Introduction
The disease burden of Campylobacter infections is considerable in Australia, where it is the most common notifiable disease [Reference Owen1]. Campylobacter infections are largely sporadic and estimated to cause about 223 000 cases of gastroenteritis each year [Reference Hall2]. The need for public health intervention is highlighted by the steady rise in notifications in Australia since the early 1990s [Reference Owen1, 3].
Case-control studies of potential risk factors for campylobacteriosis have been undertaken in several countries [Reference Wingstrand4–Reference Neimann11] including Australia [Reference Tenkate and Stafford12, Reference Stafford13]. Consumption of chicken is the most commonly identified risk factor for sporadic campylobacteriosis. However, this and other statistically significantly risk factors often do not explain the majority of cases [Reference Stafford13, Reference Rodrigues14]. Other sources of apparently sporadic campylobacteriosis are difficult to detect using current case-control study designs.
Further sources of Campylobacter infection may be detected if strains with distinct ecologies (including differential survival characteristics, and therefore specific environmental niches or host preference), with varying virulence (causing disease of variable severity) and different transmission routes could be distinguished. Numerous subtyping methods have been developed and applied to Campylobacter isolates [Reference Nachamkin, Bohachick and Patton15–Reference Hedberg19]. These methods, used to determine genomic diversity, generally have been applied to diverse collections of isolates [Reference de Boer20, Reference Lindstedt21]. Methods such as serotyping, multilocus sequence typing (MLST) and pulsed-field gel electrophoresis (PFGE) have been used to distinguish outbreak from sporadic isolates [Reference Clark18, Reference Frost, Gillespie and O'Brien22, Reference Sails, Swaminathan and Fields23]. In addition, a variety of methods have been used to compare animal and human isolates. Some studies have concluded that finding the same subtype in animal and human isolates is evidence of transmission from animals to humans or is evidence of a common source for animals and humans [Reference Hanninen17, Reference Colles24–Reference Manning28]. However, the potential to identify genotype-specific risk factors for sporadic infection is yet to be evaluated.
The aim of this study was to determine sources and risk factors of Campylobacter jejuni for specific flaA genotypes. We hypothesize that the application of flaA genotyping to isolates from a case-control study may allow detection of further sources through the examination of genotype-specific risk factors. Study isolates were collected prospectively from sporadic cases recruited into a case-control study and genotyped using flaA restriction fragment-length polymorphism (RFLP; flaA genotyping) analysis, a moderate throughput, low cost method [Reference O'Reilly, Inglis and Unicomb29] with good correlation with MLST [Reference Djordjevic30]. Genotyping data were linked to exposure data for the investigation of risk factors.
METHODS
Study base
Data and Campylobacter isolates collected for this study were drawn from an Australian case-control study that was conducted in five states between September 2001 and September 2002. A description of the study and risk factors for subjects aged ⩾5 years was reported by Stafford et al. [Reference Stafford13]. The study recruited cases and controls from five of the eight states and territories in Australia. The largest state, New South Wales, was excluded because campylobacteriosis is not notifiable there and the Northern Territory and Australian Capital Territory were not included in the study because too few cases are notified. As previously described [Reference Stafford13], each site aimed to prospectively recruit 200 cases of all ages using systematic sampling from a notifiable disease register; every second case from Tasmania and Victoria, every fourth and sixth case from Queensland and South Australia, respectively and every case from Western Australia. The sampling strategy was based on the expected number of cases notified from participating laboratories and the number required to detect significant associations with hypothesized risk factors [Reference Stafford13]. Furthermore, sample size calculations determined that about 550 isolates would probably be collected and that number would prove sufficient to detect significant associations between some genotypes and risk factors.
Cases were defined as individuals with diarrhoea (⩾3 or more loose stools in a 24 h period), who had culture-confirmed C. jejuni infection (and no other pathogen such as Salmonella, Shigella, or the enteric protozoa detected), whose isolates were flaA genotyped, whose stool samples were collected within 10 days of diarrhoea onset and who were interviewed within 30 days of onset. Telephone interviews were conducted when verbal consent was given by the study subject (or carer if the case was aged <16 years).
