Treatment of Haemophilus bacteremia with benzylpenicillin is associated with increased (30-day) mortality

Background Optimal antibiotic treatment strategies of Haemophilus infections are still needed. Therefore, 30-day case fatality rate (CFR) of Haemophilus bacteremia and efficacy of various antibiotic treatment regimes were studied. Methods All episodes of Haemophilus bacteremia in the former Copenhagen County during the period 2000-9 were included in the study. Clinical and biochemical findings and outcome were collected retrospectively from medical records. Results 105 consecutive episodes were identified (median age: 69 years, with only 4 children <16 years), 72% were due to non-typeable -, 16% to typeable H. influenzae, and 11% to other Haemophilus species. Pneumonia was the most common primary focus (in 48%), and 58% of the patients had Charlson comorbidity index > 1. Definitive antibiotic therapy was in 26 cases benzylpenicillin, in 12 cases aminopenicillins, in 50 cases cefuroxime and in 16 cases broadspectrum antibiotics, whereas 1 palliative case died without start of therapy. Whereas the use of broadspectrum antibiotics was related to the severity of the disease (admittance to ICU, need for assisted ventilation or hemodialysis, septic shock), no significant difference in clinical features was demonstrated for therapy with benzylpenicillin, aminopenicillin or cefuroxime, except benzylpenicillin was rarely administered to immunosuppressed patients. The CFR was 22% (23/105). The choice of empiric antibiotic therapy was not significantly associated with mortality (adequate vs. inadequate treatment: 23% (21/93) vs. 17% (2/12), respectively, P > 0.05). In contrast, definite antibiotic therapy with cefuroxime or aminopenicillins resulted in a significantly lower CFR than treatment with benzylpenicillin (12% (6/50) or 0% (0/12) vs. 39% (10/26), respectively, Log rank test P < 0.02). When adjustments were made for other identified risk factors in bivariate logistic regression analysis, treatment with cefuroxime was still were found to be associated with a significantly lower CFR than for benzylpenicillin: OR: 0.21 (0.06-0.69), P = 0.01 (hospital-acquired bacteremia), OR: 0.27 (0.08-0.91), P = 0.04 (polymicrobial episodes), OR: 0.16 (0.04-0.59), P = 0.006 (admittance at intensive care unit), OR: 0.22 (0.06-0.82), P = 0.02 (alcohol abuse), OR: 0.15 (0.04-0.60), P = 0.008 (altered mental state), OR: 0.22 (0.07-0.71), P = 0.01 (temperature < 38 °C), OR: 0.23 (0.07-0.79), P = 0.02 (septic shock), OR: 0.21 (0.06-0.69), P = 0.01 (mechanical ventilation). Conclusion Our results suggest that, after susceptibility testing, cefuroxime or aminopenicillins are preferable to benzylpenicillins as definitive therapy for Haemophilus bacteremia.


Background
Haemophilus bacteremia still carries a high mortality, and clinical features such as gender, age, acquisition of the bacteremic episode, and the focus of infection have recently been demonstrated to be associated with an increased risk of a fatal outcome [1]. 3rd generation cephalosporins have been the recommended therapy for childhood meningitis due to H. influenzae type B, because of a more rapid sterilization of the cerebrospinal fluid than in therapy with aminopenicillins [2] or cefuroxime, and a better outcome of the disease with 3rd generation cephalosporins compared to cefuroxime [3]. Less is known, however, about the influence of various antibiotic treatment regimes on the outcome of other Haemophilus infections including bacteremia. In Denmark therapy with benzylpenicillin has been recommended for treatment of ampicillin-susceptible Haemophilus infections and has been considered to be as effective as other recommended treatment regimes (e.g. aminopenicillins and cefuroxime). These recommendations provided in some local Danish guidelines were to our knowledge primarily based on results obtained from time-kill experiments showing benzylpenicillin to be as active as amoxicillin against various Haemophilus species [4], whereas clinical studies comparing benzylpenicillin to other recommended treatment regimes for therapy of Haemophilus bacteremia still are lacking In addition, the European Committee on Antimicrobial Susceptibility Testing (EUCAST: www.eucast.org) has stated that there still are insufficient data for H. influenzae to set clinical breakpoints for benzylpenicillin [5]. Thus, optimal antibiotic treatment strategies for infections due to Haemophilus species are still needed.
This retrospective study aims to investigate the efficacy of benzylpenicillin therapy in comparison with cefuroxime/aminopenicillins treatment in 105 consecutive episodes of bacteremia due to Haemophilus species in the former Copenhagen County during the time period 2000-2009.

