The Accuracy of Intraoperative O-arm Images for the Assessment of Pedicle Screw Postion

Abstract

Human cadaveric study. The objective of the study was to determine the accuracy of intraoperative O-arm images in determining pedicle screw position using open dissection as the gold standard. Pedicle screws are widely used in the treatment of various spinal disorders. Postoperative computed tomographic scans are the imaging gold standard to detect pedicle screw malposition. However, a second procedure is necessary if such malpositioned screws have to be revised. The O-arm is an intraoperative scanner that allows revision of a screw without having to return the patient to the operating room for a separate procedure. No previous studies have looked at the accuracy of intraoperative O-arm images in determining pedicle screw position. This factorial validation study utilized 9 cadavers in a comparison of intraoperative O-arm images and the dissection gold standard. Four hundred sixteen screws were inserted using 3-dimensional image (O-arm) guidance from C2 to S1. The screw positions were randomized into 3 groups: "IN" (fully contained within the pedicle), "OUT-lateral," or "OUT-medial." After screw insertion, O-arm images were obtained and reviewed in a blinded fashion by 3 independent observers. Dissection identified the true position of the screws. Specificity, sensitivity, positive predictive value (PPV), and negative predictive value (NPV) were calculated using dissection results as the gold standard. The interobserver reliability was also determined. The overall accuracy, specificity, sensitivity, PPV, and NPV of O-arm images for the thoracic and lumbar spine were 73%, 76%, 71%, 74%, and 72%, respectively. Accuracy of surgeon perception in the cervical spine was significantly less than in the thoracic and lumbosacral spine. There was substantial interobserver agreement between the 3 readers. Intraoperative O-arm images accurately detect significant pedicle screw violations in the thoracic and lumbosacral spine but are less accurate for the cervical spine.