Daniel R Ludwig1, Erin K Romberg2, Tyler J Fraum1, Eric Rohe3, Kathryn J Fowler4, Geetika Khanna5. 1. Mallinckrodt Institute of Radiology, Washington University School of Medicine, 510 S. Kingshighway, Box 8131, Saint Louis, MO, 63110, USA. 2. Department of Radiology, University of Washington School of Medicine, Seattle, WA, USA. 3. Saint Louis University School of Medicine, Saint Louis, MO, USA. 4. Department of Radiology, University of California, San Diego School of Medicine, San Diego, CA, USA. 5. Mallinckrodt Institute of Radiology, Washington University School of Medicine, 510 S. Kingshighway, Box 8131, Saint Louis, MO, 63110, USA. khannag@wustl.edu.
Abstract
BACKGROUND: Liver Imaging Reporting and Data System (LI-RADS) has standardized the evaluation of hepatic lesions in adults at risk of developing hepatocellular carcinoma (HCC). There is no accepted imaging algorithm for diagnosing HCC in the pediatric population. OBJECTIVE: The aim of our study was to evaluate the diagnostic accuracy and inter-rater reliability of LI-RADS version 2017 (v2017) for diagnosing HCC in a pediatric cohort. MATERIALS AND METHODS: This retrospective, Institutional Review Board-approved study involved review of all abdominal dynamic contrast-enhanced imaging at a tertiary children's hospital during a 10-year period, yielding 151 liver lesions in patients <18 years. Cases with active extrahepatic malignancy or an inadequate reference standard were excluded. Two readers independently evaluated all included hepatic lesions using LI-RADS criteria. Pathology and imaging follow-up were used as reference standards. RESULTS: A total of 41 lesions in 41 patients met criteria for evaluation (3 HCCs, 8 non-HCC malignancies, 30 benign lesions). A LI-RADS designation of definite HCC had high sensitivity (Reader 1/Reader 2: 100%, 95% confidence interval [CI] 31-100%) and high specificity (Reader 1: 84%, 95% CI: 68-93%; Reader 2: 97%, 95% CI: 85-100%) for predicting HCC. However, positive predictive value was only 33% (95% CI: 9-69%) and 75% (95% CI: 22-99%) for Reader 1 and Reader 2, respectively. For predicting any type of hepatic malignancy, a LI-RADS designation of definitely or likely malignant (i.e. not necessarily HCC) had a sensitivity of 100% (95% CI: 74-100%) and 90% (95% CI: 61-100%) for Reader 1 and Reader 2, respectively, and a negative predictive value (NPV) of 100% (95% CI: 81-100%) and 96% (95% CI: 83-99%) for Reader 1 and Reader 2, respectively. Interobserver agreement was substantial for the overall LI-RADS category (weighted κ=0.62; 95% CI: 0.38-0.86). CONCLUSION: The positive predictive value of LI-RADS v2017 for diagnosing HCC was limited by the low frequency of HCC among pediatric patients. However, a LI-RADS designation of definitely or likely malignant had high sensitivity and NPV for any type of hepatic malignancy and may serve to direct clinical management by selecting patients for tissue sampling.
BACKGROUND: Liver Imaging Reporting and Data System (LI-RADS) has standardized the evaluation of hepatic lesions in adults at risk of developing hepatocellular carcinoma (HCC). There is no accepted imaging algorithm for diagnosing HCC in the pediatric population. OBJECTIVE: The aim of our study was to evaluate the diagnostic accuracy and inter-rater reliability of LI-RADS version 2017 (v2017) for diagnosing HCC in a pediatric cohort. MATERIALS AND METHODS: This retrospective, Institutional Review Board-approved study involved review of all abdominal dynamic contrast-enhanced imaging at a tertiary children's hospital during a 10-year period, yielding 151 liver lesions in patients <18 years. Cases with active extrahepatic malignancy or an inadequate reference standard were excluded. Two readers independently evaluated all included hepatic lesions using LI-RADS criteria. Pathology and imaging follow-up were used as reference standards. RESULTS: A total of 41 lesions in 41 patients met criteria for evaluation (3 HCCs, 8 non-HCC malignancies, 30 benign lesions). A LI-RADS designation of definite HCC had high sensitivity (Reader 1/Reader 2: 100%, 95% confidence interval [CI] 31-100%) and high specificity (Reader 1: 84%, 95% CI: 68-93%; Reader 2: 97%, 95% CI: 85-100%) for predicting HCC. However, positive predictive value was only 33% (95% CI: 9-69%) and 75% (95% CI: 22-99%) for Reader 1 and Reader 2, respectively. For predicting any type of hepatic malignancy, a LI-RADS designation of definitely or likely malignant (i.e. not necessarily HCC) had a sensitivity of 100% (95% CI: 74-100%) and 90% (95% CI: 61-100%) for Reader 1 and Reader 2, respectively, and a negative predictive value (NPV) of 100% (95% CI: 81-100%) and 96% (95% CI: 83-99%) for Reader 1 and Reader 2, respectively. Interobserver agreement was substantial for the overall LI-RADS category (weighted κ=0.62; 95% CI: 0.38-0.86). CONCLUSION: The positive predictive value of LI-RADS v2017 for diagnosing HCC was limited by the low frequency of HCC among pediatric patients. However, a LI-RADS designation of definitely or likely malignant had high sensitivity and NPV for any type of hepatic malignancy and may serve to direct clinical management by selecting patients for tissue sampling.
Entities:
Keywords:
Children; Computed tomography; Diagnostic accuracy; Hepatocellular carcinoma (HCC); Inter-rater reliability; Liver; Liver Imaging Reporting and Data System (LI-RADS); Magnetic resonance imaging
Authors: Giuseppe Brancatelli; Michael P Federle; Luigi Grazioli; Rita Golfieri; Riccardo Lencioni Journal: Radiographics Date: 2002 Jul-Aug Impact factor: 5.333
Authors: Santhosh Gaddikeri; Michael F McNeeley; Carolyn L Wang; Puneet Bhargava; Manjiri K Dighe; Matthew M C Yeh; Theodore Jay Dubinsky; Orpheus Kolokythas; Neeraj Lalwani Journal: AJR Am J Roentgenol Date: 2014-07 Impact factor: 3.959
Authors: Christoph Wald; Mark W Russo; Julie K Heimbach; Hero K Hussain; Elizabeth A Pomfret; Jordi Bruix Journal: Radiology Date: 2013-02 Impact factor: 11.105
Authors: Lewis R Roberts; Claude B Sirlin; Feras Zaiem; Jehad Almasri; Larry J Prokop; Julie K Heimbach; M Hassan Murad; Khaled Mohammed Journal: Hepatology Date: 2017-11-29 Impact factor: 17.425
Authors: An Tang; Mustafa R Bashir; Michael T Corwin; Irene Cruite; Christoph F Dietrich; Richard K G Do; Eric C Ehman; Kathryn J Fowler; Hero K Hussain; Reena C Jha; Adib R Karam; Adrija Mamidipalli; Robert M Marks; Donald G Mitchell; Tara A Morgan; Michael A Ohliger; Amol Shah; Kim-Nhien Vu; Claude B Sirlin Journal: Radiology Date: 2017-11-21 Impact factor: 11.105