A. K. LaRiccia1,2, T. W. Wolff1,2, M. O’Mara1, T. V. Nguyen1, J. Hill1, D. J. Magee4, R. Patel4, D. W. Hoenninger4, M. Spalding1,3 1Ohiohealth Grant Medical Center,Trauma And Acute Care Surgery,Columbus, OH, USA 2Ohiohealth Doctors Hospital,Surgery,Columbus, OH, USA 3Ohio University Heritage College of Osteopathic Medicine,Dublin, OH, USA 4Ohiohealth,Columbus Radiology,Columbus, OH, USA
Introduction: Blunt cerebrovascular injury (BCVI) occurs in 1-2% of all blunt trauma patients. Computed tomographic angiography of the neck (CTAn) has become commonplace for diagnosis and severity determination of BCVIs. Management often escalates with injury grade and inaccurate grading can lead to both under- and over-treatment of these injuries. Several studies have investigated the sensitivity of CTAn, however, there remains a lack in understanding the inter-reader reliability. In this study, we determine the extent of variability in BCVI grades among neuro-radiologist interpretation of CTAn in traumatically injured patients.
Methods: This was a retrospective review of trauma patients with a BCVI reported on initial CTAn imaging, admitted to an urban, Level I trauma center from January 2012 to December 2017. Patients were randomly assigned for CTAn re-evaluation by two of three blinded, independent neuro-radiologists. The evaluations were compared and the variability among the BCVI grades was measured using coefficient of unalikeability (u), which can quantify variability for categorical variables on a scale of 1-100 where the higher the value, the more unalike the data. Inter-reader reliability of the radiologists was calculated using weighted Cohen’s kappa (k).
Results: In total, 228 BCVIs in 217 patients were analyzed. Seventy-six (33%) involved the carotid vessels, 144 (63%) involved only vertebral vessels, and 8 (4%) involved both. The initial grades consisted of 71 (31%) grade 1, 74 (32%) grade 2, 26 (11%) grade 3, 57 (25%) grade 4, and 0 grade 5. Interpretation variability was present in 93 (41%) of all BCVIs. Initial grade 1 injuries had the lowest occurrence of uniform consensus (u = 1) with a mean of 31% among all interpretations (see figure). Grade 4 injuries had the highest consensus (92%). Grade 2 and 3 injuries had a mean consensus of 63% and 61%, respectively. Total variability of grade interpretations (u = 100) occurred most frequently with grade 3 BCVIs (21%). No significant differences were found between carotid and vertebral injuries. Weighted Cohen’s k calculations had a mean of 0.07, indicating poor reader agreement. Treatment recommendations would have been affected in 30% of these patients, with the treatment scope downgraded in 22% and upgraded in 8%.
Conclusion: Our study revealed BCVI variability of initial radiological grade interpretation in more than a third of patients and poor reader agreement. The reliability of CTAn interpretation of BCVI grades is not uniform, potentially leads to 8% under treatment and worse neurologic outcomes. Comparisons with variability in digital subtraction angiography may be beneficial to further understand the complexity of BVCI radiologic injury grading.
