Reliability testing.

Despite the frequency of low back pain (LBP), there is often great difficulty in establishing the underlying cause of its symptoms. Many studies have demonstrated a poor correlation between structural abnormalities and LBP. Videman et al,[1] in their study of cadaveric specimens, found no correlation between previous history of LBP and presence of lumbar spine disease, whereas Deyo et al[2] report that 10% to 30% of an asymptomatic population may demonstrate pathological findings on radiographic investigation.

Because of the difficulty in formulating an accurate structural or pathological diagnosis, physical therapists usually treat according to the patient's presenting signs and symptoms. One physical sign that clinicians have associated with LBP is increased lumbar posteroanterior (PA) stiffness.[3,4] Posteroanterior stiffness is said to increase in individuals with LBP due to degeneration of articular structures, shortening of joint capsules and adjacent ligaments, or the presence of paravertebral muscle spasm.[4]

Physical therapists assess lumbar PA stiffness using manual tests such as PA mobilization, in which the therapist applies an oscillating force to the spinous process of the prone patient.[4] Stiffness is perceived by estimating the amount of force applied and the amount of displacement that results. This information is then used in con unction with information from other examination procedures to select the spinal region to be treated and the most appropriate treatment technique. These clinical decisions, however, are based on manual testing procedures shown to be unreliable[5-7] and therefore prone to error.

This poor reliability may be due, in part, to confusion about what is meant by abnormal PA stiffness. Although some authors [4,8] have proposed different grading scales for measurement of PA stiffness, many of these grading scales lack operational definitions for the categories of mobility, and consequently there may be great variation in how therapists interpret these scales.[9]

In an attempt to improve the reliability and accuracy of stiffness measurement, several new mechanical devices have been designed to collect both force and displacement data during a simulated PA mobilization. Lee and Svensson[10] have developed the "Spinal Physiotherapy Simulator" to measure responses to PA forces. This device has been found to be accurate in the measurement of a series of elastic beams, with the system tending to underestimate beam stiffness by less than 1%. The measurement procedure using the device has been shown to be reliable, with Lee and Svensson[10] reporting an intraclass correlation coefficient (ICC) of .88 for measurement of the stiffness coefficient "K" at L-3 in asymptomatic subjects. The device, however, is large and is not portable, making clinical examination of symptomatic persons difficult.

Another group of researchers[11] have also reported the development of a device for measuring lumbar PA responses. This apparatus delivers a PA force to a selected vertebral segment while measuring the displacement of the skin over the adjacent vertebral segments. This device was found to be highly reliable in the measurement of relative displacement at L-3 (ICC=.99) and at L-5 (ICC=.95) when a PA force was applied to L-4.

Although some devices show good reliability, they are large and not portable and therefore have mainly been used for investigating the PA response characteristics in asymptomatic subjects.[12,13] Research on symptomatic subjects has been delayed due to the constraints of the measurement tools. We therefore considered it important to develop a more portable device that could be easily transported to physical therapy clinics and used to measure the response to the application of PA forces in individuals with LBP. The use of this new testing device in the clinic could improve the reliability of a commonly used assessment procedure and facilitate research investigating the relationship between abnormal spinal stiffness and symptoms. This report describes the test-retest reliability of a portable stiffness testing device in measuring responses to lumbar PA forces in a group of subjects with LBP.

Method

Subjects

Twenty-two subjects (10 female, 12 male) with nonspecific LBP volunteered to participate in the reliability study. Subjects with LBP were included in the study if they reported having pain within the last 24 hours and pain when the examining physical therapist manually applied a PA force to the lumbar spine. Subjects were excluded if they had contraindications to PA stiffness testing such as a known inflammatory, metastatic, or infective disease process; osteoporosis; spondylolythesis; or signs of acute nerve root compromise.

