From 1989 to 1990, a national county-based survey of antimicrobial susceptibility among Campylobacter isolates was conducted. Sentinel clinical laboratories in 19 counties participated. The methods of this survey are described elsewhere (19,20). Briefly, the first five sporadic Campylobacter isolates identified each month were forwarded to the Centers for Disease Control and Prevention (CDC). Patients with Campylobacter infection were interviewed with a standard questionnaire, which included information about clinical illness and exposures (i.e., food, animal, and foreign travel) during the 2 weeks before illness onset. Isolates were determined to be Campylobacter by dark-field microscopic examination and hippurate hydrolysis (20–22). Hippurate-positive isolates were considered C. jejuni. All isolates with questionnaires received during the first 4 months of the study underwent susceptibility testing. Because of a shortage of reagents, a random sample of 50% of isolates with completed questionnaires received during the last 8 months of the study was further characterized. Isolates were tested for susceptibility to azithromycin, chloramphenicol, ciprofloxacin, clindamycin, erythromycin, gentamicin, nalidixic acid, and tetracycline by using broth microdilution methods (19). In 2003, a retrospective analysis of the hippurate-negative isolates was completed; these isolates were speciated by using methods described below.

NARMS for enteric bacteria is a collaboration between CDC, the Food and Drug Administration, and state and local health departments. The system monitors patterns of antimicrobial drug resistance. NARMS methods are described in detail elsewhere (23).

In brief, isolates were tested for viability, confirmed as Campylobacter, and identified to the species level by using the hippurate hydrolysis test according to published methods (21,22). Hippurate-negative Campylobacter in which the hippuricase gene could be detected by polymerase chain reaction (PCR) were identified as C. jejuni (24). Isolates that tested negative for the hippuricase gene but positive for a C. coli-specific ceuE sequence were identified as C. coli (25). Isolates that could not be identified as either C. jejuni or C. coli by these PCR assays were referred to the National Campylobacter Reference Laboratory at CDC for identification with genotypic (e.g., 16S rRNA sequencing) and phenotypic methods (21).

Isolates were tested with the E-test system (AB BIODISK, Solna, Sweden) to determine MICs for six antimicrobial agents: chloramphenicol, ciprofloxacin, clindamycin, erythromycin, nalidixic acid, and tetracycline. Beginning in 1998, azithromycin and gentamicin were also included. When available, National Committee for Clinical Laboratory Standards interpretive criteria for Enterobacteriaceae MICs were used; ciprofloxacin resistance was defined as MIC >4 µg/mL, and erythromycin resistance was defined as MIC >8 µg/mL (26). Multidrug resistance was defined as resistance to two or more of the original six antimicrobial agents.

We used a multivariable logistic regression model to assess changes in the proportion of isolates with antimicrobial drug resistance from 1997 through 2001 because the population under surveillance more than doubled from 1997 to 2001, and substantial site-to-site variation in prevalence of antimicrobial drug resistance was identified (i.e., uncertainty was found in the denominators for calculating rates). The model was for antimicrobial drug resistance as a function of year and included main effects adjustments for age categories and site-to-site variation in prevalence. Within the available data, site by year interaction was not a significant factor but because the catchment areas expanded, the hypothesis of site by year interaction could not be fully tested.

Using NARMS isolates, we conducted a retrospective case-comparison study in four NARMS sites (California, Connecticut, Georgia, and Oregon). Persons with ciprofloxacin-resistant (CipR) Campylobacter infection identified in 1997 were compared with persons in whom the diagnosis of ciprofloxacin-sensitive (CipS) Campylobacter infection was made that same year. We compared up to two CipS cases for each CipR case and matched cases by geographic site and date of stool specimen collection. All case-patients were interviewed by telephone, usually within 8 weeks of their illness onset, about demographics, clinical information, and exposures (e.g., antimicrobial drug use in the 4 weeks before illness onset, foreign travel, and consumption of poultry and raw milk in the 7 days before illness onset) with a standardized questionnaire.

