Ultrasound imaging of lumbar multifidus muscle: normal reference ranges for measurements and practical guidance on the technique

ARTICLE IN PRESS

Manual Therapy 10 (2005) 116–126 www.elsevier.com/locate/math

Original article

Ultrasound imaging of lumbar multifidus muscle: normal reference ranges for measurements and practical guidance on the technique M. Stokesa,c,, G. Rankina, D.J. Newhamb a

Institute of Complex Neuro-disability, Royal Hospital for Neuro-disability, London, UK b Centre for Applied Biomedical Research, King’s College London, UK c School of Health Professions and Rehabilitation Sciences, University of Southampton, Highfield Campus, Southampton, Hants SO17 1BJ,UK

Abstract This cross-sectional, prospective study aimed to produce normal reference data for measurements of the lumbar multifidus muscle. A total of 120 subjects, 68 females (aged 20–64 years) and 52 males (20–69 years) were studied. Bilateral transverse ultrasound images were made of multifidus at the fourth and fifth lumbar vertebrae (L4 & L5). Cross-sectional area (CSA, cm2) and linear dimensions (AP, anteroposterior; Lat, lateral) were measured and the latter expressed as a ratio (AP/Lat) to reflect shape. Relationships between CSA and anthropometric measures were examined. Multifidus CSA was larger in males (Po0.001) and age had no effect. The CSA was larger at L5 than L4 (Po0.001) and highly correlated between the two levels (males r=0.82, females 0.80). Differences in muscle shape were observed for gender, age and vertebral level. Between-side symmetry was high for size but not shape (CSA o10% difference). Linear measurements multiplied (AP  Lat) correlated highly with CSA (all groups rX0.94, Po0.0001). The AP dimension was also acceptably predictive of CSA at L4 (rX0.79). There were no clinically useful correlations between CSA and anthropometric measures. These findings provide normal references ranges for objective assessment of lumbar multifidus. This paper also addresses specific practical issues when scanning multifidus. r 2004 Elsevier Ltd. All rights reserved.

1. Introduction Ultrasound imaging of the lumbar multifidus muscle is of increasing interest to physiotherapists, both for clinical and research purposes. Clinically, the application is twofold: as an objective assessment tool for detecting abnormalities and monitoring changes during recovery (Hides et al., 1994, 1996); and for visual biofeedback during re-education of muscle contraction (Hides et al., 1998). The characteristics of multifidus for which normal data are available include cross-sectional area (CSA), linear measurements and shape, in small and relatively young populations (Hides et al., 1992, 1994). Corresponding author. Chair in Neuromuscular Rehabilitation, School of Health Professions and Rehabilitation Sciences, University of Southampton, Highfield Campus, Southampton, Hants SO17 1BJ,UK. E-mail address: [email protected] (M. Stokes).

1356-689X/$ - see front matter r 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.math.2004.08.013

Measurement of muscle size using ultrasound has provided an accurate assessment of muscle wasting in various muscles (see Stokes et al., 1997 for review). In acute low back pain (LBP), severe atrophy of multifidus was found to be selective and confined to the vertebral level and side of pain symptoms (Hides et al., 1994). The technique was also useful in demonstrating that multifidus size does not recover when pain subsides unless it undergoes specific exercises (Hides et al., 1996). For the technique to be applicable to populations other than those studied previously, normal data are required for subjects over a wider age range. Of the various skeletal muscles explored with ultrasound scanning (see Stokes et al., 1997 for a review), lumbar multifidus is one of the most difficult to image and interpret. This is mainly because its lateral border with longissimus (an erector spinae muscle) is often not clear enough to see without refinements in the procedure.

ARTICLE IN PRESS M. Stokes et al. / Manual Therapy 10 (2005) 116–126

When imaging multifidus, it is important to understand its functional anatomy, the main feature of which is the segmental arrangement of its fibres (Bogduk, 1997; Bogduk et al., 1992; Macintosh et al., 1986). Multifidus has both movement and stabilising roles. Working bilaterally with the other lumbar muscles, multifidus produces extension of the lumbar spine (Bogduk et al., 1992) and acts as a stabilizer in rotation, counterbalancing the flexion force produced simultaneously with rotation by the oblique abdominal muscles (Bogduk, 1997). Knowledge of this functional anatomy assists in imaging the muscle and using various manoeuvres for biofeedback. The present study examined multifidus size and shape in normal subjects, and sought predictive equations for estimating the expected size. The paper also provides practical details of the imaging technique to help obtain and interpret scans. Aims 1. To provide reference ranges for multifidus size and shape at vertebral levels L4 and L5 in normal males and females over a wide age range. 2. To investigate the degree of symmetry of multifidus size and shape. 3. To examine the relationship between multifidus size and various anthropometric variables to provide predictive equations for muscle size.

