- Email: [email protected]
Kinematics and movement sequencing during flexion of the lumbar spine
Clinical Biomechanics 14 (1999) 376±383
Kinematics and movement sequencing during ¯exion of the lumbar spine Michelle L. Gatton *, Mark J. Pearcy Centre for Rehabilitation Science and Engineering, School of Mechanical, Manufacturing and Medical Engineering, Queensland University of Technology, GPO Box 2434, Brisbane, Queensland 4001, Australia Received 30 September 1998; accepted 2 December 1998
Abstract Objectives. This study investigated the sequence of intervertebral joint movements and range of motion during three tasks involving lumbar ¯exion. Design. Position sensors were used to measure position and rotation of lumbar vertebrae during unconstrained ¯exion. Background. In the development of mathematical models, numerous assumptions need to be made. Few studies have attempted to assess the validity of the assumptions regarding kinematics in models of the lumbar spine. Methods. Position sensors were attached to the skin overlying the lumbar vertebrae of 14 volunteers. Volunteers performed three ¯exion tasks; unconstrained ¯exion from upright standing, with and without a mass of 5 kg held close to the body, and the transition from upright standing to a seated position. Results. Four de®nitive movement sequences were identi®ed for those subjects with consistency between replicates; `top down' motion (where the top of the lumbar spine starts to move ®rst and the bottom moves last), `bottom up' (where the bottom of the lumbar spine moves ®rst and the top moves last), `all together' (where all segments commence movement together), and `middle last' (where the middle segments of the lumbar spine are last to commence movement). Subjects not ®tting one of these sequences were categorised into a miscellaneous group. Only two subjects exhibited the same sequence for each of the three tasks, while other subjects exhibited two or three dierent sequences for the three tasks, or showed a lack of consistency for one of the tasks. Conclusions. The results from this study indicate that there is no single movement sequence exhibited by the sample population. Relevance Incorrect assumptions which are incorporated into mathematical models have the potential to in¯uence model output. Given that output from spinal models is often used to assess ergonomic issues such as safe lifting loads, validation of the assumptions is essential. Ó 1999 Elsevier Science Ltd. All rights reserved. Keywords: Kimematics; Flexion; Lumbar spine; Range of motion; Movement sequence
1. Introduction The human lumbar spine is a complex structure consisting of ®ve vertebrae, associated intervertebral discs and many attached ligaments and muscles. Each of these components is fundamental for stability and movement. However, little is known about the interconnection between components. Mathematical models are one means used to explore the various interactions between bone, muscle and ligament. In the past, output
*
Corresponding author. E-mail: [email protected]
from mathematical models has provided information on ergonomic issues such as safe lifting postures and loads. The trend in recent times has been to produce dynamic or quasi-static mathematical models, as opposed to static models, in an eort to investigate the forces experienced or generated by each of the components in the lumbar spine under speci®ed conditions. However a mathematical model is only as good as the data and assumptions used within it. The sequence of motion is an important assumption within dynamic and quasi-static models, since joint position will impact on the allocation of forces to individual muscles by altering muscle orientation and maximal force. Although much kinematic data has been reported relating to the range of movement in the
0268-0033/99/$ ± see front matter Ó 1999 Elsevier Science Ltd. All rights reserved. PII: S 0 2 6 8 - 0 0 3 3 ( 9 8 ) 0 0 1 0 9 - 0
M.L. Gatton, M.J. Pearcy / Clinical Biomechanics 14 (1999) 376±383
377
Table 1 Summary of subject characteristics
All subjects Mean (SE) Median Range Male subjects Mean (SE) Median Range Female subjects Mean (SE) Median Range
Age (yr)
Height (m)
Weight (kg)
BMIa
31.4 (2.7) 29 18±46
1.70 (0.03) 1.67 1.59±1.88
71.6 (4.2) 65 55±105
24.8 (1.4) 23.3 18.6±38.1
29.4 (3.3) 28 18±40
1.76 (0.04) 178 1.59±1.88
73.9 (5.5) 70 55±92
23.8 (1.2) 23.2 19.4±29.0
33.4 (4.4) 30 23±46
1.65 (0.02) 1.65 1.59±1.75
69.4 (6.7) 65 55±105
25.7 (2.6) 23.9 18.6±38.1
p SE Standard Error Standard deviation/ n: a 2 BMI Body Mass Index weight/height .
lumbar spine [1±5], little literature discusses the actual movements, or sequence of movements of the lumbar spine between the extremes of motion. Work by Kanayama et al. [6] suggests that during ¯exion the L3=4 , L4=5 and L5 /S1 joints do not deform simultaneously. Their study [6] implies that the spine exhibits `top±down' motion such that the L3=4 joint is deformed before the L4=5 joint and before the L5 /S1 joint. This paper reports a study of the kinematics of the lumbar spine during several ¯exion tasks with particular reference to the sequence of intervertebral joint movements. Range of motion data resulting from the study are compared to previously reported studies and factors potentially in¯uencing range or sequence of motion data are also investigated. 2. Methods A sample of 14 subjects (male female) participated in the study. All subjects volunteered to participate and signed a consent form as required for ethical clearance. None of the subjects reported any previous back injury, surgery or spinal abnormality. Two subjects reported having mild scoliosis. A summary of the subject characteristics is presented in Table 1. A Motion Star motion analysis system (Ascension Technology Corporation, Vermont, USA) was used to record the orientation and position of sensors in three dimensions whilst subjects performed a variety of tasks. Preliminary evaluation showed that this system was accurate to 0.4° in the range of measurement. The sensors were attached to the skin over the spinous processes of the lumbar vertebrae using double sided tape. The locations of the spinous processes were determined as described by Burton [7], that being the identi®cation of the L4 spinous process as the bisection of a line joining the highest points on the iliac crests. Subsequent levels
were identi®ed by counting up or down from L4 . In some instances this identi®cation was made easier by asking subjects to bend forward slightly. To ensure the sensors stayed attached during movement, bandages were bound around the subject's trunk to apply slight positive pressure to the sensors. These bandages were bound so as to apply pressure to the sensors but not limit the movement of the subject. Unfortunately, due to the size of the sensors, it was not always possible to place sensors over consecutive spinal levels. In this instance the maximum distance between sensors was two levels. An additional sensor was ®tted to a plate which was moulded to ®t over the sacrum. Fig. 1 illustrates the attachment of the sensors. Data were collected 20 times per second for 6s. Tasks performed were classi®ed as primarily uniaxial (that is in one plane) or composite (that is involving more than one primary plane of motion). Only the uniaxial ¯exion tasks will be discussed here. Subjects were asked to undertake the movements with no restriction being placed on hip motion. This approach was used to obtain
Fig. 1. Location and attachment of sensors during data collection.
