Abstract
Background.Campylobacter bacteremia is uncommon. The influence of underlying conditions and of the impact of antibiotics on infection outcome are not known.
Methods.From January 2000 through December 2004, 183 episodes of Campylobacter bacteremia were identified in 23 hospitals in the Paris, France, area. The medical records were reviewed. Characteristics of bacteremia due to Campylobacter fetus and to other Campylobacter species were compared. Logistic regression analysis was performed to identify risk factors for fatal outcome within 30 days.
Results.Most affected patients were elderly or immunocompromised. C. fetus was the most commonly identified species (in 53% of patients). The main underlying conditions were liver disease (39%) and cancer (38%). The main clinical manifestations were diarrhea (33%) and skin infection (16%). Twenty-seven patients (15%) died within 30 days. Compared with patients with bacteremia due to other Campylobacter species, patients with C. fetus bacteremia were older (mean age, 69.5 years vs. 55.6 years; P<.001) and were more likely to have cellulitis (19% vs. 7%; P=.03), endovascular infection (13% vs. 1%; P=.007), or infection associated with a medical device (7% vs. 0%; P=.02). Independent risk factors for death were cancer (odds ratio [OR], 5.1; 95% confidence interval [CI], 1.2–20.8) and asymptomatic infection (OR, 6.7; 95% CI, 1.5–29.4) for C. fetus bacteremia, the absence of prescription of appropriate antibiotics (OR, 12.2; 95% CI, 0.9–157.5), and prescription of third-generation cephalosporins (OR, 10.2; 95% CI, 1.9–53.7) for bacteremia caused by other species.
Conclusions.Campylobacter bacteremia occurs mainly in immunocompromised patients. Clinical features and risk factors of death differ by infection species.
Campylobacter jejuni subsp. jejuni is the main bacterial cause of enteroinvasive diarrhea [1–4]. As compared with Salmonella infections, C. jejuni infection is rarely complicated by bacteremia or extraintestinal localization [2, 5]. In contrast to C. jejuni, Campylobacter fetus subsp. fetus is usually isolated from blood samples and is less frequently associated with enteritis [5, 6]. The clinical features of Campylobacter bacteremia were reported in several series published during 1993–1998 [7–11], and a retrospective study of C. fetus antimicrobial susceptibility was reported in 2003 [12]. The number of cases reported in published series was quite small, even though the study periods were long. None of these reports examined factors associated with fatal outcome. In the past decade, several factors may have led to an increased incidence of Campylobacter extraintestinal localization like bacteremia, and Campylobacter susceptibility may have been affected by misuse of antibiotics, especially fluoroquinolones, in both veterinary and human medicine.
Therefore, we conducted a retrospective study of Campylobacter bacteremia diagnosed in 23 hospitals in the Paris, France, area, from January 2000 through December 2004, to increase our knowledge about the clinical, bacteriological, and therapeutic characteristics of the affected patients and to identify risk factors for death within 30 days after infection onset.
Patients and Methods
Definition of cases.Cases of Campylobacter bacteremia occurring in patients treated in 23 hospitals in the Paris area between 1 January 2000 and 31 December 2004 were collected from the relevant microbiology laboratories.
Patients and data collection.The medical records of the affected patients were reviewed. The following data were systematically extracted: demographic characteristics, clinical manifestations associated with bacteremia, underlying conditions, whether there was fatal outcome within 30 days, and microbiological data (i.e., identification to the species level, results of concomitant stool culture, and susceptibility to amoxicillin, amoxicillin–clavulanic acid, erythromycin, gentamicin, and fluoroquinolones). We also extracted the start and end dates of treatment with individual antibiotics. The nature of the empirical regimen and switches that were based on the results of culture of blood samples and susceptibility testing were recorded. The antibiotic treatment was considered to have been appropriate if the strain was susceptible to at least 1 of the drugs prescribed. C. fetus isolates were assumed to be susceptible to cefotaxime and ceftriaxone, whereas C. jejuni, Campylobacter coli, and Campylobacter lari isolates were assumed to be resistant to all third-generation cephalosporins [12–15]. All isolates were assumed to be resistant to ticarcillin and piperacillin [13, 15, 16], and treatment with these antibiotics was considered to be inappropriate.
Laboratory methods.All laboratories used continually monitored noninvasive blood culture systems—namely, BacT/Alert (Organon Technika), Bactec (Becton Dickinson) or Vital (bioMérieux). Each blood culture set included an aerobic and an anaerobic bottle that were incubated for 5 days at 37°C. Curved or spiral-shaped gram-negative rods were identified as Campylobacter species and were speciated with conventional phenotypic methods (i.e., biochemical tests, growth at 25°C and 42°C, and susceptibility to nalidixic acid and cephalotin). Susceptibility testing was based on the disk-diffusion method, as recommended by the Antibiogram Committee of the French Society of Microbiology. The species identification of some isolates was not reported by the laboratories, but the cephalotin and nalidixic acid susceptibility results were used to tentatively identify these isolates as C. fetus or as members of the group comprising C. jejuni, C. coli, and C. lari.