2. Methods Real-time ultrasound images of lumbar multifidus were taken bilaterally at vertebral levels L4 and L5. 2.1. Subjects A total of 120 subjects, 68 females (aged 20–64 years) and 52 males (aged 20–69 years), were studied (see Table 1 for demographic details). The discrepancy in numbers of subjects studied at the two vertebral levels was due to initial subjects only being scanned at L4. Subjects were either sedentary or moderately active. Sedentary subjects were in occupations involving light or no manual work and did not take part in sports. Those moderately active were in occupations involving moderate manual work but no heavy labour and/or taking part in recreational sports up to four times a week but not competitively above club level. Exclusion criteria were a history of neurological, neuromuscular, rheumatological or systemic disease; pregnancy, medication which might affect muscle size; any skin condition or wound in the area to be scanned; a lifetime history of low back pain of a severity to interfere with activities of daily living or require treatment; lifetime history of spinal or pelvic fractures, lumbar

117

surgery or any known spinal abnormality, such as scoliosis or spondylolithesis. The project received ethics approval from the Riverside Research Ethics Committee and King’s College London Research Ethics Committee. Written informed consent was obtained from subjects. 2.2. Procedure of ultrasound scanning The subject lay prone with the forehead resting just above the breathing hole in the plinth, the head in the midline and the arms supported on the plinth’s armrests. One or two pillows were placed under the hips to eliminate the lumbar lordosis. The spinous processes were palpated and marked with a chinagraph pencil. The spinous process of L5 was identified by palpating cranially from the sacrum. It is a deep, small, blunted bony point lying at the centre of the lumbosacral depression. The spinous process of L4, which is described as a comparatively large and sagitally ridged eminence (Grieve 1983), was identified by palpating cranially from L5. An Aloka SSD 1200 ultrasound scanner was used with a 5 Mz convex (50 mm footprint) transducer (Aloka Co. Ltd, Mitaka-shi, Tokyo, Japan). The transducer was first placed longitudinally over the lower lumbar spine, in the mid-line, to orientate and confirm the marks on the skin. This produced a scan of the spinous processes, which resembled the ‘Loch Ness Monster’ (as described by Nelmarı´ van Huyssteen MCSP, personal communication, with permission), seen in Fig. 1. The transducer was then rotated through 901 to lie transversely in the midline and the spinous processes and laminae were identified on a cross-sectional scan (Fig. 2). The transducer was then moved laterally to each side to image the left and right multifidus muscles (see Fig. 3 for L4 and Fig. 4 for L5). The echogenic vertebral laminae were used as a consistent landmark to identify the deep border of the muscle. In cases where it was difficult to clearly distinguish the lateral border of multifidus from longissimus, the subject was asked to contract the muscle by slightly raising the leg on the ipsilateral side and then relaxing before the image was taken. Images were captured, stored and measured off-line using an ultrasound image and analysis system (Department of Medical Physics and Bio-Engineering, St George’s Hospital, Tooting, London), consisting of computer software, a National Instruments PCI-1408 analogue frame grabber and pentium-based PC running Windows 95. 2.2.1. Ultrasound measurements The cross-sectional area or CSA (cm2) of multifidus was measured by tracing around the inner edge of the

ARTICLE IN PRESS M. Stokes et al. / Manual Therapy 10 (2005) 116–126

118 Table 1 Demographic details of subjects n Vertebral Level L4 Males 52

Age (years) Height (m) Body mass (kg) BMI (kg/m2)

Vertebral Level L5 Females 68

Males 45

Females 46

Mean

SD

Mean

SD

Mean

SD

Mean

SD

40.1 1.78 82.8 25.8

13.0 0.06 11.0 3.2

34.2* 1.65*** 62.9*** 23.0***

12.8 0.06 8.9 3.1

39.0 1.77 82.5 25.7

13.0 0.06 10.4 2.9

31.6** 1.66*** 61.8*** 22.3***

11.7 0.06 7.2 2.2

Males were significantly older, taller, heavier and had greater body mass index (BMI) than females. n=number of subjects. Significant differences between males and females: ***Po0.001 **Po0.01 *Po0.05.

muscle border with the on-screen cursor. Two linear measurements were made, defined as the greatest depth (anteroposterior, AP) and the greatest width (lateral dimension), lying perpendicular to the AP dimension. Muscle shape was expressed as a ratio of the linear measurements, with the AP divided by the lateral dimension (AP/Lat), as described by Hides et al. (1992). The length (cm) of the spinous process (SPL) and horizontal distance (cm) between the lateral edge of each lamina (bilateral lamina width) were also measured to examine their relationships with CSA.

Fig. 1. Saggital (longitudinal) view of the lumbar spine. A central image over the spinous processes (SP) resembles the ‘Loch Ness Monster’. This image is used to orientate prior to transverse scanning or moving laterally to obtain a longitudinal image over multifidus for biofeedback (usually with a linear probe, as in Fig. 6). (5 MHz Linear probe.)

