Diagnostic Test Validity of Structural Vertebral Endplate Defects

Background: Back pain is a common health problem and a substantial burden on affected individuals, their families and society. In fact, it is the single leading cause of disability worldwide. Unfortunately, the pathological mechanisms behind back pain are not well understood, hindering the development of well-targeted, effective prevention and treatment approaches. The anatomical structures referred to as the functional spinal units, each composed of an intervertebral disc, its two adjacent vertebrae, and other associated osteoligamentous structures, have been studied for several decades with limited success in identifying the source of the pain. The intervertebral disc has received the most attention, and while disc degeneration or pathology is associated with the clinical conditions of disc herniation and radiculopathy, as well as spinal stenosis, the role of the disc in common back pain is uncertain. More recently, the vertebral endplate has attracted attention as a possible contributor to back pain problems.

The vertebral endplate is a thin interface structure between the intervertebral disc and the cancellous bone of the adjacent vertebral body. Unlike the disc, the bony vertebral endplate is highly vascularized and has a plentiful neural supply, and structural defects or lesions in the endplate are relatively common in cadaveric samples and on clinical imaging. However, there is a wide range of prevalence rates from 9% to 75% for structural endplate defects across studies and conflicting findings on the association of structural endplate defects with back pain. Also, there is a lack of consistency in the use of terms and definitions for endplate defects, which limits the ability to effectively communicate findings, interpret and compare study results, and build a coherent body of knowledge on the causes and consequences of endplate defects.

Furthermore, the absence of clear definitions or descriptions for endplate defects represented by various terms is problematic for measurement reliability, which is seldom reported in studies of endplate defects. A morphologic description has been found to be more important in establishing the reliability of a standardized MRI nomenclature than the experience of the reader. However, a recent review identified few measurement methods with detailed classification and definition of structural endplate phenotypes. Among them were methods presented by Feng's and Brayda-Bruno's team. Both measurement methods stem from a previous study on endplate defect phenotypes observed in a cadaveric sample. Good to excellent intra- and inter-rater reliability was reported for each classification system in a single MRI study, but neither has been validated.

Clinical CT is one of the most common imaging modalities used for the spine and its use is expected to increase with advancements in clinical CT technology, such as the low tube potential and iterative reconstruction that decreases scan time and radiation dose while optimizing image quality. Yet, with CT, as with MRI, the varied measurement methods used to document the appearance of endplate defects on clinical imaging in the scientific literature, and the absence of validation of such methods leaves uncertainty about what the observations on imaging actually represent. It is also unclear how sensitive clinical imaging is in picking up different types and sizes of defects (e.g. diagnostic accuracy). Micro-CT has been used as a reference standard in human studies and for the characterization of morphological features of endplate defects in cadaveric samples. Micro-CT measurement is accurate, consistent and reproducible, and has the advantage of providing shape and texture information for an object within a single measurement.

Our study will address the reliability and diagnostic test validity of structural endplate defect assessments of clinical CT. Such information is critical for establishing meaningful standards for the evaluation and interpretation of endplate defects on imaging, as well as for studies of their etiology and clinical consequences.

STUDY OBJECTIVES This study seeks to evaluate and compare the reliability and validity of two methods to identify and characterize structural endplate defects on clinical imaging.

The specific objectives are to:

1. Determine the inter-rater reliability of endplate defect determinations on clinical CT scans.
2. Determine the diagnostic accuracy of assessments of clinical CT-scans for identifying endplate defects, using micro-CT as the reference standard.

MATERIALS & METHODS This is a reliability and validity study of vertebral endplate defects on clinical CT (index test) using micro-CT as the reference standard in 19 human cadavers.

Materials Source: This study comprises cadavers of 9 men and 10 women with a median age of 82 years (range 62-91) from the Anatomy lab and Western University's whole-body donation program maintained by the Department of Anatomy and Cell Biology, Schulich School of Medicine and Dentistry. We will use a sample of 19 harvested T7-S1 spines from embalmed (fixed) cadavers. Embalming consisted of arterial distribution of embalming fluid, containing a mixture of ethanol, phenol, and formalin (Wessel & Associates: Troy, MI, USA), 24- to 48- hours post-mortem).

