of the inertial sensors are received via cable-connection and temporarily stored in the DataLogger. The data transfer from the DataLogger to the PC is performed subsequently using an USB-stick. Measured data comprises accelerations and angular velocities used to calculate spatio-temporal parameters. Angular velocities were integrated to obtain information on the spatial orientation of the sensor. Sensor position is calculated
by twice integrating acceleration data while taking into account the spatial orientation of the sensor. To avoid drift, the integration process is reset following each step while factoring in the boundary conditions. Hence, the sensor velocity is forced to equal zero at the end of the step and final foot height must match initial foot height. High intra-observer reliability (ICC-range: 0.691–0.959) of the inertial sensor based system has been reported previously [20]. Briefly, data for 44 healthy subjects (age: 27.7 4.2 years) were collected in three sessions per subject (sessionto-session time: 48 h). Each subject performed three trials per session while walking for 20 m. General linear models did not reveal significant differences in gait parameters between sessions. In addition to variance and correlation analyses, Bland–AltmanPlots suggested a high reliability of the device. The validity of the RehaWatch1 system was tested in an unpublished study. Ten healthy subjects (six females and four males, mean age 27.2 9.2 years, range: 5.9– 36.8 years) were recruited for this study. Subjects performed level walking at their preferred self selected walking-speed along a 12 m-walkway (repeated trials, n = 10). All repeated measurements were recorded simultaneously with the inertial based and an electro-optical system (VICON, Oxford Metrics Ltd., Oxford, UK). The ‘‘Plug-In-Gait’’ [25] marker set was used consisting of 16 reflective markers attached to anatomical landmarks using a standardized protocol. High validity (ICC-range: 0.776–0.991) was found for cadence, stride length, stride time and walking speed. Lowest test–retest correlations (ICC < 0.75) were found for time in double support and time in single support. Hence, this study concentrates on the spatio-temporal parameters. Gait analysis depends on three main test conditions: test distance, walking speed and the shoes worn by the subjects [2]. Self-selected speed appears to be the optimal condition for gait analysis and for providing functionally relevant data [1]. In this study, subjects wore their personal walking shoes. Subjects were asked to walk straight at their self-selected speed toward a target at a 20 m-distance to ensure that a sufficient number of stable walking cycles were recorded between the start and the end of the test. Each subject performed two walking trials to adjust to the test conditions while wearing the ISGAS. Data of the third trial were used for further analysis. 2.3. Statistical analysis All statistical analyses were performed using SPSS version 18.0 for Windows (SPSS Inc., Chicago, IL, USA) and SAS version 9.2 (SASInstitute, Cary, NC, USA). Sex differences of gait parameters were tested using a general linear model. Differences of means were considered statistically significant if p-values were less than 0.05 and partial eta-squared (h2)-values were greater than 0.05. Because of the large number of cases, decisions of significance were made primarily based on h2-values. The SAS quantreg procedure was used for modeling the effects of age on the conditional quantiles (5th, 10th, 25th, 50th, 75th, 90th and 95th) of different gait parameters by means of quantile regression analysis. Quantile regression analysis [26] extends the regression model to conditional quantiles of the response variable, such as the median or the 10th and 90th percentiles. A priori defined third-order polynomials were used to calculate curve fitting regression lines. Walking speed, stride length and cadence are strongly dependent on leg length. However, because leg length was not recorded in this study, gait parameters were normalized to body height/mean body height separately for each gender (Table 1):
Table 1 Demographic and morphological characteristics of subjects (n = 1860) categorized by decades of age and gender. All values are means (one standard deviations). Age (years) All subjects 5–100
Fig. 1. Photograph of a subject wearing the inertial sensor gait analysis system (RehaWatch1, HASOMED1 GmbH): (1) data logger attached to a belt, (2) sensor cable, fixed below the knee with an elastic belt, and (3) inertial sensors (6DOF) attached to a device fixed to each foot device.