Fig. 2. Bilateral transverse scan at the level of the fourth lumbar vertebra (L4), showing the spinous process in the centre of the image and the echogenic laminae (L) appearing as bright white horizontal landmarks, either side of the base of the spinous process (SP) and beneath multifidus (M). (5 MHz curvilinear probe.)

2.2.2. Reliability of the ultrasound procedure Ten normal subjects (five males) were selected to provide reliability data for a wide range of ages (27–58 years), height (1.49–1.82 m), weight (58.8–89.9 kg) and BMI (24.1–33.4 kg/m2). Lumbar multifidus at L4 on the left side was scanned on two occasions, one week apart in a blinded fashion, at the same time of day by one operator (GR). On each day, two images were taken to establish between-scan reliability. To relocate the scanning sites accurately, the surface marking for L4 spinous process was traced onto a transparent sheet, together with bony landmarks and any permanent skin blemishes, such as freckles and scars. The CSA of multifidus, SPL and bilateral lamina width were measured. Data analysis included intraclass correlations (ICCs; Chinn, 1990; Riddle et al., 1989), and Bland and Altman tests (Bland and Altman, 1986). Muscle and anthropometric measurements were highly repeatable. The ICCs for multifidus CSA ranged between 0.98 and 1.00. Bland and Altman tests produced values of d that were close to zero and SDdiff values were very low. The 95% limits of agreement for between-scans reliability was approximately 0.25–0.5 cm2, whilst for between-days reliability 0.62–0.67 cm2.

ARTICLE IN PRESS M. Stokes et al. / Manual Therapy 10 (2005) 116–126

119

Fig. 4. Transverse scan of the right multifidus muscle at the fifth vertebral level (L5). The spinous process (SP) is shorter and the lateral edge less steep than at L4. (5 MHz curvilinear probe.)

anterior superior iliac spines (ASIS) was measured with the subject lying supine. 2.4. Analysis of data Differences in age, height, body mass and BMI were compared between males and females using independent sample t-tests.

Fig. 3. Transverse scans of the left multifidus muscle at the fourth vertebral level (L4). The round shape of the multifidus (M) muscle is clearly evident in (a) and is oval in (b). Multifidus is bordered by the lamina (L) and spinous process (SP) as indicated. The lateral border is marked by the abrupt ending of transverse fasciae of longissimus (Lo), in both cases. (5 MHz curvilinear probe.)

For L4 SPL and bilateral lamina width, ICCs ranged from 0.95 to 0.99 for between-scans and between-days. Inter-rater reliability between two of the authors (GR and MS) was previously shown to be high for scanning the anterior tibial muscles: ICC=0.92 (Rankin and Stokes, 1998).

2.4.1. Multifidus size and shape Data for multifidus size and shape were not significantly different between the right and left sides and so were averaged for each subject, and the means and SD calculated for each group. Results for males and females were compared using independent sample t tests, and for L4 and L5 using paired t tests. Subsequently, due to differences observed, data for males, females and the two vertebral levels were treated as four separate groups. The ranges of values of muscle CSA and shape (ratios described above) for 95% of the sample populations (mean+2SD) were calculated. Muscle size and shape in different age groups (detailed later) were compared using one-way ANOVA.

2.3. External anthropometric measurements

2.4.2. Degree of symmetry of multifidus size and shape The difference between sides was calculated by dividing the value from the larger side by the smaller value and expressed as a percentage (% difference= [(largest/smallest value)  100]–100). Means, SD and 95% ranges (mean72SD) for symmetry were calculated for both size and shape.

The distances (cm) between the spinous processes of C7 and L5, and between the posterior superior iliac spines (PSIS) were measured using a tape measure with the subject lying prone. The distance (cm) between the

2.4.3. Relationship between muscle area, linear and anthropometric measures The relationship between multifidus CSA and age, height, body mass, BMI and anthropometric variables

ARTICLE IN PRESS 120

M. Stokes et al. / Manual Therapy 10 (2005) 116–126

were investigated using Pearson’s correlation coefficient, and stepwise linear regression analysis. The relationships between multifidus CSA and linear dimensions, and between CSA at L4 and L5 were examined using Pearson’s correlation coefficient and linear regression.

