To determine the stability of pulsed-field gel electrophoresis (PFGE) patterns of methicillin-resistant Staphylococcus aureus in the nosocomial setting, we analyzed isolates from long-term carriers (>1 month) and from patients involved in well-defined nosocomial epidemics. The number of fragment differences between the first isolate and subsequent isolates in long-term carriers showed a bimodal distribution, with one group having 0 to 6 fragment differences and the other group having 14 to 24 fragment differences. The PFGE patterns of isolates involved in epidemics also presented a similar bimodal distribution of the number of fragment differences. Typing these isolates with another molecular method (inter-IS256 PCR) showed that isolates of the first group (i.e., with 1 to 6 fragment differences) were clonally related, whereas the second group (with 14 to 24 fragment differences) could be considered genetically different. Among long-term carriers with clonally related isolates, 74 of 84 (88%) of consecutive isolates showed indistinguishable patterns, whereas 10 of 84 (12%) showed related patterns differing by one to six fragments. Moreover, the frequency of apparition of related patterns is higher when the time between the first and the subsequent isolate is longer. During seven nosocomial epidemics lasting from 1 to 15 months, only 2 of 120 isolates (1.7%) showed a pattern which was different, although related, from the predominant one involved in each of these outbreaks.Pulsed-field gel electrophoresis (PFGE) analysis of genomic macrorestriction DNA fragments is one of the most commonly used methods for the epidemiological typing of many bacteria. When the method is used for the investigation of an epidemic, it has been proposed that the so-called "related" isolates should be considered as probably (a 1-to 4-DNA-fragment difference) or possibly (a 5-to 7-DNA-fragment difference) part of the outbreak (6, 10, 12).The mutation rate, including point mutations, genetic rearrangements, and horizontal transfers of mobile DNA elements such as phages and transposons, is different from one bacterial species to another. In a given typing system, the mutation rate will directly influence the stability of the typing patterns during the replication cycles of a given bacterial clone. Thus, for a given bacterium, the epidemiological interpretation of related PFGE patterns will depend on the time and space scales that one considers (2, 9, 11, 13). For a species which has a high mutation rate, isolates with related patterns observed over a short period of time should be considered as sharing recent epidemiological links. On the other hand, if the species has a low mutation rate, isolates with related patterns are likely to have more distant epidemiological links. To evaluate the evolution rate of typing patterns, long-term experiments using microbial in vitro cultures extended over several years have been proposed (11, 13). However, this may not properly reflect the in vivo conditions of microorganism multiplication and transmi...