Men/women
M (n = 919) W (n = 941) Subjects by age group 5.0–10.0 M (n = 112) W (n = 110) 10.1–15.0 M (n = 115) W (n = 115) 15.1–20.0 M (n = 111) W (n = 113) 20.1–30.0 M (n = 166) W (n = 166) 30.1–40.0 M (n = 93) W (n = 94) 40.1–50.0 M (n = 100) W (n = 102) 50.1–60.0 M (n = 79) W (n = 83) 60.1–70.0 M (n = 83) W (n = 86) 70.1–100.0 M (n = 60) W (n = 72)
Body mass (kg)
Body height (m)
BMI (kg/m2)
70.6 (22.7) 58.3 (17.0)
1.71 (0.17) 1.61 (0.13)
23.5 (4.98) 22.2 (4.81)
31.1 28.9 44.9 43.5 72.8 60.0 78.8 62.7 86.2 66.8 86.0 68.8 86.6 69.0 82.3 67.9 79.4 64.7
1.36 1.34 1.55 1.54 1.80 1.68 1.82 1.68 1.82 1.68 1.80 1.67 1.77 1.64 1.74 1.63 1.72 1.58
16.6 16.1 18.5 18.0 22.4 21.3 23.9 22.1 26.1 23.6 26.5 24.8 27.6 25.6 27.1 25.4 26.8 25.8
(7.17) (5.77) (12.4) (12.0) (14.0) (9.77) (10.6) (8.98) (12.9) (11.9) (13.0) (13.8) (12.5) (12.7) (13.0) (11.0) (12.3) (10.6)
(0.08) (0.04) (0.12) (0.11) (0.07) (0.06) (0.07) (0.06) (0.06) (0.06) (0.08) (0.06) (0.07) (0.07) (0.07) (0.05) (0.06) (0.06)
(2.46) (2.26) (3.21) (3.39) (3.57) (3.06) (2.59) (2.85) (3.69) (4.01) (3.35) (4.79) (3.84) (4.18) (3.72) (3.75) (3.50) (4.12)
R. Schwesig et al. / Gait & Posture 33 (2011) 673–678 normalized cadence = cadence sqrt (body height/mean body height), normalized stride length = stride length/(body height/mean body height), and normalized speed = speed/sqrt (body height/mean body height). A sex-stratified multifactorial regression model was used for each gait parameter to identify predictive variables for each gait parameter. Linear, quadratic and cubic terms were used for age, body height and body mass, respectively. A backward selection method with a significance level of 0.10 was used to find the best prediction model. Bivariate two-sided Pearson correlation was used to detect relationships between spatio-temporal parameters.
3. Results 3.1. Subjects Demographic and morphological data for all subjects are given in Table 1. The 1860 subjects included 941 women and were aged between 5 and 100 years (Table 1). The number of steps used for gait analysis ranged from 19 to 65. 3.2. Age, body height, body mass, gender and gait parameters A strong non-linear relationship between predictors (age, body height, and body mass) and gait parameters was found (Table 2). In combination, the predictors explain the greatest portion of variance for stride length (men: R2 = 0.47; women: R2 = 0.45). Men had longer stride times and walked at a lower cadence than women (mean standard deviation of stride time: men: 1.07 0.08 s; women: 1.01 0.08 s; h2 = 0.111; cadence: men: 113 9 steps/min; women: 119 9 steps/min; h2 = 0.136). This gender difference was observed in subjects aged 15 years or older. 3.3. Reference data for gait parameters Curves fitted to normalized cadence (Fig. 2a and b) data over age showed a clear difference between men and women. Male subjects in the 50th percentile showed a smaller variability than female subjects (men: 50th percentile = 109–115 steps/min; women: 50th percentile = 99–126 steps/min). Until the age of 65 years, normalized cadence slightly increased from 109 steps/
675