Controls were drawn (about one per case) from a control bank generated during a national cross-sectional survey of gastroenteritis conducted in 2001; a description of the survey is given by Hall et al. [Reference Hall31] and a description of control selection is reported by Stafford et al. [Reference Stafford13]. In brief, households were selected using random digit dialling during the gastroenteritis survey, a household member was asked to participate in the survey and subsequently consent to be part of a control bank. From this control bank, potential controls were selected and frequency-matched to cases by age bands (0–4, 5–9, 10–19, 20–29, 30–59 and ⩾60 years) in each state at a ratio of 1:1. If a person did not wish to participate or was excluded on the criteria (described below), a subsequent person was sought from the control bank. Once a control had been selected from a household, that household was no longer eligible for future selection of controls. Controls were interviewed within 30 days of interview of a notified case.
Cases and controls were excluded if they did not have a phone number, were unable to be contacted after at least six attempts, they/their parents were non-English-speakers, they could not answer questions, or if a household member had had diarrhoea or a confirmed Campylobacter infection in the 4 weeks prior to onset. In addition, cases were excluded if they were unable to recall the diarrhoea onset date or they were part of an outbreak.
Questionnaire
A standard questionnaire was administered by telephone to obtain information on a range of variables, including host factors (underlying illnesses, prior consumption of antimicrobial agents, antacids and immune suppressive therapies), overseas travel, dining locations, consumption of water and food (fruit and vegetables, meat, poultry, seafood, eggs and dairy products), animal and pet exposures, and demographics. Cases were asked additional questions about their illness and treatment. For cases, all questions related to the 7-day period before their onset of diarrhoea, except for prior use of antibiotics, antacids or immune suppressive therapies, which were based on the preceding 4 weeks. Exposure information was not collected from subjects who had travelled outside Australia during the 7-day exposure period. Controls were asked about the 7 days or 4 weeks prior to interview.
Laboratory methods
Isolates
Isolates from diarrhoeal stool cultures were stored and subsequently identified to species level by hippurate hydrolysis and PCR as described previously [Reference Burnett32, Reference Sharma33]. Those identified as C. jejuni using these methods were flaA genotyped.
flaA genotyping
Flagellin A RFLP typing (flaA genotyping) was performed according to the method described by Nachamkin et al. [Reference Nachamkin, Bohachick and Patton15]. Briefly, this involved PCR amplification of flaA, followed by digestion of products with the restriction enzyme DdeI and separation of fragments by agarose gel electrophoresis. Genotyping was performed by five laboratories, using standardized reagents and methods (including DNA extraction, PCR, restriction enzyme digestion, agarose gel electrophoresis and photography), which had been optimized by one laboratory.
Quality assurance
A set of eight isolates was distributed to the laboratories that genotyped isolates. Tiff images of flaA genotyping gels were loaded onto a BioNumerics database located at one laboratory and examined for comparability. Feedback was provided on accuracy of the patterns and image quality. Reaction or photography conditions were modified, if necessary, to produce images consistent with those from prior testing of the quality assurance set. The positive control, NCTC 11351, was included in each test run. GeneRuler™ 100 bp DNA LadderPlus (MBI Fermentas, Vilnius, Lithuania) was included in lanes 1, 5, 10, and 15 of all gels. If the gel pattern of the positive control was not compatible with those included in the BioNumerics database, gel images were to be rejected, however, this did not occur.
Assignment of flaA genotypes
Tiff images of gels were loaded onto a BioNumerics database and patterns were normalized according to molecular-weight standards on each gel. flaA genotypes were designated by a number that was assigned arbitrarily; numbered genotypes differed from each other by at least two bands and subtypes of numbered genotypes, designated by a letter (e.g. flaA-6 and flaA-6b) differed by a single band only. The numbering of genotypes was consistent with a previous Australian study [Reference O'Reilly, Inglis and Unicomb29, Reference Djordjevic30] and isolates from the previous study were included in the BioNumerics library used for analysis in this study. Genotypes were grouped together using the Dice band matching coefficient and UPGMA clustering method with a position tolerance of 1% and an optimization of 1% which clustered at >90% similarity [Reference O'Reilly, Inglis and Unicomb29]. Resultant dendrograms were checked visually by two researchers and about 10% were re-verified by the second researcher in the case of discrepancies.
