Historically, two major strata of ductal carcinoma in situ (DCIS) have
been linked to outcome, the presence or absence of comedo type and si
ze. Our initial approach in classification was dichotomous, often favo
ring the comedo type with most worrisome implications to foster agreem
ent in diagnosis. We have now tested guidelines that foster agreement
in the modified Lagios three-tiered system. Sixteen cases of DCIS were
selected, reflecting: a spectrum of histological subtypes, with speci
fic inclusion of cases in which consensus in classification using a di
chotomous (comedo/noncomedo) scheme would be difficult. Six independen
t observers reviewed a minimum of five color 35-mm slides from each ca
se at two separate occasions. The aim was to subclassify each case bas
ed on architectural pattern, nuclear grade, and presence or absence of
tumor necrosis (Modified Lagios Classification, Lagios et al, Cancer
1989). After initial review, emphasizing placement of each case into a
high-or low-grade category, there was disagreement in seven cases (44
%), confirming our aim to choose cases with uncertain cues for classif
ication. Agreement was achieved in 94% of cases by allowing re-review
with emphasis on inclusion of an intermediate-grade category. Our stud
y also suggests that pure micropapillary DCIS and apocrine DCIS warran
t independent classification as ''special type'' DCIS. Our small pilot
study sugests that, with adherence to specific criteria, most DCIS ca
ses can be easily and consistently classified into the following five
categories: (1) high grade, (2) intermediate grade, (3) low grade, (4)
pure or predominantly micropapillary, and (5) pure apocrine. Our six
observers independently reached a final concordance of 94% despite sel
ection of cases in which consensus in a dichotomous classification was
difficult. This was achieved predominantly by accepting an intermedia
te category of DCIS with intermediate nuclear features and limited nec
rosis. Confirmation of the applicability of the Modified Lagios Classi
fication awaits completion of a much larger multi-institutional study
in which statistical significance and interobserver variation can be b
etter defined. (C) 1997 by W.B. Saunders Company.