From October 2000 to September 2001, the second case-control study on acute sporadic human campylobacteriosis was conducted in Denmark. Participation in the study was voluntary and required written consent. Three groups were interviewed (computer-aided telephone interviews). The first included campylobacteriosis (CB) patients: persons with laboratory-confirmed campylobacteriosis. The second included healthy controls matched to CB patients by age, sex, and geography (6 controls per CB patient were randomly selected through the Danish Civil Registry system after receipt of a signed consent form from a CB patient). Eligible controls received a questionnaire and a consent form by mail. The time between disease onset for the CB patient and the time of interview of controls was sought diminished (mean 27 days). The third group included non-CB bacterial gastroenteritis patients (non-CB controls): patients whose specimens were culture-positive for other zoonotic bacterial infections (mainly Salmonella spp.).

Twice a week, 2–4 CB patients and 3 non-CB patients per CB patient were selected in 6 of 16 Danish counties among patients identified the previous week. An approximate match of non-CB controls to CB patients by onset date was obtained (mean 5 days apart). Children <1 year of age were omitted. Recruitment of patients increased during summer when the incidence was higher. A total of 272 persons with Campylobacter infection, 786 non-CB controls, and 2,403 healthy controls were invited for the study. The response rates were 50% for CB cases and non-CB controls and 22% for the healthy controls, respectively.

The questionnaire sought data on the following subjects: symptoms; other diseases; use of medications; use of vitamins; consumption of meat (including type, storing [frozen vs. not frozen], and handling); use of barbecue grill; consumption of rice and pasta, fruit, vegetables, cereals, bread, milk, milk products, spices and herbs, and organic products; cooking; kitchen hygiene; vacation or travel experiences; contact with ill persons; information on drinking water, swimming, household, place of residence, and whether a summer house was used; and respondent's occupation, hobbies, and socioeconomic data. More than 350 original or recoded variables were tested in the analysis.

Three analyses were conducted: A) 107 CB patients versus 178 matched healthy controls (including travel-related cases); B) 74 CB patients (domestically acquired only) versus 114 matched healthy controls; and C) 141 CB patients versus 386 non-CB controls matched to patients only on time of disease onset. Initially, variables in analyses A and B were tested by univariate conditional logistic regression (PROC PHREG, SAS Institute, Cary, NC, USA [9]). Variables with p_(univariate)<0.30 and other relevant variables were selected for multivariate conditional logistic regression analysis. Variables with p<0.01 were kept in the final models. In analysis B, 2-factor interactions between variables in final model, match variables, and other interactions relevant to the hypotheses were tested; the population attributable risk (PAR) was then estimated (10). In analysis C, univariate logistic regression (PROC GENMOD, SAS Institute [9]) was used for screening of effect of variables. Effect modification of covariates (age, sex, geography, and season) was tested in analysis C. As the modifying effect was negligible, the results from analyses without covariates is presented (Table 1).

Analysis A (full dataset) identified that travel to southern Europe (odds ratio [OR] 15.81) and outside Europe (16/107 patients exposed vs. 1/178 controls) was associated with Campylobacter infection, whereas travel to other parts of Europe was more common in controls than in patients (OR 0.068). Other risk factors identified in analysis A are listed in Tables 1 and 2. Analysis B (domestic cases only) identified eating chicken, bought fresh and not frozen in the home, to be the only significant risk factor for campylobacteriosis (OR 5.80). Some exposures, including pork prepared in large pieces (OR 0.15), eating apples or pears (OR 0.21), eating raw vegetables daily (OR 0.24), and days off (besides weekends) in the week before onset (OR 0.23) were more common in controls than patients (p<0.01). Analysis C (CB vs. non-CB patients) found that, among the variables with p<0.05 in analysis A or B, only eating chicken, bought fresh and not frozen in the home, was significantly more associated with Campylobacter infections than with other bacterial gastrointestinal infections. Contrary to this finding, travel to central and northern Europe and eating pork prepared in large pieces were less associated with Campylobacter infection than with other infections. The domestic PAR from chicken bought unfrozen was 23.8% (95% confidence interval 7.98–52.9).

