Radiographic Sensitivity and Negative Predictive Value for Acute Canine Spinal Trauma

Journal Club – Summary

Radiographic Sensitivity and Negative Predictive Value for Acute Canine Spinal Trauma. 

Kinns, J., Mai, W., Seiler, G., Zwingenberger, A., Johnson, V., Caceres, A., ValdesMartinez, A. and Schwarz, T., (2006). Radiographic sensitivity and negative predictive value for acute canine spinal trauma. Veterinary Radiology & Ultrasound, 47(6), pp.563-570 

What is the Evidence Based Veterinary Medicine (EBVM) research question being asked?  

Using sensitivity and negative predictive values, how adequate is radiography in detecting or ruling out acute spinal osseous lesions in canine trauma patients when compared to CT, and what is the interobserver variability of the results of the radiographic assessment of the spine in these patients? 


  • Acute spinal trauma can lead to unstable spinal injuries which can potentially result in temporary or permanent paralysis. 
  • Sedation can be contraindicated in patients with spinal trauma as it may release the muscular bracing that is stabilising the spine, making it more susceptible to further damage. 
  • Positioning patients without sedation can lead to images that are not of optimal diagnostic value. 
  • CT is the gold standard imaging modality for the detection of acute spinal osseus lesions. It may also be safer to use with a sedated or anesthetised patient as it can be performed with the patient strapped to a spinal board. 
  • The author introduces the concept of the 3-compartment model for assessment of unstable versus stable vertebral fractures, where an unstable fracture is one in which 2 out of 3 compartments (dorsal, middle or ventral) are fractured. 

What is the design of this study?  

  • This was a retrospective, multiple observers, blinded study. 
  • Population of 30 canine patients. 
  • Inclusion criteria were patients that presented to University of Pennsylvania between 1999 and 2005, following acute trauma, and with clinical signs attributable to lesions in the spinal column. The CT had to have been undertaken within 48 hours of survey radiographs and both studies were available.  
  • Exclusion criteria: studies that were deemed not of diagnostic value 
  • No randomisation specified for patient inclusion. 
  • No control 

Materials and Methods 

  • Canine patients between 1999 and 2005 with history of trauma and clinical signs suggestive injury to spinal cord. 
  • Patient must have had CT within 48h of radiography – both studies available 
  • Radiographs were reviewed independently, in random order by 4 observers (two board certified radiologists, one board eligible radiologist, and one radiology residentwho completed a pre-defined form for assessment. 
  • CT’s were reviewed similarly by 3 separate observers (two board certified radiologists, one radiology resident) in a predefined form and a consensus decision was made. 
  • Independent observers recorded findings using the same custom form. 
  • Accuracy matrices were calculated using CT as ‘gold standard’. 
  • Sensitivity and Negative Predictive Value (NPV) calculated for various parameters 


  • Presence of multi-observers and consensus decision reduces bias from a single observers’ result. 
  • Statistical method was clearly explained, and the results were valid with an appropriate level of significance (P<0.05). 
  • Reporting from the multiple observers was standardised using the same table. 
  • The 7 independent observers, separate for CT and radiography, provides a good degree of blinding and allows for a representative result for interobserver repeatability, and bias reduction through consensus decision for the CT images. 
  • Well written paper with clearly presented results, well explained design, with valuable conclusions with a practical applicability. 


  • Sample size was small, and no power analysis was performed to determine whether the paper had statistically significant results. 
  • Retrospective results not as statistically strong as prospective, it is more difficult to standardise the inclusion criteria. 
  • This was a blinded study, although it was not specified whether the case selector was involved in the radiographic review, as this would have compromised blinding for that reviewer. 
  • Suboptimal standardisation in terms of what type of radiographic study was being compared to CT, most but not all studies included lateral and VD, but some were only one view (the percentage of which is not reported) which would have decreased sensitivity. In 5 patients (16%) lumbar myelography was included also – which may have increased sensitivity for the detection of lesions 
  • The areas radiographed remove part of the observer blinding as it suggests the area that is of clinical interest to the original clinician, however this reflects a true clinical situation where the observer is not blinded to detailed clinical information 
  • Inclusion criteria of radiography, followed by CT and in patients with neurological symptoms, may have biased the case selection towards patients with more significant imaging lesions, this would have increased sensitivity. Alternatively, some patients with more obvious pathology of the spine on radiology may have not been referred for CT and this would then decrease sensitivity. 


  • Criticisms do not invalidate study findings. 
  • Low NPV (48%) and moderate sensitivity (72%) for radiographic detection of all vertebral fractures. 
  • Sensitivity better (82%) if any fracture was recognized, however the relevance is questionable as one missed lesion can change the treatment of prognosis for that patient. 
  • NPV of subluxation (88%), Fracture fragments in the canal (58%) and vertebral canal narrowing (51%) 
  • 17 of 32 individual fractures missed, 41% of these were unstable.  

Figure 1 Summarised Results copied from the paper 

  • Interobserver agreement was only fair to moderate for most parameters, even among experienced observers. 

Conclusion and Applicability 

This study yields important results. The moderate sensitivity (72%) and low NPV (49%) of detection for all fractures, the moderate 75% sensitivity, and 88% NPV for subluxations and the low sensitivity (57.5%), and NPV (58%) for fracture fragments in the vertebral canal were interesting and indicated that radiography cannot reliably used as a rule out tool for acute osseus lesions in the spine. The low sensitivity (58%) and low NPV (51%) for vertebral canal narrowing suggests that radiography cannot reliably be used to detect compressive lesions in trauma patients. One could extrapolate that these sensitivities and NPV’s would be lower in a general practice setting. The study suggests that patients suffering acute trauma with potential spinal trauma should be preferably referred for CT. The low interobserver agreements on radiography for all findings reemphasised the above conclusions, as these observers were well qualified and in a general practice setting this might be even lower. 

Another ‘take home’ from this paper included the introduction of the 3-compartment model to classify vertebral fractures as stable versus unstable. 

window.hsFormsOnReady = window.hsFormsOnReady || []; window.hsFormsOnReady.push(()=>{ hbspt.forms.create({ portalId: 7971300, formId: "0f1cba90-3851-49b9-8e85-314d01a6ca62", target: "#hbspt-form-1709529157000-4943214947", region: "na1", })});