clinical, morphologic, and morphometric features of cranial thoracic spinal stenosis in large and...

CLINICAL, MORPHOLOGIC, AND MORPHOMETRIC FEATURES OFCRANIAL THORACIC SPINAL STENOSIS IN LARGE AND GIANT

BREED DOGS

PHILIPPA JOHNSON, LUISA DE RISIO, ANDREW SPARKES, FRASER MCCONNELL, ANDREW HOLLOWAY

The clinical, morphologic, and morphometric features of cranial thoracic spinal stenosis were investigated inlarge and giant breed dogs. Seventy-nine magnetic resonance imaging studies of the cranial thoracic spinewere assessed. Twenty-six were retrieved retrospectively and 53 were acquired prospectively using the sameinclusion criteria. Images were evaluated using a modified compression scale as: no osseous stenosis (grade 0),osseous stenosis without spinal cord compression (grade 1), and osseous stenosis with spinal cord compression(grade 2). Morphometric analysis was performed and compared to the subjective grading system. Grades 1and 2 cranial thoracic spinal stenosis were identified on 24 imaging studies in 23 dogs. Sixteen of 23 dogs hada conformation typified by Molosser breeds and 21/23 were male. The most common sites of stenosis wereT2–3 and T3–4. The articular process joints were enlarged with abnormal oblique orientation. Stenosis wasdorsolateral, lateralized, or dorsoventral. Concurrent osseous cervical spondylomyelopathy was recognized insix dogs and other neurologic disease in five dogs. Cranial thoracic spinal stenosis was the only finding in12 dogs. In 9 of these 12 dogs (all grade 2) neurolocalization was to the T3-L3 spinal segment. The median ageof these dogs was 9.5 months. In the remaining three dogs neurologic signs were not present. Stenosis ratioswere of limited benefit in detecting stenotic sites. Grade 2 cranial thoracic spinal stenosis causing direct spinalcord compression may lead to neurologic signs, however milder stenosis (grade 1) is likely to be subclinical orincidental. C© 2012 Veterinary Radiology & Ultrasound.

Key words: canine, cranial thoracic spinal stenosis, magnetic resonance imaging (MRI), morphometric dimen-sions, vertebrae.

Introduction

SPINAL STENOSIS CAN BE focal, segmental, orgeneralized.1, 2 Although cervical and lumbosacral

spinal stenosis has been investigated widely in the dog, tho-racic spinal stenosis is uncommon.2–7

Stenosis can be acquired, congenital, or developmental.2

Acquired spinal stenosis is due to ligamentous hypertro-phy, intervertebral disc protrusion, degenerative articularchanges, synovial cysts, and malarticulation–malformationsyndromes. Congenital spinal stenoses are present at birthand have no active underlying cause.2 Congenital thoracicspinal stenosis is usually associated with hemivertebrae, butother vertebral malformations have been reported.2, 3, 6 De-velopmental stenoses are present at birth and have an activecause that remains present throughout the growth period,until the vertebrae reach maturity.2 Developmental cranial

From the Department of Radiology (Johnson, Holloway), Depart-ment of Neurology/Neurosurgery (De Risio), International Society ofFeline Medicine (Sparkes), and Department of Diagnostic Imaging, TheUniversity of Liverpool (McConnell).

Address correspondence and reprint requests to Philippa Johnson atthe above address. E-mail: [email protected]

Received January 18, 2012; accepted for publication May 3, 2012.doi: 10.1111/j.1740-8261.2012.01953.x

thoracic spinal stenosis due to vertebral malformation andarticular process hypertrophy has been documented in twoimmature dogues de Bordeaux with pelvic limb neurologi-cal deficits.7 In addition, a relationship between stenosis ofthe cranial thoracic spine and young large and giant breeddogs has been suggested; however, little is known about thisrelationship.2, 3, 8

Our objective was to investigate the clinical and morpho-logical features of cranial thoracic spinal stenosis in the dogand to compare the spinal morphometric data of dogs withcranial thoracic spinal stenosis with subjectively normalspines.

Materials and Methods

Retrospectively, dogs weighing greater than 20 kg thatunderwent magnetic resonance (MR) imaging between Jan-uary 2000 and October 2009 and that had T2-weighted(T2W) transverse and sagittal sequences of the cranialthoracic spine (T1–6) were identified. Only imaging stud-ies where the cranial thoracic spine was normal or had

Vet Radiol Ultrasound, Vol. 00, No. 0, 2012, pp 1–5.

1

2 JOHNSON ET AL 2012

FIG. 1. (A) Transverse T2-weighted image of a grade 0 site at T1–2. (B) Transverse T2-weighted image of a grade 2 site at T2–3. The spinal cord is compressedlaterally. (C) Transverse T2-weighted image of a grade 1 site at T4–5. There is attenuation of the epidural fat and subarachnoid space dorsally but no spinalcord compression. (D) Sagittal T2-weighted image of the cranial thoracic spine. Lines represent the levels that the transverse images were obtained. Obstructivehydromyelia is present within the spinal cord cranial to the grade 2 stenosis site (white arrow).

evidence of osseous stenosis were included. Twenty-sixstudies met these inclusion criteria.

Prospectively, dogs weighing greater than 20 kg under-going MR imaging for assessment of forelimb orthopedic,cervicothoracic soft-tissue, or neurologic disease betweenOctober 2009 and August 2011 were enrolled. In these dogs,T2W transverse and sagittal sequences of the cranial tho-racic spine (T1–6) were obtained in addition to sequencesevaluating the primary region of interest. Again, only imag-ing studies of subjectively normal cranial thoracic spines orthose with osseous stenosis were included. Fifty-three MRimaging studies met these inclusion criteria.

In total, 79 MR imaging studies were enrolled intothe study. One dog was scanned twice. Breeds includedLabrador retriever (n = 11), German shepherd dog(n = 9), dogues de Bordeaux (n = 9), cross-breed (n = 8),Doberman pinscher (n = 5), Rottweiler (n = 4), chow-chow(n = 4), bullmastiff (n = 4), Bernese mountain dog (n = 3),golden retriever (n = 3), Staffordshire bull terrier (n = 2),

Dalmatian (n = 2), mastiff (n = 2), Rhodesian ridgeback(n = 2), English springer spaniel (n = 2), Border collie(n = 2), and one each of Shar-pei, St. Bernard, Neapolitanmastiff, Weimeraner, boxer, and Alaskan malamute.

