Variability of polymerase chain reaction detection of the bcl-2-IgH translocation in an international multicentre study

Background: The capacity of the polymerase chain reaction (PCR) to detect very low numbers of cells bearing a t(14;18) translocation has led to its application in assessment of the results of treatment for follicular lymphoma, and suggestions that therapy might be guided by molecular studies. To test the reliability of PCR a collaborative study was undertaken to compare results from different laboratories in Europe and North America. Methods: Twenty laboratories with records of publication in molecular diagnostics were sent blood from normal donors with varying numbers of t(14;18)-bearing cells added from a cell line with a translocation in the major breakpoint region (MBR) of the bcl-2 gene. Samples contained 1000, 100, 10, 1 or 0 cells per ml of whole blood and were sent blinded in duplicate. PCR methodology varied widely, with the total number of amplification cycles between 30 and 70, and 13 different primers used for the MBR region. Twelve laboratories used nested PCR and eight single round amplification. Results: The sensitivity of nested and single round PCR was similar at 100 cells/ml but below this the nested method proved significantly more sensitive. The false positive rate was 28%, with 11 samples from 9 laboratories reported as positive when no t(14;18) cells were added. PCR product size and sequence analysis showed that false positives were due to contamination from cell-line DNA rather than background translocations in the donors. There was no significant difference in false positive rates between nested and single round techniques. Conclusion: The polymerase chain reaction to detectbcl-2-IgH rearrangements is presently carried out with widely disparate results. Further effort is required to bring forward a standard PCR protocol which can be re-tested in different laboratories to improve accuracy and reproducibility. The application of quantitative techniques such as real-time PCR may resolve many of the problems presently encountered.