| Summary: | BACKGROUND: Histological grade is an independent prognostic variable in breast cancer (BC). Previous concordance studies of BC grade have reported moderate levels of agreement; a typical finding in morphological assessment of biological variables. This study aims at investigating the impact of discordance on the prognostic value of grade and identifying the best reporting approach in borderline cases.
METHODS: A large (n=1675) well-characterised annotated cohort of BC originally graded in routine practice using glass slides was re-graded twice, by an expert breast pathologist using virtual microscopy with a three months washout period. Outcome was assessed using breast cancer specific and distant metastasis free survival (median follow-up =135 months).
RESULTS: 58% of the cases showed absolute agreement in the three separate grading sessions whereas grade 1/2 and grade 2/3 discordance were observed in 21% and 21% respectively. Absolute intra-observer agreement using virtual microscopy was observed in 77% of the cases whereas 13% and 10% showed grade 1/2 and grade 2/3 discordance respectively. Despite the concordance, outcome analysis revealed significant associations between tumour grade and patients’ outcome in the three grading sessions. Grade 1/2 and grade 2/3 discordant cases showed intermediate survival between grade 1 and grade 2 tumours and grade 2 and grade 3 tumours, respectively. Grade 1/2 discordant cases showed a worse outcome when compared with grade 1 tumours (p=0.008) but no statistical difference was identified when compared with grade 2 tumours. Similarly, grade 2/3 discordant cases showed a significant difference from grade 2 tumours (p<0.001) but no statistical difference was identified when compared with grade 3 tumours.
CONCLUSIONS: BC grade discordance is likely a reflection of biologically, and hence morphologically, borderline tumours. Cases with borderline features for grade are more likely to behave similar to the higher grade category. Repeating histological grade of borderline cases or double reporting may improve correlation with outcome
|
|---|