Comparison of different most-probable-number methods for enumeration of Listeria in poultry

To estimate levels of Listeria spp. in poultry and to select the most appropriate enumeration method for routine analysis, 40 naturally contaminated retail chicken carcasses were tested in Ponferrada (León, N.W. Spain) using the direct plate count technique and various most-probable-number (MPN) designs (UVM I [University of Vermont modified Listeria enrichment broth], Fraser enrichment broth, or both were used in 3-, 5-, and 10-tube MPN techniques). MPN estimation was obtained from the number of tubes with Listeria confirmed (after streaking on PALCAM and modified Oxford agars: "true" MPN) and from the number of dark Fraser broth tubes ("predictive" MPN). Samples were analyzed in duplicate. Low levels of Listeria were found (< 110 CFU/g). The direct plate count technique was totally ineffective for enumerating Listeria in poultry. The single-step (UVM I) and the two-step (UVM I-Fraser) MPN methods gave comparable estimations and a low number of significantly discrepant predictions. Using a single-step method with Fraser broth, lower true MPNs were obtained. The number of tubes used (3, 5, or 10) did not have a substantial influence on the results. Similar estimations, highly correlated (r = 0.538 to 0.968; P < 0.001), were found with (true MPN) and without (predictive MPN) plating confirmation when using the two-step MPN method. The statistical evaluation of the differential character of Fraser broth as part of the two-step MPN method showed high sensitivity (87.5 to 92.5%), specificity (95.2 to 98.6%), efficiency (94.2 to 97.6%), and predictive values (73.6 to 89.9% for a positive test and 98.0 to 98.9% for a negative test). Taking into account these results, we suggest the convenience of using a 3- or 5-tube two-step (UVM I-Fraser) MPN method with estimations obtained from the number of tubes with darkening, without confirmation, in order to achieve great savings in time and money.