Sensitivity for the Detection of Leukemic Cells

Dennis W. Ross, Hematology Laboratories, University of North Carolina, Chapel Hill, NC 27514 (USA) In a recent article, Cranendonk et al. [1] reach a conclusion that is at odds with the clinical experience and experimental results of many other investigators [2–5]. They state that automated flow cytochemistry (Hemalog-D; Technicon, Tarrytown, N.Y.) is not reliable for the detection of even large amounts of lym-phoblasts. This conclusion is not warranted by their own data. They formulated the following hypothesis, ‘Is it possible to recognize a leukemic relapse earlier with the aid of the HD (Hemalog) than with MDC (manual differential count)?’ In response, the authors give the number of blood counts with and without an elevated % LUC (large unstained cell) correlated with the presence or absence of lymphoblasts on the manual differential. Cranendonk et al. [1] assume that the recognition of lymphoblasts on a manual smear is an infallible test result or at least a gold standard for the recognition of early relapse of ALL. Under this assumption, from their data the sensitivity is 64% and the specificity is 82% for increased % LUC correlating with lymphoblasts [6]. To demonstrate the importance of the assumption that the manual differential represents truth let us invert the logical problem and consider % LUC as the gold standard and the presence of lymphoblasts to be a test under investigation. From the same data we calculate that the presence of lymphoblasts gives only a 19% sensitivity, but a high 97% specificity for correlating with increased % LUC. New laboratory tests are too often looked at only in terms of comparison to the older tests that they may potentially replace. The fallacy of using this type of comparison is that one must a priori decide that one test represents truth. Given this assumption, the only possible outcome is that the new test is only as good as (assuming perfect correlation), or otherwise, is inferior to the old test. To correctly compare the performance of two test strategies in diagnosing a disease, one must compare the test results directly to the presence or absence of the disease. For the monitoring of patients with leukemia, nobody advocates the complete abandonment of the blood smear. Instead, a synthesis of automated flow cytochemistry (a method which is highly sensitive for the detection of potential leukemic cells), and expert morphological review of positive specimens (a method highly specific for confirming the presence of leukemic cells) offers the optimum use of these data. This later conclusion is the experience of many published reports of users of automated flow cytochemistry and other forms of flow cytometry. References Cranendonk, E.; Abeling, N.G.G.M.; Bakker, A. Jong M.E. de; Gennip, A.H. van; Behrendt, H.: Evaluation of the use of the Hemalog D in acute lymphoblastic leukaemia and disseminated nonHodgkin’s lymphoma in children. Acta haemat. 71: 18–24(1984). Ross, D.W.; Bardwell, A.: Automated cytochemistry and the white cell differential in leukemia. Blood Cells 6: 455–470(1980).