Context
The estimated annual incidence of Campylobacter human infections in the global population varies from country to country. In the United States, it is estimated to be 880/100 000, nearly twice as high as salmonella incidence, and in England 690/100 000 (1,2). In France, there is no surveillance system for Campylobacter infections occurring in urban areas. However, two districts studies evaluate the incidence of confirmed infections to be 38/100 000 in administrative department of Charente-Maritime (3) and at 27/100 000 in the administrative department of Mayenne (unpublished data from the departmental directorate of health and social affairs, DDASS).

In Europe, an increasing number of strains resistant to quinolones and linked to the use of antibiotics in human and animal welfare has been observed since the beginning of the 1990s (4,5). A worrying increase of quinolone resistance, comparable to that in other countries, has also been observed in France since 1993 (figure), thanks to the surveillance network of Campylobacter infections based on volunteer hospital laboratories that send isolates to the national reference centre for Campylobacter and Helicobacter (Centre National de Référence des Campylobacters et Hélicobacters (CNRCH)) since 1986 (6).

Campylobacter infections are among the priority diseases for the implementation of a European surveillance network on communicable diseases. Each member state is required to provide epidemiological data on Campylobacter infections (7). In France, prior to the implementation of Campylobacter infection surveillance, a study on diagnosis practices for those infections was carried out in hospital laboratories (HL) and private laboratories (PL) by the Institut de Veille Sanitaire in collaboration with the CNRCH. The study was led within the framework of a European survey coordinated by the Robert Koch-Institut in Germany. The objective was to set up an inventory of surveillance networks and diagnostic practices of Campylobacter infections in each participating country (8).

Method
One hundred (20%) HL and 400 (10%) PL were randomly selected from the directory of national laboratories, respectively 500 HL and 4000 PL. Questionnaires were sent by post in October 2000, followed by a reminder to non-responding laboratories in December 2000.
The data collected concerned test criteria for Campylobacter, the number of stool cultures, the number of Campylobacter tests and the number of positive results in 1999, demographic data, the concept of clustered cases and travel, the diagnostic tools used, preservation delays of culture media and the characterisation of sensitivity tests to antibiotics.

Results
The response rate was 68,0% (68/100) for the HL and 50.7% (203/400) for the PL. The HL usually carry out analyses for hospitals but one third performed stool cultures at the request of general practitioners. In 1999, 77 PL (89.0%) and 64 HL (94.0%) tested for Campylobacter at least once. Thirty eight percent of PL and 37.5% of the HL always tested for Campylobacter. Testing was also motivated by the following criteria: presence of blood or mucus in stool samples (58.0% of the PL and 64,0% of the HL), request by clinician (55.0% of the PL and 62.5% of the HL), liquid stool or child's stools (45.0% of the PL and 52.0% of the HL).

In 1999, a PL tested for Campylobacter 129 times on average (56.3% of stool cultures (CI 95% [49.6% - 62.8%]) in stool samples, of which 4.7% (CI 95% [1.7% - 9.8%]) were positive versus 580 (50.5% of stool cultures, CI 95% [47.6% - 53.4%]) on average for a HL with 1.7% (CI 95% [0.8% - 3.1%]) positive results. Systematic testing for Campylobacter showed an average frequency of positive results lower than the results obtained by directed search, respectively 3.3% for PL and 1.0% for HL and 5.0% for PL and 4.0% for HL. Stool samples were most often sent without any transport medium. The minimum delay between the receipt of stool samples at the laboratory and carrying out of stool cultures was around one hour for PL and HL. The maximum delays ranged from seven hours for PL to 15 hours for HL.
A direct microscopic examination was systematically performed on stool samples by 76% of PL and 66% of HL. The laboratories (90% of PL and 66% of HL) mainly used the commercial selective culture medium Campylosel®‚ (bioMérieux). In 45% of PL and 25% of HL, the media were kept beyond the expiration dates indicated on the boxes (>8 days and up to 72 days). Micro-aerobiosis was performed by PL (86%) and HL (91%) mainly using the technique of gas generating packs (H2 and CO2) ('gas pack' special micro-aerobiosis). Enrichment media and filtration methods were seldom used to isolate Campylobacter (3% of PL and 5% of HL). The incubation temperature was generally 37°C and the mean incubation period was 48 hours. Analyses to confirm the Campylobacter species were performed by 83% of PL and 95% of HL in case of uncertain diagnosis (table 1). Eighty six per cent of HL identified Campylobacter species more frequently than PL (37%) (table 2).