Magnetic resonance imaging was performed using a 1.5T superconducting magnet.∗ Dogs were scanned under gen-eral anesthesia in lateral or dorsal recumbency using a hu-man torso or spine phased array coil, respectively. T2Wimages were acquired as fast spin-echo sequences. Trans-verse image slice thickness varied from 3 to 5 mm, interslicegap from 0.2 to 0.3 mm, field-of-view (FOV) was 14–16 cm,repetition time (TR) was 2700–5100 ms and echo time (TE)was 80–120 ms. Transverse slices were orientated perpen-dicular to the spinal cord.

A subjective MR imaging grading system, described pre-viously for assessment of cervical spinal stenosis, was mod-ified to assess the degree of osseous stenosis in the thoracic

∗GE Signa Echospeed, GE Milwaukee, Milwaukee, WI.

VOL. 00, NO. 0 FEATURES OF CRANIAL THORACIC SPINAL STENOSIS IN DOGS 3

FIG. 2. (A) Transverse T2-weighted image at T2–3. Bilateral articular process enlargement is causing dorsolateral attenuation of the vertebral canal. (B)Transverse T2-weighted image at T2–3. Moderate enlargement and medial positioning of the articular processes is causing lateralized vertebral canal stenosis.(C) Transverse T2-weighted image at T3–4. Moderate articular process enlargement and ventral positioning of the articular process is causing dorsoventralstenosis of the vertebral canal.

spine.9, 10 Transverse images at intervertebral disc spacesT1–2 to T5–6 were graded as follows: grade 0 = no compres-sion (Fig. 1A), grade 1 = subarachnoid space and epidu-ral fat attenuation by osseous stenosis without spinal cordcompression (Fig. 1C), grade 2 = spinal cord compressionby osseous stenosis (Fig. 1B). Magnetic resonance imageswere assessed subjectively using this system by one board-certified radiologist (FM) who was unaware of the signal-ment and clinical findings. For each dog a stenosis gradeof 0, 1, or 2 was given for each of the intervertebral levels.This produced three groups: a subjectively normal (grade0) group, and two stenosis groups (grade 1 or 2) determinedby the site with the highest stenosis grade.

In all the dogs with a stenosis grade of 1 or 2 a board-certified neurologist (LDR) reviewed the medical records,neurological examination, and MR imaging findings. Anyadditional regions of the spine that had MR imaging weredocumented. The presence of neurologic disease other thancranial thoracic spinal stenosis that could have contributedto the neurologic presentation was noted. For each dog inwhich cranial thoracic spinal stenosis was the only abnor-mality, and neurologic signs were consistent with cranialthoracic disease, treatment and management were docu-mented. Follow-up in these dogs was obtained by contact-ing the referring veterinarian. A standardized questionnairewas used to assess outcome after diagnosis.

The morphologic features of all sites with stenosis wereassessed again independently by another board-certified ra-diologist (AH) and a third-year radiology resident (PJJ).For sites at which the assessment differed, the images werereviewed, and consensus reached. For each site, changeswere categorized according to one of three patterns ofstenosis reflecting the direction in which the vertebral canalwas attenuated: dorsolateral (Fig. 2A), lateralized (Fig. 2B),

FIG. 3. Paramedian T2-weighted images of the cranial thoracic spine. (A)Grade 0 where the articular processes have a near horizontal orientation. (B)Grade 2 where the articular processes are elongated and have a steep obliqueorientation.

and dorsoventral (Fig. 2C). Dorsolateral stenosis was de-fined as osseous stenosis resulting in a triangular shapeof the vertebral canal or spinal cord. For lateralized anddorsoventral stenosis the vertebral canal or spinal cordwas narrowed into a flattened oval shape in the sagittaland dorsal planes, respectively. Other subjective features as-sessed were: degree of articular process enlargement (mild,moderate, or severe), symmetry (symmetric or asymmetric)and articular process orientation on paramedian images(normal-horizontal; Fig. 3A) or abnormal-oblique orien-tation (Fig. 3B). For each imaging study in which cranialthoracic spinal stenosis was the only lesion identified thepresence of spinal cord hyperintensity and obstructive hy-dromyelia was documented.

The following measurements were made in all imag-ing studies, at each vertebral level from T1 to T6. Themeasurements were made by two observers (AH andPJJ) independently using Osirix Medical Imaging software(http://www.osirix-viewer.com).


FIG. 4. Transverse T2-weighted images of a dog with grade 0 stenosis.(A) T2–3 intervertebral disc space, representing measurement of the crosssectional area of the vertebral canal (IVCa) and spinal cord (IVCo). (B)Mid-body of T2 representing measurement of the cross sectional area of thevertebral canal (MBCa).

1. IVCa = intervertebral canal area. The area was tracedusing the epidural fat–dorsal lamina interface as the dor-sal margin, the floor of the vertebral canal as the ventralmargin, and laterally defined by a line drawn perpen-dicularly from the most ventral aspect of the articularprocess across the intervertebral foramen (Fig. 4A).

2. MBCa = mid-body vertebra canal area. The area wastraced along the epidural fat–vertebral cortex interface(Fig. 4B).

3. IVCo = intervertebral spinal cord area. The area wastraced along the outer margin of the spinal cord(Fig. 4A).

To eliminate differences in vertebral size between dogs,two previously documented stenosis ratios were adaptedand calculated at each vertebral site from T1–6.11–13

1. Canal Area Ratio (canal area ratio) = IVCa/MBCa;2. Canal Occupying Ratio of spinal cord (canal occupying

ratio) = IVCo/IVCa.

After it was determined that the calculated ratios forstenosis grades of 0 and 1 were normally distributed, thedata were expressed as mean +/– standard deviations(SDs). However, data when the stenosis grade was 1 or 2 atsome point were not normally distributed. These data weretherefore expressed as median with range or interquartilerange.