3. Results Multifidus CSA was significantly greater in males than females and also larger at L5 than at L4. Age had a significant effect on shape but not size. There were statistically significant correlations between CSA with SPL and lamina width but were not considered clinically significant to enable accurate prediction of multifidus size. High correlation between L4 and L5 CSA enables the size of one to be predicted from the other using regression equations. Linear measurements were more predictive of CSA at L4 than at L5 and the multiplied dimensions gave the best results. 3.1. Multifidus size The means, SD and reference ranges (95% of population studied) for multifidus size at L4 and L5 are presented separately for males and females in Table 2. 3.1.1. Effect of gender Males had significantly greater multifidus CSA (Po0.001) but when normalized for body mass, there was not a significant gender difference (Table 3). However, it can be seen in Table 1 that males were significantly older, taller, heavier and had greater BMIs, so the two groups were analysed separately. 3.1.2. Effect of age Multifidus CSA did not correlate significantly with age. There was no significant difference in size between different 10-year age bands, for either gender or for either vertebral level. There were, however, qualitative differences observed in terms of greater echogneicity with increasing age in some cases. 3.1.3. Effect of vertebral level The CSA was significantly larger (Po0.001) at L5 than L4, by a mean of 14% in males and 21% in females (Table 4). The gender difference in this effect of vertebral level was significant (Po0.05). There were high and significant (Po0.001) correlations between the CSA at L4 and L5 in both males (r=0.82) and females (r=0.80). Regression equations for predicting multifidus CSA at L4 and L5 are shown in Table 5.

Table 2 Multifidus size and shape at the fourth (L4) and fifth (L5) vertebral levels Males

Females

P-value

L4 n CSA (cm2) Mean SD 95% Reference range

52

68

7.87 1.85 4.24–11.50

5.55 1.28 3.03–8.06

0.000

Shape ratio Mean SD 95% Reference range

1.02 0.15 0.72–1.33

1.05 0.21 0.64–1.47

0.430

L5 n CSA (cm2) Mean SD 95% Reference range

45

46

8.91 1.68 5.62–12.20

6.65 1.00 4.69–8.60

0.000

Shape ratio Mean SD 95% Reference range

1.03 0.17 0.70–1.36

0.95 0.17 0.62–1.28

0.025

Males had significantly larger muscles than the females at both levels. n=number of subjects. CSA=cross-sectional area. P values for differences between males and females.

Table 3 Multifidus size normalized for body mass Males

Females

P-value

L4 Mean SD

0.10 0.02

0.09 0.02

0.140

L5 Mean SD

0.11 0.02

0.11 0.02

0.914

Values calculated by dividing muscle cross-sectional area (cm2) by body mass (kg).

3.2. Multifidus shape The results for shape ratios of multifidus in crosssection are presented in Table 2. Different shapes in terms of muscle outline were observed at L4, such as oval, round and triangular (see Fig. 5). 3.2.1. Effect of gender The AP and lateral dimensions were almost equal in males (shape ratio close to 1.0), at both vertebral levels. In females, this shape was similar to males at L4 but at L5, the lateral dimension was greater, giving a

ARTICLE IN PRESS M. Stokes et al. / Manual Therapy 10 (2005) 116–126 Table 4 Effect of vertebral level on multifidus size and shape CSA (cm2) L4

Shape Ratio

L5

%Difference

L4

L5

Males (n=44) Mean 7.97 SD 1.80

8.89***a 1.69

13.60 17.90

1.03 0.16

1.04 0.17

Females (n=45) Mean 5.55 SD 0.99

6.64***a 1.01

20.88*b 13.42

1.02 0.17

0.95***a 0.17

***Po0.001. *Po0.05. a Paired t-test using L4 and L5 differences. b Two group t-test comparing males and females.

Table 5 Regression equations for predicting multfidus size at L4 and L5 from one another Males

Females

L4 CSA=0.19+0.88  L5CSA L5 CSA=2.74+0.77  L4CSA

L4 CSA=0.30+0.79  L5CSA L5 CSA=2.11+0.82  L4CSA

121

subdivided into age bands in Table 6. The youngest groups had the smallest shape ratios, with post hoc analysis revealing a consistent significant difference between the 20–29 year age group and those in the 30–39 and 40–49 year age groups. 3.2.3. Effect of vertebral level In females there was a significant (Po0.001) difference in shape ratio between vertebral levels L4 and L5, whilst no difference was found in males (Table 4). At L4 in females, the AP and lateral dimensions were almost equal (giving a round profile) and at L5 the lateral dimension was greater, giving a lower ratio (i.e. oval/flatter profile), as seen in Figs. 3 and 4, respectively. 3.3. Symmetry of multifidus size and shape Mean values for between-side differences in CSA for the four groups were between 7.2 and 9.6%, and for shape ratio 9.9–12.3% (Table 7). However, reference ranges and SDs were high, indicating large individual variation. There were no significant differences for symmetry between genders, age groups or vertebral levels. 3.4. Correlation between multifidus cross-sectional area and linear measurements Linear measurements were highly correlated with multifidus CSA and results for coefficients greater than 0.6 are shown in Table 8, together with predictive regression equations. The predictive value of the AP dimension was higher at L4 than at L5 and the best results (X0.94) were found when the linear measurements were combined i.e. multiplied (Table 8). 3.5. Correlations between multifidus size and anthropometry

