Selection of Cases and Controls

Cases and controls were selected from the national MRSA database at RIVM. Case-patients were defined as persons carrying NT-MRSA who provided the first isolate from a cluster of 1 particular referring laboratory (index-patient) in the period January 2003 to September 2005. Cases were considered to be secondary to an index case when the strain was isolated within 3 months after the previous isolate with the same PFGE typing result. Controls were persons who carried MRSA that was typable with PFGE and who also fulfilled the index-patient definition. Controls were derived from the laboratories that provided cases and were selected at random. Twice as many controls as case-patients were selected.

Collection of Epidemiologic Background Information

Data were collected by questionnaires that were sent to the referring laboratories. The questionnaire contained items about patient characteristics (birth date, sex, postal code, presence or absence of infection, hospital admission dates, profession, profession of partner, profession of parents, and contact with animals, e.g., pigs, cows, horses, chickens, cats, or dogs) and microbiologic data (isolation date, source of culture, medical specialty). All data were collected and entered into the database without our knowing whether it concerned a case or control.

Initially, 41 cases and 82 controls from 26 different laboratories were selected from the national database. The response rate was 98% (40 cases and 81 controls). During workup, 5 cases and 5 controls were excluded for the following reasons: the confirmation test of the isolate indicated that it was not methicillin resistant (1 case), or the case did not fulfill the case definition because it was not the first case from a cluster (4 cases). Since 2 of these cases were from laboratories that had no other case in the study, the accompanying controls were excluded (n = 3). Two controls were identified outside the study period. Finally, 35 cases and 76 controls from 24 different laboratories were included in the analysis.

Molecular Typing and Susceptibility Testing

All MRSA isolates were typed by PFGE (8). All isolates from case-patients and 74 isolates from controls were typed by spa typing (10). Multilocus sequence typing (MLST) was performed on all case isolates, as well as on 1 strain of each spa type of the control isolates (n = 37) (11). PCR of the staphylococcal chromosome cassette (SCCmec) was performed according to Zhang et al. on all isolates from case-patients and 74 isolates from controls (12). The presence or absence of PVL genes (lukS-PV/lukF-PV) was determined in all case isolates and in 71 control isolates. The PVL genes were detected by PCR according to the method of Lina et al. (13). The susceptibility to antimicrobial agents was tested for 32 case isolates and 74 control isolates, according to CLSI guidelines that used Mueller-Hinton agar and multipoint inoculation (14).

Geographic Distribution

Figure 1 shows the geographic distribution of NT-MRSA and typable MRSA isolates, plotted over the density of the pig and human populations, respectively. The density of NT-MRSA isolates corresponds to the density of pig farming, whereas the density of typable strains corresponds to the density of the human population. The density of cattle farms is more or less identical to the density of pig farms.

Open in a separate window

Figure 1

A) Number of nontypable methicillin-resistant Staphylococcus aureus (NT-MRSA) isolates per municipality received at the National Institute for Public Health and the Environment (RIVM), Bilthoven, the Netherlands, January 2003–June 2005. The background color represents the density of pigs per km² in 2003. B) Number of typable MRSA per municipality received at the RIVM January 2003–June 2005. The background color represents the population density per km² (source: CBS Statline).

Epidemiologic Data

Results of the univariate analysis are shown in Table 1. Comparable values were observed for the baseline characteristics of sex and age. Case-patients more often lived in rural areas and indicated more frequent contact with pigs or cattle than did controls. Controls were more often associated with healthcare facilities.

Among case-patients, MRSA was more frequently found in clinical samples (an unexpected finding) compared with controls, whose MRSA was found more often by targeted screening in nose, throat, and perineum. Among persons infected by MRSA, respiratory tract infections were more frequent in case-patients, whereas skin and soft tissue infections predominated in controls.

Multivariate analysis that used a model with the variables describing type of residence (rural vs. other) and contact with pigs, cattle, cats, and dogs (yes, no, or unknown) showed that contact with pigs and contact with cattle were independent statistically significant variables. The adjusted odds ratios (OR) for pigs and cattle were 9.4 (95% confidence interval [CI] 1.8–47.7) and 13.5 (95% CI 1.0–179.3), respectively.

Molecular Typing

Thirty-two of 35 case-patients had MLST sequence type (ST) 398; 1 had ST 9; and the remaining 2 had ST 752 and 753, closely related to 398 (Figure 2). Among case-patients, the most frequent spa types were t108, t011, and t034 (Table 2). These MLST and spa types were not found among the controls. Twenty-two different STs and 37 different spa types were found in the controls (Table 2 and Figure 2).

Open in a separate window

Figure 2

Genetic relatedness of methicillin-resistant Staphylococcus aureus from cases and controls, represented as a minimum spanning tree based on multilocus sequence typing (MLST) profiles. Each circle represents a sequence type, and numbers in the circles denote the sequence type. The size of the circle indicates the number of isolates with this sequence type. The number under and right of the lines connecting types denotes the number of differences in MLST profiles. The halos surrounding the circles indicate complexes of sequence types that differ by <3 loci.

SCCmec typing showed that in isolates from cases SCCmec types III, IV, and V were found, whereas in isolates from controls all SCCmec types were found (Table 2). For 11 cases and 33 controls, the SCCmec type could not be determined. There was no difference in the presence of the PVL genes (Table 2).

We thank Ing. H. Giesbers for constructing Figure 2; R. Bosboom, A. Demeulemeester, B. v. Dijke, B. Hendrickx, R. Hendrix, C. Hol, E. IJzerman, S. Kuijpers, A. Lampe, M.K.E. Nohlmans-Paulssen, N. Meessen, A. Möller, M. Peeters, D. Potters, P. Rietra, P. Schneeberger, F. Sebens, L. Spanjaard, A. Troelstra, C. Vandebroucke-Grauls, C. Verduin, and M. Wulf for answering the questionnaires; M. van Santen-Verheuvel and E. Spalburg for performing spa, MLST, and SCCmec typing and susceptibility testing; C. Klaassen for SCCmec typing; and L. Schouls for analysis of the MLST data.