378
M.L. Gatton, M.J. Pearcy / Clinical Biomechanics 14 (1999) 376±383
Fig. 2. Flexion tasks performed by subjects (a) unloaded ¯exion, (b) loaded ¯exion, and (c) sitting.
data from `natural' body movements. Flexion from the standing position (hereafter referred to simply as ¯exion) was measured on four occasions, two replicates in an unloaded state, and two replicates in a loaded state where a 5kg barbell was held close to the body. The sitting task, in which subjects were asked to sit on a stool then stand up again, was also repeated twice. Fig. 2 provides a schematic illustration of these tasks. Changes in the angle of sensors from the original upright posture were calculated using the methodology of Pearcy et al. [8] Standard statistical techniques such as linear regression, analysis of variance (A N O V A ) and student's t-test were used to analyse continuous data while Fisher's Exact test was used to test for dierences in proportions for categorical data. In order to sequence the spinal movement, data were smoothed using a moving average with a window size of 5 data points. This smoothing helped to reduce the amount of small, erratic segmental movement exhibited by some subjects before and during the initial stages of motion. A segment was declared to have started motion when the average rate of change in the angle (or velocity) between adjacent sensors over two consecutive time periods was greater than or equal to 5° =s for each joint encompassed by the sensors (that is, if the sensors were separated by two intervertebral joints the average velocity over two consecutive time periods had to be greater than or equal to 10° =s). Sequencing was achieved by ordering the start of motion of the spinal segments. Segments needed to start motion at least 0.1 s apart to be declared as starting at dierent times.
along with the ranges presented hereafter, was obtained using data from 13 of the 14 subjects tested. One male subject had to be omitted from this section of the analysis due to a reporting discrepancy in sensor locations. There was a statistical dierence in the range of spinal motion between males and females during unloaded ¯exion, with males having a signi®cantly higher mean range of spinal ¯exion than females (66.0° vs 51.6°; P 0.008). No signi®cant relationship with age was detected (P > 0:6). There was a signi®cant change in spinal motion with subject height such that the range of spinal ¯exion increased with height (P 0.03). Fig. 3 illustrates the data on which this relationship was developed. The regression model has an R2 value of 0.38, exhibits a random residual plot with no apparent patterning and can be expressed mathematically as, Range of spinal flexion degrees ÿ63:79 0:72 Subject height cm: The 95% con®dence interval for the mean voluntary range of motion in loaded ¯exion from L1 to S1 was 54.3°±66.4°. This value is not signi®cantly dierent from the voluntary range of motion obtained during unloaded ¯exion for all subjects combined (P 0.154), and for males and females analysed separately (P 0.866 for males and P 0.109 for females). There appears to be an increasing trend in the range of spinal ¯exion (during loaded ¯exion) with height (R2 0:28; P 0.06). The mean spinal ¯exion from L1 to S1 in the seated position was 27.6° with a standard error of 2.3°. When considered as a proportion of the subject's full range of ¯exion (during unloaded ¯exion), the mean was 49.6% with the associated 95% con®dence interval being 40.5% to 58.8%. 3.2. Movement patterns Consistency in movement patterns was measured by comparing the two replicates of each task obtained from each subject. Consistency was judged based on the
3. Results 3.1. Range of motion The mean voluntary range of motion from L1 to S1 during unloaded ¯exion was 58.3° with a 95% con®dence interval for the mean of 51.4°±65.2°. This value,
Fig. 3. Range of spinal ¯exion with subject height for unloaded ¯exion.
M.L. Gatton, M.J. Pearcy / Clinical Biomechanics 14 (1999) 376±383
sequence of movement between intersegmental levels. A sequence is de®ned as the order in which intersegmental levels commence movement. Consistency was categorized as `consistent' if the movement sequence was identical between replicates, `almost consistent' if a segment changed it's position relative to the adjacent sensors in the sequence such that it starts moving at the same time as the adjacent sensor when it previously started before or after, or vice versa, and `not consistent' if there was a change in the ordering of the sequence (see Fig. 4). Sequences for all 14 subjects were analysed for the forward ¯exion tasks, while only sequences from 13 subjects were analysed for the sitting task due to a
Fig. 4. Examples of consistency categories. The shaded areas represent the ®rst two time intervals satisfying the movement criterion. Subject 1 exhibits a `consistent' sequence since the ordering of sensors has not changed between replicates. Subject 2 has `almost consistent' sequences since in the second repetition sensor 2 no longer moves with sensor 1 instead moving with (and not after) sensor 3. Subject 3 shows sequences which are `not consistent' since sensor 1 has changed its position from moving ®rst to moving last relative to the other sensors.