Analysis of normally distributed data was undertakenwith parametric tests (ANOVA with Tukey’s posttest pair-wise analysis, Pearson’s correlation coefficient). Analysisof other data, and comparison of normally and nonnor-mally distributed data, were performed using nonparamet-ric tests (Kruskal–Wallis with Dunn’s pairwise posttest

analysis, Spearman Rank correlation coefficient). A P-value of <0.05 was considered significant.

For comparative purposes, reference ranges were calcu-lated for the canal area ratio and canal occupying ratioat each of the five vertebral levels from the grade 0 imag-ing studies. These ranges were based on the mean ±2 SD.These reference ranges were then applied to the data ob-tained from the grades 1 and 2 stenotic sites to see if thecanal area ratio or canal occupying ratio fell outside thecalculated reference range in these dogs at the differentvertebral sites. Any canal area ratio value (at the appropri-ate vertebral site) below the reference range was classifiedas positive (stenosis) and any canal occupying ratio valueabove the reference range was similarly classified as stenosis.The agreement between the subjective stenosis scores andthe objective data derived from the canal area ratio andcanal occupying ratio data were evaluated. Any missingdata points were excluded from the analyses. Interobserveragreement for the calculated ratios was assessed with theuse of Pearson’s correlation coefficients and Bland–Altmanplots.

Results

There were 55 studies graded as 0 and 24 graded as 1 or2. Thirteen of the stenotic studies had a maximum grade of2 and 11 had a maximum grade of 1.

Eighteen breeds had normal imaging studies: Labradorretriever (n = 11), German shepherd dog (n = 8), cross-breed (n = 8), Doberman pinscher (n = 4), Bernese moun-tain dog (n = 3), golden retriever (n = 3), Rottweiler (n =3), chow-chow (n = 2), Dalmatian (n = 2), English Springerspaniel (n = 2), Rhodesian ridgeback (n = 2), Border collie(n = 2), and one each of Alaskan malamute, boxer, doguesde Bordeaux, bullmastiff, and Weimeraner. There were 4 fe-males, 16 neutered females, 10 males and 25 neutered males.The mean age was 77 months (median 83, range 5–156).

The 24 MR imaging studies graded as 1 or 2 were ac-quired in 23 dogs (one dog underwent imaging twice, 10months apart). Affected breeds were: dogues de Bordeaux(n = 8), bullmastiff (n = 3), chow-chow (n = 2), Stafford-shire bull terrier (n = 2), mastiff (n = 2), and one eachof Doberman, German shepherd dog, Neapolitan mas-tiff, Rottweiler, Shar Pei, and St. Bernard. The majorityof grades 1 and 2 dogs were male (12 intact and 9 neutered)with only two females (both intact) in this group. The meanage was 23 months (median 14, range 4–105). In all but twostudies, more than one vertebral site was stenotic (grade of1 or 2).

In addition to the cranial thoracic spine, MR imaging inthe dogs with grade 1 and 2 thoracic stenosis included thecervical spine (n = 9), lumbar spine (n = 7), lumbosacralspine (n = 2), and both cervical and lumbar spine (n = 5).Ten of the 23 dogs with grade 1 or 2 cranial thoracic spinal


stenosis had additional spinal lesions identified, includingosseous cervical spondylomyelopathy (n = 6), subluxationof the caudal thoracic spine (n = 1), extradural mass (n = 1),arachnoid diverticulum (n = 1), degenerative lumbosacralstenosis (n = 1). When the medical and neurologic recordswere assessed, one epileptic dog had tetraparesis and ataxiaattributed partly to high-dose phenobarbitone administra-tion. Due to the concurrent disease, the significance of thecranial thoracic stenosis could not be determined in thisgroup of dogs.

In the remaining 12 dogs cranial thoracic spinal stenosiswas the only neurologic lesion identified. Of these 12 dogs,three (one with grade 2 and two with grade 1 stenosis), hadno neurologic deficits. The remaining nine had neurolog-ical deficits consistent with a cranial thoracic localizationand were considered to have primary cranial thoracic spinalstenosis. All of these dogs had grade 2 stenosis. Breeds af-fected were dogues de Bordeaux (n = 3), chow-chow (n =2), bullmastiff (n = 1), Rottweiler (n = 1), St. Bernard (n =1), and German shepherd dog (n = 1). The median age was9.5 months (range 4 months to 6 years). All nine dogs pre-sented with ambulatory (n = 7) or nonambulatory (n = 2)paraparesis. The neurologic localization was to the T3-L3spinal cord segments and the type of general propriocep-tive ataxia was suggestive of a cranial thoracic localization.Proprioceptive deficits were present in the pelvic limbs ofall dogs, ranging from delayed (n = 6) to absent (n = 3).Spinal hyperesthesia was not identified in any of these ninedogs however two dogs had aggressive behavior limitingthis assessment.

Of the nine dogs with primary cranial thoracic spinalstenosis, four were lost to long-term follow-up. Oneseverely paraparetic, 4-month-old dogues de Bordeaux waseuthanized at the time of diagnosis, three were managedconservatively, and one surgically. Of the three dogs man-aged conservatively; one 6-month-old dogues de Bordeauxwas euthanized 6 months after diagnosis due to progressivepelvic limb ataxia and paresis, one Rottweiler improvedneurologically, although was still ataxic 2 years after diag-nosis, and one German shepherd dog had static neurologicsigns with persistent pelvic limb proprioceptive deficits overthe following 5 years. One chow-chow was managed surgi-cally and underwent a modified dorsal laminectomy at T2–4 but died of aspiration pneumonia and cardiopulmonaryarrest 72 hours after surgery. Postmortem or histopatho-logic confirmation of stenosis was not performed inany dog.

Of the three dogs with subclinical cranial thoracic spinalstenosis, one dogues de Bordeaux underwent a second MRimaging study 10 months after the initial imaging studydue to development of ambulatory tetraparesis. Cervicalspondylomyelopathy with articular process hypertrophywas present at that time. Subjectively, the thoracic steno-sis differed between the two imaging studies. The T2–3 site

FIG. 5. Frequency and distribution of stenotic sites along the cranialthoracic spine on 24 MR imaging studies of 23 dogs with grade 1 or 2stenosis using the subjective grading system.

was graded 2 initially and 1 subsequently. The T4–5 site,initially graded 0, was graded 1 at the second MR study.