Identification of Haemophilus bacteremia episodes
108 episodes of Haemophilus bacteremia were identified from 1 January 2000 to 31 December 2009. Three bacteremic episodes caused by H. haemolyticus, H. parainfluenzae and H. influenzae were considered contaminations, because the patients had no clinical evidence of bacteremia, and they survived without receiving any antibiotic therapy. Two patients had two episodes of Haemophilus bacteremia with different primary infection foci two and five years apart, respectively. These patients were included with both episodes. Thus the study compromised 105 episodes of Haemophilus bacteremia in 103 patients.

Antibiotic treatment
Empiric antibiotic therapy at time of blood culturing was in 37 cases benzylpenicillin (15 of these in combination with gentamicin), in 7 cases aminopenicillins (3 of these in combination with gentamicin), in 41 cases cefuroxime (1 of these in combination with ampicillin, 8 with gentamicin and 1 with ciprofloxacin), in 5 cases piperacillintazobactam (3 of these in combination with gentamicin and 2 with ciprofloxacin), in 5 cases a 3rd generation cephalosporin (2 in combination with ampicillin), in 1 case meropenem, in 1 case ciprofloxacin, and in 1 case gentamicin alone. Dosage of antibiotics followed local guidelines (e.g. benzylpenicillin 1.2 g qid, ampicillin 1-2 g qid, cefuroxime 1.5 g tid for adults except for treatment of CNS infections and endocarditis that were treated with higher doses). 3 patients were initially treated with antibiotics not considered adequate (dicloxacillin, erythromycin), whereas 4 patients did not receive empiric antibiotic treatment. According to the susceptibility testing, empiric antibiotic therapy was not considered adequate in 2 additional episodes (1 treated with cefuroxime and 1 treated with penicillin), whereas 96 patients received adequate empiric antibiotic treatment).
Definitive antibiotic therapy chosen according to the susceptibility testing of the pathogen was in 26 cases benzylpenicillin (9 of these in combination with gentamicin), in 12 cases aminopenicillins (3 of these in combination with gentamicin), in 50 cases cefuroxime (10 of these in combination with gentamicin and 1 with ciprofloxacin), and in 15 cases broadspectrum antibiotics (piperacillin-tazobactam (n = 7; 4 of these in combination with gentamicin and 3 with ciprofloxacin), 3rd generation cephalosporin (n = 5), ciprofloxacin (n = 3), meropenem (n = 1)). 1 case with inoperable mouth cancer, who died 6 days after blood culturing, was palliative and did not receive any antibiotic therapy. The definitive antibiotic therapy was changed in 42 cases, whereas it was a continuation of the empiric antibiotic therapy in 62 cases, including the 5 cases where the patient died before telephone contact was made by a physician of the Department of Clinical Microbiology with the notification of the positive blood culture and the subsequent results of susceptibility testing. Among these 5 cases only one case, who was treated with benzylpenicillin, did not receive adequate definitive antibiotic therapy.