379
Table 2 Frequency of consistency patterns for unloaded and loaded ¯exion, and sitting
Unloaded Loaded Sitting
Consistent
Almost consistent
Not consistent
4 5 4
7 8 6
3 1 3
problem with data collection from one subject during the sitting trial. The allocation of subjects to a consistency class is given in Table 2 for unloaded and loaded ¯exion, and sitting. A comparison of the consistency of sequences for the three tasks indicates that there is no statistical dierence in the frequency of consistency categories under the dierent conditions (P > 0.6). Although a total of seven inconsistent sequences were recorded, these inconsistencies were spread amongst the subjects with no individual having more than one inconsistent sequence recorded for the three tasks (see Table 3). No single variable such as age, height, weight, BMI or sex was found to signi®cantly dierentiate those subjects having inconsistent and consistent movement sequences (P > 0:05). Examining the subject trials where `consistent' or `almost consistent' movement patterns were detected, a variety of movement sequences were observed. These sequences can be broadly classi®ed into four groups; `top down' (where the top of the lumbar spine is the ®rst to move and the bottom is the last to move), `bottom up' (where the bottom of the lumbar spine is ®rst to move and the top is last to move), `all together' (where all segments start to move together) and `middle last' (where the middle of lumbar spine is last to commence movement) (see Fig. 5). Sequences not ®tting one of these categories were classi®ed in a miscellaneous group. Examples of miscellaneous sequences are (L3 ±L4 then T12 ±L1 then L1 ±L3 then L4 ±S1 ) and (T12 ± L1 then L3 ±L5 then L5 ±S1 then L1 ±L3 ). Table 3 indicates which movement sequence group each subject was allocated to for the three tasks. Little consistency in the sequence group was found within individual subjects across the three tasks with only two subjects exhibiting the same sequence during all three tasks. A more common feature was to have the same sequence for two of the tasks and one task with a different or inconsistent sequence (7 subjects). Table 4 provides a summary of the frequency of each of the identi®ed movement sequences. When spinal ¯exion is disaggregated into individual rotations between sensors (something akin to investigating intersegmental rotation), the relative contribution of each segment can be investigated. Fig. 6 illustrates the proportioning of spinal ¯exion into segmental units for
380
M.L. Gatton, M.J. Pearcy / Clinical Biomechanics 14 (1999) 376±383
Table 3 Movement sequence category for individual subjects during unloaded and loaded ¯exion and sitting (subjects showing inconsistent sequences between replicates are indicated by `±', with `*' representing missing data)
Subject Subject Subject Subject Subject Subject Subject Subject Subject Subject Subject Subject Subject Subject
1 2 3 4 5 6 7 8 9 10 11 12 13 14
Unloaded ¯exion
Loaded ¯exion
Sitting
T O ± B A O A ± ± M A A O A
T A O B B O ± O B M M M M A
± A B ± M O A O B M T ± B *
A: all lumbar levels start motion together; B: `bottom up' motion; M: middle lumbar level is last to start moving; O: movement sequence does not ®t one of other groups; T: `top down' motion.
4. Discussion
Fig. 5. Sequencing intervertebral motion during forward ¯exion of the trunk. The S1 sensor (solid line) is an indicator of the amount of hip rotation. At point A, the T12 and L2 segments deviate from the L4 and S1 curves indicating that the T12 ±L2 and L2 ±L4 segments are starting to ¯ex. Also at point A the T12 curve deviates from the L2 curve meaning that the T12 ±L2 segment is ¯exing quicker than the L2 ±L4 segment. Flexion of the L4 ±S1 segment does not occur until after point B when the L4 curve deviates from the S1 curve. Therefore, the movement can be sequenced as T12 ±L2 then L2 ±L4 then L4 ±S1 , or `top down'.
three of the subjects sampled. Negative values indicate that this segment extended during the voluntary ¯exion movement. The patterns exhibited in Fig. 6 are indicative of the patterns seen in other subjects, namely that there is a wide variety of curve shapes and that no one segment appears to dominate movement in all subjects. For some subjects there appears to be a constant proportioning once movement has commenced, but this is not the case for all subjects. The curves in Fig. 6 do not start at zero ¯exion since at the very start of movement, the calculation of segmental contribution is based on very small rotations resulting in the proportions being large and highly variable.
The 95% con®dence interval for the mean range of spinal ¯exion from L1 to S1 obtained from this study appears to be in agreement with data reported in other studies [1,4]. The mean value of 58.3° compares well with results from McGregor et al. [4] who obtained a mean voluntary range of motion of 56.6° using a sample of 103 males and 100 females. Pearcy [9] reported ¯exion angles from L1 to S1 in a seated posture for 10 males subjects. The mean ¯exion reported was 44% of the maximum voluntary ¯exion. This value lies within the 95% con®dence interval for the mean proportion of maximum ¯exion in a seated posture obtained from this study (40.5±58.8%). The results from this study indicate that the range of spinal ¯exion in this subject population was not in¯uenced by age. Other studies have had mixed results when relating range of motion to age. It would appear that studies which have reported age dierences in spinal ¯exion include subjects from a wider age group compared to this study. For instance, the study by Gracovetsky et al. [1] which reported a decrease in spinal ¯exion with age, sampled people aged 19±64 yr with half the sample being 41 years or older (compared to the age range of this study of 18-46 years with median age of 29 yr). A signi®cant relationship was found between the range of spinal ¯exion and sex, with males having a larger range than females. This result is in agreement with the results reported by Hindle et al. [2]. An interesting result is the signi®cant relationship between subject height and range of spinal ¯exion. This relationship suggests that taller subjects have a larger
M.L. Gatton, M.J. Pearcy / Clinical Biomechanics 14 (1999) 376±383
381
Table 4 Summary of movement sequences during unloaded and loaded forward ¯exion and sitting Sequence
`Top down' `Bottom up' `All together' `Middle last' `Other'
Frequency of sequence Unloaded ¯exion
Loaded ¯exion
Sitting
1 1 5 1 3
1 3 2 4 3
1 3 2 2 2
Fig. 6. Proportional allocation of rotation to segmental units. Positive value represents ¯exion of joint while negative value represents extension of joint.
