individuals. This was performed in a number of studies for flow cytometry (Kern et al.; 6 San Miguel et al. Instead, we look at bone marrows that are regenerating after heavy chemotherapy. The expression of presumed leukemiaspecific genes or the frequency of presumed leukemia specific phenotypes can be very different in such samples compared to steady-state healthy bone marrow.For instance, the gene CSPG4 was identified as a possible MRD marker. 4 It was highly expressed in AML samples but not in healthy bone marrow. However, we found that it was useless as an MRD marker because expression of CSPG4 was also high in bone marrow that was free of leukemia but was regenerating after chemotherapy. 4 Therefore, to determine the real sensitivity and specificity of an MRD analysis, it is mandatory to analyze a large number of leukemia-free, regenerating bone marrows.A second problem of analyzing serial dilutions of leukemic cells in samples of healthy bone marrow is that this does not really simulate the situation of minimal residual disease. The important clinical question is the amount of leukemic stem cells that is still present in the patient. If the leukemic phenotype that is used for flow cytometry is present on the majority of leukemic cells but not on leukemic stem cells, the sensitivity could be lower than anticipated. If the expression of a leukemia-associated gene, like WT1, is particularly high in leukemic stem cells, the sensitivity might be much better than anticipated and vice versa.
ConclusionsMonitoring MRD has become a strong diagnostic tool in acute leukemia. It is widely used for clinical decision-making in acute lymphoblastic leukemia, chronic myeloid leukemia and acute promyelocytic leukemia. Various methods have been developed to monitor MRD in AML and we should proceed to put them into clinical practice.Sensitivity is an important issue in the comparison of these methods. When the term is used, it should always be clear whether we are talking about the LPT or the real sensitivity and specificity of our diagnostic test.The LPT is easy to determine and for most methods of monitoring MRD, is in the range of 10 À4 -10 À6 . These figures have a very limited meaning because for clinical decisionmaking LPT cannot be used as the cutoff for high versus low MRD. At such a low threshold, the specificity of monitoring MRD (likelihood of a negative test result in a truly negative patient) is too low due to false positive healthy and regenerating bone marrows.If we use, for instance, a threshold of 10 À3 for clinical decision-making, it does not matter whether the LPT is 10 À4 or 10 À6 . What really matters is the sensitivity and specificity of the diagnostic test at the threshold of 10 À3 . We should put less effort into determining LPTs and more effort into determining sensitivity and specificity of our methods at the clinically relevant thresholds.Because of the many issues that influence the quality of each method of monitoring MRD, the best way of really comparing two methods is to simultaneously apply ...