Statistical analyses
Logistic regression analysis was used to compare demographic characteristics and host factors between study cases and (a) study controls, (b) cases that did not have isolates flaA genotyped and (c) cases notified through the national surveillance system.
Logistic regression analysis was also used to identify potential risk factors for specific flaA genotypes. The exposures reported for cases of each major flaA genotype were compared to those for all study controls combined, in order to increase the power of hypothesis testing. A final model for each genotype was constructed by including all exposure variables with P<0·1 (in univariate analyses) and using backwards stepwise elimination, controlling for confounders (demographic and host factors). Models were tested for goodness of fit and compared using the likelihood ratio test. Genotype-specific population-attributable fractions (PAF) were calculated for each risk factor from final models for each of the major flaA genotypes and for the group comprising ‘other’ flaA genotypes.
To allow for the possibility that some food and environmental exposures were location-specific, we included terms for the interaction between exposure and state (as a categorical variable) in the logistic regression models. The significance of multiple interaction terms was tested using the likelihood ratio test. Only significant interactions are reported.
Multinomial regression was applied to data on cases to explore differences in exposure variables for the major flaA genotypes, using ‘other’ flaA genotypes, comprising the remaining study cases, as the reference group, controlling for confounders. This type of analysis was also used to compare demographic and clinical characteristics between cases infected with each of the major flaA genotypes. Results are expressed as relative risk ratios (RRRs), as is appropriate for this case-only analysis [Reference Gould and Sribney34].
Analyses were performed using Stata version 9.1 (Stata Corporation, College Station, TX, USA).
RESULTS
Recruitment of cases and controls
During the study period there were 8847 Campylobacter notifications in the five participating states and, of these, 1019 cases were recruited and interviewed (12%). There were 590 (58%) cases for which an isolate was stored, subsequently found to be C. jejuni, and flaA genotyped, representing 7% of notifications. There were no significant differences in age distribution between 590 study cases and cases notified in 2001 but fewer males were included (52·5% among cases vs. 54·4% for notifications, P=0·03). When cases for whom isolates were genotyped were compared to cases that did not have an isolates genotyped for demographic, host factor and clinical characteristics, a greater number of the former were educated to school level only [37% vs. 28%; odds ratio (OR) 1·5, 95% confidence interval (CI) 1·1–2·0] but no further differences were detected.
A total of 967 controls were recruited and there were statistically significant, but modest, differences between study cases and controls with respect to sex, income, place of residence and use of acid-reducing medications (Table 1).
Table 1. Comparison of characteristics of study cases with controls
OR, Odds ratio (shown in bold where P<0·05); CI, confidence interval; Ref., reference category.
flaA genotype distribution
Among 590 isolates, there were 61 different flaA genotypes, of which five accounted for 325 (55%; an image of the electrophoretic patterns of the major genotypes is shown in the Fig.), and 21 (4%) were single isolates. The five major flaA genotypes comprised flaA-6b (n=129, 22%), flaA-6 (n=70, 12%), flaA-10 (n=48, 8%), flaA-2 (n=43, 7%), flaA-131 (n=35, 6%); the remaining 265 study cases comprised the ‘other’ genotype group used in case-only analyses. Some geographic differences were noted for the major genotypes; flaA-10, and -131 were identified in all states, flaA-6b and -2 were found in four of the five states but flaA-6b, the most common genotype, was not found in South Australia and flaA-6 was found in South Australia and Queensland only. Of the major flaA genotypes, only flaA-2, and -10 were detected among the 13 isolates from overseas travellers (data not shown). The full descriptive epidemiology of flaA genotypes and detection of clusters will be reported separately.
Fig. flaA genotype profiles of the five major flaA genotypes.