min to about 115 steps/min (interdecile range: 97–124 steps/ min) for men and from 113 steps/min to about 125 steps/min (interdecile range: 99–134 steps/min) for women. Between 65 and 90 years cadence declined from 115 to 109 steps/min (interdecile range: 94–123 steps/min) for men and from 125 to 99 steps/min (interdecile range: 95–134 steps/min) for women. Normalized stride length showed a two peak pattern (Fig. 2c and d), and non-normalized (Fig. 3c) stride length showed a single peak pattern. The age of occurrence of minimum and maximum normalized stride length for male subjects was 90 and 6 years, respectively (50th percentile: minimum = 0.95 m; maximum = 1.60 m) and for female subjects 100 and 5 years, respectively (50th percentile: minimum = 0.88 m; maximum = 1.48 m). Stride time data over age showed a local maximum and a local minimum (Fig. 3b). Stride time increased from the age of 5 years (50th percentile: 0.96 s), reached a local maximum at the age of 28 years (50th percentile: 1.07 s), decreased to a local minimum at the age of 64 years (50th percentile: 1.00 s) and increased beyond the local minimum until the age of 100 years (50th percentile: 1.31 s) (Fig. 3b). While non-normalized walking speed showed a single peak pattern (Fig. 3a), walking speed normalized for body height showed two peaks for male and female subjects (Fig. 2e and f). The age of occurrence of the maxima during childhood was the same for male and female subjects (male: 1.37 m/s at 6 years; female: 1.32 m/s at 5 years). Walking speed decreased during adolescence (male: 1.26 m/s at 22 years; female: 1.30 m/s at 13 years). The second maximum was reached during adulthood (male: 1.38 m/s at 56 years; female: 1.43 m/s at 55 years). Especially above the age of 70 years, the decrease in walking speed occurs more rapidly: male subjects reach their minimum of 0.79 m/s at 90 years, while female subjects reach their minimum of 0.64 m/s at 100 years. The largest and smallest variations (percent difference between minimum and maximum of the 5th and 95th percentiles) across the ages were found for normalized walking speed (98%) and normalized stride length (89%), respectively.
Table 2 Results of multifactorial regression model (method: backward selection) for spatio-temporal gait parameters by sex. Dependent variable
Walking speed (m/s)
Stride time (s)
Stride length (m)
Cadence (steps/min)
Men
Women
Predictor variables
R2
Predictor variables
R2
Body height squ Age lin Age squ Age cb Body mass squ Body height cb Age lin Age squ Age cb Body mass squ Body height squ Age lin Age squ Age cb Body mass squ Body mass cb Body height lin Age lin Age squ Age cb Body mass squ
0.171
Body height lin Age lin Age squ Age cb Body mass squ Body height lin Age lin Age squ Age cb Body mass squ Body height lin Age squ Age cb Body mass squ Body mass cb
0.291
Body height lin Age lin Age squ Age cb Body mass squ
0.172
List of abbreviations: lin, linear; squ, squared; cb, cubed. Units of predictor variables: m (body height); years (age); kg (body mass).
0.223
0.472
0.243
0.156
0.445
R. Schwesig et al. / Gait & Posture 33 (2011) 673–678
676
Normalized cadence = cadence * sqrt(body height/mean body height) a) men
b) women 170
150
..
Normalized cadence [steps/min]
Normalized cadence [steps/min]
170
.
. . . . . . . . .... .. . ...... .. .. . . . ... .. . .. .. ... .. .. . . . ... . . . . . .. . . .. ............... .................................... .. . ...................... ......... .. .. ............ ..................... ... . ....... ...... ......... ........... ........ ....... ... ....... ... .. .. ... ......... ... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ................... ......................... .... . .............. ....... .. ............. . .... .. ..... .... . . .................... ............ ............ . .. .. .... .. .. ... . ... . .. . .. .... . . . . . ............ .. ...... ..... ..... ..... .. . . . . .... . . . .... . .. ............. .. .. .. . . . . . .... . . . ...... .. . .. .
130
110
90
.
