Clinical predictors of significant findings on EUS for the evaluation of incidental common bile duct dilation.

BACKGROUND AND AIMS: While endoscopic ultrasound(EUS) is highly accurate for the evaluation of common bile duct (CBD) dilation, the yield of EUS in patients with incidental CBD dilation is unclear. METHODS: Serial patients undergoing EUS for incidental, dilated CBD (per radiologist, minimum >6mm objectively) from two academic medical centers, without active pancreaticobiliary disease or significantly elevated LFTs were evaluated. Multivariable logistic regression identified predictors of EUS with significant findings and a novel prediction model was derived from one center, internally validated with bootstrapping, and externally validated at the second center. RESULTS: Of 375 patients evaluated, 31 (8.3%) had significant findings including 26 choledocholithiasis, 1 ampullary adenoma, and 1 pancreatic mass. Predictors of significant findings with EUS included: age ≥70(OR 3.7, 95CI 1.5-10.0), non-biliary-type abdominal pain without chronic pain(OR 6.1, 95CI 2.3-17.3), CBD diameter ≥15 mm or ≥17mm with cholecystectomy(OR 6.9, 95CI 2.7-18.7), and prior ERCP(OR 6.8, 95CI 2.1-22.5). A point-based novel clinical prediction model was created: age ≥70=1, non-biliary-type abdominal pain without chronic pain=2, prior ERCP=2, CBD dilation=2. A score <1 had 93% (development) and 100% (validation) sensitivity and predicted a <2% chance of having a significant finding in both cohorts while excluding the need for EUS in ∼30% of both cohorts. Conversely, a score >4 was >90% specific for the presence of significant pathology. CONCLUSION: Less than 10% of patients undergoing EUS for incidental CBD dilation had pathologic findings. This novel, externally validated, clinical prediction model may reduce low-yield, invasive evaluation in nearly a third of patients.

Identification of human short introns.

Canonical pre-mRNA splicing requires snRNPs and associated splicing factors to excise conserved intronic sequences, with a minimum intron length required for efficient splicing. Non-canonical splicing-intron excision without the spliceosome-has been documented; most notably, some tRNAs and the XBP1 mRNA contain short introns that are not removed by the spliceosome. There have been some efforts to identify additional short introns, but little is known about how many short introns are processed from mRNAs. Here, we report an approach to identify RNA short introns from RNA-Seq data, discriminating against small genomic deletions. We identify hundreds of short introns conserved among multiple human cell lines. These short introns are often alternatively spliced and are found in a variety of RNAs-both mRNAs and lncRNAs. Short intron splicing efficiency is increased by secondary structure, and we detect both canonical and non-canonical short introns. In many cases, splicing of these short introns from mRNAs is predicted to alter the reading frame and change protein output. Our findings imply that standard gene prediction models which often assume a lower limit for intron size fail to predict short introns effectively. We conclude that short introns are abundant in the human transcriptome, and short intron splicing represents an added layer to mRNA regulation.