ORIGINAL COMMUNICATIONS

Quantitative Analysis of Diurnal Variation in Volumeand Water Content of Lumbar Intervertebral Discs

NEIL ROBERTS,1* DAVID HOGG,2 GRAHAM H. WHITEHOUSE,1 AND PETER DANGERFIELD2

1Magnetic Resonance and Image Analysis Research Centre, University of Liverpool, Liverpool, United Kingdom2Department of Human Anatomy and Cell Biology, University of Liverpool, Liverpool, United Kingdom

The Cavalieri method of modern design stereology has been used in combination withmagnetic resonance imaging (MRI) to obtain unbiased estimates of the volume of the lumbarintervertebral discs immediately at the end of a day of normal activity and again following anight’s rest. In addition, pixel-by-pixel mapping of the T2-relaxation time has been used tocharacterize objectively the tissues of the intervertebral discs. The mean increase in height ofseven female subjects of average age 21 years (range 19–23 years) measured with astediometer was 19.3 mm (range 8–26 mm). Image analysis showed that the mean overnightincrease in volume of lumbar discs was 1300 mm3 (range 100–2700 mm3). The increase involume of the disc was accompanied by an increase in the T2-relaxation time of the nucleuspulposus. This suggests that the change in disc volume is most probably caused by apreferential increase in the water content of the nucleus pulposus. Clin. Anat. 11:1–8,1998. r 1998 Wiley-Liss, Inc.

Key words: Cavalieri method; intervertebral disc; lumbar spine; magnetic reso-nance imaging (MRI); sleep; stereology; T2-relaxation time; volume

INTRODUCTION

The daily variation in fluid content of the interver-tebral discs causes changes in stature, which lengthensduring sleep and shortens during day time (DePuky,1935; Reilly et al., 1984; Adams et al., 1987; Krag et al.,1990; Wing et al., 1992). Magnetic resonance imaging(MRI) has demonstrated diurnal variation in the watercontent of discs (Isherwood et al., 1987; Boos et al.,1993) and the present study builds on this work. Animage analysis method has been developed for estimat-ing disc volume and the water content of the discs hasbeen assessed by mapping the MR T2-relaxationtimes within the disc. These methods have beenapplied to a group of female volunteers in order toestimate the magnitude of the diurnal variation in thevolume and water content of the lumbar discs.

MATERIAL AND METHODS

The local Ethical Committee approved this study,which involved seven female volunteers with a meanage of 21 years (range 19–23 years) and mean weight of61 kg (range 47–88 kg). None had any significanthistory of low back pain and all were physically active

for more than 30 min, two or three times per week.Each subject was scanned once between 6 pm and 12midnight and again between 9 am and 12 noon thefollowing day. The subjects were instructed to rise at 8am on the day of the evening scan and to ensure thatthey had continually been upright (sitting vertically orstanding) for approximately ten hours prior to theevening scan. They were requested to go about theirnormal activities during the day, and not to lie supine.The morning scan took place after more than ninehours’ bed rest within the Magnetic Resonance andImage Analysis Research Centre.

Estimation of Intervertebral Disc Volumeby the Cavalieri Method

The lumbar spine was imaged using a 1.5 T SIGNAwhole body MR imaging system (General Electric,Milwaukee, USA). With random starting position aseries of 4 mm thick, contiguous Fast Spin Echo (FSE)images was obtained through the lumbar spine in the

*Correspondence to: Dr. Neil Roberts, Magnetic Resonance and ImageAnalysis Research Centre, University of Liverpool, P.O. Box 147,Liverpool, L69 3BX, United Kingdom. E-mail: neil@liverpool. ac.uk

Received 15 November 1996; Revised 19 March 1997

Clinical Anatomy 11:1–8 (1998)

r 1998 Wiley-Liss, Inc.