The subjects had a mean age of 27.4 years (SD=9.36, range=19-58) and had experienced pain for an average of 22.5 months (SD=37.95, range=0.1-120). The mean McGill pain questionnaire score [14] of the group was 11. 04 (SD = 7.31, range=2-32). The group included subjects with both constant and intermittent LBP. Subjects are described in detail in Table 1. Stiffness assessment was performed at the most painful lumbar spinal level, with 15 subjects tested at L-5, 4 subjects tested at L-4, 1 subject tested at L-3, and 2 subjects tested at L-2. Informed consent was obtained prior to testing.

Table 1. Subject Data (N=22)

x SD Range

Age (y) 27.4 9.36 19-58 Height (cm)

172.59 12.09 153.5-191.0 Moss (kg)

71.41 13.55 44-97 Body mass index (kg/cm2)

23.84 2.76 17.74-29.41 Duration of symptoms (mo)

22.46 37.95 0.1-120 McGill pain questionnaire score

11.04 7.31 2-32

Equipment

A portable stiffness device was developed to measure PA responses in the lumbar spine (Fig. 1). This device applies a PA force to the spine and measures the amount of resultant displacement. The force and displacement data are then used to calculate two indices of the response to PA forces.

The device consists of a testing bed, a small metal pad (indenter) that applies the force to the subject, a mechanical head (height= 395 mm, breadth = 145 mm, depth = 195 mm) that controls the movement of the indenter and measures the applied force and resultant displacement, and a small control box that houses the central processing unit(*) containing an analog-to-digital converter board and memory chips. The control box is connected to a laptop computer[dagger] for operation control and data collection. The specially designed bed consists of a rigid wooden surface that is hinged in the middle for portability and detachable aluminum legs. The mechanical head is attached to a steel bar that sits over the testing bed. The position of the mechanical head, and thus the indenter, can be adjusted in several ways. The height of the bar may be adjusted to enable assessment of patients of varying dimensions. The mechanical head also rotates about the bar to allow the force to be applied in a cephalic or caudal direction for testing different lumbar levels.

The components housed in the mechanical head are shown in Figure 2. A small reversible servomotor[double dagger] is stepped down via two pulleys to reduce the speed and increase the torque output of the motor. This step-down enables a maximum force production of 300 N at a maximum frequency of 2 Hz. An inextensible cable travels around the lower pulley and is attached to the top of the indenter. Movement of this cable produces movement of the indenter, with the reversibility of the motor producing both upward and downward movement.

Displacement of the indenter is calculated by measuring the rotation of the pulley using an optical encoder[sections] and then converting this angular displacement to linear displacement of the indenter.

Force is indirectly measured using a conductive plastic linear potentiometer.[parallel] As the indenter moves downward, applying a force to the patient, an equal and opposite force is provided upward by the prone patient. This force causes a small amount of compression of the spring. The amount of spring compression is measured by the linear potentiometer. The signal from the linear potentiometer is input to the control box, where the signal undergoes low-pass filtering and analog-to-digital conversion, and is then input to the computer with the digital displacement data via the serial port. For each cycle, force and displacement data are displayed graphically on the computer screen.

Operation of the stiffness device is controlled by a laptop computer via the central processing unit. Using the computer menu, it is possible to move the indenter to make contact with the skin, initiate cycling of the indenter, and retract the indenter. The number of loading cycles, the speed of cycling, and the maximum force and displacement achieved during cycling can be specified by the operator.

The device includes a number of safety features that ensure that the applied force remains within the limits of safety. These safety features include a default setting, which ensures that a maximum force of 105 N is not exceeded, and circuitry that shuts down the device should a fault in the measurement of force occur.

The accuracy of the device was evaluated by measuring the stiffness of a series of elastic beams and comparing the values with those calculated using weights to apply a known force and a dial indicator to accurately measure displacement. This testing indicated that the device underestimated the stiffness of all beams, with a maximum error of 2.5% (Appendix).