Fig. 5. Triangular shape of multifidus at the fourth lumbar vertebral level (L4). The shape of this muscle suggests hypertrophy, as it was seen in some of the more physically active subjects. The different connective tissue patterns between multifidus and longissimus are evident. (5 MHz curvilinear probe.)

significantly smaller ratio than in males (mean shape ratio=0.95, Po0.05). 3.2.2. Effect of age There were significant differences in multifidus shape at L5 between age groups in both genders, who were

In both genders and at both vertebral levels, there were significant positive correlations (Po0.05–0.001) between CSA and spinous process length (r=0.38–0.60) and laminar width (r=0.36–0.52). There were no significant correlations between CSA with height, body mass or BMI, except in females at L4 there was a significant (Po0.05) but weak correlation with body mass (r=0.26). There were no significant correlations between multifidus size with the distance between C7 and L5, the ASISs or PSISs. The highest correlation coefficient was that for multifidus CSA and SPL at L4 in males (r=0.60; Po0.0001).

ARTICLE IN PRESS M. Stokes et al. / Manual Therapy 10 (2005) 116–126

122

Table 6 Shape ratios for multifidus at L4 and L5 in different age groups Age Band Group (Years)

1 20–29

2 30–39

3 40–49

4 50–69

Males L4 n Mean (SD) 95% range L5 n Mean (SD) 95% range

13 0.95(0.16) 0.63–1.27 13 0.89(0.11) 0.67–1.10

16 1.05(0.12) 0.81–1.30 14 1.03(0.11) 0.82–1.25

7 1.02(0.20) 0.63–1.42 4 1.22(0.26) 0.70–1.73

16 1.06(0.14) 0.77–1.34 14 1.12(0.14) 0.85–1.38

Females L4 n Mean (SD) 95% range L5 n Mean (SD) 95% range

35 1.01(0.23) 0.56–1.47 27 0.90(0.11) 0.68–1.12

11 1.03(0.17) 0.70–1.37 8 1.03(0.18) 0.68–1.39

11 1.15(0.21) 0.74–1.56 5 1.16(0.26) 0.65–1.67

11 1.10(0.16) 0.79–1.40 5 0.91(0.09) 0.74–1.08

P-values for differences between age groups (e.g. group 1 versus group 2) Males 1v2 1v3 L4 NS NS L5 0.0022 0.0019

1v4 NS 0.0001

2v3 NS 0.0484

2v4 NS NS

3v4 NS NS

Females L4 L5

NS NS

NS NS

NS NS

NS NS

NS 0.0124

NS 0.0005

n=number of subjects. SD=standard deviation. NS=not significant 40.05.

Table 7 Symmetry of multifidus size and shape for males and females at L4 and L5 Males

Females

% difference between sides

L4 (n=52)

L5 (n=45)

L4 (n=68)

L5 (n=46)

CSA Mean (SD) 95% range

9.6 (8.1) 0–25.9

8.1 (5.5) 0–19.0

7.2 (7.0) 0–21.2

7.2 (6.5) 0–20.1

Shape ratio Mean (SD) 95% range

10.7 (8.4) 0–27.5

9.9 (9.0) 0–27.9

12.3 (13.0) 0–38.3

12.2 (9.2) 0–30.6

There were no significant differences between genders or the two vertebral levels. CSA=cross-sectional area. 95% range=lower and upper limits of range of values for 95% of the sample population.

4. Discussion

4.1. Multifidus size

Reference ranges for multifidus size and shape were produced for normal subjects of different ages. Muscle size was influenced by gender and vertebral level but not age. The CSA can be estimated from the muscle’s linear measurements or the CSA of the muscle at the adjacent vertebral level. None of the anthropometric measures were predictive of muscle size.

4.1.1. Effect of gender The larger muscles in males were expected and appear to be due to differences in body mass. The mean CSA at L4 in females agreed with findings by Hides et al. (1992, 1994) but the mean size of almost 8 cm2 in males was larger than previously reported (6.2 cm2), perhaps due to the present males being heavier (mean mass 82.8 kg

ARTICLE IN PRESS M. Stokes et al. / Manual Therapy 10 (2005) 116–126

123

Table 8 Correlation between multifidus cross-sectional area (CSA) and linear measurements Dimensions

Males

Females

Regression equation

r

Regression equation

r

L4 CSA vs. Combined AP

0.84  (AP  Lat)+0.18 3.85  AP-3.78

0.96*** 0.80***

0.73  (AP  Lat)+0.72 2.73  AP-1.54

0.95*** 0.70***

L5 CSA vs. Combined AP

0.70  (AP  Lat)+0.63 2.93  AP-0.61

0.95*** 0.66***

0.69  (AP  Lat)+1.03 not applicable

0.94*** 0.54***

***Po0.0001. AP=anterior-posterior linear dimension, combined=AP multiplied by lateral dimension (AP  Lat), r=correlation coefficient.