Of the 24 MR imaging studies with cranial thoracicspinal stenosis, a total of 68 stenotic sites were identified.Sites of stenosis were associated with variation in the po-sitioning and angulation of the articular process joints to-gether with variable degrees of irregular hypointense bul-bous enlargement. The distribution of the grade 1 and 2sites is shown in Fig. 5. In all but two MR imaging studies,more than one vertebral site was scored as stenotic (scoreof 1 or 2). A mean of 72% of assessed sites were stenotic(median 75%, range 20–100%).

Forty-eight sites had grade 1 stenosis with the highestfrequency at T3–4 (n = 11) and T4–5 (n = 11). The patternof stenosis for vertebral sites was dorsolateral (n = 29),lateralized (n = 13), and dorsoventral (n = 6). Ten of 48sites were asymmetric. Articular process enlargement wasonly mild or moderate and abnormal oblique orientationof the articular process joints was present at 17 sites.

Twenty sites had grade 2 stenosis. The most frequent sitesof grade 2 stenosis were T2–3 (n = 10) and T3–4 (n = 7).No grade 2 stenosis was present at T5–6. The pattern ofstenosis was: lateralized (n = 8), dorsolateral (n = 10), anddorsoventral (n = 2); 9/20 grade 2 sites were asymmetric.Articular process enlargement was mild to severe. Abnor-mal oblique orientation of the articular process joints waspresent at all grade 2 sites.

For imaging studies (n = 12) where cranial thoracic spinalstenosis was the only neurologic lesion identified, three hadintramedullary T2-hyperintensity at the site of most severestenosis. Mild obstructive hydromyelia was present cranialto the most severe site of stenosis in six dogs (Fig. 1D). Nei-ther of the grade 1 studies in this group had intramedullaryT2-hyperintensity or obstructive hydromyelia. In six studiesassessment of cord intensity was not possible.

Upon initial analysis (not shown) of the 55 MR imagingstudies assessed subjectively as having no cranial thoracicspinal stenosis, there was no significant difference between


TABLE 1. Summary Data from the 55 MR Studies without SubjectiveEvidence of CTSS (Stenosis Grade 0); Differences between Vertebral

Sites Were Assessed Using ANOVA and Tukey’s Analysis Was Used toIdentify Pairs of Data that Differed Significantly the ANOVA Showed

Significant Differences

Pairs of DataVertebral Mean Ratio Significantly Different

Site # ± SD∗ ANOVA (Tukey’s Analysis)

CAR† T1–2 54 0.839 ± 0.139 P < 0.0001 T1/2 vs. T3/4T2–3 55 0.793 ± 0.111 T1/2 vs. T4/5T3–4 55 0.956 ± 0.151 T1/2 vs. T5/6T4–5 53 1.013 ± 0.158 T2/3 vs. T3/4T5–6 28 1.062 ± 0.160 T2/3 vs. T4/5

T2/3 vs. T5/6T3/4 vs. T5/6

COR‡ T1–2 54 0.253 ± 0.056 P = 0.0014 T1/2 vs. T2/3T2–3 55 0.292 ± 0.050 T2/3 vs. T5/6T3–4 55 0.277 ± 0.051T4–5 53 0.269 ± 0.060T5–6 28 0.246 ± 0.048

∗Standard deviation (SD).†Canal area ratio (CAR).‡Cord occupying ratio (COR).

FIG. 6. Mean and standard deviation of canal area ratio (CAR) for grade0 MR studies (n = 55).

different breeds, when breeds with four or more dogs werecompared, and no significant relationship between age andeither canal area ratio or canal occupying ratio. However,using pooled data there was a highly significant differencein the canal area ratio (P < 0.0001) and canal occupy-ing ratio (P = 0.0014) values at different vertebral sites(Table 1, Figs. 6 and 7).

Summary data and analysis of the canal area ratio andcanal occupying ratio of the 24 MR imaging studies withcranial thoracic spinal stenosis (stenosis grade of 1 or 2 atone or more thoracic vertebral levels) are shown in Tables 2and 3, Figs. 8 and 9, together with the comparative data forthe 55 dogs without cranial thoracic spinal stenosis. Thecanal area ratio appeared to give a better discriminationbetween the different subjective levels of stenosis than thecanal occupying ratio. Indeed, median canal occupying ra-tio values did not differ significantly between MR imaging

FIG. 7. Mean and standard deviation of canal occupying ratio (C0R) forgrade 0 MR studies (n = 55).

TABLE 2. Summary of the Canal Area Ratio (CAR) Data from the 79MR Studies with and without Subjective Evidence of CTSS; Differencesbetween Groups of Dogs Was Assessed with the Kruskal–Wallis Test andCorrelation between CAR and SG was assessed with the Spearman Rank

Correlation Coefficient

SG∗ # Median RangeKruskal–Wallis

Test

Spearman RankCorrelation SG

vs. CAR†

T1–2 0a‡ 54 0.848 0.532–1.2540b§ 9 0.865 0.649–1.151 P = 0.0007 rs = –0.4751 10 0.666 0.530–0.796 P < 0.00012 1 0.307 0.307–0.307

T2–3 0a 55 0.770 0.590–0.9960b 5 0.622 0.579–0.897 P < 0.0001 rs = –0.6671 9 0.588 0.502–0.759 P < 0.00012 10 0.489 0.246–0.606

T3–4 0a 55 0.937 0.665–1.4200b 3 0.874 0.733–0.929 P < 0.0001 rs = –0.6401 12 0.676 0.537–0.999 P < 0.00012 7 0.501 0.252–0.684

T4–5 0a 53 0.996 0.770–1.4820b 7 1.021 0.747–1.164 P = 0.0006 rs = –0.4791 11 0.717 0.497–1.143 P < 0.00012 2 0.330 0.229–0.430

T5–6 0a 28 1.064 0.793–1.4230b 6 0.87 0.543–1.422 P = 0.0181 rs = –0.4051 7 0.623 0.522–1.185 P = 0.00022 0 – –

∗Stenosis Grade (SG).†Canal area ratio (CAR).‡0a = Dogs with no subjective evidence of CTSS.§0b, 1, 2 = Stenosis grade at vertebral level of dogs with subjective evidenceof CTSS.

studies with different stenosis grades at three of the fivevertebral sites, and based on the Spearman rank correla-tion coefficient there was a low level of correlation betweenthe stenosis grade and canal occupying ratio, which wasnonsignificant for two of the five vertebral sites.