Procedure

To investigate the test-retest reliability of the stiffness device, two measurements of PA responses were obtained for each subject on a single day. All subject testing was conducted by the same examiner (JL), who had performed over 20 tests using the new stiffness assessment device prior to commencing the reliability study.

Prior to testing, anthropometric data and a brief history of LBP were collected from each subject. Subjects were asked to complete a McGill Pain Questionnaire[14] regarding the typical amount of pain experienced over the preceding 24 hours.

Posteroanterior testing was then performed as follows. Subjects were asked to lie prone on the testing table while the examiner manually applied a PA force to each lumbar level. The most painful spinal level was identified and marked. This level was identified after first locating L-5 using the method described by Grieve.3(p5) the most painful level was defined as the level that best reproduced the subject's presenting pain or the level with the greatest amount of local pain that could logically refer pain to the symptomatic area. The same markings were used for both the initial test and the repeat test. The position adopted by the subject during both tests was standardized by having the subject lie with arms by the side and knees over a pillow.

The mechanical head was tilted in either a cephalic or caudal direction, depending on the level to be tested, to enable a force to be applied perpendicular to average vertebral body axis directions. The reason for directing the force in such a manner was to mimic the manual clinical assessment of PA stiffness as performed by physical therapists. The directions used for L2-5 were cephalad 5.5 degrees for both L-2 and L-3, caudad 4.5 degrees for L-4, and caudad 16 degrees for L-5. The directions used for L3-5 have been used in previous research[15] and are within the range used by experienced physical therapists.[16] The indenter was then moved into contact with the skin at the level to be tested. Several preliminary cycles were performed to establish the appropriate displacement for the individual subject, to familiarize the subject with the measurement procedure, and to precondition the viscoelastic spinal tissues[17] prior to data collection. Data were then collected for five loading cycles to a force of 105 N at a frequency of 0.5 Hz. This magnitude of force and the frequency of cycling were chosen to try to replicate the manual assessment of PA stiffness. During data collection, subjects were asked to hold their breath at functional residual capacity to standardize the lung volume at which stiffness testing was performed.[18]

After a period of 5 minutes, the subject was retested. Between tests, the mechanical head was removed and the subject lay prone on the testing table. The head was then repositioned over the tested level, and the procedure was repeated exactly.

Data Analysis

For each test occasion, five loading cycles were obtained; however, the force-displacement data for four loading cycles only (cycles 2-5) were analyzed. Cycle 1 was discarded due to transient behavior on start-up. The remaining four cycles were averaged, and a mean force-displacement curve obtained.

Two measures were then calculated to quantify the main characteristics of the force-displacement curve obtained when measuring the lumbar spine. These characteristics were first, the nonlinear toe region, where small forces produce relatively large displacements, and second, the linear region of higher stiffness for forces above 20 to 30 N (Fig. 3). In previous studies evaluating PA stiffness in the spine,[12,13] only the linear portion of the force-displacement curve has been investigated. These studies report a coefficient of stiffness ("K"), calculated as the slope of a regression line fitted to the linear region of higher stiffness. The nonlinear part of the loading cycle, however, may also be important in judgments of abnormal PA stiffness and therefore was also analyzed in our study. The nonlinear region was quantified by calculating the displacement of the indenter from 0.5 N to 30 N. Thus, the two measures calculated in this study were the stiffness coefficient (K) and the displacement from 0.5 to 30 N (D30).

Intraclass correlation coefficients (ICC[2,1]) were calculated (using software developed at The University of Sydney) for the stiffness measures to determine the test-retest reliability. Due to the small number of subjects tested at L2-4, ICCs were calculated by combining the data from all levels. The percentage of agreement was also calculated. To detect any systematic changes in stiffness due to repeated testing, a t test for paired samples was calculated to compare the initial and second measurements. The significance level was set at P<.05.