70 0
10
150
. .. . .... . . . .. . . . .. .. . . . .. .. .. .. . ... .... .... ....... ....... ..... ............... ...... .. . . .. . . ............. . . . ... .. . .. .. .... ..... . ..... ...... . .. . . .. .. ... .... . ....... .... .. . ....... ....... ..... . ... .... . ...... .. ..... . . . ............. ............... .......... ........ . ..... ..... .. . .. ... . .... ... ... . ..... . . . .................................................................. ......... .......... ....... ................ ............... ... . .... ... ... . . . . . .. . ... ...... . ...................... ........................... .. ... . . . . . . ......... . ....... ..... . .. . . . .. . . . .. . . . ................ .. .... .. . .. . . . . . . . ...... .. . . . . . . . .. .. . . . .. . . .
130
110
90
70 20
30
40
50
60
70
80
90
100
0
10
20
30
40
Age [years]
50
60
70
80
90
100
Age [years]
Normalized stride length = stride length / (body height/mean body height) c) men 2.1
2.1
. . . . . .. . . . .. . . .......... . ... . . . . ... . . ... . . . ... .. . . . ............... ........... ..... . .. .. .. ......... ... . ... .... .. .. . ....... . .......................................................... ......................... .......................... .... ............................ .. . . ................ .......................................... ....... ................. .. . ...... .... ...... ............. . . ... ... ... .. ... .. .. .. . .. . . .. .... .. . ...... .... . . . .. ............... ........................ .. .. . . .. ..... ... .... . . ... . ... ..... ... ... ..... . . ... .. . . . ............ ............. ...... ...... . ... . . . . . . .. . . .. . .. . . . . .
1.7 1.5 1.3 1.1 0.9 0.7
. . . . . .. . .. . . . . ... . . . . . .. . .. . .. .. . ...................... ................... ....... ... ...... ........ ... .................. ....... ... ... . . . . . .................................................... ........... . ................ ....... .. .... ..... ............... ... ... .. .................................................................................................................................. ............................................ .. .. .. ..... . . . . ............ ................... .. . .. ...... ... .............. . ....... .. . .. ...... .. . . .. ... .. . ........ . . . . .. . .. . .. . . . . . . . .. .. ... .. . . . . . . . . ... . .
1.9 1.7 Normalized stride length [m]
1.9
Normalized stride length [m]
d) women
1.5 1.3 1.1 0.9 0.7
0.5
0.5
0.3
0.3
0.1
.
0.1 0
10
20
30
40
50
60
70
80
90
100
0
10
20
30
40
Age [years]
50
60
70
80
90
100
Age [years]
Normalized speed = speed / sqrt(body height/mean body height) e) men . . . .. . . . . . . . .... . . . . . . . .......... ... . .. . .. . .. . .. . ... .. .. ... . . . . . . . . . . ...... ........ . ... .. . ... . ...... . . ... .. .... . . . . . ..... . . . .. . . . . ................... .. ................ ....... ......... ......................... ...... .... ...................... . .. .............. .. ............... ......... ...... .......................... . ... ................. ......... .. ......................... ....... ...... ... . ..... . ... .... . . ... .. . . . . ............. .. ........................................... . .. .... .. ...... ..... .... ...... ... . .... . . .... . .. ................ . . . .... .. . . . ..... .. ... ...... .... . ............ ...... . ........ .... . . . . . . .. . . . . . . . . . . .. . . ... . .. . .. . . . . . .. . .. ... . . . . . . . . . . .
1.8 1.6 1.4 1.2 1.0 0.8 0.6
2.0
. . . . . . .. . . . .. . .. . . . .. . .. . .... . .. . . . ... ... ... .. .. ... .. .... .. . . ..... .. .... .. ..... ... .. .... .. . ............ . ....................... . ..... ................. ........ . . ... ........ ............ .......... . ........... ... ..................... ..... ........ . .. ..... ... .. ..... . ..... ...... .. . . .......................................................................... ........... ......... .. ......... .............. ... .... .... . . . . . . . . . . . . . . . . . ............................................................ .. ..... ... .. ...... ...... ........... . ... .. .... ..... . . ............... ............ .. . . .... ....... ... . ... . . ... .. . .. .. ... . .. . . . .. . ... .. . . .. .. . . . ... . .. . . . ... . .. ... . . . . . . .. .. .. . . .. . . .. .