The agreement between the stenosis grade (positive be-ing a grade of 1 or 2, or negative being a grade of 0) and thecanal area ratio or canal occupying ratio data was calcu-lated (Table 4). For canal area ratio, the overall agreement


TABLE 3. Summary of the Canal Occupying Ratio (COR) Data from the79 MR Studies with and without Subjective Evidence of CTSS;

Differences between Groups of Dogs Was Assessed with theKruskal–Wallis Test and Correlation between COR and SG Was

Assessed with the Spearman Rank Correlation Coefficient

SG∗ # Median RangeKruskal–Wallis

Test

Spearman RankCorrelation SG

vs. COR†

T1–2 0a‡ 54 0.239 0.135–0.4590b§ 9 0.225 0.152–0.319 P = 0.081 rs = 0.2261 10 0.263 0.169–0.374 P = 0.0532 1 0.362 0.362–0.362

T2–3 0a 55 0.298 0.179–0.4080b 5 0.276 0.239–0.360 P = 0.008 rs = 0.3781 9 0.375 0.260–0.452 P = 0.00062 10 0.374 0.230–0.505

T3–4 0a 55 0.269 0.159–0.4210b 3 0.318 0.259–0.360 P = 0.0009 rs = 0.4121 12 0.309 0.172–0.442 P = 0.00022 7 0.425 0.300–0.547

T4–5 0a 53 0.265 0.135–0.4240b 7 0.273 0.144–0.392 P = 0.097 rs = 0.1851 11 0.266 0.186–0.632 P = 0.1172 2 0.664 0.535–0.792

T5–6 0a 28 0.239 0.158–0.3350b 6 0.246 0.148–0.317 P = 0.540 rs = 0.2811 7 0.272 0.217–341 P = 0.2812 0 – –

∗Stenosis Grade (SG).†Canal area ratio (CAR).‡0a = Dogs with no subjective evidence of CTSS.§0b, 1, 2 = Stenosis grade at vertebral level of dogs with subjective evidenceof CTSS.

TABLE 4. Summary of the Agreement between Stenosis Grade and CARor COR Data in the 24 MR Studies with Subjective CTSS (Data Pooled

for All Five Vertebral Sites)

CAR∗ CAR COR† CORPositive Negative Positive Negative

SG‡ positive 36 33 22 47SG negative 1 29 1 29Agreement 65.7% 51.5%

∗Canal area ratio (CAR).†Canal occupying ratio (COR).‡Stenosis grade (SG).

FIG. 8. Box and whisker plot of the canal area ratio (CAR) for the grade0 and grade 1 and 2 groups at each vertebral level. *Significant differencebetween groups present (Kruskal–Wallis). The specific grading groups thatwere significantly different from each other are labeled with letters.

FIG. 9. Box and whisker plot of the canal occupying ratio (COR) forthe grade 0 and grade 1 and 2 groups at each vertebral level. *Significantdifference between groups present (Kruskal–Wallis). The specific gradinggroups that were significantly different from each other are labeled withletters.

(pooling data for all five vertebral sites) was only moder-ate at 66%, with this varying between 55% and 71% forindividual vertebral sites. For canal occupying ratio, theoverall agreement was poorer at 52%, varying between 46%and 55% for individual vertebral sites.

FIG. 10. Bland–Altman plots of canal area ratio (CAR) (A) and canal occupying ratio (COR) (B) data comparing observer difference and mean. Blackdotted line = mean bias. Red dotted lines = 95% limits of agreement.


Pearson’s correlation coefficients and Bland–Altmanplots were used to assess the interobserver agreement forthe calculated ratios (Fig. 10). The correlation between ob-servers for canal area ratio was moderate (r2 = 0.754), butconsiderably better than for canal occupying ratio (r2 =0.577), although neither was characterized by a high levelof correlation.

Based on the Bland–Altman plots, there was little evi-dence of consistent bias between observers, that is, one ob-server consistently measuring canal area ratio or canal oc-cupying ratio higher or lower than the other observer. Meanbias and 95% limits of agreement were +0.006 (−0.216 to+0.229) for canal area ratio data, and −0.013 (−0.1340 to+0.103) for canal occupying ratio. However, there was ev-idence of increasing variance in the measures between thetwo observers with increasing values for both ratios.

If percentage difference between the two observers wascalculated rather than absolute differences, the mean biasand 95% limits of agreement were 10.0% (−24% to +26%)for canal area ratio data, and −4% (−39% to +31%) forcanal occupying ratio data.

Discussion

Dogs with MR features of cranial thoracic spinal steno-sis were predominantly young, heavy, male dogs with aconformation typified by that of Molosser breeds. Osseous-associated cervical spondylomyelopathy in giant breed dogsalso tends to occur in young, predominantly male dogs andhas several morphologic features in common with cranialthoracic spinal stenosis.8, 14–17 This form of cervical spondy-lomyelopathy has multiple proposed etiologies includinggenetic, congenital vertebral malformation, developmentalabnormalities, and acquired changes, some of which maybe applicable to cranial thoracic spinal stenosis.4, 18–20

Molosser breeds are descendants of ancient mastiff-typedogs, which originated in ancient Greece and Babylonia.They became the root stock of the Alpine Mastiff fromwhich modern breeds including the various European mas-tiffs (English Mastiff, German Mastiff, dogues de Bor-deaux, Neapolitan Mastiff), St. Bernard, and Rottweilerstem. Although these modern descendants demonstrate ashared morphology from the common ancient ancestor andare closely related phylogenetically, they are all geneticallydistinct.21 The significance of genetic factors as an etiologyfor cranial thoracic spinal stenosis is therefore uncertain.