sagittal plane. The images were acquired with arepetition time (TR) of 3000 ms, echo time (TE) of102 ms and echo train length of eight, in a time of 3min 48 sec. The acquisition matrix comprised 256readings of 192 phase encodings with 4 excitations perphase encoding. The Field of View (FOV) of eachimage is 20 cm. This sequence yielded proton density(PD) weighted images which gave a clear outline tothe intervertebral discs by virtue of the low signal fromthe surrounding vertebral end plates and spinal liga-ments. The PD images were transferred to ANALYZEsoftware (Mayo Foundation, Minnesota, USA) run-ning on a SPARC 10 workstation (SUN Microsystems,California, USA). The volume of each disc was esti-mated separately using the Cavalieri method. Applica-tion of this approach for obtaining estimates of struc-ture volume from MR images in vivo has beendescribed elsewhere (Roberts et al., 1993; Roberts etal., 1994; Light et al., 1995). The method requires that,beginning from a uniform random starting positionwithin the sectioning interval, the structure of interestis exhaustively sectioned with a systematic series ofMR image planes. Accordingly, a mathematically unbi-ased estimate of the volume of the structure isobtained by multiplying the sum of the cross-sectional(transect) areas of the structure on consecutive MRimages by the sectioning thickness. Estimates of thetransect areas were obtained by manual point-count-ing techniques, where the spacing between test pointswas 6 pixels (4.69 mm). The coefficient of error (CE)on the volume estimate was computed as described inRoberts et al. (1994). Typically, only five systematicMR images are required to provide a Cavalieri esti-mate of the volume of the disc, with a CE of less than5%.

Characterization of Lumbar Intervertebral Discsby MR T2 Relaxometry

The T2-relaxation time is a measure of the magni-tude of processes contributing to the transverse decayof the MR signal as a result of interactions at theatomic and molecular level within tissues. The param-eter correlates most powerfully with the bulk watercontent of the tissue and, in general, will be moreprolonged the wetter the tissue, although a variety ofother relaxation mechanisms (e.g., chemical exchange,diffusion) exists (see, for example, Elster, 1994). In thepresent study T2-relaxation times were computed on apixel-by-pixel basis from a series of four sagittal spinecho images acquired in the mid-line with echo timesof 20, 40, 60, and 80 ms, and constant TR of 2000 ms.The images were obtained using a FOV of 20 cm, slicethickness of 5 mm and an acquisition matrix of 256readings of 128 phase encodings with two excitations

per phase encoding, in a time of 8 min 40 sec. Theimages were transferred to the SPARC 10 workstationand parametric maps of the T2 relaxation time werecomputed by linear least squares fitting techniquesaccording to the formula

I 5 Io exp (2TE/T2)

where I is the signal intensity of the image acquired atTE and Io is a constant proportional to the relativeproton density in each voxel (Barbosa et al., 1994).Histograms of the T2-relaxation times of the pixelswithin the boundary of the intervertebral disc wereextracted using ANALYZE software (MAYO Founda-tion, Minnesota, USA) and transferred to S-plus soft-ware (StatSci, Washington, USA) for statistical analy-sis.

Stediometry

Standing height was measured using a portablestediometer accurate to 1 mm. The subjects weremeasured immediately before both the evening andmorning MR scans. Three measurements were taken,the mean being recorded.

RESULTS

The diurnal variations in the height of the sevensubjects are shown in Table 1. The average overnightincrease in height was 19.3 mm (range 8–26 mm).Relative to the evening measurement this correspondsto a mean percentage increase in height of 1.17%(SD 5 0.34%). The subjects are numbered 1–7 inorder of height change (subject 1 having the largestheight change, but also the smallest overall height).Although there was little variation in height betweensubjects 4–7, there was a large range of height change.

Unbiased estimates of volume were obtained for 34of the 35 lumbar intervertebral discs in the 7 subjects.For one subject an artifact in the region of the L1-L2disc on the evening scan meant that a reliable estimateof the diurnal variation in volume could not beobtained for this disc. The predicted CEs on theestimates of disc volume were always less than 5%.Figure 2 is a histogram of the volume changes in the 34lumbar discs between evening and morning. Themean volume change for each of the five lumbar disclevels is presented in Table 2. The overnight changein disc volume is significant at disc levels L1-L2,L4-L5 and L5-S1 (p , 0.05). It increases from levelL1-L2 to L3-L4 and decreases at L4-L5 and L5-S1.The mean volume of the 34 discs in the evening (6100mm3) is highly significantly different (p , 0.0003)

2 Roberts et al.

from the mean volume in the morning (7400 mm3).The mean volume increase, over the five levels is 1300mm3, with a range of 100–2700 mm3. The 95 percentconfidence interval for the difference in means is500–2,000 mm3.