Mortality and definitive antibiotic therapy
When therapy with cefuroxime or aminopenicillins was chosen after result of susceptibility testing, the CFR was significantly lower than for therapy with benzylpenicillin (12% (6/50) or 0% (0/12) vs. 39% (10/26), respectively or to therapy with broadspectrum antibiotics 38% (6/16) (Log rank test, P < 0.02, see Figure 1). The CFR was not significantly associated with the use of mono-or combination therapy (19% (14/74) vs. 27% (8/30), respectively, P = 0.4). When adjusting for other identified risk factors in bivariate Logistic regression analysis (see Table 2), treatment with cefuroxime was still associated with a significantly lower CFR than for benzylpenicillin: OR  Table 2 (e. g. combination therapy with gentamicin/ciprofloxacin, age, gender, Charlson index, smoking, immunosuppression, focus of infection, type of Haemophilus species) in bivariate analysis, treatment with benzylpenicillin was still associated with a significantly higher CFR than therapy with cefuroxime/aminopenicillins (data not shown, P < 0.05). Beside this, if all analyses were performed for H. influenzae only or if the patient, who were treated with benzylpenicillin and did not receive adequate definitive antibiotic therapy, was excluded from the analyses, treatment with benzylpenicillin was still associated with a significantly higher CFR than treatment with cefuroxime/aminopenicillins (data not shown, P < 0.05).

Discussion
In the present study we found that the CFR was significantly higher when treatment with benzylpenicillin was continued as definitive therapy after the result of susceptibility testing, when compared to treatment with cefuroxime/aminopenicillins. When adjusting for other identified risk factors or other co-factors in bivariate Logistic regression analysis, a significantly higher risk of dying was still observed after therapy with benzylpenicillin as compared to cefuroxime. Moreover, empiric treatment with cefuroxime/aminopenicillins resulted in lower CFR compared to therapy with benzylpenicillin, although the difference did not reach statistical significance (P = 0.06). To our knowledge this is the first study investigating the treatment efficacy of benzylpenicillin in Haemophilus bacteremia. Given the lack of research in this area, EUCAST has considered that there are insufficient data for H. influenzae to set clinical breakpoints for benzylpenicillin, whereas other guidelines until recently have suggested 1/1 mg/L (SRGA) or 1/4 mg/L (The Norwegian Working Group on Antibiotics (www.unn.no/afa) as the breakpoints (S≤/R>) for benzylpenicillin [5]. Interestingly, the MIC values and distribution for H. influenzae and H. parainfluenzae for benzylpenicillin, aminopenicillin and cefuroxime are comparable, and time-kill experiments have demonstrated benzylpenicillin to be as active as amoxicillin against various Haemophilus species [4]. In contrast, the pharmacokinetic profile is more favorable with aminopenicillins/cefuroxime than with benzylpenicillin due to the former's lower protein binding (17-40% vs.~45-65%, respectively) and longer serum elimination half-life (T ½ : 1.1-1.4 vs. 0.5-0.75 hrs, respectively) (www.eucast. org, [6,7]). Indeed, EUCAST has calculated significantly lower target attainment rates (i.e. time above the   Our study has some important limitations. Our study was not a randomized study comparing the treatment efficacy of benzylpenicillin and cefuroxime/aminopenicillins for Haemophilus bacteremia, and the retrospective study design could potentially have influenced the strength of the statistical calculations due to missing values. Our data collection, however, was almost complete regarding the clinical characteristics and antibiotic therapy. Multivariable analysis for significant risk factors for progression to death could not be performed including all the risk factors identified in the univariate analysis, because there was a relatively low number of outcomes compared to number of variables included in the analysis [8]. The definitive antibiotic therapy was therefore only adjusted for single individual risk factors in separate bivariate analyses. Whereas the epidemiology of Haemophilus infections has been studied extensively since the introduction of the   H. influenzae type b vaccine in childhood vaccination programs [9][10][11] demonstrating a virtual elimination of invasive H. influenzae type b infections [12][13][14][15][16], less is known about the clinical presentation of Haemophilus infections in the post-vaccination era. In the present study, we found that the clinical presentation (e.g. focus of infection) of Haemophilus bacteremia still followed observations from the pre-vaccination period and was closely related to the Haemophilus species. Almost all bacteremic episodes with a lung focus were due to nontypeable H. influenzae; meningitis and epiglottis were primarily observed among episodes due to typeable H. influenzae, whereas endocarditis was solely due to H. parainfluenzae. Non-typeable H. influenzae accounted for most of the bacteremic episodes (3 out of every 4 episodes), and in accordance with previous studies, patients infected with non-typeable H. influenzae had an higher age, were more often immunosuppressed, and had a high CFR [17][18][19][20]. Almost all healthcare-related bacteremic episodes were due to non-typeable H. influenzae, whereas patients with typeable H. influenzae bacteremia had the youngest median agea relationship which has also been demonstrated previously [20]. These patients were more frequently transferred to an intensive care unit and received mechanical ventilation, but had a lower CFR than the other patient groups. This was most likely due to a high number of patients with epiglottitis, who all required assisted ventilation, but all survived the bacteremic episode.