The five major flaA genotypes (flaA-2, -6, -6b, -10, and -131) were analysed separately for risk factors.
Patient characteristics and symptom profile for the major flaA genotypes
Comparison of cases due to each of the five major flaA genotypes with those due to all ‘other’ genotypes using multinomial regression showed no differences in the following characteristics (proportions for all study cases are shown in parentheses): proportion of males (53%), those with cramps (89%), persistent diarrhoea (12%), those that were hospitalized (13%), those treated with anti-diarrhoeal medications (48%) or intravenous fluids (12%). A significantly higher proportion of subjects with flaA-2, compared to ‘other’ genotypes, had fever (RRR 2·3, 95% CI 1·1–5·4), bloody diarrhoea (RRR 2·2, 95% CI 1·1–4·4) and >20 bowel motions in a 24 h period (RRR 5·4, 95% CI 1·1–25·7).
Sources of C. jejuni and risk factors
Subjects that had travelled internationally were included in the study but exposure information was not collected. When compared to controls, overseas travel was significantly associated with flaA-10 disease (OR 14·5, 95% CI 2·3–85·7). A total of 66 exposure variables were examined in univariate analyses of cases infected with the major flaA genotypes who acquired their infections locally, compared to controls.
In final multivariate models, constructed to explain exposures associated with locally acquired disease (Table 2), infection with flaA-6b, the most common flaA genotype, was independently associated with consumption of barbecued chicken, offal, paté and exposure to pet chickens. The second most common flaA genotype, flaA-6, was independently associated with consuming chicken and exposure to farm birds (ducks, geese, etc.). Infection with flaA-10 was associated with undercooked beef, offal, paté, exposure to young pet chickens (aged <6 months), and poor food handling. flaA-131 was associated with chicken (meat and pets) and poor food handling, and flaA-2 was associated with consumption of undercooked chicken, ham, exposure to puppies and poor food-handling practices (Table 2). When all ‘other’ flaA genotypes combined (comprising 56 flaA genotypes) were compared to controls, disease among this group was associated with undercooked chicken, offal, and bottled water (Table 2). Of the seven flaA-10 case-patients that consumed offal, three specified lamb, one chicken and the remaining three cases did not indicate a species. Of the seven flaA-6b case-patients that consumed offal, two specified lamb, the remaining gave no details on species. Among flaA-6b and -10 cases that ate offal, there was no geographic or time clustering. While the odds ratios for some of the risk factors implicated may be of borderline statistical significance, some (e.g. offal and contact with pet chickens and farm birds) were unusually high. In the final multivariate models for each of the major flaA genotypes, 74% of flaA-2 cases, 71% of flaA-6 cases, 30% of flaA-6b cases, 57% of flaA-10 cases, 30% of flaA-131 cases, and 16% of ‘other’ flaA genotypes were attributed to significant exposures (Table 2).
Table 2. Final multiple logistic regression modelsFootnote * for exposures associated with major Campylobacter jejuni flaA genotypes
OR, Odds ratio, CI, confidence interval, n.c., 95% CI unable to be calculated due to small numbers.
* Logistic regression models controlled for age, sex (and state, when appropriate).
† Population-attributable fraction shown in bold within square brackets with 95% CI, determined for cases infected with each flaA genotype; blank cells indicate that no statistically significant association was found for the respective flaA genotype and exposure.
‡ Exposure period of 7 days.
§ Model for this flaA genotype included state.
∥ Model for this flaA genotype included consumption of acid-reducing medications.
¶ Model for this flaA genotype included chronic gastrointestinal illness and immunosuppressive therapies.
# Model for this flaA genotype included age and place of residence.
** All remaining flaA genotyped cases. Model for this flaA genotype included chronic gastrointestinal illness.
For the case-only comparisons, we used multinomial regression analyses to determine distinct exposures among the major flaA genotypes, as shown in Table 3. flaA-2, flaA-6b and flaA-10 differed from the comparison group in their exposure to various types of non-poultry meats. Significant exposures associated with flaA-10 were different from those associated with flaA-6b and flaA-131, and reflected exposures associated with disease due to those flaA genotypes in the final models (Table 2). flaA-2 infected cases were significantly more likely to consume ham (Tables 2 and 3). Poultry meat exposures did not differ between flaA types in multinomial regression analyses (Table 3), but were significantly associated with disease due to flaA-2, -6, -6b and -131 genotypes when compared to healthy controls (Table 2).