The mean values of the evening and morningT2-relaxation times of the 34 lumbar discs of the sevenvolunteers are presented in Table 3. A t-test shows nosignificant difference (p 5 0.2) between the meanevening and morning T2 values when the whole of the

disc is considered, but the situation changes when oneconsiders just the central region, i.e., the nucleuspulposus, of the disc. The diurnal variation in T2relaxation time illustrated for the L3-L4 disc of onesubject in Figure 3 is typical. In the evening imagealmost all the pixels have T2 values less than 100 ms(i.e., are represented by the colors light green, blue orpink). However, by morning several hundred pixelspossess T2 values greater than 100 ms (i.e., are yellowor dark green). These pixels all lie within the nucleus

TABLE 1. Diurnal Variations in the Height of the Seven Subjects

Subjectnumber

Eveningheight (mm)

Morningheight (mm)

Change inheight (mm)

% Changeof total height

1 1556 1582 26 1.672 1680 1705 25 1.493 1787 1811 24 1.344 1640 1661 21 1.285 1612 1628 16 0.996 1630 1645 15 0.927 1619 1627 8 0.49Mean 1646.3 mm 1665.6 mm 19.3 mm 1.17%

SD 5 29.5 mm SD 5 30.3 mm SD 5 2.7 mm SD 5 0.34%

Fig. 1. The average overnight increase in height and the percentage change in height are plotted forthe seven female subjects with mean age of 21 years.

Intervertebral Disc Diurnal Variation 3

pulposus, indicating that there has been a preferentialincrease in the water content of this region alone.

Negligible changes in volume (i.e., ,100 mm3) andwater content occurred within an intervertebral disc iden-tified by a radiologist (GHW) as both degenerate andprolapsed. Several other discs which showed signs ofdegeneration also exhibited reduced diurnal variation.

DISCUSSION

According to Krag et al. (1990), the first report thatbody height will be affected by a change from recum-

bency to upright posture was by Bishop (1852). How-ever, Wales (1995) states that the first documentedmeasurement of the diurnal variation was in 1724 bythe Reverend Joseph Wasse from Anyho in Northamp-tonshire, who measured a loss of up to six-tenths of aninch (i.e., 15 mm) in the height of young farm laborers,which he correctly assumed was due to compression ofthe intervertebral discs. DePuky (1935) was the first tosuggest that the variations were due to changes in thewater content of the discs. He measured the height ofover 1,000 people at three different times of day andreported an average daily oscillation of 15.7 mm

Fig. 2. A histogram of the diurnal variation in the volumes of the 34 lumbar intervertebral discs forwhich data were obtained in the seven subjects.

TABLE 2. The Mean and Standard Deviation of the Change in Disc Volume for the Lumbar Intervertebral Discs in SevenSubjects After Rest*

Disclevel

Numberof discs

Mean eveningdisc volume

Mean morningdisc volume

Mean volumechange

Standarddeviation p-value

Range of changein volume

L1/2 6 4.5 5.6 11.1 0.5 0.0173 0.4–2.0L2/3 7 6.0 7.4 11.4 0.4 0.0570 1.0–2.0L3/4 7 6.5 8.0 11.5 0.4 0.0563 0.1–2.2L4/5 7 7.4 8.7 11.3 0.8 0.0366 0.4–2.7L5/S1 7 5.6 6.9 11.3 0.4 0.0133 0.6–1.8All levels 34 6.1 7.4 11.3 0.6 0.0003 0.1–2.7

*All values are given in cm3.

4 Roberts et al.

(approximately 1%). Adams et al. (1987) postulatedthat the increased amount of fluid within the discsafter a period of rest results in the spine being moreresistant to flexion, and hence more vulnerable toprolapse. The range of flexion increases during the dayas fluid is lost by the discs.

More detailed information on the diurnal variationwas provided by Reilly et al. (1994), who reported that54% of the average height loss of 10.3 mm (1.1% ofbody stature) over a 24-hr period in a group of eightmales occurred within an hour of rising and 80% wasrecovered within the first three hours of sleep. Similarobservations were made by Krag et al. (1990). Goodeand Theodore (1983) and Krag et al. (1990) pointedout that variation in posture could in part account fordiurnal height changes. However, using a photo-graphic technique Wing et al., (1992) reported that40% of the diurnal variation in height occurred in thelumbar spine without any change in the lumbar lordosisand 40% in the thoracic spine without reduction of thekyphosis. This left 20% to be accounted for bychanges in the cervical spine or lower extremity joints.