Seventy five per cent of HL and 32% of PL systematically tested Campylobacter strains for antibiotic sensitivity (table 3). The Muller Hinton medium with horse or sheep blood and the agar diffusion technique at 37°C or 42°C were most frequently used.
Among variables of epidemiological interest, the patient's age and sex, and the sample date were almost always collected and recorded by private and hospital laboratories. Hospital laboratories had little information on the place of residence. Other information such as diarrhoea, hospital admission, clustered cases, and travel abroad were less often collected by either type of laboratory (table 4).

Discussion
The laboratories' response rate was satisfactory (51% for PL and 68% for HL) for this type of study. A selection bias may, however, have been the source of an over-representation of laboratories testing for Campylobacter in stool samples. Our study shows that over one third of laboratories tested for this bacteria systematically. Unlike salmonellae, survival and growth of Campylobacters require particular micro-aerobiosis conditions, more expensive, and testing for them can be difficult. This could explain why all laboratories do not perform this test systematically. Results of the national quality control carried out in 1991 (9) showed that only one third of private laboratories isolated Campylobacter among six bacterial mixtures. This study shows that the average frequency per laboratory of stool cultures with Campylobacter isolation (4.7% of tests for PL and 1.7% of tests for HL) is similar to the one in the study carried out in Charente-Maritime (France), in which 3.4% of tests were positive (5). Campylobacter infection most frequently causes acute gastroenteritis that rarely calls for hospital admission, which explains a higher rate of positive results in PL than in HL.

In general, samples were sent to laboratories without any transport medium. If stool culture is carried out within a reasonable delay or if stool samples are kept under satisfactory conditions before stool culture, the absence of a transport medium should not affect the viability of Campylobacters. The average delay shown in this study between the reception of stool samples at the laboratory and the performance of stool culture was one hour, and should not affect the viability of Campylobacter provided the delay between sample taking and reaching the laboratory is itself quite short. However, an unsuitable preservation mode of samples by patients (such as non-refrigeration) before transport to laboratories can be the source of decreasing viability of Campylobacters. It was not possible to verify this information, however. The quality of selective media decreases with time, and using media that are not at the peak of their efficacy can contribute to the absence of isolation of those bacteria. This study shows that only 50% of private laboratories and 75% of hospital laboratories respected expiry dates indicated by manufacturers. Considering the difficulties in the culture and diagnosis of Campylobacter, the use of reference documents could be enlarged and completed by training for diagnosis techniques.
For surveillance purposes, some data is necessary to describe spatial and temporal trends of Campylobacter infections as well as some of the patient characteristics. This study shows that data on the place of residence, the concept of clustered cases, and information about travel abroad in the 10 days preceding disease onset were rarely collected. This last piece of information would allow comparison of the rate of autochthonous infections with those acquired in another country. All the same, an increase in the number of strains or the concept of clustered cases should lead to sending all strains systematically to the CNRCH in order to contribute to the exploration of a link between them with the help of molecular epidemiology, and to start an investigation. The deficit in the species diagnosis of private laboratories could be compensated for by sending strains to other laboratories that perform this diagnosis or to the CNRCH.
In addition to need to know the antimicrobial resistance of Campylobacter strains on a therapeutic level, knowledge of the strains' resistance profile isolated in animals will contribute to a better understanding of the epidemiology of Campylobacter infections.

Conclusion
A significant number of laboratories performed Campylobacter diagnosis with comparable methods. Not all laboratories, however, test systematically for Campylobacter, and some diagnosis practices should be improved, in particular, decreasing the preservation delay of the media used for the culture of Campylobacter. In terms of laboratory diagnosis practices, this study showed the feasibility of Campylobacter infections surveillance in urban areas. A study on the willingness of laboratories to participate in the surveillance of Campylobacter infections in town was carried out in November 2001.

Acknowledgements

The authors wish to thank the private and hospital laboratories that participated in this study.