Comparative genomic hybridization, a widely used method for screening for g
enomic imbalances, suffers from a lack of standardized evaluation. In order
to compare a recently proposed data-driven procedure with the commonly use
d fixed cutoff values, we tested 257 events by both procedures as well as b
y fluorescence in situ hybridization using selected probes. With the data-d
riven procedure, a much higher fraction (42/218 vs. 8/218) of false positiv
e results was obtained, whereas a higher sensitivity with respect to the de
tection of imbalances (30/39 vs. 19/39) was reached. Based on the significa
ntly higher positive likelihood ratios and the positive predictive value, w
e strongly recommend the use of fixed diagnostic thresholds, because the al
ternative procedure generates an unacceptably high portion of incorrectly s
cored chromosomal imbalances. (C) 2000 Wiley-Liss, Inc.