1.8 1.6 Normalized speed [m/s]
2.0
Normalized speed [m/s]
f) women
1.4 1.2 1.0 0.8 0.6
0.4
0.4 0
10
20
30
40
50
60
70
80
90
100
Age [years]
0
10
20
30
40
50
60
70
80
90
100
Age [years]
Fig. 2. (a–f) Age and sex dependent presentation of the body height normalized cadence (a and b), normalized stride length (c and d) and normalized speed (e and f) based on quantile regression analysis. Various lines describe the 5th, 10th, 25th, 50th, 75th, 90th and 95th quantiles (bottom-up).
4. Discussion In this study, reference data for gait parameters across all ages were presented. Results of the quantile regression analysis (Figs. 2a–f, 3a–c) showed that analyzing age and gender aspects
using age groups (e.g. decades) is not sufficient. Because of the large variation within each age group especially during childhood and adolescence and at advanced age, an ordinal scale for the predicting parameter age should not be used. Hence, comparison of this data set with results of previous studies [1,8] prove
R. Schwesig et al. / Gait & Posture 33 (2011) 673–678
677
Fig. 3. (a–c) Age dependent presentation of (a) walking speed, (b) stride time and (c) stride length across the entire life span using quantile regression analysis. In addition, bivariate two-sided Pearson correlations are given. Various lines describe the 5th, 10th, 25th, 50th, 75th, 90th and 95th quantiles (bottom-up).
challenging because those studies typically reported mean values for age groups (decades) or gender specific mean values. In general and in agreement with many previous studies [2,6,8,10], men took larger steps (stride length: women: 1.42 m; men: 1.47 m; h2 = 0.021), showed longer stride time (women: 1.01 s; men: 1.07 s; h2 = 0.111) and had lower cadence than women (women: 119 steps/min; men: 113 steps/min; h2 = 0.109). In particular, middle-aged women tend to walk at a higher cadence compared to men. To date, the largest studies on reference gait data [18,19] with respect to size of study population and age range reported a mean self-selected walking speed of 1.37 m/s (women: 1.28 m/s; men: 1.43 m/s) across all ages (20–85 years). In our study, nonnormalized mean walking speed across all ages was 1.31 m/s (women: 1.33 m/s; men: 1.30 m/s) and was within the range reported by Murray et al. [18,19]. In general, women walked slightly faster than men. However, in contrast to previous studies, the gender difference in this study was not significant (h2non-normalized ¼ 0:004; h2normalized ¼ 0:004). According to Kirtley et al. [27], a self-selected walking speed of less than 1 m/s indicates a disability or pathology. In our study, non-normalized walking speed was lower than 1 m/s only in subjects aged 93 years or older (50th percentile: 0.94 m/s). In addition, walking speed lower than 1 m/s was only observed in the 5th percentile (age 16 and 69 years), 10th percentile (age 7 and 73 years) and 25th percentile (age 80 years). Hence, the results of our study support the use of walking speeds of 1.0 m/s and lower as an indicator for disability or pathology. Normal mean (non-normalized) stride length of 1.41 m as reported by
Murray et al. [18,19] was reached already at the age of 16 years (50th percentile) with an average stride length of 1.45 m across all ages (men: 1.47 m; women: 1.42 m). The same investigators reported a norm value for non-normalized cadence of 113 steps/ min. In our study, a mean non-normalized cadence across all ages of 116 steps/min was found (men: 113 steps/min; women: 119 steps/min). A cadence of 113 steps/min was reached at 22 years and at 81 years (50th percentile). Results of the multifactorial regression analysis stratified by gender (Table 2) and those of the quantile regression analysis (Figs. 2a–f and 3a–c) suggest a non-linear influence of the predictors on gait parameters. In addition, the results of our study (r = 0.076) did not support earlier reports [1,8] of age dependency of walking speed (r = 0.58 in women, r = 0.49 in men). Similar observations