versus 72.8 kg). Also, in the above studies of Hides et al., subjects were excluded if they took part in sports or fitness training involving the back muscles in the previous three months, whilst many of the present subjects were more active. Multifidus size at L5 has only been studied in young females and the mean CSA of 7.1 cm2 compares favourably with the present 6.6 cm2 (Hides et al., 1995). 4.1.2. Effect of age Age did not have a significant effect on multifidus size but the quality of the muscle may have altered. Changes in water and fat content occur with age and produce changes in signal intensity on MRI scans (Tsubahara et al., 1995). On ultrasound images, muscle tissue appears relatively black because it only reflects small amounts of the ultrasound beam; this is termed low echogenicity. In contrast, connective tissue and bone show high echogenicity and appear white. Infiltration with fatty or fibrous tissue may increase the echogenicity of muscle making it appear ‘whiter’, as observed in some of the present older subjects but a reliable method of quantifying these changes has yet to be developed. Further research is needed to fully evaluate changes that occur in multifidus with age and activity, in terms of the size and quality of the muscle. 4.1.3. Effect of vertebral level The greater size of multifidus at L5 than L4 agreed with Hides et al., (1995; L4 mean approx. 5 cm2, L5, 7 cm2). A cadaver study demonstrated that multifidus muscle bulk progressively increased from L2 caudally to S1 (Amonoo-Kuofi, 1983). An interesting finding was the high correlation between muscle size at the two levels, which is useful for predicting one from the other (see below). 4.2. Multifidus shape The shape ratio was only influenced by age at L5, with the value increasing with age, indicating that the muscle

became more ovoid in the AP direction. The mean shape ratio in the youngest group of males at L4 (0.95) agreed with that found in a similar group (0.91) by Hides et al. (1992). Multifidus shape is not always regular, particularly in subjects with relatively large muscles, which can appear more triangular than ovoid (see Fig. 5). The medial and inferior (deep) borders are confined by the spinous process and lamina, so multifidus can only hypertrophy in a lateral or superior (superficial) direction, which may explain the more triangular shape of hypertrophied muscles (see Fig. 5). In such cases, the shape ratio is misleading as it would tend to suggest a round shape. It may be more appropriate to describe a triangular shaped muscle using three measurements i.e the superior, medial and lateral borders. The clinical relevance of multifidus shape has yet to be explored but preliminary observations suggest that it may reflect changes in muscle tone. Subjects with acute LBP had a significantly rounder multifidus muscle at the affected level, possibly indicating muscle spasm (Hides et al., 1994), but this was not investigated formally. 4.3. Prediction of normal values and assessment of abnormality Assessment of abnormal multifidus size and shape can be made by comparison with the normal 95% reference ranges (Table 2) for similar populations, in terms of gender, age and activity level. The ability to predict the CSA of L4 and L5 from each other (Table 5) is potentially useful but dependent on one level being normal. In acute unilateral LBP, muscle wasting was found to be isolated to the affected segmental level (Hides et al., 1994) but this may not be the case for chronic LBP or other conditions affecting the paraspinal muscles. 4.3.1. Correlation between muscle area and linear measurements Linear measurements can be made quickly and easily, requiring less skill than that for tracing round the

ARTICLE IN PRESS 124

M. Stokes et al. / Manual Therapy 10 (2005) 116–126

muscle border, and not all ultrasound scanners have a facility to measure area. Linear measurements are therefore more applicable for clinical use than CSA, provided they predict area accurately. The combined linear measurements (AP  Lat) were highly correlated with CSA at both vertebral levels and in both genders (Table 8). The present results for L4 and L5 (ranging from r=0.94 to 0.96) compare favourably with those found at L4 by Hides et al. (1992, males r=0.98, females 0.93; & 1994, males 0.97, females 0.92) in young healthy subjects. However, it is known that this correlation weakens when muscle becomes atrophied (Hides et al., 1994) and so cannot be assumed in all situations. The AP dimension was also highly correlated with area at L4 and could thus be used as a quick clinical measure to estimate area but the relationship at L5, although statistically significant, was not as strong. Multiplying the linear measurements provides similar two-dimensional values to area (cm2) and is obviously preferable to a single measurement if area cannot be measured. 4.3.2. Symmetry of multifidus size and shape Between-side comparisons are often used clinically to assess unilateral abnormalities. Symmetry of multifidus CSA (Table 7) was lower than in previous studies (Hides et al., 1992, 1994), possibly due to variation in the activity levels between study populations. If the higher degree of asymmetry was due to physical activity, it could be hypothesized that the dominant side would be larger. However, although 82% of males and 90% of females were right handed, there was no significant difference in CSA between sides and no trends of asymmetry related to handedness were seen. Subjects with acute LBP had much higher degrees of asymmetry (mean approximately 30%, range 15–46%; Hides et al., 1994) than the present subjects, despite the large normal variation, so asymmetry does appear to be a potentially useful way of detecting abnormality. In agreement with Hides et al. (1992), symmetry of shape was poorer than that of size, with large individual variation. Results at L5 were similar to those at L4. Asymmetry of muscle shape ratio was not investigated in LBP subjects and due to the large values in normal subjects, it is unlikely that this form of assessment would detect abnormality accurately. 4.3.3. Relationships between multifidus size and anthropometry The low correlations between multifidus size and general anthropometric measurements of height, body mass or BMI were contrary to those reported by Hides et al. (1992), who found significant positive correlations between L4 CSA and body mass and height. In the previous study of males, the correlation coefficient for CSA and mass was 0.78 compared to 0.14 in the present

