Interpretation of Post Procedural ERCP Spot Films by Radiologists Is Prone to Significant Error

Abstract

Background: ERCP has evolved from a diagnostic to a predominantly therapeutic procedure, where decisions regarding interventions need to be made immediately. Over time, radiology support during ERCP has declined due to time and other constraints. Radiologists are now rarely present during the procedure and typically ‘post read’ saved spot films. Post reading cannot guide decision making during this dynamic test and it is unclear whether it is as accurate as a dynamic read. This study was undertaken to evaluate discrepancies between endoscopists’ dynamic reading of fluoroscopic images and spot films and radiologists’ post reading of spot films alone. Methods: We reviewed the charts of 108 consecutive patients who underwent ERCP at a major US university hospital, performed by two experienced therapeutic endoscopists. Radiologists were not present during ERCP. 100 patients for whom both endoscopy and radiology reports were available (endoscopy n=108; radiology n=100) were analyzed for clinically relevant discrepancies. All 100 cases were also subsequently ‘post read’ in a blinded fashion by two endoscopists and an experienced GI radiologist to arrive at a consensus interpretation, against which the discrepancies were compared. Results: The radiology post read (RPR) missed 65 significant findings in 43 patients, while the endoscopy dynamic read missed 2 significant findings in 2 patients. RPR correctly identified only 20 of 34 patients with CBD stones (Sensitivity=59%; 95% CI = 41-75%), 18 of 24 biliary strictures (75%; 53-89%), 46 of 65 cases of biliary dilation (71%; 58-81%), 41 of 56 biliary stents (73%; 59-84%), 5 of 10 pancreatic ductal (PD) strictures (50%; 20-80%), 6 of 15 PD stents (40%; 17-67%) and 14 of 15 cases of PD dilation (93%; 66-100%). RPR identified 100% of bile leaks (3 of 3), PD stones/calcifications (13 of 13) and pancreatic cysts (2 of 2). RPR identified only 61 of 143 intraductal endoscopic interventions. Conclusions: 1. Radiology interpretation of post ERCP spot films does not guide decision making during this dynamic test and adds little to the endoscopist's read of the study. 2. The significant error rate in radiology reports is likely due to an interpretation based solely on spot films, whereas endoscopists have access to dynamic intraprocedural fluoroscopic and endoscopic images. 3. Cost analysis of endoscopist interpretation alone versus the current practice of dual interpretation will be presented. Background: ERCP has evolved from a diagnostic to a predominantly therapeutic procedure, where decisions regarding interventions need to be made immediately. Over time, radiology support during ERCP has declined due to time and other constraints. Radiologists are now rarely present during the procedure and typically ‘post read’ saved spot films. Post reading cannot guide decision making during this dynamic test and it is unclear whether it is as accurate as a dynamic read. This study was undertaken to evaluate discrepancies between endoscopists’ dynamic reading of fluoroscopic images and spot films and radiologists’ post reading of spot films alone. Methods: We reviewed the charts of 108 consecutive patients who underwent ERCP at a major US university hospital, performed by two experienced therapeutic endoscopists. Radiologists were not present during ERCP. 100 patients for whom both endoscopy and radiology reports were available (endoscopy n=108; radiology n=100) were analyzed for clinically relevant discrepancies. All 100 cases were also subsequently ‘post read’ in a blinded fashion by two endoscopists and an experienced GI radiologist to arrive at a consensus interpretation, against which the discrepancies were compared. Results: The radiology post read (RPR) missed 65 significant findings in 43 patients, while the endoscopy dynamic read missed 2 significant findings in 2 patients. RPR correctly identified only 20 of 34 patients with CBD stones (Sensitivity=59%; 95% CI = 41-75%), 18 of 24 biliary strictures (75%; 53-89%), 46 of 65 cases of biliary dilation (71%; 58-81%), 41 of 56 biliary stents (73%; 59-84%), 5 of 10 pancreatic ductal (PD) strictures (50%; 20-80%), 6 of 15 PD stents (40%; 17-67%) and 14 of 15 cases of PD dilation (93%; 66-100%). RPR identified 100% of bile leaks (3 of 3), PD stones/calcifications (13 of 13) and pancreatic cysts (2 of 2). RPR identified only 61 of 143 intraductal endoscopic interventions. Conclusions: 1. Radiology interpretation of post ERCP spot films does not guide decision making during this dynamic test and adds little to the endoscopist's read of the study. 2. The significant error rate in radiology reports is likely due to an interpretation based solely on spot films, whereas endoscopists have access to dynamic intraprocedural fluoroscopic and endoscopic images. 3. Cost analysis of endoscopist interpretation alone versus the current practice of dual interpretation will be presented.