Predominant spa types showed a wide geographic distribution. The average distance between the locations from which the same spa types were sampled was smaller for MRSA isolates than MSSA isolates, suggesting a higher degree of geographical clustering of MRSA. Moreover, genetic diversity was much lower for invasive MRSA than MSSA and differed considerably between countries. Spatial-scan statistics corroborated a fundamental difference between MRSA and MSSA with respect to regional dissemination. The majority of isolates that formed regional clusters were MRSA, and 13 of the 15 major MRSA spa types (defined as more than ten isolates in the database) occurred in geographical clusters. They were typically hospital acquired and no more than three clusters overlapped in the same region. Conversely, of the 27 major MSSA spa types, only five showed significant geographical clustering and only three consisted of MSSA alone. Thus, invasive MRSA clones in Europe display a typical epidemic behaviour and have a predominantly regional distribution. To unravel the dynamics of spread of these epidemic MRSA requires the present type of survey to be repeated every few years.
The emergence of MRSA occurs by the acquisition of the methicillin resistance determinant (SSCmec
) by MSSA strains. MRSA strains typically emerge from the most prevalent MSSA strains and it is a rare event, although new findings suggest that it is more frequent than originally suggested 
. There are thus fewer MRSA clones compared to MSSA clones and they are very young on evolutionary time scales (less than 50 y old) and have had little time to diversify since they arose, whereas MSSA are much older and thus more diverse. MRSA clones also expand because of the selective forces exerted by heavy antibiotic use in hospitals and conditions that favour transmission within and between hospitals, which constrains their diversity. In contrast, MSSA will be subject to much weaker selection leading to neutral genetic drift that maintains their diversity. Geographic spread of MRSA clones will be facilitated by repeated hospital admissions and referrals of MRSA carriers 
who typically belong to an older and thus less mobile segment of the population. The broader distribution of MSSA clones may reflect dissemination by hosts with different travel patterns than MRSA carriers as well as the longer time that MSSA clones have had to spread.
The present survey set several precedents in the field of molecular epidemiology of S. aureus. First, it brought together reference laboratories from 26 European countries adopting a standardised quality-controlled DNA sequence-based typing method; second, it provided a contemporaneous and comprehensive population snapshot of S. aureus isolates from invasive disease using an agreed sampling frame; third, it utilised modern spatial scan statistics to identify geographical clustering; fourth, it provided the first public domain Web-based interactive mapping tool for future public health research; and finally, it consolidated a collaborative framework for the continuation of this important European surveillance initiative.
All Member States of the European Union except Estonia, Lithuania, Luxemburg, and Slovakia participated and variously achieved a country-wide enrolment of diagnostic laboratories. In the run-up to this study, a successful effort was undertaken to agree on a single molecular typing approach to scale up the typing capacity, and improve quality assessment, by introducing proficiency testing for SRLs that intended to participate in advanced S. aureus
surveillance. This effort provided the basis for the execution of a mutually agreed protocol using a standardized sampling frame and a quality-controlled genetic typing approach 
, based on the sequencing of the variable region of the S. aureus spa
. Multilocus sequence typing (MLST) was also carried out on many strains allowing most of the prevalent spa
types to be related to MLST sequence types. However, given the scope and ambition of this investigation, it is not surprising that the study still suffered from several operational shortcomings.
In order to keep the amount of work manageable for the participating SRLs, it was decided to include about 20 laboratories that were demographically and geographically representative of each of the participating countries. This number is arbitrary and cannot equally represent small and large countries alike. Thus the precision of spatial scan statistics is reduced in areas where the density of laboratories is low. Laboratory enrolment based on population size would provide a more appropriate sampling strategy but would also impose a proportional and frequently unacceptable amount of work on SRLs in large countries if the sample size from small countries should remain meaningful. Even medium-sized countries such as Bulgaria, Finland, Greece, and Romania were unable to enrol the intended number of laboratories mainly for technical and logistic reasons. Naturally, the number of laboratories and isolates varied between countries and geo-demographic representation could be improved in future investigations. The original intention was to collect equal numbers of successive MSSA and MRSA in all laboratories during the 6-mo sampling interval; this proved to be unrealistic, especially in countries where MRSA levels are low. As a result, Norway and Iceland could not provide any MRSA, whereas Sweden, Denmark, and the Netherlands each provided fewer than ten isolates. Cyprus and Malta had only one participating laboratory but since both provide the microbiological service for the whole of the respective island population (for Cyprus, only the Republic of Cyprus), they were entitled to submit up to 20 isolates. Nevertheless, even taking these potential problems into account, the simultaneous collection of 2,890 isolates from patients with invasive S. aureus
infection treated in 450 hospitals during a 6-mo study interval is unparalleled and remains unmatched by previous investigations, which have drawn their conclusions from convenience samples of predominantly MRSA collected by laboratories for different clinical or biological reasons. The current collection includes one-third MRSA and thus over-represents the natural population of MRSA causing invasive disease, which was 22% in the European Union in 2007 
. All isolates were collected from laboratories and hospitals participating in EARSS for which estimates of the overall catchment population are known. Thus, the present sample of hospitals catered for approximately 22 million people, totalling 4.4% of the citizens living in the European Union.