Three NARMS-participating state health departments (Georgia, Maryland, and Minnesota) participated in a survey of retail chicken products. From January to June 1999, each site purchased a convenience sample of 10 whole broiler chickens per month from supermarkets located within the state. State public health laboratories at each site tested the samples for Campylobacter. Carcass rinse samples were centrifuged, and pellets were incubated in enrichment broth and plated onto Campylobacter blood agar plates according to methods published elsewhere (27); neither media contained quinolone or macrolide antimicrobial agents. Campylobacter isolates were forwarded to CDC for species identification and antimicrobial susceptibility testing according to NARMS methods.

Two hundred ninety-eight patients were interviewed, and their Campylobacter isolates were tested. Of these isolates, 289 (97%) were C. jejuni, 8 (3%) were C. coli, and 1 (0.3%) was a C. lari. None were resistant to ciprofloxacin, and 3 (1%) of 294 were resistant to nalidixic acid (MIC > 32 µg/mL); 1 isolate was C. lari, which is inherently resistant to nalidixic acid (28), and 2 were C. jejuni. The C. lari isolate was resistant to ofloxacin (MIC = 8 µg/mL), intermediately resistant to norfloxacin (MIC = 8 µg/mL) but susceptible to ciprofloxacin (MIC = 2 µg/mL).¹ The two nalidixic acid-resistant C. jejuni isolates were susceptible to ciprofloxacin (MIC = 0.5 µg/mL) and norfloxacin and ofloxacin (MIC < 2 µg/mL). The proportion of the isolates resistant to tetracycline was 42% (124/295). The resistant proportion for the other antimicrobial agents tested were as follows: erythromycin 3% (8/295), clindamycin 2% (6/295), azithromycin 2% (5/294), chloramphenicol 0% (0/295), and gentamicin 0% (0/295). Travel history was available for 296 patients with Campylobacter infection; 23 (8%) patients traveled outside of the United States in the week before illness onset. Of the persons with available information, 32 (11%) of 295 had taken an antimicrobial agent in the 30 days before illness onset, 46 (15%) of 298 were hospitalized, and 241 (81%) of 297 were treated with an antimicrobial agent for their illness. Among the 234 persons for whom treatment data were available, the most common agents used for treatment were erythromycin (62%), ciprofloxacin (19%), and trimethoprim-sulfamethoxazole (5%). Of the three patients with nalidixic acid-resistant infections, none traveled outside the United States, and none were treated with a quinolone or fluoroquinolone in the month before illness.

From 1997 to 2001, a total of 1,932 presumptive Campylobacter isolates were received at CDC through NARMS; 193 (10%) were excluded because they were not viable, 104 (5%) were not in accordance with the one-a-week sampling method, 39 (2%) were determined not to be Campylobacter, 22 (1%) were duplicates, and 21 (1%) were contaminated cultures. Of the 1,553 (80%) isolates further characterized and included in this analysis, 1,471 (95%) were C. jejuni, 63 (4%) were C. coli, 7 (0.4%) were C. upsaliensis, 5 (0.3%) were C. fetus, 2 (0.1%) were C. lari, and 5 (0.3%) were undetermined (i.e., determination by 16S study did not identify a species). Forty-five percent of case-patients were female; the median age was 33 years (range <1–96). Among 1,439 isolates with known source of specimen collection, 1,426 (99%) were from stool samples, and 13 (1%) were from blood samples. Among blood isolates, eight were C. jejuni, two were C. fetus, two were C. upsaliensis, and one was C. lari.

The results of susceptibility testing among Campylobacter isolates by species are shown in Table 1. Resistance to ciprofloxacin among all Campylobacter isolates was 13% in 1997 and 19% in 2001. Resistance to erythromycin among all Campylobacter isolates was 2% in 1997 and 2% in 2001. The results of antimicrobial susceptibility testing by year for isolates of the most common species, C. jejuni, are shown in Table 2.