Results

The mean K values for L2-5 were 12.78, 17.46, 15.94, and 15.14 N/mm, respectively (Tab. 2). High test-retest reliability was observed for measurements of the slope of the linear region of the force-displacement curve (K), with an ICC (2,I) value of .96. The ICC 95% confidence interval was .91 to.98. On 90% of occasions, the test-retest difference was less than 1.8 N/mm. No difference was found between the initial and second measurements, indicating that there was no systematic effect produced by repeat testing (P=.82).

Table 2. Slope of the Linear Portion of the Force-Displacement Curve (K) (in Newtons per Millimeter) and Length of the Low-Stiffness or "Toe" Region (D30) (in Millimeters) for Test 1 and 2 for Each Subject

K(N/mm) D30(mm) Level Test 1 Test 2

Test 1 Test 2 L-2 10.84 10.44 3.49

3.55

14.72 12.66 4.49 4.35

L-2X 12.78 11.55 3.99 3.95

L-3 17.46 17.30 4.83 4.45

L-3X 17.46 17.30 4.83 4.45

L-4 8.87 8.95 8.28 8.53

26.65 27.71 3.58 3.35

12.39 12.48 6.04 5.56

15.83 15.43 4.96 6.04 L-4X 15.94

16.14 5.71 5.87

L-5 12.94 13.43 5.84 6.13

15.06 13.77 6.66 6.64

14.56 14.80 5.05 5.13

14.12 13.82 4.67 4.98

12.22 11.46 5.85 5.81

11.36 11.31 5.55 6.87

15.26 15.40 5.21 5.02

19.47 18.77 4.40 3.99

19.90 18.11 2.67 3.19

14.15 14.66 5.70 5.50

11.09 10.86 8.20 8.08

14.98 14.20 4.61 4.90

17.12 16.76 5.61 5.81

22.50 24.07 5.57 5.52

12.32 16.09 7.58 7.56

L-5X 15.14 15.17 5.54 5.68

The mean displacement (D30) values for L2-5 were 3.99, 4.83, 5.71, and 5.54 mm, respectively (Tab. 2). The ICC value obtained was high (ICC[2,1]=.89). The ICC 95% confidence interval was .76 to .95. On 90% of occasions, the test-retest difference was less than 1.1 mm. No difference was found between the initial and second measurements, indicating that there was no systematic effect on the displacement produced by repeated testing (P=.32).

There was generally very high consistency in both K and D30 values between successive cycles of loading, that is, between cycles 2 to 5 of the test. Tables 3 and 4 show typical variations in responses between cycles.

Table 3. Slope of the Linear Portion of the Force-Displacement Curve (K) (in Newtons per Millimeter) for Cycles 2 to 5 for Typical Subjects at Each Tested Level

Subject Cycle 2 Cycle 3 Cycle 4 Cycle 5

A (L-2) 11.79 10.76 10.53 10.37 B (L-3) 16.95

17.85 16.83 16.85 C (L-4) 8.81 8.96

8.88 8.81 D (L-5) 11.14 10.90 10.64

11.73

Table 4. Length of the Low-Stiffness or "Toe" Region (D30) (in Millimeters) for Cycles 2 to 5 Typical Subjects at Each Tested Level

Subject Cycle 2 Cycle 3 Cycle 4 Cycle 5

A (L-2) 3.23 3.46 3.64 3.63 B (L-3) 4.66

4.81 4.94 4.91 C (L-4) 7.96 8.42

8.31 8.44 D (L-5) 8.30 8.31 8.26

7.93

Discussion

Our results demonstrate that measurements of responses to PA forces in subjects with LBP made using the stiffness testing device have high reliability when performed by the same experienced examiner. Although several previous studies have demonstrated high reliability using a mechanical device to measure the response to lumbar PA forces in asymptomatic subjects,[10,11] our study is the first to demonstrate reliable measurement in subjects with LBP.