The behavior of the disc is largely dependent onthe composition and organization of proteoglycanmacromolecules and collagen within it. Proteoglycanshave a high osmotic pressure and tend to draw in fluid.The collagen network acts to restrain the proteoglycanswelling which is mainly within the nucleus pulposus,providing the disc with strength and flexibility. Theproteoglycan content and water concentration of thenucleus pulposus decreases with advancing age, ascollagen replaces the gelatinous tissue of the nucleus.This provides the age-related degeneration of thediscs, revealed by decreased signal within T2-weighted MR images of them.

Generally it is believed that fluid changes withinthe disc are dependent on a diffusion gradient forwater and nutrients from the adjacent vertebral bodies(Johnstone and Bayliss, 1995), and that fluid exchangerelated to mechanical loading may be relatively unim-portant for disc nutrition. However, studies using dogsindicate that exercise, with resulting spinal motion, is

an important contributor to disc nutrition. Further-more, the report that the collagen to proteoglycan ratioin the disc annuli of rats that were flown in space for 14days (COSMOS 2044) was significantly elevated com-pared to normal controls suggests that the health of thediscs may be adversely affected by continuous unload-ing. In particular, an increase in the diffusion distancerequired for nutrients to reach the center of the discmay compromise the marginal nutritional status of thedisc (see LeBlanc et al., 1994).

MRI studies are unique in that they enable directmeasurement of the diurnal variation of the discs. Thefirst report of the use of MRI to visualize the diurnalvariation of the lumbar discs was by Isherwood et al.(1987). The first detailed study of this phenomenonwas by Boos et al. (1993). A study group of fivevolunteers presented for MR imaging after a night’srest. They lay down for a further hour in the Fowlerposition (supine with legs on a block and knees flexed)to allow a more rapid recovery of the fluid loss whichwould have occurred between awakening and arrivingat the MR unit. Each subject was then positioned inthe MR scanner without standing up again. Thesubjects then worked an eight-hour day, and the MRexamination was repeated in the evening. The fivesubjects in the control group followed a similar proce-dure except that the second MR examination wasperformed after a break of only 10 min. The eveningT1-relaxation time and proton density values in thediscs were significantly less than the values beforestarting work. No difference in the T2 values of thelumbar discs was reported, although the mean T2within the vertebral bodies increased significantlybetween morning and evening.

Le Blanc et al. (1994) used MRI to investigate theeffects of prolonged bed rest. Bed rest, overnight orlonger, resulted in an increase in disc area on mid-sagittal images by an average of 22%, but this wasassociated with only a modest increase in T2-relaxation time. After five weeks of bed rest, the discarea only returned to baseline values after several daysof ambulation. After 17 weeks of bed rest, the disc arearemained above the baseline for six weeks. Theauthors suggest that the spinal ligaments which nor-mally limit disc expansion had stretched due toprolonged tension from the expanding discs.

The problem of using Region Of Interest (ROI)cursors to measure relaxation times to describe theaverage properties of regions of heterogeneous tissuehas been highlighted by Jenkins et al. (1989). Boos etal. (1993) and LeBlanc et al. (1994) used an ROI cursorto obtain the mean T2-relaxation time for the nucleuspulposus and annulus fibrosus taken together. Inthe present study T2 relaxation times have been com-

TABLE 3. Mean and Standard Deviation of the T2-RelaxationTime (in Milliseconds) of the Intervertebral Discs in the SevenSubjects at Evening and Morning

Disc

EVENING MORNING

Mean T2Standarddeviation Mean T2

Standarddeviation

L1/2 64.61 7.51 71.18 9.53L2/3 67.70 7.68 78.45 17.56L3/4 75.06 8.47 84.30 24.56L4/5 74.59 3.99 82.09 11.32L5/S1 62.35 11.88 61.88 14.51

Intervertebral Disc Diurnal Variation 5

puted on a pixel-by-pixel basis and we have been ableto demonstrate (see Figure 3) that increases in the sizeof the disc were almost always accompanied by apreferential rise in the T2-relaxation times of thenucleus pulposus region.