were made for stride length (r = 0.119) and cadence (r = 0.060) where relevant correlations were not found. It is possible that these discrepancies between results are caused by difference in age range between studies (20–86 years and 20–98 years vs. 5–100 years) and by the lack of information on children, adolescents and the elderly in previous studies. Ashton-Miller [28] observed a 1%-decrease in walking speed per year in persons above the age of 60 years. Those results support our observation. Between the age of 60 years (50th percentile: 1.40 m/s) and the age of 90 years (50th percentile: 1.01 m/s), walking speed decreased by 28% corresponding to a yearly change of 0.93%. However, when considering the decline in walking speed from the maximum walking speed which was found for subjects at 50.1 years (50th percentile: 1.42 m/s), the annual decline in walking speed would be 0.73%. It is important to note
678
R. Schwesig et al. / Gait & Posture 33 (2011) 673–678
that all of these values are mean values and that there is no linear relationship between walking speed and age. Moreover, above the age of 70 years, the decrease in cadence, stride length and walking speed and increase in stride time occurs more rapidly. Grieve and Gear [29] observed a very strong linear relationship between cadence and walking speed (r > 0.95) within one subject. In contrast, in our study we found a moderate relationship between cadence and the square root of walking speed (r = 0.544) for the entire population. A strong relationship between stride length and the square root of walking speed was found (r = 0.843). These relationships were independent of age and gender (rage = 0.118; rsex = 0.084). Possible reasons for the discrepancies between the results of these studies include different age ranges (1–35 years vs. 5–100 years) and different methods employed in these studies. The clinical importance of our study is the extensive data base that provides reference gait data that may be used in the diagnosis or evaluation of orthopaedic or neurological patients with conditions, such as osteoarthritis, and/or diseases, such as cerebral palsy or Parkinson’s disease. Frequently, information on spatiotemporal parameters is sufficient for these patient populations, especially when considering the trade-off between effort and time spent on data collection and processing and the limited additional information as it relates to the specific disorder or pathology. 5. Conclusion The presented reference data on gait parameters of men and women aged 5–100 years (self-selected walking speed, 20 m straight corridor, subjects wearing their own comfortable shoes) forms the foundation for future studies that aim to evaluate gait pathologies and disorders of older adults. In particular, specific gait parameters can now be more precisely evaluated not only in comparison to age groups but also at specific time points throughout a person’s life span. The inclusion of additional data especially on subjects above the age of 70 years is desirable to improve the precision of the information extracted from the database. The data collected in this study has been integrated as reference database into the RehaWatch1 gait analysis system. Since there are no data on the comparison of different gait analysis systems the assignability and comparability of the reference data is limited. Further studies including different gait analysis systems are necessary to provide information about reliability and comparability of gait parameters. Acknowledgements The authors thank Christoph Hintze for his help with data acquisition and Dr. Annegret Mu¨ndermann for her writing assistance. The authors would like to express their appreciation to all participating institutions (n = 62) and subjects (n = 1860) for their participation in this study. Funding for this study was provided by Investitionsbank Sachsen-Anhalt, Germany. Conflict of interest statement The authors do not have any conflicts of interest.
