The uniqueness of the
complementarity determining region III (CDRIII) has been utilized successfully in the last decade for development of a patient specific, molecular, polymerase chain reaction (PCR)-based assay for determining
minimal residual disease in various lymphoid
malignancies. There are various approaches for carrying out this test. i) CDRIII primers are used to amplify the corresponding
DNA from the same patient and quantitation of the amplified CDRIII bands is done by generation a standard curve of known amounts of purified patient's
tumor DNA, followed by a linear regression analysis to quantitate the results. ii) CDRIII primers are used to amplify a serially-diluted patient's sample (unknown), with replicate points. According to Poisson equation, replicate points in each dilution can be either all positive, all negative, or 'mixed', negative and positive. The quantitation, according to this approach is done by determination of the dilution point where there are 'mixed' lanes plus the flanking 'all negative' and 'all positive' lanes, assuming that the test can always detect one
tumor cell in 100,000 cells. In this communication we show evidence that the use of the Poisson method can lead to an underestimation of the amount of
tumor cells, due to the great variability in the priming and amplification among the various CDRIII primers. This variation is inherent to the size, C/G ratio, melting point of each primer, etc. In a simulated statistical model we show that the magnitude of error in the Poisson method could reach 1-2 logs. In contrast, using the standard curve for each patient and regression analysis eliminate these problems.