range of spinal ¯exion than shorter subjects. Biomechanically this appears feasible if one assumes that ligament strain is a limiting factor in spinal ¯exion. Taller people in general would be expected to have slightly longer torsos compared to shorter people, resulting in longer ligament lengths. If the limiting ligament strain is constant then it would be necessary for a longer ligament to change length more than a shorter ligament in order to obtain the same strain, resulting in an increased rotation. Some simple calculations enable this hypothesis to be tested to assess if the magnitude of rotation change is similar to that seen experimentally. We can investigate the eect of subject height change in a mathematical model of the lumbar spine simply by adjusting the initial height of the intervertebral discs. A 10 mm change in subject height approximately represents a 0.45 mm change in height of the lumbar spine. Spreading this change equally over the joints of the lumbar spine results in an increase in trunk rotation of 0.8°. This change in comparable to the observed relationship between height and range of spinal rotation where a 1cm height change results in a 0.72° increase in spinal ¯exion. The relationship between subject height and range of spinal ¯exion raises some interesting questions in relation to apparent dierences in range of spinal ¯exion between males and females. This study reported a signi®cantly larger range of spinal ¯exion for males than females. Taking into account that the mean height of males is 0.11 m larger than that for females, one must ask whether the dierence in range of spinal ¯exion between the sexes is not, at least partially, a remnant of the height dierence. Unfortunately this study is not large enough to further investigate this issue since there is not enough overlap between males and females of similar height to enable a valid comparison. The analysis of movement patterns highlighted the fact that individuals can perform a simple task like ¯exion using a number of dierent strategies. This raises an interesting problem when trying to model the kine-
382
M.L. Gatton, M.J. Pearcy / Clinical Biomechanics 14 (1999) 376±383
matics of the spine since there appears, from the limited sample of this study, to be no dominant movement pattern for any of the tasks considered. Kanayama et al. [6] suggest that during forward ¯exion the spine moves from the top ®rst. Although we did record this sequence, `top down' motion was only recorded in a small proportion (9%) of trials. The issue of movement sequencing is also complicated by the fact that approximately half of the sample group exhibited inconsistent movement sequences during one of the three tasks performed. Hence these people can perform a speci®c task twice using dierent movement sequences each time. This is an interesting ®nding given the ongoing debate into lumbar muscle recruitment strategies. If one assumes that muscle activity aects the rotation of individual intervertebral joints, it would be reasonable to conclude that if there is one `optimal' muscle recruitment strategy, then the spine would also move using only one sequence. This is obviously not the case for the tasks performed as part of this study. To investigate this idea, further research needs to be conducted during which more physically challenging ¯exion tasks are performed to place more demands on the muscles of the spine. One feature common to several subjects is the rotation of one level in the opposite direction to the overall rotation at some point during ¯exion (Fig. 6). Pearcy [3] experienced the same phenomenon when he recorded an extension of the L5 /S1 joint in one subject at full ¯exion. Past mathematical models of the lumbar spine have made various assumptions about the kinematics of the spine. For instance, McGill [10] allocates bending between the rib cage and pelvis to individual lumbar levels using the relationship Ria aia Ba where Ria is the rotation of the ith vertebra about axis a, aia is the percentage of rotation attributed to the ith vertebra about axis a and Ba is the total lumbar bending about axis a. For ¯exion McGill [10] uses aL1 aL2 13:2%, aL3 21%, aL4 29% and aL5 23:6%. This relationship implies that all lumbar levels start to move together and that there is a constant relationship between the angle of rotation at each lumbar level. The results of this study indicate that although approximately 27% of trials exhibited the sequence `all together', few of these cases showed constant proportioning of intersegmental angle. There also appears to be no constant relationship between the angles of rotation at each lumbar level across the entire range of ¯exion (Fig. 6). In some subjects there appears to be a constant proportioning between levels after movement has commenced, however even in these subjects it does not appear that any one level dominates movement. For instance, as illustrated in Fig. 6, the dominant segment in subject 10 is L1 ±L3 , while in subject 5 L3 ±L5 dominates and there is no dominant
segment in subject 13. Hence the assumption of a simple relationship between intervertebral movements is not supported by the results of this study. 5. Conclusion This study investigated the kinematics of the lumbar spine during unconstrained ¯exion of the lumbar spine, as well as the movement sequencing of the lumbar vertebrae between the extremes of motion. The results relating to range of motion are in general agreement with other published studies. This study has detected a relationship between subject height and range of spinal ¯exion and discusses a possible reason for this relationship as well as considering the implications of such a relationship. The study was instigated to investigate the sequencing of movement within the lumbar spine. To this end we have identi®ed a number of movement strategies and highlighted the wide dierences between subjects. In short, these results, although obtained from a small sample, indicate that the movement assumptions in mathematical models need to be carefully reviewed. Acknowledgements The authors wish to thank all the people who volunteered for this study, Dr. Pei Lai Cheng for his assistance with data collection and Dr. Graeme Pettet for his valuable comments. References [1] Gracovetsky S, Newman N, Pawlowsky M, Lanzo V, Davey B, Robinson L. A database for estimating normal spinal motion derived from noninvasive measurements. Spine 1995;20(9):1036± 46. [2] Hindle RJ, Pearcy MJ, Cross AT, Miller DHT. Three-dimensional kinematics of the human back. Clin Biomech 1990;5(4):218±28. [3] Pearcy MJ. Stereo radiography of lumbar spine motion. Acta Orthop Scand 1985;56(Suppl 212):212. [4] McGregor AH, McCarthy ID, Hughes SP. Motion characteristics of the lumbar spine in the normal population. Spine 1995;20(22):2421±28. [5] Panjabi MM, Oxland TR, Yamamoto I, Crisco JJ. Mechanical behavior of the human lumbar and lumbosacral spine as shown by three-dimensional load-displacement curves. J Bone Joint Surg[Am]1994;76A(3):413±24. [6] Kanayama M, Tadano S, Kaneda K, Ukai T, Abumi K, Ito M. A cineradiographic study on the lumbar disc deformation during ¯exion and extension of the trunk. Clin Biomech 1995;10(4):193± 99. [7] Burton AK. Regional lumbar sagittal mobility: measurements by ¯exicurves. Clin Biomech 1986;1(1):20±26.
M.L. Gatton, M.J. Pearcy / Clinical Biomechanics 14 (1999) 376±383 [8] Pearcy MJ, Gill JM, Whittle MW, Johnson GR. Dynamic back movement measured using three-dimensional television system. J Biomech 1987;20(10):943±949. [9] Peacy MJ. Twisting mobility of the human back in ¯exed postures. Spine 1993;18(1):114±19.
383
[10] McGill SM, Norman RW. Paritioning of the L4-L5 dynamic moment into disc, ligamentous and muscular components during lifting. Spine 1986;11(7):666±78.