A developmental etiology is the most likely explana-tion for the cranial thoracic spinal stenosis in this studyas dogs with clinical signs presented with osseous morpho-logic changes at an early age (median 9.5 months), oftenbefore skeletal maturity of the spine. This is earlier thanin giant-breed dogs with cervical spondylomyelopathy, forwhich a median age of 2.5 years is reported.8 Interestinglythe single dog that was scanned twice in this study had

MR findings of cranial thoracic spinal stenosis recognizedat an early age but with more significant osseous cervicalspondylomyelopathy developing later.

In common with the osseous form of cervical spondy-lomyelopathy, sites of cranial thoracic spinal stenosis wereassociated with enlargement and malformation of the ar-ticular process joints.8, 9, 15 Congenital malformation andasymmetry of the articular processes are factors that affectmobility and increase stress on the articular surface.4, 19

These stressors have been related to development of theligamentous hypertrophy and osteoarthritis observed inosseous-associated cervical spondylomyelopathy and areimplicated in the development of thoracic stenosis inhumans.16, 19, 22–24 The orientation of the articular processjoints in the cranial thoracic spine in normal dogs producesa shallow angle with the dorsal plane.25, 26 At most sites ofcranial thoracic spinal stenosis this angle was subjectivelysteeper and the articular processes more elongated both intransverse and paramedian images. Such abnormalities inarticular process geometry may contribute additional stresson these joints and are likely to play a role in the develop-ment of cranial thoracic spinal stenosis in dogs.

The developmental condition of achondroplasia hasbeen associated with spinal stenosis in humans and dogsand is speculated to be linked with cranial thoracic spinalstenosis.3, 7, 27–29 However, although some morphologic fea-tures correlate with achondroplasia, the stenosis in cranialthoracic spinal stenosis is segmental rather than general-ized and abnormal maturation of cartilage was not notedas a concurrent abnormality associated with cranial tho-racic spinal stenosis in this study.27, 29 The reason for thissegmental distribution of stenosis lesions in the cranial tho-racic spine remains uncertain. An association between so-matotype, and smaller vertebral canal diameter has beenrecognized in humans as a factor in spinal stenosis.30 Simi-larly, in the dog a variation in vertebral canal diameter hasbeen related to breed, size, and conformation.31, 32 Theseosteological studies demonstrated that the cranial thoracicvertebral canal diameter was significantly smaller in largebreed dogs compared to smaller breeds.31 In the Germanshepherd dog, computed tomographic morphometry con-firmed marked narrowing of the vertebral canal diameterat T2–4, though morphometric data for other breeds arelacking.32 In this study the canal area ratio in grade 0 stud-ies indicated that the greatest narrowing between mid-bodyand intervertebral levels was at T2–3, and T3–4, whichwere also the most frequent sites of stenosis in the dogswith primary clinical cranial thoracic spinal stenosis. Con-genital narrowing of the vertebral canal in the cranial tho-racic spine compared to the remainder of the thoracic spinein Molosser-type dog breeds may contribute to segmentalstenosis secondary to developmental or acquired changes.31

However, the number of grade 0 studies of affected breedswas insufficient to allow breed-specific data to be generated.


The modified compression scale used in this study, hasbeen used previously to detect cervical stenosis in Dober-man pinschers.9 However, the presence of spinal cord com-pression did not always correlate to clinical signs, with 4of 16 clinically normal dogs graded as having spinal cordcompression.9 Indeed, in the cervical spine, clinical signsdo not necessarily occur in dogs where the spinal cord issubjectively deformed and compressed as long as the cross-sectional area of the spinal cord is preserved.9, 11 In cranialthoracic spinal stenosis the majority of dogs graded withspinal cord compression (grade 2) had neurologic signs as-sociated with the lesion. Therefore, differentiating betweenthe presence of spinal cord compression and shape change isimportant. This disparity between the cervical and thoracicspine may be due to the difference in spinal mobility withinthese two regions. The cervical spine is highly mobile andcervical spondylomyelopathy is usually linked to dynamicfactors, whereas the cranial thoracic spine is relatively im-mobile and stenosis is therefore static and absolute.4, 5, 33

Still, three dogs with cranial thoracic spinal stenosis hadno neurologic signs and therefore caution should be ex-ercised to avoid overinterpretation of the osseous changeswhere there is a change in spinal cord shape without evi-dence of spinal cord compression and to correlate imagingfindings with clinical status. Spinal cord signal intensitychange has been correlated to clinical status in cervicalspondylomyelopathy, and may be a more accurate way ofidentifying clinically significant lesions.34 IntramedullaryT2-hyperintensity was present in some dogs with neuro-logic signs in this study. However this finding was difficult toassess and was less marked than with the intensity changesobserved in cervical spondylomyelopathy, presumably dueto the absence of dynamic factors.

Six dogs with cranial thoracic spinal stenosis had con-current osseous-associated cervical spondylomyelopathyand one subclinical dog subsequently developed cervicalspondylomyelopathy. All had a Molosser-type breed con-formation. Stenotic sites at the T1 and T2 level have beenreported in 10% of dogs with cervical spondylomyelopa-thy and the necessity to extend the FOV up to the thirdthoracic vertebrae has been recommended.8 In this study,sites of stenosis in dogs with cervical spondylomyelopa-thy were observed caudal to T2, and therefore the aboverecommendation should be to extend the FOV to includethe cranial thoracic spine up to T6 in dogs with cervicalspondylomyelopathy, especially in Molosser breeds. Whenperforming neuroimaging in dogs with a T3-L3 neurolocal-ization, the potential for a lesion within the cranial thoracicspine should not be ignored.7 The nine dogs with primaryclinical cranial thoracic spinal stenosis in this study hadneurologic localization to the T3-L3 spinal cord segmentsand truncal ataxia suggestive of a cranial thoracic lesion.Grade 2 stenosis was identified at T2–3 in most of these

dogs highlighting the importance of including the cranialthoracic spine in dogs with a T3-L3 neurologic localization.

