A sufficient sensitivity of PCR is a prerequisite for its use in the diagnosis of infectious diseases. We have used PCR for detecting gene elements of Borrelia burgdorferi, mycobacteria and Bordetella pertussis. With all these microbe groups, difficulties were encountered in achieving the demanded sensitivity with the primer pairs primarily selected. An extensive testing of various reaction parameters did not improve the sensitivity. Subsequently, we synthesized more primers derived from slightly different positions of the original target sequences. When the original and new primers were tested in possible combinations, some primer pairs reached 100-fold to 1000-fold higher sensitivity than the primary pairs. We conclude that in optimizing the sensitivity of PCR, more emphasis should be put on testing of several primer pairs than on the extensive screening of reaction parameters. Thus far, a trial-and-error approach has to be used, because there is no means to predict the sensitivity properties of a selected primer pair.