Statistical analysis.Descriptive statistics were expressed as percentages for categorical variables and as mean ± SD and median and interquartile range for continuous variables, as appropriate. Demographic and clinical characteristics (table 1) were compared between patients with bacteremia due to C. fetus and patients with bacteremia due to other Campylobacter species, by use of the χ2 test for categorical variables or the Mann-Whitney U test for continuous variables, after exclusion of patients whose isolates were not speciated. Associations between fatal outcome within 30 days and candidate variables were assessed as ORs and 95% CIs. The tested variables are listed in table 2. ORs determined by univariate analysis were calculated by means of simple regression analysis. A multiple logistic regression model (Hosmer and Lemeschow backward procedure) was constructed using variables that were statistically significantly associated with fatal outcome within 30 days (P<.10) in univariate analyses, to calculate adjusted ORs. All results with P values <.05 were considered to be statistically significant. Data on patients who had died >30 days after presentation of an unknown cause were consequently removed from the analysis, because the impossibility of excluding an association between their death and the presence of factors related to bacteremia might have biased the identification of risk factors for fatal outcome. The SPSS software package, version 11 (SPSS), was used for statistical analysis.
Demographic characteristics, underlying conditions, and clinical features of patients with Campylobacter bacteremia due to Campylobacter fetus and to other identified isolates.
Demographic characteristics, underlying conditions, and clinical features of patients with Campylobacter bacteremia due to Campylobacter fetus and to other identified isolates.
Risk factors of death within 30 days after bacteremia diagnosis, by Campylobacter species.
Risk factors of death within 30 days after bacteremia diagnosis, by Campylobacter species.
Results
During the 5-year study period, 183 episodes of Campylobacter bacteremia were diagnosed in 178 patients (5 patients presented with recurrent episodes of bacteremia) in the 23 participating hospitals. Clinical characteristics and rates of resistance by species are listed in table 1. C. fetus was the most commonly identified species (94 [53%] of 178 patients). Among the 77 remaining non–C. fetus isolates (43%), precise identification to the species level was made as follows: 54 C. jejuni isolates, 16 C. coli isolates, and 2 C. lari isolates. Most cases involved male patients and elderly patients. Only 37 patients (21%) were not immunocompromised. Cancer (67 patients [38%]) and liver disease (69 patients [39%]) were the leading underlying conditions. Gastrointestinal tract (66 patients [37%]) and skin (28 patients [16%]) infections were the most frequent manifestations, but there was no symptom other than fever in 74 patients (42%). The rate of resistance to fluoroquinolones was 32%, whereas none of the isolates was resistant to imipenem, 1% were resistant to amoxicillin–clavulanic acid, and 3% were resistant to gentamicin.
Compared with patients with bacteremia due to other Campylobacter species, patients with C. fetus bacteremia were older, were more likely to be male, and were more likely to have both implanted medical devices and no identified cause of immunosuppression; they were more likely to have received a diagnosis of skin, endovascular, or medical device infection than were patients with bacteremia due to other Campylobacter species. C. fetus bacteremia tended to be less frequently associated with febrile diarrhea. Resistance to amoxicillin and to erythromycin was less frequent in C. fetus than in other species.
Three patients died >30 days after presentation. Twenty-seven (15%) of the remaining 175 patients died within 30 days after the date the first culture-positive blood sample was obtained. Death occurred after a mean (±SD) of 11.5±9 days and in <72 h in 8 cases. The risk of death was higher among patients who had received a smaller mean number of appropriate antibiotics (0.88 drugs for patients who died vs. 1.34 drugs for survivors; P<.001) and among those with no appropriate antibiotic treatment (87.5% vs. 12%; P<.001). None of the 12 patients who were treated with imipenem died within 30 days. Risk factors of death differed by Campylobacter species (table 2). In patients with C. fetus bacteremia, death was associated with cancer, splenectomy, lack of any clinical manifestations of bacteremia other than fever (in particular, the absence of diarrhea), and treatment with fluoroquinolones. In patients with bacteremia due to other Campylobacter species, the risk of death was higher among those who did not receive any appropriate antibiotics and among those who were prescribed a third-generation cephalosporin. Third-generation cephalosporins were deleterious, either as a treatment given after identification of Campylobacter infection or given alone or in association with fluoroquinolones or macrolides, before test results had confirmed the presence of Campylobacter species. Among patients who received treatment with a third-generation cephalosporin, mortality tended to be lower among those with C. fetus bacteremia than among those with bacteremia due to another Campylobacter species (18% vs. 62%; OR, 6.6; 95% CI, 0.84–87.72; P=.08). In multivariate analysis, factors that remained independently associated with death were cancer and isolated fever for patients with C. fetus bacteremia, the absence of treatment with appropriate antibiotics, and the prescription of third-generation cephalosporins for bacteremia due to other Campylobacter species (table 2).