Table 3. Multinomial logistic regression analysisFootnote * for the major Campylobacter jejuni flaA genotypes compared to ‘other’ flaA genotypes
* Multinomial models controlled for age and sex.
† All cases with flaA genotypes apart from the five major flaA genotypes.
‡ Relative risk ratio (RRR) with 95% confidence intervals (CI) given in parentheses. RRR and 95% CIs given in bold indicate that a statistically significant difference was found for the respective flaA genotype when compared to ‘other’ flaA genotypes.
Discussion
Case-control studies of campylobacteriosis have consistently identified chicken consumption as the most commonly associated risk factor [Reference Kapperud7, Reference Friedman9, Reference Neimann11, Reference Stafford13, Reference Effler35]. Similarly we found chicken associated with disease due to four out of five major flaA genotypes, suggesting that chicken harbour a range of C. jejuni genotypes. Interventions aimed at minimizing chicken contamination are needed to reduce the burden of disease, and have been initiated in Iceland and Denmark [Reference Stern36, Reference Krause37]. Here we have attempted to determine whether separate analysis of C. jejuni genotypes for risk factors may provide insights into further sources of this important disease.
We demonstrated the benefit of using molecular methods to more specifically define cases of campylobacteriosis to study possible risk factors for infection. We found that illness due to C. jejuni genotype flaA-10 was independently associated with undercooked beef consumption. In both case-control and case-only comparisons undercooked beef was associated with flaA-10 disease (Tables 2 and 3). Non-poultry meat has not previously been identified as a risk factor for campylobacteriosis in Australia [Reference Stafford13] (L. E. Unicomb et al., unpublished results). Consumption of raw milk and/or contact with calves have been implicated in four Australian outbreaks [Reference Dalton38] (OzFoodNet Outbreak Register, M. Kirk, personal communication, July 2006) and exposure to non-poultry meats and bovine husbandry have been associated with Campylobacter illness in case-control studies conducted in other countries [Reference Schonberg-Norio5, Reference Potter, Kaneene and Hall8, Reference Friedman9, Reference Neimann11]. By way of comparison, the case-control study from which subjects in this study were drawn had 881 cases and 833 controls aged >5 years. It found undercooked chicken, offal, ownership of domestic chickens aged <6 months, and domestic dogs aged <6 months as risk factors [Reference Stafford13].
Disease caused by flaA-2 was associated with exposure to ham in both case-control and case-only comparisons (Tables 2 and 3). This finding was unexpected, since pigs are predominantly (but not exclusively) infected with C. coli [Reference Jensen39]. Previously, a ham-containing salad has been implicated in a C. coli outbreak [Reference Ronveaux40] and it has been detected as a risk factor in a case-control study conducted in the United States [Reference Friedman9]. It is unclear how processed meats such as ham could be the source of disease; contamination at retail outlets from raw meats may occur. Alternatively, this finding may reflect cross-contamination of ham and other foods during preparation in the home.
Consumption of offal (from a variety of species) was associated with disease due to flaA-6b, -10 and ‘other’ genotypes, and was also reported for all Campylobacter species and genotypes in an Australian case-control study [Reference Stafford13]. Furthermore, duck liver consumption has been implicated in one Australian outbreak (OzFoodNet Outbreak Register, M. Kirk, personal communication, July 2007).