Morphometric evaluation of the spine is used to char-acterize anatomic changes, improve understanding of dis-ease, help define normal populations, and allow objec-tive comparisons to be made.35 Stenosis ratios have beenused extensively in people and the dog to allow compar-isons to be made between spine segments and betweenindividuals.9, 11–13, 36 In this study, morphometric evaluationin the form of stenosis ratios was used to evaluate a popu-lation of large breed dogs that were subjectively normal onMR imaging (grade 0 studies). This evaluation identifiedsignificant differences between vertebral levels indicatingthat a comparison can only be made between different dogsfor the same vertebral level. Morphometric analysis of dogswith evidence of stenosis on MR imaging (grades 1 and 2)found that both the stenosis ratios were poor at detectingstenotic sites. At best there was only moderate correlationwith the subjective grading system. The median canal oc-cupying ratio values were significantly different betweengrade 0 and grades 1 and 2 stenosis at only two ofthe five vertebral levels. This is similar to another studywhere Doberman pinschers with clinical cervical spondy-lomyelopathy only had a significant difference in the canaloccupying ratio compared to normal foxhounds at theC7 level.11, 12 The reason the canal occupying ratio is of-ten not significantly different at stenotic sites may be dueto the presence of spinal cord atrophy, as observed inDoberman pinchers with cervical spondylomyelopathy.9

Therefore, both the spinal cord and vertebral canal cross-sectional areas are reduced at sites of stenosis.

Interobserver variability, for both the canal occupying ra-tio and canal area ratio was sufficiently large that clinicallysignificant differences may still arise in measurements be-tween different observers evaluating the same MR images.This is similar to the assessment of interobserver variabilityof the canal occupying ratio in cervical spondylomyelopa-thy where wide limits of agreement were reported.11 Acause for this variability between observers may be re-lated to the assumptions that have to be made regard-ing vertebral landmarks and the boundary of the verte-bral canal. The methodology used to define the area ofthe cranial intervertebral foramen in this and other stud-ies was similar, but it is important to recognize that atthis level assumptions regarding the lateral margin de-fined by the intervertebral foramen are inevitable.11 Fur-thermore although the agreement between linear measure-ments made on CT and MR images in the dog has beendescribed, validation of MR imaging area measurementshas not been performed.37 Overall, stenosis ratios were oflimited benefit in the detection of stenotic sites, correlatedpoorly to the subjective grading system and had concern-ing levels of interobserver variability. Therefore caution


is recommended in the use and interpretation of stenosisratios.

None of the dogs with cranial thoracic spinal stenosishad pathologic confirmation of the condition and onlyone had surgical confirmation. Determination of a ref-erence standard when assessing imaging tests for spinalstenosis poses difficulties. Surgical confirmation can beconsidered as a gold standard however the limitationis that there may only be absolute confidence when se-vere stenosis is present. Furthermore not all affected in-dividuals undergo surgery and normal patients are notsubjected to surgical confirmation of a normal vertebralcanal.35

Our methods resulted in a lack of age- and breed-matched populations in the grade 0 and stenosis groups.In the cervical spine, breed-significant differences in verte-bral body and canal ratios have been identified and it hasbeen suggested that data collected from one breed cannot beextrapolated to other breeds.22 To assess the significance ofthis limitation in the thoracic spine, differences in the steno-sis ratios between breeds and ages were assessed in the grade0 group. Although significant differences in the stenosis ra-tios between breeds and ages in the grade 0 groups were notidentified there was an insufficient number of many of theMolosser breeds in this group. Also, although the grade 0

dogs used for the morphometric comparison had no subjec-tive evidence of bony stenosis, they could not be considereda true control group as a normal neurologic status was notdetermined.

In conclusion, cranial thoracic spinal stenosis is a pri-marily developmental osseous stenosis in predominantlyyoung, male dogs with a Molosser-type conformation.Stenotic sites were observed most commonly at the T2–3and T3–4 levels. Enlargement, and abnormal oblique ori-entation of the articular process joints are observed andarticular process malformation is likely to be an underly-ing cause. Of the 12 dogs having cranial thoracic spinalstenosis as the only identified neurologic lesion, nine had aT3-L3 neurologic localization with pelvic limb propriocep-tive deficits and truncal ataxia. The presence of spinal cordcompression in cranial thoracic spinal stenosis (grade 2)caused neurologic deficits in the majority of dogs. Howeverspinal cord shape change was observed without the pres-ence of spinal cord compression or neurologic signs andthis should be interpreted with caution. Cranial thoracicspinal stenosis can be observed concurrently with cervi-cal spondylomyelopathy in Molosser-type breeds and it istherefore recommended to screen the cranial thoracic spinefor stenosis in all Molosser-type dogs diagnosed with cer-vical spondylomyelopathy.

REFERENCES

1. Naylor A. Factors in the development of the spinal stenosis syndrome.J Bone Joint Surg Br 1979;61-B:306–309.

2. Westworth DR, Sturges BK. Congenital spinal malformations in smallanimals. Vet Clin North Am Small Anim Pract 2010;40:951–981.

3. Bailey CS, Morgan JP. Congenital spinal malformations. Vet ClinNorth Am Small Anim Pract 1992;22:985–1015.

4. da Costa RC. Cervical spondylomyelopathy (Wobbler syndrome) indogs. Vet Clin North Am Small Anim Pract 2010;40:881–913.

5. Meij BP, Bergknut N. Degenerative lumbosacral stenosis in dogs. VetClin North Am Small Anim Pract 2010;40:983–1009.

6. Knecht CD BW, Raffe MR. Stenosis of the thoracic spinal canal inEnglish bulldogs. J Am Anim Hosp Assoc 1979;15:181–183.

7. Stalin CE PJ, Smith PM, Jeffery ND. Thoracic stenosis causing lateralcompression of the spinal cord in two immature Dogues de Bordeaux. VetComp Orthop Traumatol 2009;22:59–62.

8. da Costa RC, Echandi RL, Beauchamp D. Computed tomographymyelographic findings in dogs with cervical spondylomyelopathy. Vet RadiolUltrasound 2012;53:64–70.