study. It is difficult to explain these markedly different findings other than by the fact that the present subjects were considerably heavier and probably more active. Although body mass and height will influence the forces acting on the spine, and therefore to some extent affect lumbar muscle activity, multifidus is not a weight bearing muscle and a direct linear relationship between multifidus and mass or height would not be expected. Of the anthropometric measures, SPL in males at L4 had the highest correlation coefficient (0.6), which was not considered sufficient to produce a clinically useful regression equation for predicting multifidus CSA. 4.4. Practical issues in the technique of imaging multifidus Multifidus is one of the most difficult muscles to image using ultrasound. Due to the curved shape of the muscle, a convex transducer is preferable to a linear one, since more of the sound beams are perpendicular to the muscle-fascia interface and thus reflected back to the transducer. 4.4.1. Identifying landmarks and the lateral border of multifidus Care is needed to find the correct echogenic landmark. For example, an image over a facet joint will result in a smaller muscle size than over the lamina, as evident in Fig. 6. To ensure a consistent technique, images are taken at the lowest point between the facet joints, first using a longitudinal orientation to aid locating the point of true muscle depth. Accurate relocation of the external scanning site for repeated scans (as described in ‘Methods’) is also important for reliability. It is particularly difficult to image the lateral border of multifidus, which lies adjacent to longissimus. Connective tissue within normal muscle surrounds its fascicles and appears as irregular white straight or curved white lines. The fascicular structure of multifidus can sometimes be clearly identified on a scan, as the connective tissue pattern differs to that of longissimus (see Fig. 5). In some cases, although a clear border cannot be identified, it can be visualized where these different patterns meet. The shape of the muscle at L4 can vary from the round or oval shapes seen in Fig. 3, to the triangular shape in Fig. 5, which appears to occur in highly trained individuals with hypertrophied muscles. Furthermore, muscle tissues lying either side of the border move differently during contraction, assisting identification of the border. Specifically, the fascicles of multifidus move in relation to each other with a swirling action, giving a lava lamp effect (as described by Judith Pearson MCSP, personal communication, with permission). This can be demonstrated by asking the subject to lift the ipsilateral leg gently.

ARTICLE IN PRESS M. Stokes et al. / Manual Therapy 10 (2005) 116–126

125

Ultrasound imaging in routine physiotherapy practice: Skill of the operator can have a marked effect on the quality of images and their reliability needs to be established, following a period of practice. Adoption of ultrasound imaging of muscle by physiotherapists into routine clinical practice should involve training, including technical and safety aspects, preferably by a sonographer, to ensure the technique is used appropriately. Adherence to safety guidelines produced by the BMUS should help secure a place for muscle imaging as a recognized application in the field of medical ultrasound. 5. Conclusions Fig. 6. Biofeedback of lumbar multifidus—longitudinal view using split screen facility. The facet joints (F) can be used as landmarks for the lower borders of the muscles. During contraction (right panel), the muscle becomes thicker and the angle of the fibres becomes steeper, providing feedback. Right multifidus of female aged 46 years. (5 MHz linear probe.)