The high reliability coefficients obtained for the stiffness coefficient (K) and the displacement (D30) contrast markedly with the poor reliability found for manual assessment of PA stiffness.[5-7] Although therapists have proposed a relationship between spinal symptoms and PA stiffness, the poor reliability of manual assessment has meant that this relationship has not been investigated. The development of a device that reliably measures the response of the lumbar spine to PA forces may enable investigation of this proposed relationship. In addition, the portability of the new device will facilitate testing of a clinically relevant population.

Bench testing has demonstrated that the device is accurate in its measurement of PA stiffness when tested on a series of elastic beams. The accuracy of the device in measuring the response of the lumbar region to PA forces, however, is yet to be examined, although the device appears to have good face validity, as it closely replicates the manual assessment procedure. The issue of whether either manual or mechanical assessments of PA stiffness actually reflect properties of the spine relevant to spinal disorders is not yet known. Validity issues such as the ability of measurements obtained using the device to discriminate between symptomatic and asymptomatic persons, to classify patients, and to guide treatment need to be investigated. Criterion-related validity, determined by comparing measurements with those obtained using other established mechanical devices,[10] also needs to be investigated.

In our study, several decisions were made about how to characterize the force-displacement curve obtained when applying a PA force to the lumbar region. It is currently unclear how physical therapists interpret force and displacement data to make decisions about stiffness. We chose, therefore, to measure two variables that we believe best described the key features of the force-displacement relationship obtained when applying a PA force to the lumbar spine. These two features were the length of the nonlinear toe region and the gradient of the higher-stiffness linear region. Although the stiffness coefficient (K) has been quantified by previous researchers, our study also described a measure of the low-stiffness region, that is, the displacement (D30) to an arbitrary low force of 30 N. The good reliability for this variable (ICC=.89), together with its value in describing the low-stiffness region of the PA response, suggests that D30 may be a useful variable to measure when evaluating the force-displacement relationship during the application of a PA force to the spine.

Conclusion

The reliability of measurements of aspects of the PA force-displacement relationship using a new measurement device was determined in subjects with LBP. Reliability coefficients were high. Bench testing also demonstrated a high degree of accuracy of measurements made with the device on beams of varying stiffness. The clinical significance of the PA force-displacement relationship remains to be established.

Appendix

Accuracy Testing

In order to assess the accuracy of the stiffness assessment device, three aluminum beams of varying stiffness were tested with the new device and by applying known forces to the beams and accurately measuring the resultant displacement. This Appendix reports the results of accuracy testing.

Method

Three regular aluminum channel sections of approximately 1.8 m each were tested. The beams were supported on a steel frame for testing. Stiffness of the beams was established by using five deadweights (measured using an electronic scale(a) to 0.001 kg) to apply known forces to the beam and measuring the resultant displacements as each weight was applied. Weights were applied in increments to a total of 100 N. The displacements were measured using a dial indicator.(b) From these force-displacement data, the stiffness of the beams was calculated. The stiffness values of the three beams were similar in magnitude to the stiffness values encountered when testing the lumbar spine.

The beams were then tested using the new stiffness assessment device. Force-displacement data were collected for I cycle at 0.5 H, to a maximum force of approximately 100 N. The data were adjusted using regression equations. The data were low-pass filtered with a cutoff frequency of 10 Hz. Linear regression was then performed on the force-displacement relationship. The linear stiffness of each beam was calculated by determining the gradient of the force-displacement line.

Results

The Table below shows the stiffness values obtained for each beam using the deadweights and dial indicator, the stiffness values obtained using the new stiffness assessment device, and the percentage of measurement error. The Table shows that the difference in the stiffness values is very small. The measurement device underestimated the true stiffness for all measurements with a maximum error of 2.5%. Thus, a high level of accuracy has been found for the stiffness assessment device.