Kinematics and movement sequencing during ¯exion of the lumbar spine Michelle L. Gatton *, Mark J. Pearcy Centre for Rehabilitation Science and Engineering, School of Mechanical, Manufacturing and Medical Engineering, Queensland University of Technology, GPO Box 2434, Brisbane, Queensland 4001, Australia Received 30 September 1998; accepted 2 December 1998
Abstract Objectives. This study investigated the sequence of intervertebral joint movements and range of motion during three tasks involving lumbar ¯exion. Design. Position sensors were used to measure position and rotation of lumbar vertebrae during unconstrained ¯exion. Background. In the development of mathematical models, numerous assumptions need to be made. Few studies have attempted to assess the validity of the assumptions regarding kinematics in models of the lumbar spine. Methods. Position sensors were attached to the skin overlying the lumbar vertebrae of 14 volunteers. Volunteers performed three ¯exion tasks; unconstrained ¯exion from upright standing, with and without a mass of 5 kg held close to the body, and the transition from upright standing to a seated position. Results. Four de®nitive movement sequences were identi®ed for those subjects with consistency between replicates; `top down' motion (where the top of the lumbar spine starts to move ®rst and the bottom moves last), `bottom up' (where the bottom of the lumbar spine moves ®rst and the top moves last), `all together' (where all segments commence movement together), and `middle last' (where the middle segments of the lumbar spine are last to commence movement). Subjects not ®tting one of these sequences were categorised into a miscellaneous group. Only two subjects exhibited the same sequence for each of the three tasks, while other subjects exhibited two or three dierent sequences for the three tasks, or showed a lack of consistency for one of the tasks. Conclusions. The results from this study indicate that there is no single movement sequence exhibited by the sample population. Relevance Incorrect assumptions which are incorporated into mathematical models have the potential to in¯uence model output. Given that output from spinal models is often used to assess ergonomic issues such as safe lifting loads, validation of the assumptions is essential. Ó 1999 Elsevier Science Ltd. All rights reserved. Keywords: Kimematics; Flexion; Lumbar spine; Range of motion; Movement sequence
1. Introduction The human lumbar spine is a complex structure consisting of ®ve vertebrae, associated intervertebral discs and many attached ligaments and muscles. Each of these components is fundamental for stability and movement. However, little is known about the interconnection between components. Mathematical models are one means used to explore the various interactions between bone, muscle and ligament. In the past, output
*
Corresponding author. E-mail: [email protected]
from mathematical models has provided information on ergonomic issues such as safe lifting postures and loads. The trend in recent times has been to produce dynamic or quasi-static mathematical models, as opposed to static models, in an eort to investigate the forces experienced or generated by each of the components in the lumbar spine under speci®ed conditions. However a mathematical model is only as good as the data and assumptions used within it. The sequence of motion is an important assumption within dynamic and quasi-static models, since joint position will impact on the allocation of forces to individual muscles by altering muscle orientation and maximal force. Although much kinematic data has been reported relating to the range of movement in the
0268-0033/99/$ ± see front matter Ó 1999 Elsevier Science Ltd. All rights reserved. PII: S 0 2 6 8 - 0 0 3 3 ( 9 8 ) 0 0 1 0 9 - 0
M.L. Gatton, M.J. Pearcy / Clinical Biomechanics 14 (1999) 376±383
377
Table 1 Summary of subject characteristics
All subjects Mean (SE) Median Range Male subjects Mean (SE) Median Range Female subjects Mean (SE) Median Range
Age (yr)
Height (m)
Weight (kg)
BMIa
31.4 (2.7) 29 18±46
1.70 (0.03) 1.67 1.59±1.88
71.6 (4.2) 65 55±105
24.8 (1.4) 23.3 18.6±38.1
29.4 (3.3) 28 18±40
1.76 (0.04) 178 1.59±1.88
73.9 (5.5) 70 55±92
23.8 (1.2) 23.2 19.4±29.0
33.4 (4.4) 30 23±46
1.65 (0.02) 1.65 1.59±1.75
69.4 (6.7) 65 55±105
25.7 (2.6) 23.9 18.6±38.1
p SE Standard Error Standard deviation/ n: a 2 BMI Body Mass Index weight/height .
lumbar spine [1±5], little literature discusses the actual movements, or sequence of movements of the lumbar spine between the extremes of motion. Work by Kanayama et al. [6] suggests that during ¯exion the L3=4 , L4=5 and L5 /S1 joints do not deform simultaneously. Their study [6] implies that the spine exhibits `top±down' motion such that the L3=4 joint is deformed before the L4=5 joint and before the L5 /S1 joint. This paper reports a study of the kinematics of the lumbar spine during several ¯exion tasks with particular reference to the sequence of intervertebral joint movements. Range of motion data resulting from the study are compared to previously reported studies and factors potentially in¯uencing range or sequence of motion data are also investigated. 2. Methods A sample of 14 subjects (male female) participated in the study. All subjects volunteered to participate and signed a consent form as required for ethical clearance. None of the subjects reported any previous back injury, surgery or spinal abnormality. Two subjects reported having mild scoliosis. A summary of the subject characteristics is presented in Table 1. A Motion Star motion analysis system (Ascension Technology Corporation, Vermont, USA) was used to record the orientation and position of sensors in three dimensions whilst subjects performed a variety of tasks. Preliminary evaluation showed that this system was accurate to 0.4° in the range of measurement. The sensors were attached to the skin over the spinous processes of the lumbar vertebrae using double sided tape. The locations of the spinous processes were determined as described by Burton [7], that being the identi®cation of the L4 spinous process as the bisection of a line joining the highest points on the iliac crests. Subsequent levels
were identi®ed by counting up or down from L4 . In some instances this identi®cation was made easier by asking subjects to bend forward slightly. To ensure the sensors stayed attached during movement, bandages were bound around the subject's trunk to apply slight positive pressure to the sensors. These bandages were bound so as to apply pressure to the sensors but not limit the movement of the subject. Unfortunately, due to the size of the sensors, it was not always possible to place sensors over consecutive spinal levels. In this instance the maximum distance between sensors was two levels. An additional sensor was ®tted to a plate which was moulded to ®t over the sacrum. Fig. 1 illustrates the attachment of the sensors. Data were collected 20 times per second for 6s. Tasks performed were classi®ed as primarily uniaxial (that is in one plane) or composite (that is involving more than one primary plane of motion). Only the uniaxial ¯exion tasks will be discussed here. Subjects were asked to undertake the movements with no restriction being placed on hip motion. This approach was used to obtain
Fig. 1. Location and attachment of sensors during data collection.
