Confusion generated by the validation of the application of an enzyme linked immunosorbent assay and a polymerase chain reaction method for the detection of Aeromonas salmonicida in field samples

Abstract Two protocols designed to detect Aeromonas salmonicida , one based on an enzyme linked immunosorbent assay (ELISA) and one on a polymerase chain reaction (PCR), had previously performed well in extensive laboratory-based validation studies. These proxy methods were used together with direct culture to examine 223 kidneys of wild salmonid fish and 35 sediment samples collected from their habitat. Direct culture failed to detect the organism in any of the kidney samples. ELISA generated positive results from 23.3% of the kidneys and PCR from 20.6% but the degree of agreement between the two proxy methods was not greater than would have been predicted by chance ( p >0.05). A total of 60% of the samples generating a positive response in ELISA were negative by PCR and 56% of PCR-positive samples were negative by ELISA. The overall concordance was, therefore, only 27%. With respect to the sediment samples, ELISA generated positive results from 53% and PCR from 33%. Again, the degree of agreement between the two proxy methods was not significantly greater than chance and the concordance was only 24%. The data generated by the two proxy methods investigated in this work was employed in a comparative validation of their application to the detection of A. salmonicida in these sample matrices. The hypothesis that both the PCR and ELISA protocols used in this work can validly be used in this way is clearly inconsistent with the data obtained. This analysis could not, however, discriminate between the hypotheses that neither of the methods possessed validity for these applications or that one of them possesses considerably greater validity than the other. The extensive laboratory-scale validation studies to which the two proxy methods used in this work have been submitted are discussed. It is argued that these laboratory studies provided no grounds for predicting the lack of validity that was detected when these proxy methods were applied to field samples. The results presented in this work provide strong evidence that laboratory-scale studies are not adequate to validate the application of proxy methods to field samples.