The reproducibility of acute lymphoblastic leukemia phenotype determinations: evaluation of monoclonal antibody and conventional hematopoietic markers

Variability in interpretation of immunohistologic findings in lymphoproliferative disorders by hematopathologists. A comprehensive statistical analysis of interobserver performance

Eight hematopathologists independently reviewed 56 consecutive cases of benign and malignant lymphoproliferative disorders (LPD) to determine: (1) the degree of interobserver agreement on the interpretation of immunologic findings on fresh-frozen sections alone and on that of the immunologic findings in conjunction with corresponding hematoxylin and eosin (H & E)-stained histologic sections; (2) whether prior knowledge of morphologic characteristics influences the interpretation of immunohistologic sections; (3) whether immunologic phenotype could be predicted reliably based solely on study of histologic sections; and (4) the significance of immunologic data as an aid in the interpretation of histologic sections. The study was carried out in three independent review sessions consisting of (1) review of immunohistologic sections only, (2) review of the same immunohistologic sections together with histologic sections, and (3) review of the histologic sections alone. A consensus diagnosis was defined as agreement of five or more pathologists on the final diagnosis and identification of the immunophenotype. When the authors compared the total number of major disagreements in the first review session with those in the second, the accuracy of the determination of immunophenotype in the second session was clearly superior (P less than 0.05). Similarly, the total number of major disagreements in the second review session was significantly lower than that in the third review session (P less than 0.001). When histologic diagnoses in the second session were compared with those in the third session, it became apparent that the immunologic data helped the pathologist to correct major misinterpretations in 14 cases (25%). This study is the first to demonstrate quantitatively that (1) knowledge of morphologic features influences and greatly enhances the accuracy of the interpretation of immunologic findings, (2) the immunophenotype of LPD cannot be predicted based on morphologic findings alone, and (3) immunologic findings improve the accuracy of interpretation of histologic findings in situations in which a diagnosis cannot be made from morphologic features only.