The prevalence of ciprofloxacin-resistant Campylobacter ranged from 0% (0/14) in Tennessee in 1999 to 26% (14/53) in Georgia in 2001. By using a multivariate logistic regression model and controlling for age and site-to-site variation in prevalence, the proportion of all Campylobacter isolates resistant to ciprofloxacin and nalidixic acid in 2001 was significantly higher than the proportion of isolates resistant to ciprofloxacin in 1997 (data only shown for ciprofloxacin in Table 3). The remaining antimicrobial drugs had no statistically significant change in resistance over time (data not shown).

Fifty-one percent of Campylobacter isolates were resistant to >1 drug, 18% were resistant to >2 drugs, and 10% were resistant to >3 drugs. The most common multidrug resistance (i.e., >2 drugs) pattern included ciprofloxacin, nalidixic acid, and tetracycline.

Sixteen (57%) of 28 ciprofloxacin-resistant Campylobacter (CipR) case-patients and 31 ciprofloxacin-sensitive (CipS) case-patients were interviewed. The median age was 46 years (range 9–76 years) for CipR patients and 24 years (range 1–87 years) for CipS patients (Wilcoxon rank-sum, p = 0.08). CipR patients did not differ significantly from CipS patients in terms of sex (40% vs. 42% female, p = 0.1), race (87% vs. 77% white, p = 0.08), and place of residence (87% vs. 61% urban/suburban areas, p = 0.06). Five (31%) CipR patients were hospitalized for gastroenteritis compared with 1 (3%) CipS patient (matched odds ratio [mOR] = 13.6, 95% confidence interval [CI] 1.4 to 130.1). Eight (57%) CipR case-patients reported having bloody diarrhea compared with eight (30%) CipS patients (mOR = 3.2, 95% CI 0.8 to 12.1). Seven (44%) of 16 CipR patients compared with 1 (3%) of 31 CipS patients traveled to a foreign country in the 7 days before illness onset (mOR = 23.3, 95% CI 2.5 to 215.6); 5 (71%) of 7 CipR patients traveled to Europe while 1 CipS patient traveled to the Caribbean. Among all case-patients, 35 of 47 reported treatment with an antimicrobial agent for their illness. Of those who recalled the name of the antimicrobial drug, 75% reported taking a fluoroquinolone, 16% reported taking a macrolide, and 8% took trimethoprim-sulfamethoxazole (TMP/SMX). One CipR patient and one CipS patient took fluoroquinolones between onset of illness and collection of stool specimens. Among the eight CipR patients who did not travel and did not take fluoroquinolones between illness onset and stool specimen collection, seven (87%) consumed poultry in the 7 days before infection; this finding was not statistically different from that in CipS patients. No other exposures were significantly associated with ciprofloxacin-resistant infection, including having pets, drinking raw milk, or being exposed to a farm (data not shown).

Among the 180 retail chicken products purchased, representing 18 domestic brand names from 22 grocery stores, Campylobacter was isolated from 80 (44%) samples. Sixty-two (77%) were C. jejuni, 16 (20%) were C. coli, and 2 (2%) were undetermined (i.e., determination by 16S study did not identify a species). The prevalence of Campylobacter isolated was 33% (20 of 60) in Georgia, 37% (22 of 60) in Maryland, and 63% (38 of 60) in Minnesota. This difference among sites was in part due to the difference in isolation rates of C. coli; 14 (87%) of the 16 C. coli isolates came from retail chickens purchased in Minnesota. A ciprofloxacin-resistant strain of Campylobacter was identified in 10% of the 180 retail chicken products tested, and an erythromycin-resistant strain was identified in 2% of chicken products (Table 4). The distribution of ciprofloxacin MICs in Campylobacter species of retail chicken and human isolates was similar. For both human and poultry Campylobacter isolates, MICs were predominantly <0.5 or >32 µg/mL with few intermediate phenotypes (Figure 1).