In the present study the intervertebral disc wasregarded as bounded by the signal voids corresponding

to the end-plates of the associated vertebral bodies,and also of the spinal ligaments, on PD-weighted MRimages. No attempt was made to differentiate the discin such a manner that volume and T2-relaxation timechanges could be separately presented for the nucleuspulposus and annulus fibrosus. Although frequentlythese regions could be identified on the MR images,

Fig. 3. Histograms and pixel-by-pixel maps of T2-relaxation time computed for the L3-L4intervertebral disc of one of the subjects. The upper and lower panels refer to the data obtained in theevening and morning, respectively. The pseudocoloring is such that light green, blue, pink, yellow and darkgreen correspond to ranges of T2 values increasing from,49 ms, 50 to 69 ms, 70 ms to 89 ms, 90 ms to109 ms and 110 ms to 130 ms, respectively.

6 Roberts et al.

their boundaries were blurred to the extent that anymeasurements reported specifically for them wouldhave been unacceptably subjective. Qualitative assess-ment indicates that the changes in volume and T2-relaxation time appear to be an order of magnitudegreater in the nucleus pulposus than in the annulusfibrosus.

Like Boos et al. (1994), we have observed thatdegenerate discs show less pronounced diurnal varia-tion than normal ones, and endorse their suggestionthat study of the diurnal variation may have a clinicalrelevance if differences in the magnitude of thephenomenon were to be found between, for example,cohorts of patients with back pain and asymptomaticcontrols. Ideally these investigations should be quanti-tative. In a previous study we used a statistical analysisof pixel-by-pixel T2-relaxation time maps to set thresh-olds to distinguish between, and quantify, the relativeamounts of fat and muscle in the limbs of controls andpatients with muscular dystrophy (Phoenix et al.,1996). This approach could be developed to provideobjective definitions of the range of T2 values for anormal nucleus pulposus and annulus fibrosus atevening and morning, with respect to which theevening and morning values, and diurnal changes, ofpatients with low back pain could be quantified.

Botsford et al. (1994) used volume-acquired (i.e.,3D) MR images to measure simulated diurnal varia-tion in the volume of the lower three lumbar interver-tebral discs in eight normal young males. The com-bined volume of these discs was on average 16.2% lessfollowing four hours of standing and three hours ofsitting than following six hours of rest. This is similarto our value of 16.7% for the percentage change in thevolume of the lower three lumbar intervertebral discs.

Our study is the first to describe the combinedapplication of MR relaxometry and volumetric mea-surement in the evaluation of the diurnal variation ofthe lumbar discs, and the first to study subjects beforeand after several hours of sleep. It is also the first studyin which disc volumes have been estimated usingunbiased stereological methods, and represents a fur-ther refinement in methodology for the non-invasiveassessment of disc function. A preferential increase inthe T2-relaxation time of the nucleus pulposus wasobserved in the morning compared to the evening MRscans, indicating that the volume increase is probablydue to an increase in free water within this region ofthe disc.

The rapid changes in height on adopting a newposture led Reilly et al. (1984) to suggest that thediurnal variation is ‘induced by the compressive load-ing or relaxation of the disc rather than by any

exogenous rhythm.’ This idea paved the way for theoccupational studies carried out by Troup et al. (1985);they suggested that the magnitudes of gains in heightin positions of rest were proportional to the ratingstheir subjects gave for their feelings of relaxation andcomfort, and proposed that measurements of staturemay therefore provide a means of assessing the effectsof work on the spine, and could be used to determinethe optimal temporal pattern for work and recovery.MR studies of the discs similar to those reported hererepresent a direct and probably more sensitive meansof studying the effects of work on the spine.

Moreover, generally we are interested to learnwhether the magnitude of the diurnal variation is ameasure of the health of the intervertebral discs, andto know whether it is possible for the nutrition of theintervertebral discs to a improved by, for example, asubject ‘‘learning to use themselves better’’ (Alex-ander, 1932; Park, 1995), or obtaining physiotherapy ormedication.

REFERENCESAdams, M.A., P. Dolan and W.C. Hutton 1987 Diurnal variations

in the stresses on the lumbar spine. Spine 12:130–137.Alexander, F.M. 1932 The use of the self. London: E.P. Dutton

and Company.Barbosa, S., L.D. Blumhardt, N. Roberts, T. Lock and R.H.T.