Donor information: Only the anonymized cadaver ID number, cause of death, age, and sex of the donor were available from Western's donation program.

Spine imaging Clinical CT was acquired (scanned at 1.25mm x 1.25mm on a standard algorithm and then reformatted into an axial, coronal and sagittal series at 3mm x 2mm) on all fully intact cadavers prior to the onset of this study by the Schulich School of Medicine & Dentistry at the University of Western Ontario, in accordance with the Anatomy Act of Ontario and guidelines of Western's Committee for Cadaveric Use in Research.

Micro-CT (µCT) of all cadaveric specimens was performed at Robarts Research Institute at a peak voltage of 80 kVp and a tube current of 50 mA. The X-ray projections were reconstructed into a single three-dimensional volume with isotropic voxel spacing of 154 µm. In preparation for µCT the lower thoracic and lumbar spine (T7 through S1) was removed from each cadaver. The ribs were dissected 1-2 cm lateral to the costovertebral joints and the soft tissue associated with the spine was preserved and remained intact during micro-CT.

Assessment of endplate defects Assessment of clinical CT scans: Three raters will independently assess the sagittal images of the clinical CT for endplate defect presence, phenotype classification, and defect size and location, blinded to all prior assessments. Endplate defect measurement methods by Brayda-Bruno's and Feng's teams will be strictly used. First, measurement according to Brayda-Bruno will be used to assess the 19 clinical CTs of the spines. Two weeks after the Brayda-Bruno assessment, evaluations according to Feng will be performed on sagittal images to determine the type and location of the defects, as well as to measure the anteroposterior and transverse diameter of the defects and endplates. Across all measurements, endplates will be labelled with reference to the disc. All endplate defect measurements will be observed from the immediate normal slice before the defect to the first normal slice after the defect. Feng's initial measurement protocol used only sagittal MR images to estimate the transverse diameter by dividing the number of images with endplate defects by the total number of sagittal images containing the endplate. However, due to the large number of sagittal images in CT compared to MRI, the current study will use the transitional distance covered by the defects in relation to the image to determine the transverse diameter. Artefact including osteophytes will be excluded when taking an endplate measurement. Each rater will assess the images in the same order (Brayda-Bruno's followed by Feng's measurements). All measurements will be taken in millimetres rounded up to the nearest decimal point. All assessments will be entered into a pilot-tested excel data entry form. After all assessments, a panellist with the three raters will discuss each rating for a consensus on the single index test that will be compared to the reference standard.

Raters, training, and Consensus The clinical-CT scans will be assessed independently by two radiologists with experience in musculoskeletal imaging and a physical therapist, a Ph.D. student studying endplate defects. A master's student in Clinical Anatomy has assessed and documented endplate defects on micro-CT, the reference standard.

To ensure a consistent understanding of the nomenclature in each classification system, an evaluation manual consisting of measurement descriptions and figures was prepared based on the published articles introducing each of the classification systems. For the purpose of training and to identify practical issues in the evaluation process; four joint training sessions were organized for the raters to evaluate training sets of clinical CT images and discuss practical issues. Each rater independently rated three clinical CT images (66 endplates). Results were collected and analyzed, conflicting domains were noted, discussed and clarified during meetings. All the raters indicated satisfaction with the understanding and ability to use each measurement method.

Statistical Analysis Data will be imported into Stata/IC 14 (Stata Corp LP) for statistical analysis. Inter-rater reliability among the three raters (objective 1) of the clinical CT measurements will be determined using Cohen's kappa statistic for categorical variables (presence or absence of defect and the type of the defect) or Intraclass Correlation Coefficients (ICC) for continuous variables (e.g., anteroposterior and transverse diameters.). ICC and Kappa values will be interpreted as poor (< 0.40), fair (0.41-0.60), good (0.61-0.80), and excellent (0.81-1.0). Discrepancies will then be reviewed and resolved through consensus. For objective 2, for each endplate level, a 2x2 table will be used to calculate percent agreement in detecting endplate defects between the clinical CT and micro-CT.