Flow-cytometric detection of minimal residual disease (MRD) has proven in several single-institute studies to have an independent prognostic impact. We studied whether this relatively complex approach could be performed in a multicenter clinical setting. Five centers developed common protocols to accurately define leukemia-associated (immuno)phenotypes (LAPs) at diagnosis required to establish MRD during/after treatment. List mode data files were exchanged, and LAPs were designed by each center. One center, with extensive MRD experience, served as the reference center and coordinator. In quarterly meetings, consensus LAPs were defined, with the performance of centers compared with these. In a learning (29 patients) and a test phase (35 patients), a mean of 2.2 aberrancies/patient was detected, and only 1/63 patients (1.6%) had no consensus LAP(s). For the four centers without (extensive) MRD experience, clear improvement could be shown: in the learning phase, 39-63% of all consensus LAPs were missed, resulting in a median 30% of patients (range 21-33%) for whom no consensus LAP was reported; in the test phase, 27-40% missed consensus LAPs, resulting in a median 16% (range 7-18%) of 'missed' patients. The quality of LAPs was extensively described. Immunophenotypic MRD assessment in its current setting needs extensive experience and should be limited to experienced centers.