Histologic grading of breast cancer: Linkage of patient outcome with level of pathologist agreement

In the histologic grading of invasive breast cancer with the Nottingham modification of the Scarff-Bloom-Richardson grading scheme (NSBR), it has been found that when pathologists disagree, they tend not to disagree by much. However, if tumor grade is to be used as an important parameter in making treatment decisions, then even this generally small degree of pathologist variability in assessing grade needs to be correlated with patient outcome. Findings from the Nottingham/Tenovus Primary Breast Cancer Study were used for patient outcome data. Kaplan-Meier survival curves were constructed for NSBR scores grouped according to the level at which pathologists tend to agree in assessing grade, from a reproducibility perspective. For example, if a given tumor were assessed by several pathologists as having either an NSBR score of 5 or 6, then what is the correct score-the intermediate-grade Score 6 assessments or the low-grade Score 5 assessments? By "regrouping" the Nottingham outcome data such that data from patients with Score 5 tumors are grouped with patients having Score 6 tumors (a 5-6 group), then the level in which the pathologists agreed with each other (that the tumor was either score 5 or 6) is better matched with patient outcome. In response to the above example, it was not surprising to find that patients with Score 5-6 tumors had a probability of survival between the established low and intermediate NSBR final combined grades. However, it is the discussion of this approach that highlights that optimal use of grading requires awareness of the level of pathologist agreement and understanding the value of pathologists' reaching consensus in assessments. Also, knowledge of possible clinical decision thresholds can help in providing relevant interpretations of grading results.