Despite the above limitations, the sample provides a realistic insight into the epidemiology of S. aureus
currently causing invasive infection in Europe. The age distribution and all-cause mortality was consistent with the expected range 
. High frequencies of invasive infections were ascertained in the very young (infants and under 5-y-olds) and older age groups with males clearly more at risk than females. Patients with MRSA were older than patients with MSSA and were 2.4 times more likely to have their infection attributed to hospital acquisition (p
<0.0001). Invasive MRSA carried a higher all-cause mortality after 14 d. This finding is most likely confounded by a host of variables that distinguish MRSA patients from MSSA patients in fundamental ways. All-cause mortality was independent of spa
type, indicating that this study did not identify any single spa
type that stands out with respect to hyper-virulence or other factors that would increase the risk of fatal outcome after 14 d. While a laboratory-based cross-sectional study is limited in its ability to control for many of the crucial confounders such as comorbidity and disease severity scores and thus may not detect subtle differences in virulence properties between different clonal lineages at the patient level, this sample provides an unbiased estimate of the frequencies of specific spa
types that have previously been reported to cause outbreaks and have attracted considerable public health attention. Few MRSA isolates carried the PVL-toxin genes and this could be an indication that many CO-MRSA were originally hospital-acquired. Only six of the PVL-positive CO-MRSA isolates, which made up 0.5% of the overall sample, had spa
types consistent with the CA-MRSA clones most frequently reported in Europe (t008/ST8/SCCmec
IV and t044/ST80/SCCmec
. These values when compared to the overall numbers of MRSA are small but still warrant attention since PVL-positive CA-MRSA are more commonly associated with skin and soft tissue infections and are rarely found in blood stream infections 
from which the majority of the sample isolates were drawn. Livestock-associated spa
types belonging to MLST sequence type ST398 
made up 0.4% of the overall sample. However, none were methicillin-resistant indicating a low rate of human systemic infections induced by MRSA variants of this clone despite the increasing interest and concern about such isolates by health authorities.
With the decision to utilize spa
typing as a common platform to address the geographic abundance of S. aureus
clones, various potential problems that might affect observations need to be taken into account. First, a single sequence of approximately 280 base-pairs under potential immune selection may be a relatively weak indicator for the genetic background of a genome that is approximately 10,000-times larger, even in a species such as S. aureus
that is evolving in a predominantly clonal manner. Furthermore, convergent evolution could occur as a result of the high mutability of the repeat region of the spa
gene used in spa
. Finally, spa
typing on its own it may not be sufficiently discriminatory to distinguish between MRSA isolates given that only five spa
types accounted for almost half (48.1%) of all 967 MRSA isolates examined in this survey. The regional clusters found in this study provide a good indication that homoplasy is not a major problem that hinders the recognition of clonal dissemination on a geographic scale. Moreover, when comparing agr
type, toxin gene, and antibiotic susceptibility profiles within and between different spa
types, a significant within-spa
-type consistency and between-spa
-type discordance supports the notion that in most cases spa
typing provides a convenient and valid marker for the major clones and clonal lineages 
In order to explore the geographic distribution of different spa
types, and the genetic and phenotypic detail of isolates in the European sample, the reader is encouraged to explore the purpose-built interactive Web application, SRL-Maps, at http://www.spatialepidemiology.net/srl-maps/
. This application contains the public domain data made available to the scientific and public health community by all SRLs that participated in the study. It also illustrates the potential of such a communication platform. The underlying database is fully searchable and the map-based application is a template for the future addition of further epidemiological and biological information. We believe that this approach to spatial epidemiology will become a rich resource for future enquiry into the population dynamics of infectious agents and their evolution.
In conclusion we provide evidence that the major MRSA clones in Europe occur predominantly in geographical clusters. This also indicates that MRSA, rather than spreading freely in the community, diffuses through regional health care networks. This important finding suggests that control efforts aimed at interrupting the spread within and between health care institutions may not only be feasible but ultimately successful and should therefore be strongly encouraged. We also showed that an international surveillance network sharing decentralised typing results on a Web-based platform can provide crucial information for clinicians, diagnostic microbiologists, and infection control teams on the dynamics of S. aureus spread, and especially the spread of MRSA isolates, to provide early warning of emerging strains, cross-border spread, and importation by travel.