Gel patterns for flaA-6 and flaA-6b genotypes differed by the size of one band (Fig.), probably from an insertion or deletion in the flaA gene. flaA-6 was detected in a previous study from January 1999 to July 2001 in New South Wales [Reference O'Reilly, Inglis and Unicomb29, Reference Djordjevic30] and has been detected in a subsequent study in South Australia from November 2005 to March 2006 (B. Coombs, personal communication, December 2006). Geographic and temporal differences in flaA-6 and -6b distribution suggest that one may be a variant of the other. When flaA-6 and -6b were analysed as a single genotype (controlling for the potential confounders age and sex) similar variables were significantly associated with disease, in univariate models, as those for the more common flaA-6b alone (data not shown). The two genotypes were analysed separately as we could not assume that they are variants without further laboratory investigations.
Clinical manifestations differed slightly among those infected with the major flaA genotypes with flaA-2 infections apparently more severe than those due to other genotypes; this suggests that there may be differences in virulence between flaA genotypes. While the differences between flaA-2 study cases and others were small, they were consistent for symptoms that resulted in missing school, work, recreation or other activities. We have previously reported that flaA genotypes closely predicted MLST; 88% of flaA-2 isolates, when tested by MLST, belonged to clonal complex (CC) 48 [Reference Djordjevic30]. Among CC 48 human isolates on the international MLST database (http://pubmlst.org/campylobacter; accessed 11 July, 2007), those included had caused Guillain–Barré syndrome, Miller–Fisher syndrome and systemic disease in addition to gastroenteritis. Genotype-specific differences in symptoms by age could not be explored in this study as numbers in each age group were small.
Our findings should be considered in the light of study limitations. Selection bias in recruitment of controls is possible, since people who spend more time at home would be more easily contactable, but several factors were controlled for in analyses. Measurement biases may have occurred as we relied on information from participants that was not validated. However, this applied to both cases and controls. Interviewer bias may have occurred as interviewers knew which interviewees were cases and controls; and recall bias was possible since cases potentially would have better recall than controls. Study cases were selected from notified cases, which include those with more severe disease. Study cases for whom isolates were not genotyped, had similar characteristics to those that were genotyped, a greater number of the former were educated to school level only; this small difference may have had an impact on exposures. We conducted many hypothesis tests, so it is plausible that some statistically significant differences may have arisen by chance; of the 66 exposure variables examined for each subtype, an average of three are expected due to chance alone (using the 5% level of significance). Our findings therefore need to be replicated by further studies to confirm our results.
Detecting genotype-specific risk factors may be better determined by enrolling cases based on their isolate subtype results, focusing on the more common subtypes, thereby including sufficient sample size to test hypotheses. This would further be enhanced by studies of the distribution of flaA genotypes of non-human C. jejuni isolates; such data on Australian non-human isolates are not available. Information on genotypes for other countries is available from the international MLST database. As stated above, flaA genotypes can closely predict MLST CCs: 96% of Australian flaA-6 isolates belong to CC 257, 91% of flaA-10 to CC 354 and 88% of flaA-2 to CC 48 [Reference Djordjevic30]. Among the data from various countries on the MLST database, these clonal complexes have been detected in a variety of non-human samples and countries: CC 257 (flaA-6) was detected from bovine, avian (poultry and other avian), ovine and porcine samples, CC 354 (flaA-10) from bovine, poultry (including environmental) and ovine samples and CC 48 (flaA-2) from bovine, avian (poultry, poultry environment and other avian), ovine, water, sand and domestic pet samples. While these data could potentially be useful in supporting findings from the case-control and case-case analyses, only one of the 124 non-human isolates from the CC 257, 354 and 48 isolates on the database was from Australia. Determining flaA genotypes from systematically collected, non-human sources in Australia may be useful to identify potential reservoirs of genotypes.
flaA genotyping is a gel-based method which has limitations; it requires standardization to achieve comparable results across laboratories but is rapid and cheap. Ideally, this method would be automated to achieve ‘high throughput’ status. There are reports of instability in the flaA locus [Reference Harrington, Thomson-Carter and Carter41] but the majority of C. jejuni isolates are apparently genetically stable in this region over time [Reference Fitzgerald26, Reference Burnens42, Reference On43] and that instability may be strain specific. Sequence-based methods such as MLST remove dependence on visual and potentially subjective, genotype assignment but require expensive equipment to handle medium to high numbers of isolates and reagents are costly [Reference O'Reilly, Inglis and Unicomb29, Reference Dingle44]. Isolates from this study were genotyped retrospectively, but ‘real time’, genotyping of notified case isolates would be preferable, to enable rapid detection of temporal clusters using a library of common flaA genotypes and should be feasible.
