Session VIII - Pelvis
Magnetic Resonance Venography: Interobserver Variability in the Identification of Pelvic Venous Thrombosis
Steven J. Morgan, MD; Michael S. Stover, MD; Daniel J. Stackhouse, MD; Michael J. Kelley, MD; Brian A. Howard, MD; Michael J. Bosse, MD, Carolinas Medical Center, Charlotte, North Carolina
Purpose: The purpose of this study was to determine the degree of interobserver variability in the interpretation of magnetic resonance venography (MRV) of the pelvic venous system for the presence of pelvic thrombosis in asymptomatic individuals with pelvic and acetabular fractures.
Methods: The MRV images of the pelvic vasculature in thirty-two patients with pelvic and acetabular fractures awaiting definitive fixation were reviewed independently by three separate radiologists experienced in the interpretation of MRV. They were asked to determine whether a thrombus was present, absent, or that the scan was uninterpretable. The original interpretation of the patient's MRV served as the fourth observer group. Data were analyzed with standard statistical means, and the level of agreement was assessed with kappa statistics for each category. A weighted kappa statistic was calculated for overall agreement. A kappa value of 1.0 indicates perfect agreement while a kappa value of 0.0 would be expected with random guessing.
Results: All of the studies were considered interpretable by the three observers; one study was considered uninterpretable in its initial reading. The four observers evaluating 32 MRV images had complete agreement on 22 (68.8 percent) of the readings. Of those 22 cases with complete agreement, the readers concurred that no DVT was present. The overall kappa value was 0.26472 (Strength of Agreement - Fair). The individual kappa statistics for each of the categories as well as the overall kappa statistic are listed in the table
| MRV Interpretation | Kappa Value |
Strength of Agreement |
| No Thrombus Present | 0.27893 |
Fair |
| Thrombus Present | 0.26734 |
Fair |
| Unable To Read | -0.0078740 |
Poor |
| Overall Weighted Kappa | 0.26472 |
Fair |
Discussion: Deep venous thrombosis (DVT) and pulmonary embolus are common complications in trauma patients; those with pelvic and acetabular fractures are believed to be at increased risk for DVT involving the pelvic venous structures. Magnetic resonance venography (MRV) has recently been proposed as the imaging modality of choice for identification of pelvic DVT in patients who undergo surgery for pelvic and acetabular fractures. The clinical utility of a test depends on the consistency of its interpretation. Observer variability is an important source of error in imaging studies. If observer variability is high, either the technique should not be introduced or improvements must be made and the technique re-evaluated in further variability studies before introduction into general use.
The low consistency in interpretation or increased interobserver variation as represented by the low percentage of absolute agreement (68%) and the decreased overall kappa statistic suggesting only fair agreement is a cause for concern. Extrapolation of this data would suggest that 32% of the patients evaluated for DVT in this manner would receive a different interpretation based on the reviewing radiologist. This is a significant dilemma considering that vena caval filter placement is recommended to prevent pulmonary embolism in fracture patients awaiting definitive stabilization of their fractures when diagnosed with DVT in the thigh or pelvis region pre-operatively.
Conclusion: The high degree of observer variability in the interpretation of MRV in patients with pelvic and acetabular fractures being evaluated for asymptomatic pelvic DVT precludes its routine use in this group of patients.