9. da Costa RC, Parent JM, Partlow G, Dobson H, Holmberg DL,Lamarre J. Morphologic and morphometric magnetic resonance imagingfeatures of Doberman Pinschers with and without clinical signs of cervicalspondylomyelopathy. Am J Vet Res 2006;67:1601–1612.

10. Kasai Y, Uchida A. New evaluation method using preoperative mag-netic resonance imaging for cervical spondylotic myelopathy. Arch OrthopTrauma Surg 2001;121:508–510.

11. De Decker S, Gielen IM, Duchateau L, et al. Morphometric dimen-sions of the caudal cervical vertebral column in clinically normal DobermanPinschers, English Foxhounds and Doberman Pinschers with clinical signsof disk-associated cervical spondylomyelopathy. Vet J 2012;191:52–57.

12. Okada Y, Ikata T, Katoh S, Yamada H. Morphologic analysis ofthe cervical spinal cord, dural tube, and spinal canal by magnetic resonanceimaging in normal adults and patients with cervical spondylotic myelopathy.Spine (Phila Pa 1976) 1994;19:2331–2335.

13. Mayhew PD, Kapatkin AS, Wortman JA, Vite CH. Association ofcauda equina compression on magnetic resonance images and clinical signsin dogs with degenerative lumbosacral stenosis. J Am Anim Hosp Assoc2002;38:555–562.

14. Lewis DG. Cervical spondylomyelopathy (‘Wobbler’ syndrome) in thedog: a study based on 224 cases. J Small Anim Pract 1989;30:657–665.

15. Lipsitz D, Levitski RE, Chauvet AE, Berry WL. Magnetic resonanceimaging features of cervical stenotic myelopathy in 21 dogs. Vet Radiol Ul-trasound 2001;42:20–27.

16. Gutierrez-Quintana R, Penderis J. MRI features of cervical articularprocess degenerative joint disease in Great Dane dogs with cervical spondy-lomyelopathy. Vet Radiol Ultrasound 2012;53:304–311.

17. Denny HR, Gibbs C, Gaskell CJ. Cervical spondylopathy in the dog:a review of thirty-five cases. J Small Anim Pract 1977;18:117–132.

18. Abramson CJ, Dennis R, Smith KC, Platt SR. Radiographic diagnosis:lateralized vertebral osseous compression causing cervical spondylomyelopa-thy in a Great Dane. Vet Radiol Ultrasound 2003;44:56–58.

19. Trotter EJ, deLahunta A, Geary JC, Brasmer TH. Caudal cervicalvertebral malformation-malarticulation in Great Danes and Doberman Pin-schers. J Am Vet Med Assoc 1976;168:917–930.

20. Selcer RR, Oliver JE. Cervical spondylopathy: Wobbler syndrome indogs. J Am Anim Hosp Assoc 1975;11:175–179.

21. Parker HG, Kim LV, Sutter NB, et al. Genetic structure of the pure-bred domestic dog. Science 2004;304:1160–1164.

22. Drost WT, Lehenbauer TW, Reeves J. Mensuration of cervical verte-bral ratios in Doberman pinschers and Great Danes. Vet Radiol Ultrasound2002;43:124–131.

23. Lim A DUP. Single-level bilateral facet joint hypertrophy causingthoracic spinal canal stenosis. J Clin Neurosci 2009;16:1363–1365.

24. Barnett GH HRJ, Little JR, Bay JW, Sypert GW. Thoracic spinalcanal stenosis. J Neurosurg 1987;66:338–344.

25. Evans HE. The skeleton. In: Evans HE (ed): Miller’s anatomy of thedog. Philadelphia: W. B. Saunders, 1993;174–176.


26. Baines EA, Grandage J, Herrtage ME, Baines SJ. Radiographic defini-tion of the anticlinal vertebra in the dog. Vet Radiol Ultrasound 2009;50:69–73.

27. Nelson MA. Spinal stenosis in achondroplasia. Proc R Soc Med1972;65:1028–1029.

28. Wynne-Davies R, Walsh WK, Gormley J. Achondroplasia andhypochondroplasia. Clinical variation and spinal stenosis. J Bone Joint SurgBr 1981;63B:508–515.

29. Misra SN, Morgan HW. Thoracolumbar spinal deformity in achon-droplasia. Neurosurg Focus 2003;14:e4.

30. Nightingale S. Developmental spinal canal stenosis and somatotype.J Neurol Neurosurg Psychiatry 1989;52:887–890.

31. Breit S. Osteological features in pure-bred dogs predisposing to tho-racic or lumbar spinal cord compression. Res Vet Sci 2002;73:87–92.

32. Dabanoglu I KME, Turan E, Ocal MK. Morphometry of the thoracicspine in German Shepherd dog: a computed tomographic study. Anat HistolEmbryol. 2004;33:53–58.

33. Takeuchi T, Abumi K, Shono Y, Oda I, Kaneda K. Biomechanical roleof the intervertebral disc and costovertebral joint in stability of the thoracicspine. A canine model study. Spine (Phila Pa 1976) 1999;24:1414–1420.

34. da Costa RC, Parent J, Dobson H, Holmberg D, Partlow G. Com-parison of magnetic resonance imaging and myelography in 18 Dobermanpinscher dogs with cervical spondylomyelopathy. Vet Radiol Ultrasound2006;47:523–531.

35. Kent DL, Haynor DR, Larson EB, Deyo RA. Diagnosis of lumbarspinal stenosis in adults: a metaanalysis of the accuracy of CT, MR, andmyelography. AJR Am J Roentgenol 1992;158:1135–1144.

36. Laurencin CT LSJ, Senatus P, Botchwey E, Jones TR, Koris M.,Hunter J. The stenosis ratio: a new tool for the diagnosis of degenerativespinal stenosis. J Surg Invest 1999;1:127–131.

37. De Decker S, Gielen IM, Duchateau L, Polis I, Van Bree HJ, VanHam LM. Agreement and repeatability of linear vertebral body and canalmeasurements using computed tomography (CT) and low field magneticresonance imaging (MRI). Vet Surg 2010;39:28–34.

clinical, morphologic, and morphometric features of cranial thoracic spinal stenosis in large and...

Documents