Discussion
The study of this large series of 178 patients (183 episodes) emphasizes the differences between clinical presentation and prognostic factors by Campylobacter species. Campylobacter bacteremia occurred mainly in elderly and immunocompromised patients. The overall mortality rate was high. Patients with C. fetus bacteremia had more frequently received a diagnosis of skin, endovascular, or medical device infection, compared with patients with other Campylobacter bacteremia. The resistance rate to fluoroquinolone was 32%. Resistance to amoxicillin and to erythromycin was less frequent with C. fetus than with other Campylobacter species. Failure to administer appropriate antibiotics was strongly associated with fatal outcome. Independent risk factors for death were cancer and isolated fever among patients with C. fetus bacteremia, whereas risk factors were the absence of appropriate antibiotic treatment and the prescription of third-generation cephalosporins among patients with bacteremia due to other Campylobacter species.
Human Campylobacter infections are well documented [5]. C. jejuni is isolated mainly from fecal culture and causes acute enteroinvasive diarrhea, whereas bacteremia occurs in <1% of cases [5]. In contrast, C. fetus is isolated mainly by blood culture, but only rarely by stool culture [6, 12]. When stool specimens were available for culture for our patients, the cultures were less likely to isolate C. fetus than C. jejuni. However, this finding should be cautiously interpreted, because many laboratories do not routinely screen stool specimens for C. fetus. The presence of C. fetus in fecal specimens is, therefore, rare but probably also underestimated.
Most cases of Campylobacter bacteremia involved males and patients with underlying conditions, in keeping with previous reports [7–11, 17]. However, no cause of immunosuppression could be found in 21% of our patients; therefore, age itself could be a major risk factor.
C. fetus was the leading cause of bacteremia among the patients in our study, which represents the largest recent clinical series of C. fetus bacteremia since previous reports [12, 17, 18]. This is consistent with some studies [19, 20], but it is in striking contrast to the main recent clinical case reviews [7–11], in which C. jejuni and C. coli infections were far more frequent than they were among our patients. This difference could reflect improvements in recovery of these fastidious organisms through use of automated blood culture systems [21] since the beginning of our study or a preponderance of C. fetus in this population of elderly patients. The 30-day mortality rate associated with Campylobacter bacteremia was 15% in this series, similar to that observed in previous reports [8–10, 18].
Cellulitis was the leading extraintestinal manifestation of bacteremia, especially in patients with C. fetus infection. Cases of Campylobacter cellulitis have been reported elsewhere [9, 10, 17, 18, 22, 23]. Many of the patients with cellulitis received successful treatment with amoxicillin–clavulanic acid and were discharged from the hospital before culture results were available. The frequency of Campylobacter cellulitis could, therefore, be underestimated [22, 23]. Like some strains of Salmonella species, C. fetus has a predilection for the vascular endothelium [20] and for the site of medical devices. C. fetus has been known to infect prosthetic joints [24] and prosthetic heart valves [25]. In our series, all but 1 endovascular infection and all medical device–related infections were due to C. fetus. The infection rate was considered to be 20% among bacteremic patients with a medical device; but infection at the site of a medical device could have been underestimated and should therefore be carefully considered in patients with C. fetus bacteremia.
It is uncertain whether all episodes of Campylobacter bacteremia need to be treated: C. jejuni bacteremia might be transient in immunocompetent patients, and transient C. fetus bacteremia can resolve without antimicrobial chemotherapy [5, 17]. The present study clearly demonstrates the need for an appropriate antibiotic treatment in a population of elderly and immunocompromised patients: 88% of our patients who received no appropriate antibiotics died within 30 days after presentation.
However, there is no consensus on the optimal antibiotic regimen; to our knowledge, there have been no published controlled clinical trials and only anecdotal cases and clinical case reviews. Although our study was descriptive and not designed to assess antibiotic efficacy, several results are of particular interest with respect to treatment.
Third-generation cephalosporins are widely used for empirical treatment of community-acquired infections in immunocompromised patients and are considered to be alternatives to fluoroquinolones for enteroinvasive bacterial diarrhea. We observed a strong association between fatal outcome and prescription of a third-generation cephalosporin for patients with Campylobacter bacteremia due to species other than C. fetus, a potentially important finding. Disk-diffusion or Etest susceptibility data for third-generation cephalosporins were not available in our series. Several studies of Campylobacter susceptibility to cephalosporins have been reported over the past 30 years [12–16, 26–31]. With use of the breakpoints of the National Committee on Clinical Laboratory Standards, few strains of C. jejuni and C. coli seem to be susceptible to cefotaxime [13, 15]. In contrast, recent reports [12, 14] state that most C. fetus isolates were susceptible to cefotaxime. The bactericidal activity of cefotaxime was nonetheless weaker than that of ampicillin, gentamicin, and imipenem [30, 31]. These reports and our findings do not support the use of such third-generation cephalosporins for the treatment of bacteremia involving species other than C. fetus.