This is the first study to suggest the value of flaA genotyping for identifying strain-specific risk factors for Campylobacter. Case-control analyses using logistic regression and case-only analyses using multinomial regression were employed to detect risk factors for C. jejuni flaA genotypes among a selection of notified cases drawn from five Australian states over a 12-month period. Differences were detected for symptom profile, geographic distribution and exposures between flaA genotypes. The value of flaA genotyping is therefore worthy of further investigations in studies with a larger sample size and in other settings, and particularly in the course of outbreak investigations. The ability of flaA genotyping to detect clusters and outbreaks among apparently sporadic notified cases will be assessed in a separate report (L. E. Unicomb et al., unpublished observations).
ACKNOWLEDGEMENTS
This study was funded under the OzFoodNet programme of work, which is an initiative of the Australian Government Department of Health and Ageing. Hunter New England Population Health is a unit of the Hunter New England Area Health Service, supported by funding from NSW Health through the Hunter Medical Research Institute. This article made use of the Campylobacter jejuni Multi Locus Sequence Typing website (http://pubmlst.org/campylobacter/) developed by Keith Jolley and Man-Suen Chan and sited at the University of Oxford [Reference Jolley, Chan and Maiden45]. The development of this site has been funded by the Wellcome Trust.
APPENDIX. Australian Campylobacter Subtyping Study Group (listed in alphabetical order):
Penny Adamson (Flinders Medical Centre, South Australia), Rosie Ashbolt (Public and Environmental Health Service, Department of Health and Human Services, Hobart), Kellie Cheung (Institute of Clinical Pathology and Medical Research, Westmead, New South Wales), Barry Combs (Department of Human Services, Adelaide, South Australia), Craig Dalton (Hunter New England Population Health, Newcastle, New South Wales), Steve Djordjevic (Elizabeth Macarthur Agricultural Institute, Camden, New South Wales), Robyn Doyle (Institute of Medical and Veterinary Science, Adelaide, South Australia), John Ferguson (Hunter New England Health Service, Newcastle, New South Wales), Lyn Gilbert (Institute of Clinical Pathology and Medical Research, Westmead, New South Wales), Rod Givney (Department of Human Services, Adelaide, South Australia), David Gordon (Flinders Medical Centre, Bedford Park, South Australia), Joy Gregory (Department of Human Services, Melbourne, Victoria), Geoff Hogg (Microbiological Diagnostic Unit, University of Melbourne, Parkville, Victoria), Tim Inglis (Division of Microbiology & Infectious Diseases, PathWest, Nedlands, Western Australia), Peter Jelfs (Institute of Clinical Pathology and Medical Research, Westmead, New South Wales), Martyn Kirk (Department of Health and Ageing, Canberra, Australian Capital Territory), Karin Lalor (Department of Human Services, Melbourne, Victoria), Jan Lanser (Institute of Clinical Pathology and Medical Research, Westmead, New South Wales), Lance Mickan (Institute of Medical and Veterinary Science, Adelaide, South Australia), Lyn O'Reilly (Division of Microbiology & Infectious Diseases, PathWest, Nedlands, Western Australia), Rosa Rios (Microbiological Diagnostic Unit, Parkville, Victoria), Minda Sarna (Department of Health, Perth, Western Australia), Hemant Sharma (Hunter New England Health Service, Newcastle New South Wales), Helen Smith (Queensland Health Scientific Services, Coopers Plains, Queensland), Leanne Unicomb (Hunter New England Population Health and National Centre for Epidemiology and Population Health, Australian National University, Canberra, ACT), Mary Valcanis (Microbiological Diagnostic Unit, University of Melbourne, Parkville, Victoria).
DECLARATION OF INTEREST
None.
