Reverse transcription (RT)-polymerase chain reaction (PCR) raises unique methodological matters that may hamper the reliability of the procedure, especially when results should direct therapeutic decisions. One of these matters is represented by the RT step. The present study shows that differences in complementary DNA (cDNA) preparations purposely containing increasing amounts of retrotranscribed RNA were not disclosed by nonquantitative RT-PCR by two different housekeeping genes, leading to fictitious results when the expression of a given gene was quantitatively assessed. To overcome this problem, the following are proposed: 1) to evaluate the efficiency of RT step through the quantification, by competitive RT-PCR, of the expression levels of the housekeeping gene beta2-microglobulin (beta2M); 2) to normalize each cDNA preparation to be comprised within 1 standard deviation of the mean value of beta2M absolute level (3.14 +/- 1.14 attomoles/microg RNA) found by analyzing 33 cell lines of hematopoietic origin. To validate this strategy in a clinical setting, serial cDNA samples from patients were checked by conventional and quantitative RT-PCR for beta2M. Again, only a quantitative evaluation of beta2M levels was allowed to unveil significant differences, otherwise undetected, in the efficiency of RT reactions among these cDNA samples. Normalization of samples to obtain cDNA preparations containing comparable beta2M levels, eventually led to an increased sensitivity in the detection of PML-RARalpha fusion transcripts. This approach seems of great value for the monitoring of minimal residual disease in serial patient samples when a tumor-specific marker is available.