Materials and Methods

Case ascertainmentAll diagnostic clinical laboratories in Hawaii routinely culture diarrheal stool specimens for Campylobacter. These laboratories were requested to report any isolations of Campylobacter during the period 1 May 1998 through 15 September 1998 to the Hawaii Department of Health, and isolates were forwarded to the Hawaii State Laboratories Division for identification of species [8]

A case patient was defined as a person with diarrheal illness from whom Campylobacter was isolated from a stool specimen. Case patients were excluded if they did not speak English, had traveled outside Hawaii in the 28 days before illness, had chronic diarrhea (defined as >10 days of illness before stool specimen culture), were a household contact of a case patient previously identified, or were part of a recognized outbreak of C. jejuni infections

Case-control studyOne age- and telephone exchange–matched control subject was selected for each case patient by sequential addition or subtraction of 1 to the patient’s home telephone number. Potential control subjects were excluded if they had histories of diarrhea, were unable to communicate in English, were not present in Hawaii during the exposure period of the case patient, or were household contacts of a person with confirmed Campylobacter infection

All case patients and control subjects were interviewed by telephone, using a standardized questionnaire, regarding demographics, medical history, food consumption and handling practices, and other exposures associated with Campylobacter infection in previous investigations. Patients were asked about potential exposures that occurred during the 7-day period before the illness began or the preceding 28-day period when inquiring about medications and illness. Control subjects were asked about the same time period for each question as the matched case patient. Case patients and matched control subjects were interviewed a median of 6 and 9 days, respectively, after the stool sample culture was obtained

Statistical analysesMatched univariate analyses for independent exposure variables were computed with Epi Info software (version 6.04; Centers for Disease Control and Prevention). Young children who did not consume any table food were excluded from univariate analyses examining table food exposures but were included in analyses of nondietary exposures (e.g., contact with animals). Conditional logistic regression analyses, controlled for matching, were done by using LogXact 4 software (Cytel Software). All risk factors for which the P value for matched odds ratio was <.05 in univariate analyses were included in the initial regression model. Reverse stepwise procedures were done, and risk factors for which the P value for the adjusted odds ratio was >.1 were removed from the model

Results

Case patient enrollmentOf the 338 Campylobacter infections reported during the study period, 211 (62%) patients were enrolled in the case-control study. Reasons that patients were not enrolled included refusal or failure to finish the interview (6% of all reported infections), inability to locate and interview the patient within 21 days of stool culture report (9%), failure to enroll a matched control subject within 28 days (4%), or the case patient having met one of several exclusion criteria, including having been outside Hawaii in the 28 days before onset (9%), diarrhea for >10 days before stool culture (5%), or other (5%). There were no significant differences between enrolled case patients and those not enrolled with regard to age group, sex, or county of residence, and the ethnic distribution of enrolled case patients reflected that of Hawaii’s population

A total of 199 (94%) of the isolates from enrolled case patients were identified as C. jejuni. Isolates were unavailable for the remaining 12 (6%) patients

Case-control studyThere were no significant differences in composition of the patient and control subject groups with regard to race or ethnicity, education level, household size, and annual household income, although patients were more likely to be male than were control subjects (P<.05)

The proportion of case patients and control subjects who reported exposures to various food and animals is shown in table 1. In univariate analyses, several exposures were significantly associated with illness, including eating chicken prepared “outside the home,” which was defined as any chicken that was not prepared at the respondent’s household

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Table 1

Frequency of dietary, environmental, and medication exposures or health conditions reported by 211 patients with confirmed Campylobacter-associated diarrhea and matched control subjects, Hawaii, 1998

Eating chicken prepared by a commercial food establishment (a subset of eating any chicken outside the home) was also significantly associated with illness. The mean number of restaurants patronized by patients during the exposure period (χ̅=3.5) was not significantly greater (P=.14) than that for control subjects (χ̅=3.0)

Consumption of chicken prepared in the home, beef jerky, or steak was found to be inversely associated with illness. These exposures also were inversely associated with other exposures, such as eating chicken outside the home (P<.05)

Table 1 also shows the number of persons who self-reported certain medical conditions or pharmaceutical exposures. Nineteen patients and 10 control subjects reported taking antibiotics at any time in the 4 weeks before the case patients’ illness onset date (P<.05). Patients most frequently reported taking amoxicillin (n=7), followed by cephalexin (n=3), cefaclor (n=2), ciprofloxacin (n=1), erythromycin (n=1), minocycline (n=1), tetracycline (n=1), penicillin (n=1), and loracarbef (n=1); one patient could not recall the specific antibiotic taken. Amoxicillin was also the antibiotic most frequently taken by control subjects (n=5) during the exposure period, followed by penicillin (n=2), cefaclor (n=1), and an unknown antibiotic (n=2)

