Comparison of Five Pathological Tumor Regression Grading Systems for Rectal Cancer Following Chemoradiation: Correlation Coefficient and Intra-Rater Reliability.

Abstract

Objective: To determine the correlation among five different types of tumor regression grading (TRG) systems. Test-retest reliability analyses were conducted at two time points to assess the internal validity and consistency of these five TRG systems.

Methods: A test-retest study was performed in 34 pathologically confirmed rectal adenocarcinoma specimens. All patients underwent pre-operative CRT followed by total mesorectal resection. Each specimen was examined twice to examine the variability of test-retest measurements. Every specimen was examined according to the 5 different TRG systems (Dworak, Mandard, Ryan, AJCC, modified Ryan). The time interval between the initial assessment and the repeat assessment was 3 weeks by the same pathologist who was not allowed to know the results of his initial measurements.

Result: For TRG systems comparing therapy-induced fibrosis in relation to residual tumor, a very strong correlation among them was found, with correlation coefficient values ranging from 0.964 to 1. The modified Ryan TRG system determines the degree of tumor regression based solely on the quantity of residual viable cancer cells only (not fibrosis). The system had lower correlation coefficient values, ranging from 0.549 to 0.617. The present study revealed an excellent intra-rater correlation coefficient of 0.947 (95% CI: 0.895-0.974) for the Mandard and Dworak TRG systems, 0.918 (95% CI: 835-0.959) for the Ryan TRG system, 0.957 (95% CI: 0.913-0.978) for the AJCC TRG system, and 0.934 (95% CI: 0.867-0.967) for the modified Ryan TRG system.

Conclusion: TRG systems with different scales categorizing tumor regression based on residual tumor and fibrosis revealed a strong to very strong correlation among them. The modified Ryan system, which categorizes tumor regression based solely on the quantity of residual viable cancer cells (not fibrosis), resulted in discrepancies in interpretations and lower correlation values. The present study revealed an excellent intra-rater correlation coefficient with high internal validity.