Adjustment of the angle of the transducer by tilting it in the cephalad-caudad and medial-lateral directions can help to improve the definition of muscle borders, whilst at the same time maintaining a sharp image of the echogenic lamina. Positioning of the subject: Subject positioning is important to standardize muscle length and maintain a flat lumbar spine. The standard prone position for imaging multifidus may not be suitable for people with e.g. severe LBP, neurological disorders or in post-natal women. It was found that measurements from scans taken in side lying did not differ from those taken in prone (Coldron et al., 2003), so this alternative position can be used where necessary. Visual biofeedback tool: The use of ultrasound for biofeedback has been described for facilitating lumbar mulitifidus (Hides et al., 1998) and is attracting increasing clinical interest. The muscle can be imaged transversely (observing the ‘lava lamp’ effect described above) or longitudinally over the facet joints, as illustrated in Fig. 6. Various manoeuvres are used to ilicit a contraction e.g. in forward lean standing or walk standing, the subject is asked to lift the ipsilateral leg and/or contralateral arm. Richardson et al. (1998) suggested palpating the muscle and asking the subject to swell out the muscle beneath the examiner’s fingers. As the muscle contracts, the subject can observe it on the screen as it becomes thicker and the angle of fibres become steeper (Fig. 6). The effectiveness of subject posture and the manoeuvre used vary between individuals, so it is a matter of trying different techniques to find one that suits each person. The British Medical Ultrasound Society Safety Guidelines (website: www.bmus.org) advise that exposure time is kept to the minimum necessary.

Assessment of multifidus size can be made by comparison with the 95% reference ranges reported. Separate data are needed for each gender and vertebral level but not for different age groups up to 69 years. Further data are needed for populations of different activity levels and specific sports, and people over 70 years. Changes in quality of muscle tissue with age require investigation. Shape varied considerably amongst normal subjects, suggesting that it may be inappropriate to describe a typical multifidus shape. The mean between-side difference in CSA was higher than previously reported but was still much lower than the degree of asymmetry found previously in acute LBP patients. Regression equations to predict L4 multifidus CSA from L5, and vice versa, could be used clinically, provided wasting appears to be isolated to one vertebral level. Similar relationships may exist between muscles at other levels but require investigation. Linear measurements of multifidus provide an estimate of CSA and are simple and quick to use. Ultrasound imaging is being used increasingly for research, clinical assessment and biofeedback. If ultrasound is to be adopted for use in routine physiotherapy practice, it is important that the methodology for obtaining and measuring images is standardized, to ensure the technique is robust and reliable. Acknowledgments The authors thank the subjects at the Royal Hospital for Neuro-disability who took part in the study, Dr Anthony Swan for statistical advice and data analysis, and the Neuro-disability Research Trust for financial support. References Amonoo-Kuofi HS. The density of muscle spindles in the medial, intermediate and lateral columns of human intrinsic postvertebral muscles. Journal of Anatomy 1983;136:509–19.

ARTICLE IN PRESS 126

M. Stokes et al. / Manual Therapy 10 (2005) 116–126

Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986;1:307–10. Bogduk N. Clinical anatomy of the lumbar spine and sacrum, 3rd ed. Edinburgh: Churchill Livingstone; 1997. Bogduk N, Macintosh JE, Pearcy MJ. A universal model of the lumbar back muscles in the upright position. Spine 1992;17:897–913. Chinn S. The assessment of methods of measurement. Statistics in Medicine 1990;9:351–62. Coldron Y, Stokes M, Cook K. Lumbar multifidus muscle size does not differ whether imaging is performed in prone or side lying. Manual Therapy 2003;8:161–5. Grieve GP. Common vertebral joint problems. Edinburgh: Churchill Livingstone; 1983. Hides JA, Cooper DH, Stokes MJ. Diagnostic ultrasound imaging for measurement of the lumbar multifidus muscle in normal young adults. Physiotherapy Theory and Practice 1992;8:9–26. Hides JA, Richardson CA, Jull GA. Magnetic resonance imaging and ultrasonography of the lumbar multifidus muscle. Spine 1995; 20:54–8. Hides JA, Richardson CA, Jull GA. Multifidus muscle recovery is not automatic following resolution of acute first episode low back pain. Spine 1996;21:2763–9.

Hides JA, Richardson CA, Jull GA. Use of real-time ultrasound imaging for feedback in rehabilitation. Manual Therapy 1998;3:125–31. Hides JA, Stokes MJ, Saide M, et al. Evidence of lumbar multifidus muscle wasting ipsilateral to symptoms in patients with acute/ subacute low back pain. Spine 1994;19:165–72. Macintosh JE, Bogduk N, Valencia F, et al. The morphology of the human lumbar multifidus. Clinical Biomechanics 1986;1: 196–204. Rankin G, Stokes MJ. Reliability of assessment tools in rehabilitation: an illustration of appropriate statistical analysis. Clinical Rehabilitation 1998;12:187–99. Riddle DL, Finucane SD, Rothstein JM, et al. Intrasession and intersession reliability of hand-held dynamometer measurements taken on brain-damaged patients. Physical Therapy 1989;69: 182–94. Stokes M, Hides J, Nassiri D. Musculoskeletal ultrasound imaging: diagnostic and treatment aid in rehabilitation. Physical Therapy Reviews 1997;2:73–92. Tsubahara A, Chino N, Akaboshi K, et al. Age-related changes of water and fat content in muscles estimated by magnetic resonance (MR) imaging. Disability and Rehabilitation 1995;17: 298–304.