Table. Stiffness Values (in Newtons per Millimeter) Obtained Using Weights and Dial Indicator (Method 1) and the New Stiffness Device (Method 2), and the Percentage of Error (%E)

Beam Method 1 Method 2 %E

1 6.91 6.86 1.5 2

7.55 7.36 2.5 3 15.9

15.65 1.5

(*)Nebula Electronics Pty Ltd. 12 Karimbla Rd, Miranda, New South Wales, Australia. [dagger]Toshiba 386Dx, Model T3200, Protech Australia, 110 Christie St, St Leonards, New South Wales, Australia. [double dagger]Airpax Model 9904 120 13316, Philips Electronics, 15 Blue St, North Sydney, New South Wales, Australia. [sections]Hewlett Packard, HEDS-550, 17-23 Talavera Rd. North Ryde, New South Wales, Australia. [parallel]Sakae, Model 15FLP15A, RS Components Pty Ltd. 129 Beaconsfield St, Silverwater, New South Wales, Australia. (a)Data Hytech, Scales Pty Ltd, 3 Reserve St. Preston, Victoria, Australia 3072. (b)Model 2050-11, Mitutoyo Corp, 31-19 Shiba5-Chome, Minato-ku, Tokyo 108, Japan.

References

[1] Videman T, Nurminen M, Troup J. Lumbar spinal pathology in cadaveric material in relation to history of back pain, occupation, and physical loading. Spine. 1990;15:728-738.

[2] Deyo RA, Haselkorn J, Hoffman R, Kent DL. Designing studies of diagnostic tests for low back pain or radiculopathy. Spine. 1994;19:2057S-2065S.

[3] Grieve G. Mobilization of the Spine: Notes on Examination, Assessment, and Clinical Method. Edinburgh, Scotland: Churchill Livingstone; 1984.

[4] Maitland G. Vertebral Manipulation. London, England: Butterworth & Co (Publishers) Ltd; 1986.

[5] Matyas T, Bach T. The reliability of selected techniques in clinical arthrometrics. Australian Journal of Physiotherapy. 1985;31:175-199.

[6] Binkley J, Stratford P, Gill C. Interrater reliability of lumbar accessory motion mobility testing. Phys Ther. 1995;75:786-795.

[7] Maher C. Adams R. Pain and stiffness assessment reliability in clinical manual lumbar spine examination. Phys Ther. 1994;74:801-811.

[8] Jull G, Bullock M. A motion profile of the lumbar spine in an aging population assessed by manual examination. Physiotherapy Practice. 1987;3: 70-81.

[9] Riddle DL. Measurement of accessory motion: critical issues and related concepts. Phys Ther. 1992;72:33-42.

[10] Lee M, Svensson N. Measurement of stiffness during simulated spinal physiotherapy. Clin Phys Physiol Meas. 1990;1 1:201-207.

[11] Lee R, Evans J. Load-displacement-time characteristics of the spine under posteroanterior mobilization. Australian Journal of Physiotherapy. 1992;38: 115-123.

[12] Lee M, Latimer J, Maher C. Manipulation: investigation of a proposed mechanism. Clin Biomech. 1993;8:302-306.

[13] Lee M, Esler M, Mildren J, Herbert R. The effects of extensor muscle activation on the response to lumbar posteroanterior forces. Clin Biomech. 1993;8:115-119.

[14] Melzack R. The McGill pain questionnaire: major properties and scoring methods. Pain. 1975;1:277-299.

[15] Lee M, Liversidge K. Posteroanterior stiffness at three locations in the lumbar spine. J Manipulative Physiol Ther. 1994; 1 7:511-516.

[16] Viner A, Lee M. Direction of manual force applied during assessment of stiffness in the lumbosacral spine. J Manipulative Physiol Ther. 1995;18:441-447.

[17] Loebl W. The assessment of mobility of metacarpophyseal joints. Rheum Phys Med. 1972; 1 1:365-378.

[18] Beamont A, McCrumb C, Lee M. Differences in the postero-anterior stiffness of the lumbar spine during tidal breathing and breath holding. In: Proceedings of the Seventh Biennial Conference of the Manipulative Physiotherapist Association of Australia, Sydney, New South Wales, Australia. 1991:244-251.