Detection of multiple-strain carriers: the value of re-sampling

Bacteria may colonize a carrier with more than one strain of a species at any one time. Attempts to determine multiple colonization are labour intensive because of the large number of colonies per carrier which need to be tested. A possible solution -- in which only 3 colonies per carrier are initially tested and only multiple-strain carriers are re-sampled -- was recently described. We evaluated the accuracy of the "re-sampling method" devised by Cespedes et al. with 500,000 stochastic simulations per scenario. Re-sampling is acceptable where > or = 5 colonies are initially tested or where one strain predominates over others in colony counts. The method introduces bias towards overestimation which decreases with the number of available carriers, increases with the proportion of truly multiple carriers, with decreasing number of colonies available for testing, and with decreasing number of colonies tested per carrier. Initial testing of 5-8 colonies tested with re-sampling are adequate for a large study (>100 carriers), or a small study where it is suspected that no strain predominates over the other in colony counts. Testing 9-20 colonies with re-sampling is necessary for small studies where one strain predominates over others. Re-sampling is unnecessary where >20 colonies are tested.