References [1] Auvinet B, Berrut G, Touzard C, Moutel L, Collet N, Chaleil D, et al. Reference data for normal subjects obtained with an accelerometric device. Gait Posture 2002;16:124–34. ¨ berg T, Karszina A, O ¨ berg K. Basic gait parameters: reference data for normal [2] O subjects, 10–79 years of age. J Rehabil Res Dev 1993;30:210–23. [3] Nigg BM, Fisher V, Ronsky JL. Gait characteristics as a function of age and gender. Gait Posture 1994;2:213–20. [4] Ostrosky KM, VanSwearingen JM, Burdett RG, Gee Z. A comparison of gait characteristics in young and old subjects. Phys Ther 1994;74:637–44. [5] Blanc Y, Balmer C, Landis T, Vingerhoets F. Temporal parameters and patterns of the foot roll over during walking: normative data for healthy adults. Gait Posture 1999;10:97–108. [6] Cho SH, Park JM, Kwon OY. Gender differences in three dimensional gait analysis data from 98 healthy Korean adults. Clin Biomech (Bristol Avon) 2004;19:145–52. [7] Kimura T, Kobayashi HNE, Hanaoka M. Effects of aging on gait patterns in the healthy elderly. Anthropol Sci 2007;115. [8] Senden R, Grimm B, Heyligers IC, Savelberg HH, Meijer K. Acceleration-based gait test for healthy subjects: reliability and reference data. Gait Posture 2009;30:192–6. [9] Menz HB, Latt MD, Tiedemann A, Mun San KM, Lord SR. Reliability of the GAITRite walkway system for the quantification of temporo-spatial parameters of gait in young and older people. Gait Posture 2004;20: 20–5. [10] Winter DA, Patla AE, Frank JS, Walt SE. Biomechanical walking pattern changes in the fit and healthy elderly. Phys Ther 1990;70:340–7. [11] Norlin R, Odenrick P, Sandland B. Development of gait in the normal child. J Pediatr Orthop 1981;1:261–6. [12] Todd FN, Lamoreux LW, Skinner SR, Johanson ME, St HR, Moran SA, et al. Variations in the gait of normal children. A graph applicable to the documentation of abnormalities. J Bone Joint Surg Am 1989;71:196–204. [13] Wheelwright EF, Minns RA, Law HT, Elton RA. Temporal and spatial parameters of gait in children. I: normal control data. Dev Med Child Neurol 1993;35:102–13. [14] Sutherland TH, Olshen R, Biden EN, Wyatt MP. The development of mature walking. London: MacKeith Press; 1988. [15] Schwartz MH, Rozumalski A, Trost JP. The effect of walking speed on the gait of typically developing children. J Biomech 2008;41:1639–50. [16] Samson W, Desroches G, Cheze L, Dumas R. 3D joint dynamics analysis of healthy children’s gait. J Biomech 2009;42:2447–53. [17] Beck RJ, Andriacchi TP, Kuo KN, Fermier RW, Galante JO. Changes in the gait patterns of growing children. J Bone Joint Surg Am 1981;63:1452–7. [18] Murray MP, Drought B, Kory RC. Walking patterns of normal men. J Bone Joint Surg Am 1964;46:335–60. [19] Murray MP, Kory RC, Sepic SB. Walking patterns of normal women. Arch Phys Med Rehabil 1970;51:637–50. [20] Schwesig R, Kauert R, Wust S, Becker S, Esperer HD, Leuchte S. Reliability of the novel gait analysis system RehaWatch. Biomed Tech 2010;55:115. [21] Negard NO, Kauert R, Andres S, Schauer T, Raisch J. Gait phase detection and step length estimation of gait by means of inertial sensors. In: 3rd European medical & biological engineering conference, IFMBE European conference on biomedical engineering. 2005. p. 11. [22] Perry J. Ganganalyse. Norm and Pathologie des Gehens. Mu¨nchen, Jena: Urban & Fischer; 2003. [23] Beckers D, Deckers J. Ganganalyse and Gangschulung. Therapeutische Strategien fu¨r die Praxis. Berlin u.a.: Springer; 1997. [24] Robinson RO, Herzog W, Nigg BM. Use of force platform variables to quantify the effects of chiropractic manipulation on gait symmetry. J Manipulative Physiol Ther 1987;10:172–6. [25] Kadaba MP, Ramakrishnan HK, Wootten ME. Measurement of lower extremity kinematics during level walking. J Orthop Res 1990;8:383–92. [26] Koenker R, Bassett G. Regression quantiles. Econometrica 1978;46:33–50. [27] Kirtley C, Whittle MW, Jefferson RJ. Influence of walking speed on gait parameters. J Biomed Eng 1985;7:282–8. [28] Ashton-Miller JA. Age-associated changes in the biomechanics of gait and gaitrelated falls in older adults. In: Hausdorff JM, Alexander NB, editors. Gait disorders: evaluation and management. Boca Raton: Taylor & Francis; 2005. p. 63–100. [29] Grieve DW, Gear RJ. The relationships between length of stride, step frequency, time of swing and speed of walking for children and adults. Ergonomics 1966;9:379–99.