Campylobacter isolates are not regularly susceptible to penicillins [12–14, 16]. The β-lactamase enzyme commonly encountered in C. jejuni strains [13]—but rarely encountered in C. fetus strains [12]—is inhibited by clavulanic acid but not by tazobactam or sulbactam [13, 16]; this is why amoxicillin–clavulanic acid remains effective against almost all strains [26]. In recent reports from Quebec, all strains of C. jejuni were susceptible to amoxicillin–clavulanic acid, imipenem, and meropenem (MIC(90), ⩽0.06 µg/mL) [26], and all strains of C. fetus were susceptible to imipenem (MIC(90), ⩽0.06 µg/mL) [12]. Imipenem has been suggested for the treatment of Campylobacter bacteremia [5, 6, 23]. Imipenem and amoxicillin–clavulanic acid seem to be the preferential β-lactam agents for treatment of Campylobacter infection. Of note, none of the patients treated with imipenem in our study died.
Fluoroquinolones and macrolides remain the standard choice for empirical treatment of Campylobacter enteritis [5]. However, 32% and 8% of our bloodstream isolates were resistant to fluoroquinolones and erythromycin, respectively, in keeping with the results of a recent international survey [3]. Empirical treatment of C. fetus bacteremia with fluoroquinolones was associated with fatal outcome. This, together with the increasing frequency of resistance to this class of antibiotics [32, 33], prompts us to discourage the use of fluoroquinolones for empirical treatment of Campylobacter bacteremia.
Gentamicin has very low MICs for C. fetus [12, 14, 29–31], C. jejuni, and C. coli [28]. No gentamicin-resistant C. fetus strains were identified in a recent report from Quebec [12]. On the basis of clinical experience and microbiological data, many experts advocate the use of a regimen that includes gentamicin to treat severe Campylobacter bacteremia and endovascular infection [5, 8, 20].
Failure to administer appropriate antibiotics is strongly associated with fatal outcome, but several features make it difficult for the clinician to give an appropriate and prompt treatment at the time of presentation. First, Campylobacter remains an uncommon cause of bloodstream infection and is a fastidious bacterium. Second, clinical findings of Campylobacter bacteremia lack of specificity. Eventually, third-generation cephalosporins and fluoroquinolones, which are widely prescribed in this particular population of immunocompromised patients with fever, seem to be deleterious in treatment of Campylobacter bacteremia. Empirical use of fluoroquinolones is hampered by the high rate of resistance and by its association with an increased risk of death among patients with C. fetus bacteremia; conversely, the prescription of third-generation cephalosporins is deleterious in bacteremia that is attributable to Campylobacter species other than C. fetus. Therefore, we advocate that third-generation cephalosporins and fluoroquinolones be avoided as soon as Campylobacter species have been isolated from culture of blood samples, pending precise species identification and susceptibility pattern.
Other Members of the CAMPYL Study Group
Sophie Coignard (Laboratoire de Microbiologie, Hôtel Dieu, Paris), Yannick Costa (Laboratoire de Microbiologie, Hôpital de Lagny-Marne la Vallée, Lagny), Frederic Faibis (Laboratoire de Microbiologie, Hôpital de Meaux, Meaux), Agnès Ferroni (Laboratoire de Microbiologie, Hôpital Necker, Paris), Beate Heym (Laboratoire de Bactériologie, Hôpital Ambroise Paré, Boulogne), Patrick Honderlic (Laboratoire de Bactériologie, Hôpital Foch, Suresnes), Françoise Jaureguy (Laboratoire de Microbiologie, Hôpital Avicenne, Bobigny), Lea Lebrun (Laboratoire de Microbiologie, Hôpital Antoine Béclère, Clamart), Claudette Muller (Laboratoire de Microbiologie, Hôpital Bichat, Paris), Marie-Hélène Nicolas-Chanoine (Laboratoire de Microbiologie, Hôpital Beaujon, Clichy), Isabelle Poilane (Laboratoire de Microbiologie, Hôpital Jean Verdier, Bondy), Claire Poyart (Laboratoire de Bactériologie, Hôpital Cochin), Laurent Raskine (Laboratoire de Microbiologie, Hôpital Lariboisière, Paris), and Hoang Vu Thien (Laboratoire de Microbiologie, Hôpital Trousseau, Paris).
Acknowledgments
Potential conflicts of interest.All authors: no conflicts.
Comments