378
M.L. Gatton, M.J. Pearcy / Clinical Biomechanics 14 (1999) 376±383
Fig. 2. Flexion tasks performed by subjects (a) unloaded ¯exion, (b) loaded ¯exion, and (c) sitting.
data from `natural' body movements. Flexion from the standing position (hereafter referred to simply as ¯exion) was measured on four occasions, two replicates in an unloaded state, and two replicates in a loaded state where a 5kg barbell was held close to the body. The sitting task, in which subjects were asked to sit on a stool then stand up again, was also repeated twice. Fig. 2 provides a schematic illustration of these tasks. Changes in the angle of sensors from the original upright posture were calculated using the methodology of Pearcy et al. [8] Standard statistical techniques such as linear regression, analysis of variance (A N O V A ) and student's t-test were used to analyse continuous data while Fisher's Exact test was used to test for dierences in proportions for categorical data. In order to sequence the spinal movement, data were smoothed using a moving average with a window size of 5 data points. This smoothing helped to reduce the amount of small, erratic segmental movement exhibited by some subjects before and during the initial stages of motion. A segment was declared to have started motion when the average rate of change in the angle (or velocity) between adjacent sensors over two consecutive time periods was greater than or equal to 5° =s for each joint encompassed by the sensors (that is, if the sensors were separated by two intervertebral joints the average velocity over two consecutive time periods had to be greater than or equal to 10° =s). Sequencing was achieved by ordering the start of motion of the spinal segments. Segments needed to start motion at least 0.1 s apart to be declared as starting at dierent times.
along with the ranges presented hereafter, was obtained using data from 13 of the 14 subjects tested. One male subject had to be omitted from this section of the analysis due to a reporting discrepancy in sensor locations. There was a statistical dierence in the range of spinal motion between males and females during unloaded ¯exion, with males having a signi®cantly higher mean range of spinal ¯exion than females (66.0° vs 51.6°; P 0.008). No signi®cant relationship with age was detected (P > 0:6). There was a signi®cant change in spinal motion with subject height such that the range of spinal ¯exion increased with height (P 0.03). Fig. 3 illustrates the data on which this relationship was developed. The regression model has an R2 value of 0.38, exhibits a random residual plot with no apparent patterning and can be expressed mathematically as, Range of spinal flexion degrees ÿ63:79 0:72 Subject height cm: The 95% con®dence interval for the mean voluntary range of motion in loaded ¯exion from L1 to S1 was 54.3°±66.4°. This value is not signi®cantly dierent from the voluntary range of motion obtained during unloaded ¯exion for all subjects combined (P 0.154), and for males and females analysed separately (P 0.866 for males and P 0.109 for females). There appears to be an increasing trend in the range of spinal ¯exion (during loaded ¯exion) with height (R2 0:28; P 0.06). The mean spinal ¯exion from L1 to S1 in the seated position was 27.6° with a standard error of 2.3°. When considered as a proportion of the subject's full range of ¯exion (during unloaded ¯exion), the mean was 49.6% with the associated 95% con®dence interval being 40.5% to 58.8%. 3.2. Movement patterns Consistency in movement patterns was measured by comparing the two replicates of each task obtained from each subject. Consistency was judged based on the
3. Results 3.1. Range of motion The mean voluntary range of motion from L1 to S1 during unloaded ¯exion was 58.3° with a 95% con®dence interval for the mean of 51.4°±65.2°. This value,
Fig. 3. Range of spinal ¯exion with subject height for unloaded ¯exion.
M.L. Gatton, M.J. Pearcy / Clinical Biomechanics 14 (1999) 376±383
sequence of movement between intersegmental levels. A sequence is de®ned as the order in which intersegmental levels commence movement. Consistency was categorized as `consistent' if the movement sequence was identical between replicates, `almost consistent' if a segment changed it's position relative to the adjacent sensors in the sequence such that it starts moving at the same time as the adjacent sensor when it previously started before or after, or vice versa, and `not consistent' if there was a change in the ordering of the sequence (see Fig. 4). Sequences for all 14 subjects were analysed for the forward ¯exion tasks, while only sequences from 13 subjects were analysed for the sitting task due to a
Fig. 4. Examples of consistency categories. The shaded areas represent the ®rst two time intervals satisfying the movement criterion. Subject 1 exhibits a `consistent' sequence since the ordering of sensors has not changed between replicates. Subject 2 has `almost consistent' sequences since in the second repetition sensor 2 no longer moves with sensor 1 instead moving with (and not after) sensor 3. Subject 3 shows sequences which are `not consistent' since sensor 1 has changed its position from moving ®rst to moving last relative to the other sensors.
