In a previous criterion-based pathology report review of 717 cases of non-Hodgkin's lymphoma in an Australian population-based epidemiologic study, a WHO category could be assigned in 91% of cases, but confidence in this classification was high in only 57.5%. Given this lack of confidence, a pathology review was done in a subset of 315 cases, with the aims of assigning a WHO classification category and the corresponding International Classification of Diseases for Oncology, Third Edition code in all cases previously unclassified or classified with low confidence and testing the accuracy of report review in assigning a confident WHO classification. After pathology review, 10 cases were ineligible (not non-Hodgkin's lymphoma, 3.2%) and 99% (301 of 305) of the remainder were assigned a WHO classification, with high confidence in 87% (261 of 301). There was 78% overall agreement between the WHO classification assigned by report review and pathology review, with 92% agreement when there was high confidence in the report review classification and 69% agreement when there was low confidence. Eighteen percent of follicular lymphomas and 23% of diffuse large B-cell lymphomas were reclassified. The pathology review increased the accuracy of WHO classification by an estimated 12.5% in the 694 cases who were still eligible in the study. Although a potential error rate of 7.5% remained, reviewing more cases, or not reviewing any cases classified with high confidence, would have produced only a small change in accuracy. Criterion-based pathology report review of all cases followed by selective pathology review in cases classified with low confidence is recommended as a cost-saving and accurate strategy for pathology review in large epidemiologic studies.