Only in model B (domestic patients) were 2-factor interactions examined. The risk from fresh chicken was significantly increased (p<0.05) in summer (vs. winter) and when preparing whole chicken (vs. cuts). The risk was reduced (p<0.05) by frequently eating fruits, raw vegetables, high-fiber cereals, vitamins (p = 0.050), and acidified milk products (p = 0.070). Eating turkey bought fresh and chicken in general interacted borderline significantly with season (chicken: higher risk in summer [p = 0.078], turkey: higher risk in winter [p = 0.056]). A borderline significant interaction between risk from chicken cuts and barbecuing was found (p = 0.0502). Finally, the apparent protection from eating apples or pears was stronger in the cold season (p = 0.043).

We found that the main domestic risk factor for campylobacteriosis is eating chicken meat that is bought fresh and subsequently not frozen in the home. Eating other poultry meat products and eating previously frozen chicken meat were borderline significant risk factors.

Adding the case-case approach to the risk factor study (CB patients vs. non-CB patients) was expected to highlight risk factors or potentially protective factors, which are specific for campylobacteriosis. Only exposure to unfrozen chicken remained a significant risk factor for campylobacteriosis in the case-case study. The study findings strongly support the contribution of fresh poultry specifically as a source of human campylobacteriosis. In contrast, true common factors for both case groups were expected to be reduced or disappear. Also the apparent effect of factors associated with willingness to participate as a control in the case-control studies was expected to be eliminated. Several significant risk factors from the case-control studies were insignificant or markedly reduced in the case-case study (e.g., apparently protective factors [certain fruits and vegetables] and risk factors [travel, certain types of bread, and fresh turkey]).

The results of the present study are consistent with the hypothesis that a marked increase in the consumption of fresh chicken has been a major driving force behind the increasing incidence of human campylobacteriosis in Denmark during the 1990s. Bacteriologic investigation of fresh and frozen chicken collected at retail outlets in Denmark has shown that the number of viable Campylobacter bacteria in fresh samples exceeds that of previously frozen chicken. In a survey of chicken meat in retail stores, 194 (79.8%) of 243 samples of frozen chicken harbored <0.4 thermophilic Campylobacter bacteria per gram, whereas 134 (46.4%) of 289 samples of fresh chicken were below this level (11). This result is because the freezing process reduces the number of viable Campylobacter organisms. In the 1990s, the national consumption of poultry meat increased by ≈40% (1991: 63,900 tons, 1998: 93,200 tons) (12). The increase was observed for almost all types of chicken and turkey products but most markedly in fresh cuts. In the same period, the incidence of campylobacteriosis increased by >400%, from 20 to 86 cases per 100,000 inhabitants. The bacteriologic data, which show higher loads of Campylobacter in fresh poultry, suggest that the exposures to Campylobacter spp. have increased much more than the general increase in poultry consumption and thus explains why the increase in human disease incidence has exceeded the increase in poultry consumption.

The Danish broiler industry, in collaboration with governmental institutions, introduced a voluntary control program in 2002–2003, whereby among other initiatives, flocks of chicken are tested for Campylobacter spp. immediately before slaughter (13). Positive flocks are, to the extent that doing so is logistically feasible, used to produce frozen products, whereas Campylobacter-free flocks are primarily used to produce fresh chicken. In the winter, the prevalence of Campylobacter-free flocks is sufficiently elevated to enable a near complete separation, but in the summer, when the flock prevalence is high, Campylobacter-positive flocks are also included in the fresh product line to some extent. In 2002, the incidence of human campylobacteriosis dropped 5% from the year before and in 2003 another 19%, possibly as a result of the control program (Figure). Thus, the program appears to have a positive effect, which lends further support to the hypothesis.

In conclusion, the results of this study support the hypothesis that fresh chicken is the main risk factor for domestically acquired campylobacteriosis in Denmark. This risk is significantly increased in the summer, when the incidence of infected broiler flocks peak, and when whole chickens are prepared. Travel to southern Europe and travel outside Europe, respectively, were also significant risk factors. The marked increase in consumption of fresh poultry during the 1990s may explain, at least in part, the increased incidence of human campylobacteriosis in Denmark in this period.