379
Table 2 Frequency of consistency patterns for unloaded and loaded ¯exion, and sitting
Unloaded Loaded Sitting
Consistent
Almost consistent
Not consistent
4 5 4
7 8 6
3 1 3
problem with data collection from one subject during the sitting trial. The allocation of subjects to a consistency class is given in Table 2 for unloaded and loaded ¯exion, and sitting. A comparison of the consistency of sequences for the three tasks indicates that there is no statistical dierence in the frequency of consistency categories under the dierent conditions (P > 0.6). Although a total of seven inconsistent sequences were recorded, these inconsistencies were spread amongst the subjects with no individual having more than one inconsistent sequence recorded for the three tasks (see Table 3). No single variable such as age, height, weight, BMI or sex was found to signi®cantly dierentiate those subjects having inconsistent and consistent movement sequences (P > 0:05). Examining the subject trials where `consistent' or `almost consistent' movement patterns were detected, a variety of movement sequences were observed. These sequences can be broadly classi®ed into four groups; `top down' (where the top of the lumbar spine is the ®rst to move and the bottom is the last to move), `bottom up' (where the bottom of the lumbar spine is ®rst to move and the top is last to move), `all together' (where all segments start to move together) and `middle last' (where the middle of lumbar spine is last to commence movement) (see Fig. 5). Sequences not ®tting one of these categories were classi®ed in a miscellaneous group. Examples of miscellaneous sequences are (L3 ±L4 then T12 ±L1 then L1 ±L3 then L4 ±S1 ) and (T12 ± L1 then L3 ±L5 then L5 ±S1 then L1 ±L3 ). Table 3 indicates which movement sequence group each subject was allocated to for the three tasks. Little consistency in the sequence group was found within individual subjects across the three tasks with only two subjects exhibiting the same sequence during all three tasks. A more common feature was to have the same sequence for two of the tasks and one task with a different or inconsistent sequence (7 subjects). Table 4 provides a summary of the frequency of each of the identi®ed movement sequences. When spinal ¯exion is disaggregated into individual rotations between sensors (something akin to investigating intersegmental rotation), the relative contribution of each segment can be investigated. Fig. 6 illustrates the proportioning of spinal ¯exion into segmental units for
380
M.L. Gatton, M.J. Pearcy / Clinical Biomechanics 14 (1999) 376±383
Table 3 Movement sequence category for individual subjects during unloaded and loaded ¯exion and sitting (subjects showing inconsistent sequences between replicates are indicated by `±', with `*' representing missing data)
Subject Subject Subject Subject Subject Subject Subject Subject Subject Subject Subject Subject Subject Subject
1 2 3 4 5 6 7 8 9 10 11 12 13 14
Unloaded ¯exion
Loaded ¯exion
Sitting
T O ± B A O A ± ± M A A O A
T A O B B O ± O B M M M M A
± A B ± M O A O B M T ± B *
A: all lumbar levels start motion together; B: `bottom up' motion; M: middle lumbar level is last to start moving; O: movement sequence does not ®t one of other groups; T: `top down' motion.
4. Discussion
Fig. 5. Sequencing intervertebral motion during forward ¯exion of the trunk. The S1 sensor (solid line) is an indicator of the amount of hip rotation. At point A, the T12 and L2 segments deviate from the L4 and S1 curves indicating that the T12 ±L2 and L2 ±L4 segments are starting to ¯ex. Also at point A the T12 curve deviates from the L2 curve meaning that the T12 ±L2 segment is ¯exing quicker than the L2 ±L4 segment. Flexion of the L4 ±S1 segment does not occur until after point B when the L4 curve deviates from the S1 curve. Therefore, the movement can be sequenced as T12 ±L2 then L2 ±L4 then L4 ±S1 , or `top down'.
three of the subjects sampled. Negative values indicate that this segment extended during the voluntary ¯exion movement. The patterns exhibited in Fig. 6 are indicative of the patterns seen in other subjects, namely that there is a wide variety of curve shapes and that no one segment appears to dominate movement in all subjects. For some subjects there appears to be a constant proportioning once movement has commenced, but this is not the case for all subjects. The curves in Fig. 6 do not start at zero ¯exion since at the very start of movement, the calculation of segmental contribution is based on very small rotations resulting in the proportions being large and highly variable.
The 95% con®dence interval for the mean range of spinal ¯exion from L1 to S1 obtained from this study appears to be in agreement with data reported in other studies [1,4]. The mean value of 58.3° compares well with results from McGregor et al. [4] who obtained a mean voluntary range of motion of 56.6° using a sample of 103 males and 100 females. Pearcy [9] reported ¯exion angles from L1 to S1 in a seated posture for 10 males subjects. The mean ¯exion reported was 44% of the maximum voluntary ¯exion. This value lies within the 95% con®dence interval for the mean proportion of maximum ¯exion in a seated posture obtained from this study (40.5±58.8%). The results from this study indicate that the range of spinal ¯exion in this subject population was not in¯uenced by age. Other studies have had mixed results when relating range of motion to age. It would appear that studies which have reported age dierences in spinal ¯exion include subjects from a wider age group compared to this study. For instance, the study by Gracovetsky et al. [1] which reported a decrease in spinal ¯exion with age, sampled people aged 19±64 yr with half the sample being 41 years or older (compared to the age range of this study of 18-46 years with median age of 29 yr). A signi®cant relationship was found between the range of spinal ¯exion and sex, with males having a larger range than females. This result is in agreement with the results reported by Hindle et al. [2]. An interesting result is the signi®cant relationship between subject height and range of spinal ¯exion. This relationship suggests that taller subjects have a larger
M.L. Gatton, M.J. Pearcy / Clinical Biomechanics 14 (1999) 376±383
381
Table 4 Summary of movement sequences during unloaded and loaded forward ¯exion and sitting Sequence
`Top down' `Bottom up' `All together' `Middle last' `Other'
Frequency of sequence Unloaded ¯exion
Loaded ¯exion
Sitting
1 1 5 1 3
1 3 2 4 3
1 3 2 2 2
Fig. 6. Proportional allocation of rotation to segmental units. Positive value represents ¯exion of joint while negative value represents extension of joint.
