Poster Presentation The Pancreas Summit 2025

Evaluating reliability of structured computed tomography scan reporting in pancreatic cancer staging: Insights from a quality assurance study exploring inter- and intra-observer agreement and variability. (#7)

Faizah Alam 1 , Lin Li 2 3 , Jessica Kasza 3 , Samantha Ellis 4 , Mark Goodwin 5 , Bruno di Muzio 4 , Jan F Gerstenmaier 4 , Hyun S Ko 6 , Kelvin Lim 5 , Lianne Ionna 1 2 , John Zalcberg 2 3 4 , Charles HC Pilgrim 2 3 4
  1. Monash Health, Melbourne, VIC, Australia
  2. UGICR , Monash University, Melbourne, VIC , Australia
  3. Monash University, Melbourne, VIC, Australia
  4. Alfred Health, Melbourne, VIC
  5. Austin Health, Melbourne, VIC
  6. Peter MacCallum Cancer Centre, Melbourne, VIC

Introduction: Pancreatic cancer remains a significant cause of cancer-related mortality, with surgical resection the major curative option. Despite the pivotal role of Computed Tomography (CT) in initial staging, there is notable inconsistency in resectability classification. In response, the 2017 International Consensus definition for borderline resectable pancreatic ductal adenocarcinoma (PDAC) led to a pilot study exploring feasibility of structured CT synoptic reporting templates. Building upon this, the presented project evaluates inter-observer and intra-observer agreement in relation to the standardised synoptic report, shown to improve consistency and likely reduce variability to thereby enhance treatment decisions. 

Methods: Sixty dedicated pancreatic protocol CT scans were independently reported by six radiologists using the standardised synoptic report. Variability was indicated by percentage agreement between anatomical findings and final resectability status. Gwet’s Agreement Coefficients (AC) were calculated to determine inter-observer agreement and Cohen’s Kappa for intra-observer agreement. 

Results: A total of 22320 scan parameters were collected, with six radiologists individually rating 60 scans. Twenty-six scans had previously been rated in the pilot study by four radiologists and were blindly re-rated. Full inter-observer agreement of anatomical resectability classification was seen in 21.67% (13/60) scans, with 78.33% (47/60) showing majority inter-observer agreement. The intra-observer agreement of anatomical resectability classification was 65.38% (17/26). The highest level (very good) of agreement was seen regarding biliary stent presence (Gwet’s AC, 95% CI) (0.96 (0.93, 1.00)) and the least (poor) significant agreement was seen describing jejunal vein involvement (0.17 (0.00, 0.35)). Moderate inter-observer and intra-observer agreement for overall resectability status was demonstrated with Gwet’s AC 0.40 (95% CI 0.29, 0.51) and Cohen’s Kappa 0.45 (95% CI 0.20, 0.69). 

Discussion: This study highlights synoptic reporting variability pertaining to PDAC resectability status, with moderate inter- and intra-observer agreement elucidating inherent complexities, particularly regarding jejunal vein assessment. There were 358 synoptic reports (N=360) with 100% completeness, demonstrating the improvement of synoptic over free-text reporting. Future efforts to increase the rate of synoptic reporting within radiology practice and its inclusion in multidisciplinary team discussions, should be made to enhance the reliability and comprehensiveness of pre-operative CT staging, optimising pancreatic cancer management and patient care.