Histopathological assessment of ductal carcinoma in situ (DCIS), a non-obligate precursor of invasive breast cancer, is characterized by considerable inter-observer variability. Previously, post hoc dichotomization of multi-categorical variables was used to determine the 'ideal' cut-offs for dichotomous assessment. The present international multi-center study evaluated inter-observer variability among 39 pathologists who performed upfront dichotomous evaluation of 149 consecutive DCIS.All pathologists independently assessed nuclear atypia, necrosis, solid DCIS architecture, calcifications, stromal architecture and lobular cancerization in one digital slide per lesion. Stromal inflammation was assessed semi-quantitatively. Tumor-infiltrating lymphocytes (TILs) were quantified as percentages and dichotomously assessed with a cut-off at 50%. Krippendorff's alpha (KA), Cohen's kappa and intraclass correlation coefficient were calculated for the appropriate variables.Lobular cancerization (KA = 0,396), nuclear atypia (KA = 0,422) and stromal architecture (KA = 0,450) showed the highest inter-observer variability. Stromal inflammation (KA = 0,564), dichotomously assessed TILs (KA = 0,520) and comedonecrosis (KA = 0,539) showed slightly lower inter-observer disagreement. Solid DCIS architecture (KA = 0,602) and calcifications (KA = 0,676) presented with the lowest inter-observer variability. Semi-quantitative assessment of stromal inflammation resulted in a slightly higher inter-observer concordance than upfront dichotomous TILs assessment (KA = 0,564 versus KA = 0,520). High stromal inflammation corresponded best with dichotomously assessed TILs when the cut-off was set at 10% (kappa = 0,881). Nevertheless, a post hoc TILs cut-off set at 20% resulted in the highest inter-observer agreement (KA = 0,669).Despite upfront dichotomous evaluation, the inter-observer variability remains considerable and is at most acceptable, although it varies among the different histopathological features. Future studies should investigate its impact on DCIS prognostication. Forthcoming machine learning algorithms may be useful to tackle this substantial diagnostic challenge.