range of spinal ¯exion than shorter subjects. Biomechanically this appears feasible if one assumes that ligament strain is a limiting factor in spinal ¯exion. Taller people in general would be expected to have slightly longer torsos compared to shorter people, resulting in longer ligament lengths. If the limiting ligament strain is constant then it would be necessary for a longer ligament to change length more than a shorter ligament in order to obtain the same strain, resulting in an increased rotation. Some simple calculations enable this hypothesis to be tested to assess if the magnitude of rotation change is similar to that seen experimentally. We can investigate the eect of subject height change in a mathematical model of the lumbar spine simply by adjusting the initial height of the intervertebral discs. A 10 mm change in subject height approximately represents a 0.45 mm change in height of the lumbar spine. Spreading this change equally over the joints of the lumbar spine results in an increase in trunk rotation of 0.8°. This change in comparable to the observed relationship between height and range of spinal rotation where a 1cm height change results in a 0.72° increase in spinal ¯exion. The relationship between subject height and range of spinal ¯exion raises some interesting questions in relation to apparent dierences in range of spinal ¯exion between males and females. This study reported a signi®cantly larger range of spinal ¯exion for males than females. Taking into account that the mean height of males is 0.11 m larger than that for females, one must ask whether the dierence in range of spinal ¯exion between the sexes is not, at least partially, a remnant of the height dierence. Unfortunately this study is not large enough to further investigate this issue since there is not enough overlap between males and females of similar height to enable a valid comparison. The analysis of movement patterns highlighted the fact that individuals can perform a simple task like ¯exion using a number of dierent strategies. This raises an interesting problem when trying to model the kine-
382
M.L. Gatton, M.J. Pearcy / Clinical Biomechanics 14 (1999) 376±383
matics of the spine since there appears, from the limited sample of this study, to be no dominant movement pattern for any of the tasks considered. Kanayama et al. [6] suggest that during forward ¯exion the spine moves from the top ®rst. Although we did record this sequence, `top down' motion was only recorded in a small proportion (9%) of trials. The issue of movement sequencing is also complicated by the fact that approximately half of the sample group exhibited inconsistent movement sequences during one of the three tasks performed. Hence these people can perform a speci®c task twice using dierent movement sequences each time. This is an interesting ®nding given the ongoing debate into lumbar muscle recruitment strategies. If one assumes that muscle activity aects the rotation of individual intervertebral joints, it would be reasonable to conclude that if there is one `optimal' muscle recruitment strategy, then the spine would also move using only one sequence. This is obviously not the case for the tasks performed as part of this study. To investigate this idea, further research needs to be conducted during which more physically challenging ¯exion tasks are performed to place more demands on the muscles of the spine. One feature common to several subjects is the rotation of one level in the opposite direction to the overall rotation at some point during ¯exion (Fig. 6). Pearcy [3] experienced the same phenomenon when he recorded an extension of the L5 /S1 joint in one subject at full ¯exion. Past mathematical models of the lumbar spine have made various assumptions about the kinematics of the spine. For instance, McGill [10] allocates bending between the rib cage and pelvis to individual lumbar levels using the relationship Ria aia Ba where Ria is the rotation of the ith vertebra about axis a, aia is the percentage of rotation attributed to the ith vertebra about axis a and Ba is the total lumbar bending about axis a. For ¯exion McGill [10] uses aL1 aL2 13:2%, aL3 21%, aL4 29% and aL5 23:6%. This relationship implies that all lumbar levels start to move together and that there is a constant relationship between the angle of rotation at each lumbar level. The results of this study indicate that although approximately 27% of trials exhibited the sequence `all together', few of these cases showed constant proportioning of intersegmental angle. There also appears to be no constant relationship between the angles of rotation at each lumbar level across the entire range of ¯exion (Fig. 6). In some subjects there appears to be a constant proportioning between levels after movement has commenced, however even in these subjects it does not appear that any one level dominates movement. For instance, as illustrated in Fig. 6, the dominant segment in subject 10 is L1 ±L3 , while in subject 5 L3 ±L5 dominates and there is no dominant
segment in subject 13. Hence the assumption of a simple relationship between intervertebral movements is not supported by the results of this study. 5. Conclusion This study investigated the kinematics of the lumbar spine during unconstrained ¯exion of the lumbar spine, as well as the movement sequencing of the lumbar vertebrae between the extremes of motion. The results relating to range of motion are in general agreement with other published studies. This study has detected a relationship between subject height and range of spinal ¯exion and discusses a possible reason for this relationship as well as considering the implications of such a relationship. The study was instigated to investigate the sequencing of movement within the lumbar spine. To this end we have identi®ed a number of movement strategies and highlighted the wide dierences between subjects. In short, these results, although obtained from a small sample, indicate that the movement assumptions in mathematical models need to be carefully reviewed. Acknowledgements The authors wish to thank all the people who volunteered for this study, Dr. Pei Lai Cheng for his assistance with data collection and Dr. Graeme Pettet for his valuable comments. References [1] Gracovetsky S, Newman N, Pawlowsky M, Lanzo V, Davey B, Robinson L. A database for estimating normal spinal motion derived from noninvasive measurements. Spine 1995;20(9):1036± 46. [2] Hindle RJ, Pearcy MJ, Cross AT, Miller DHT. Three-dimensional kinematics of the human back. Clin Biomech 1990;5(4):218±28. [3] Pearcy MJ. Stereo radiography of lumbar spine motion. Acta Orthop Scand 1985;56(Suppl 212):212. [4] McGregor AH, McCarthy ID, Hughes SP. Motion characteristics of the lumbar spine in the normal population. Spine 1995;20(22):2421±28. [5] Panjabi MM, Oxland TR, Yamamoto I, Crisco JJ. Mechanical behavior of the human lumbar and lumbosacral spine as shown by three-dimensional load-displacement curves. J Bone Joint Surg[Am]1994;76A(3):413±24. [6] Kanayama M, Tadano S, Kaneda K, Ukai T, Abumi K, Ito M. A cineradiographic study on the lumbar disc deformation during ¯exion and extension of the trunk. Clin Biomech 1995;10(4):193± 99. [7] Burton AK. Regional lumbar sagittal mobility: measurements by ¯exicurves. Clin Biomech 1986;1(1):20±26.
M.L. Gatton, M.J. Pearcy / Clinical Biomechanics 14 (1999) 376±383 [8] Pearcy MJ, Gill JM, Whittle MW, Johnson GR. Dynamic back movement measured using three-dimensional television system. J Biomech 1987;20(10):943±949. [9] Peacy MJ. Twisting mobility of the human back in ¯exed postures. Spine 1993;18(1):114±19.
383
[10] McGill SM, Norman RW. Paritioning of the L4-L5 dynamic moment into disc, ligamentous and muscular components during lifting. Spine 1986;11(7):666±78.