Conclusion
Our results suggest that, after susceptibility testing, cefuroxime/aminopenicillins is preferable to benzylpenicillins as definitive therapy for H. influenzae bacteremia. Blood culturing was performed at the attending physician's decision comprising 1 venipuncture per episode (40 ml of blood distributed in two aerobic and two anaerobic bottles for adults and 8 ml in pediatric bottles for small children; BACTEC W , Becton-Dickinson, Sparks, MD, USA). In case of a positive blood culture, a physician of the Department of Clinical Microbiology notified by telephone the treating physician of the patient concerned, giving the results of the Gram staining and the motility examination, and also supplying recommendations about antibiotic therapy, diagnostic and therapeutic procedures etc. Another contact with the treating physician was made the following day to provide the results of susceptibility testing and bacterial identification. The positive blood culture was sub-cultured, and the identification of the various Haemophilus strains was done using standard methods [21]; subsequent serotyping was performed using serotype-specific anti-sera and/or was tested at the National Reference Laboratory at Statens Serum Institut. Susceptibility testing was performed by disc diffusion methods (Oxoid, Hanks, UK), according to zone diameter break-points set out by the Swedish Reference Group for Antibiotics (SRGA: www.srga.org). β-lactamase production was determined by use of Cefinase™ paper discs (Becton Dickinson, NJ, USA) to identify resistance to benzylpenicillin and aminopenicillins. Cefuroxime-resistance was confirmed by subsequent MIC determination (E-test W , bioMérieux, Marcy l'Etoile, France).

Data collection
Clinical and laboratory information were collected retrospectively from medical records. Hospital admission and discharge records including date of death were confirmed in the National Hospital Registration Database. The study was approved by the Danish Data Protection Agency (Record 2008-41-2688) and followed the guidelines of the local scientific committee.

Definitions
The bacteremic episode was defined as communityacquired or hospital-acquired according to CDC criteria [22]. In addition, a healthcare-related group was defined for patients who had been hospitalized within 30 days prior to the bacteremic episode, or who regularly visited the hospital (e.g. for chemotherapy or hemodialysis) [23]. Polymicrobial episodes were defined as isolation of a separate significant agent from blood cultures within two days of the Haemophilus bacteremic episode. Hospital departments were divided into medical -, surgical -, intensive care unit (ICU) or other departments (e.g. pediatric, gynaecological and ear-nose-throat departments). The focus of the infection was registered on the basis of clinical, radiological and microbiological findings.
Comorbidity at time of hospitalization was assessed for each individual patient using the Charlson comobidity index scores: 0 points = low, 1-2 points = medium and >2 = high [24]. Patients were defined as being immunosuppressed if they had cancer or had received immunosuppressive therapy. Alcohol abuse was defined as an alcohol intake of more than 14 units per week for women and more than 21 units per week for men. Patients were considered to have either never smoked or smoked at least once. Shock was registered if a patient