ACCURACY AND AGREEMENT OF AN INTERNATIONAL PANEL FOR THE DIAGNOSIS OF PANCREATIC CYSTIC LESIONS USING EUS-GUIDED NEEDLE-BASED CONFOCAL LASER ENDOMICROSCOPY
Presentation Number: 78View Presentation
AuthorBlock: Jorge Machicado1, Bertrand Napoleon2, Anne Marie Lennon3, Damien Meng Yew Tan4, Stephen P. Pereira5, Samer Eldika6, Rahul Pannala7, Mohit Girotra8, Pradermchai Kongkam9, Helga Bertani10, Yunlu Feng11, Sijie Hao12, Nathan Zhongning13, Vincent Valantin14, Sarah Leblanc15, Alice Hinton16, Somashekar G. Krishna17
1Mayo Clinic Health System, Eau Claire, Wisconsin, United States; 2Ramsay Generale de Sante, Paris, Île-de-France, France; 3Johns Hopkins Medicine, Baltimore, Maryland, United States; 4Singapore General Hospital, Singapore, Singapore; 5University College London, London, London, United Kingdom; 6Stanford University, Stanford, California, United States; 7Mayo Clinic Arizona, Scottsdale, Arizona, United States; 8University of Miami School of Medicine, Miami, Florida, United States; 9Bumrungrad International Hospital, Bangkok, Bangkok, Thailand; 10Azienda Ospedaliero Universitaria Citta della Salute e della Scienza di Torino, Torino, Piemonte, Italy; 11Peking Union Medical College Hospital, Dongcheng-qu, Beijing, China; 12Huashan Hospital Fudan University, Shanghai, Shanghai, China; 13Qilu Hospital of Shandong University Qingdao, Qingdao, Shandong, China; 14Hopital Europeen Georges Pompidou, Paris, Île-de-France, France; 15Hopital Cochin Service de Cancerologie, Paris, Île-de-France, France; 16The Ohio State University, Columbus, Ohio, United States; 17The Ohio State University Wexner Medical Center, Columbus, Ohio, United States;
Background: EUS-guided needle based confocal laser endomicroscopy (nCLE) can diagnose pancreatic cystic lesions (PCLs) with high accuracy and substantial interobserver agreement (IOA) and intraobserver reliability (IOR); however, prior agreement and reliability studies had small patient populations (n<35), used few observers (n≤6), and did not use histopathology as the ‘reference standard’. Thus, we sought to assess the IOA, IOR, and performance of EUS-nCLE for differentiating PCLs in a larger cohort of patients and observers.
Methods: EUS-nCLE videos representative of the most common types of PCLs were enrolled from subjects recruited at three prospective studies [INDEX (single center), CLIMB (multicenter), and CONTACT (multicenter)] conducted in the US and France. Videos were edited to a representative clip of <1 minute. An international panel of 13 endosonographers (US=4, Europe=4, Asia=5) with experience in nCLE (median case volume of 50) and blinded to the histopathological diagnosis, participated in the study. Observers reviewed the same set of nCLE videos in different sequences between two phases separated by a 2-week washout period. Observers were first queried on PCL classification (mucinous vs. non-mucinous) and then on specific PCL diagnosis. Solid pseudopapillary neoplasm (SPN) and cystic neuroendocrine tumors (NET) have similar nCLE imaging features and were grouped together. Fleiss’s kappa (k) for IOA and Cohen’s k for IOR were interpreted using Landis and Koch interpretation.
Results: Observers reviewed 76 unique nCLE videos in subjects with confirmed histopathology (IPMN=37; mucinous cystic neoplasm (MCN)=14; serous cystadenoma (SCA)=9; pseudocyst=3; NET/SPN=13), most of which were rated as good quality (65%). The IOA (k=0.82) and IOR (k=0.82) were almost perfect to differentiate mucinous vs. non-mucinous PCLs (Table 1). For the diagnosis of individual cyst types, IOA was greatest for SCA (almost perfect; k=0.85), followed by IPMN (substantial, k=0.72), and NET/SPN (substantial, k=0.73). The IOA was moderate for MCN (k=0.47), and pseudocyst (k=0.57).
Compared to histopathology, observers differentiated mucinous vs. non-mucinous PCLs with high diagnostic indices (sensitivity (SN)=95%, specificity (SP)=94%, and accuracy (Acc)=95%). Among non-mucinous PCLs, diagnostic indices were higher for SCA (SN=95%, SP=99%, Acc=98%) compared to NET/SPN (SN=81%, SP=99%, Acc=96%), and pseudocyst (SN=87%, SP=96%, Acc=96%).
The interpretation of nCLE images represented by diagnostic accuracy of differentiating mucinous from non-mucinous PCLs (correlation coefficient R=0.152, p=0.62) was not dependent on individual EUS-nCLE case volume (Figure 1).
Conclusion: Compared to histopathology, independent blinded observers achieved high diagnostic accuracy, and excellent agreement and reliability for EUS-nCLE imaging guided diagnosis and differentiation of PCLs.