A total of 28 patients and 17 control subjects reported taking medications to reduce the stomach acidity (e.g., over-the-counter antacids, H₂-blockers, or proton-pump inhibitors) at any time in the 4 weeks before the case patients’ illness onset date (P<.05)

The results of the matched logistic regression model, controlling for sex, are presented in table 2. Consuming commercially prepared chicken or antibiotics remained significant independent predictors of illness with all risk exposures significant in univariate analyses included in the model and also in the model created when those risk factors for which the P value was >.1 were removed. The population-attributable fraction was estimated to be 18% for commercially prepared chicken and 10% for prior antibiotic use

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

Table 2

Results of conditional regression analysis that controlled for sex and matched study design among 211 patients with confirmed Campylobacter-associated diarrhea and matched control subjects, Hawaii, 1998

Discussion

To our knowledge, this is the first report to suggest that prior antibiotic exposure may be a risk factor for C. jejuni infection and the first reported association between Campylobacter infection and commercially prepared chicken in the United States

The observation that Campylobacter infection appears to be associated with prior use of antibiotics for treatment of an unrelated illness is very intriguing. Previous investigations have identified consuming antibiotics as a risk for Salmonella infection, and 2 distinct hypotheses have been proposed to explain this phenomenon [9, 10]. The first is that antibiotics provide a selective advantage to drug-resistant organisms by lowering the infectious dose required to produce illness or by facilitating conversion of asymptomatic colonization into clinical illness. The second proposed explanation is that persons who take antibiotics may have increased risk of enteric infection because of prolonged alteration of the colonic bacterial flora, which results in decreased resistance to infection when they are subsequently exposed [9]. Unfortunately, we were not able to further investigate these hypotheses in our study because we did not record the exact dates of prior antibiotic consumption and because the isolates were not subjected to antimicrobial sensitivity testing. We note, though, that 89% (17/19) of the patients had discontinued or completed their antibiotics before onset of their C. jejuni illness

The association observed between C. jejuni infection and commercially prepared chicken is important because this exposure is common and because most state or local health departments have the authority to regulate foodhandling practices within food establishments. Current strategies to reduce the incidence of foodborne Campylobacter infections include educating foodhandlers on the need to cook poultry thoroughly, to keep raw and cooked foods separate, and to avoid recontamination of the poultry after cooking. Because the extent to which staff members in commercial establishments adhere to these recommendations is likely to be variable, public health agencies must be committed to enforcing compliance with the Food Code [11]. In addition, industry should give serious consideration to eliminating Campylobacter contamination from poultry before sale by using ionizing irradiation [12]

Most case-control studies have limitations, and ours is no exception. First, we did not enroll every patient reported during the study period, and differential enrollment theoretically could have introduced bias. Examining the available data does not suggest, however, that there were substantial demographic differences between patients who were enrolled and those who were not. Second, it is possible that patients, as a result of their illness, may have been better able to remember exposures than were control subjects, which resulted in recall bias. It is difficult to postulate credibly, however, that recall bias could account for the identification of novel, yet plausible, risk exposures such as prior antibiotic use, as was found in our study. Third, we collected only a limited amount of information on the extent to which survey respondents routinely accessed health care services. If people being treated for an unrelated illness are more likely to seek care should they develop diarrhea than those not under treatment, it is possible that persons under a physician’s care (and perhaps receiving antibiotics) are overrepresented among diagnosed Campylobacter infections. Data from our study, albeit limited, do not indicate that persons under a physician’s care for any of several major health conditions, including diabetes, immunosuppression, or cancer, were significantly overrepresented among the case patients; however, future investigations should examine this hypothesis with greater rigor by documenting the proportion of respondents under a physician’s care for all causes. Finally, because it is a well-recognized source of C. jejuni the food histories pertaining to poultry were intentionally very detailed. Resources did not permit obtaining such comprehensive information on all other food items included in the questionnaire, and it is possible that associations between illness and other food items were missed as a result. Differential ascertainment of details in the food histories also means that comparisons among food items in logistic regression analyses should be interpreted with caution. Two other studies, however, both from New Zealand, have also found an association between Campylobacter illness and commercially prepared chicken [13, 14]. In our opinion, the agreement of these separate studies conducted by different investigators and across countries, greatly reduces the likelihood that the association is spurious

Although our study did not identify any unique, indigenous, food preference or handling practices that can explain the high incidence of Campylobacter infection in Hawaii, they may still exist. With amusement and a touch of irony, we note that other investigators have recently identified stir-fried chicken as the source of a Campylobacter outbreak at a “Hawaiian theme restaurant” in Wales [15]

In summary, eating commercially prepared chicken and taking antibiotics for an unrelated illness both were significantly associated with C. jejuni infection in Hawaii. These findings are based on relatively limited data and should be considered promising but preliminary. Additional work is needed to elucidate more fully the risk factors for C. jejuni so that appropriate public health intervention strategies can be developed and implemented