73 Sense and nonsense of yT-staging on MRI after chemoradiotherapy in rectal cancer 5 gists, and ten general radiologists with no specific track record in rectal MRI). Readers were asked to review the restaging MRIs (alongside the primary staging MRIs) of the 90 study cases and define the ymrT-stage as ymrT0, ymrT1-2, ymrT3ab, ymrT3cd, ymrT4a or ymrT4b, in line with the structured reporting template for restaging published by the European Society of Gastrointestinal and Abdominal Radiology (ESGAR) [2]. Readers were also instructed to estimate the degree of tumour regression by assigning an mrTRG score (mrTRG1 = no/minimal fibrosis, mrTRG2 = dense fibrotic scar without macroscopic tumour signal, mrTRG3 = fibrosis predominates but with obvious areas of tumour signal, mrTRG 4 = tumour signal predominates with little/minimal fibrosis, mrTRG 5 = tumour signal only, including cases with tumour progression) [16]. Readers were blinded to each other’s results and the final treatment and response outcomes. Statistical analysis and standard of reference Statistical analyses were performed using R statistics version 4.1.0 (2021) and IBM SPSS version 27 (2020). Group interobserver agreement (IOA) was calculated using Krippendorff’s alpha considering the ordinal nature of the staging outcomes and bootstrapping was used to calculate the 95% confidence interval (CI). Concordance between ymrT-stage and ypT-stage was calculated as the percentage of correct diagnoses (ymrT = ypT), understaging (ymr < ypT) and overstaging (ymrT > ypT). Results were compared between patients with predominant fibrosis at histopathology (pTRG1-3) and patients with predominant tumour (pTRG4-5), using Mandard’s tumour regression grade as a standard of reference [17]. In patients included in a W&W program, a recurrence free follow-up interval of >2 years was used as a surrogate endpoint to establish a complete response (TRG1). Results were analysed separately for the expert and non-experts readers. The originally assigned ymrT-stages were grouped as ymrT0, ymrT1, ymrT2, ymrT3 (= ymrT3ab and ymrT3cd), and ymrT4 (=ymrT4a and ymrT4b) to enable one-on-one correlation with histopathology where yT-substages were not consistently available. Associations between concordance, under- and overstaging rates and the different pTRG and expert/ non-expert subgroups were tested usinga multilevel logistic regression model with a random effect to model the correlation between observations within a same patient. To do so, 3 binary dummy variables were created to state if an observation was overstaged (1 if overstaged, 0 if not), understaged (1 if understaged, 0 if not) or correctly staged (1 if correct, 0 if not). These variables were then used as the outcome of 3 different logistic regressions where the experience level of the reader and pTRG subgroups were used as covariates. To study overstaging, patients with ypT4 stage were removed as they cannot be overstaged; to study understaging, ypT0 patients were removed. A significance threshold of p<0.05 was used throughout the analyses.
RkJQdWJsaXNoZXIy MjY0ODMw