Edwards 1994 Magnetic resonance relaxation time mappingin multiple sclerosis: normal appearing white matter and theinvisible lesion load. Mag. Res. Imaging 12:33–42.

Bishop, J. 1852 Researches into the Pathology and Treatment ofDeformities in the Human Body. London: Higley and SonsPublishers.

Boos, N., A. Wallin, T. Gbedegbegnon, M. Aebi and C. Boesch1993 Quantitative MR imaging of lumbar intervertebraldiscs and vertebral bodies: influence of diurnal water con-tent variations. Radiology 188:351–354.

Botsford, D.J.S., S.I. Esses and D.J. Ogilvie-Harris 1994 In vivodiurnal variations in the intervertebral disc volume andmorphology. Spine 19:935–940.

DePuky, P. 1935 The physiological oscillation of the length ofthe body. Acta. Orthop. Scandinav. 6:338–347.

Elster, A.D. 1994 Questions and Answers in Magnetic Reso-nance Imaging. Mosby, pp. 278.

Goode, J.D. and B.M. Theodore 1983 Voluntary and diurnalvariation in height and associated contour changes in spinalcurves. Eng. Med. 12:99–101.

Isherwood, I., D.I. Prendergast, D.S. Hickey and J.P.R. Jenkins1987 Quantitative analysis of intervertebral disc structure.Acta Radiol (Suppl) 369:492–495.

Jenkins, J.P.R., M. Stehling, G. Silverwright, D.S. Hickey, V.F.Hillier and I. Isherwood 1989 Quantitative magnetic reso-nance imaging of vertebral bodies: A T1 and T2 study.Magn. Res. Imaging 7:17–23.

Johnstone, B. and M.T. Bayliss 1995 The extracellular matrix ofthe intervertebral disc: proteoglycan biochemistry. In Lum-

Intervertebral Disc Diurnal Variation 7

bar Spine Disorders: current concepts. R.M. Aspden andR.W. Porter (eds.). Singapore: World Scientific, pp. 242.

Krag, M.H., M.C. Cohen, L.D. Haugh and M.H. Pope 1990Body height changes during upright and recumbent posture.Spine 15:202–207.

Le Blanc, A.D., H.J. Evans, V.S. Schneider, R.E. Wendt andT.D. Hedrick 1994 Changes in intervertebral disc cross-sectional area with bed rest and space flight. Spine 19:812–817.

Light, K., N. Roberts, G.H. Whitehouse and R.H.T. Edwards1995 Unbiased and efficient estimation of bladder volumewith MR imaging. J. Mag. Res. Imaging 5:33–41.

Park, G. 1995 The art of changing. Ashgrave Press.Phoenix, J., D. Betal, N. Roberts, T.R. Helliwell and R.H.T.

Edwards 1996 Objective quantification of muscle and fat inhuman dystrophic muscle by magnetic resonance imageanalysis. Muscle and Nerve 19:302–310.

Reilly, R., A.R. Tyrell and J.D.G. Troup 1984 Circadian varia-

tion in human stature. Chronobiology International 1:121–126.

Roberts, N., L.M. Cruz-Orive, N.M.K. Reid, D.A. Brodie, M.W.Bourne and R.H.T. Edwards 1993 Unbiased estimation ofhuman body composition by the Cavalieri method usingmagnetic resonance imaging. J. Microscopy 171:239–253.

Roberts, N., A.S. Garden, L.M. Cruz-Orive, G.H. Whitehouseand R.H.T. Edwards 1994 Estimation of fetal volume bymagnetic resonance imaging and stereology. Brit. J. Radiol.67:1067–1077.

Troup, J.D.G., T. Reilly, J.A.E. Eklund and P. Leatt 1985Changes in stature with spinal loading and their relation tothe perception of exertion or discomfort. Stress Medicine1:303–307.

Wales, J. 1995 Night growth. New Scientist 145:81.Wing, P.C., I.K.Y. Tsang, L. Susak, F. Gagnon and R. Gagnon

1992 Diurnal change in the profile, shape and range ofmotion of the back. Spine 17:761–765.

8 Roberts et al.