An 8-year survey of strains identified in blood cultures in a clinical haematology unit



The aim of our study was to determine the epidemiological profile and the antibiotic susceptibility of bacteria and fungi identified from blood cultures in the patients of the clinical haematology unit. A retrospective study was carried out over an 8-year period (2003–2010) in the clinical haematology unit of the Percy Military Medical Center. During this period, we collected 723 isolates: Gram-negative bacilli (70.8%) and Gram-positive cocci (18.7%). The four most commonly isolated species were Escherichia coli (18.5%), Pseudomonas aeruginosa (14.8%), Stenotrophomonas maltophilia (6.2%) and Staphylococcus epidermidis (5.4%). The rate of methicillin-resistant Sthaphylococcus aureus was 6.45% and that of coagulase-negative staphylococci 61.2%. No resistance to glycopeptides was observed. In E. coli, as in the Klebsiella-Enterobacter-Serratia group, a 27% resistance to fluoroquinolones was observed. Concerning P. aeruginosa, the phenotypes were distributed over penicillinase (23.4%) and cephalosporinase (13.1% were resistant to ceftazidime). The impermeability rate of imipenem was 9.3%. The aggressiveness and duration of haematological treatments explains why infections remain one of the main complications of neutropenia. The emergence of new or unusual bacteria is highly likely. Antibiotic selective pressure and long periods of hospitalization could explain the emergence of multiresistant bacteria. As a consequence, epidemiological surveillance is indispensable.


Infections are serious and frequent complications of cytotoxic chemotherapy-induced neutropenia. Seventy per cent of patients with agranulocytosis or who receive haematopoietic stem cells transplantation develop a fever [1]. The mortality rate is high: 5% in the case of bacteraemias associated with Gram-positive cocci (GPC) and 18% in the case of those associated with Gram-negative bacilli (GNB) [2]. For 30 years, the epidemiology of bacterial infections in these patients has evolved. In the seventies, and until the mid-eighties, the reports from the febrile neutropenic antibiotherapy study group of the EORTC (European Organisation for Research and Treatment of Cancer) showed that in 60% to 70% of cases, bacteraemia in this population was related to GNB [3]. By the end of the 1980s, the proportions reversed to a greater percentage of GPC, before tending towards a balanced situation following the year 2000. The study previously performed in the onco-haematology department of our hospital revealed the variability of these data [4]. Almost 10 years later, it appeared necessary to review the blood culture flora sampled in onco-haematology and to compare the results with the literature. Our aim was to improve the understanding of the impact of various therapeutic scenarios on the flora, and to optimize probabilistic antibiotherapy during episodes of febrile neutropenia.


We performed a retrospective study, between 1 January 2003 and 31 December 2010, in the Clinical Haematology Department of the Military Medical Centre Percy. This is a 23-bed department, with 10 rooms devoted to intensive care. It handles acute and chronic haematological malignancies and also undertakes procedures related to chemotherapy, autografting and allografting. All of the studied blood culture isolates were validated by the physician and communicated to the clinician. Ten millilitres of venous blood were inoculated into aerobic and anaerobic bottles using the BacT/Alert 3D system (BioMérieux®, Marcy l'Etoile, France). The bottles were incubated at 37°C with constant stirring for 6 days, before being considered as sterile. All positive cultures were subcultured on enriched media: blood agar, boiled blood agar, Sabouraud® agar and Schaedler® agar (BioMérieux®), under anaerobic conditions whenever necessary. An initial examination was carried out followed by another examination after Gram staining. Based on examination, a biochemical identification using the API system® (BioMérieux®) was inoculated directly from the blood culture broth. An antibiogram was also carried out, following the guidelines of the Antibiogramme Committee of the French Society for Microbiology (CA-SFM). An antifungigram was systematically carried out using the E-test® technique (BioMérieux®), whenever the isolate contained a fungus. The inhibition zone or E-test® diameters were read by a SIRSCAN® analyzer (i2a®, Perols, France).

The data were gathered using the ‘epidemiological’ module of the SIR® software (i2a®), with individualized analyses of the species and their resistance profiles. The bacteria belonging to commensal flora (coagulase-negative staphylococcus and Corynebacteria sp.) were retained if they were isolated at least twice with the same antibiotype. Duplicates (even isolates with the same sensitivity profile, isolated several times in the same patient over a period of at least 5 days) were excluded.


The colistin–amphotericin B combination is used in our hospital for selective digestive decontamination (SDD). During allogeneic bone marrow transplantation (BMT), patients receive primary antibioprophylaxis based on a combination of ciprofloxacin and a piperacillin/tazobactam. In the case of febrile neutropenia, the most urgent issue is antibacterial therapy, based on an empirical broad-spectrum antibiotherapy. The protocol applied in our haematology department relies on a dual empirical antibiotherapy according to the Infectious Diseases Society of America (piperacillin/tazobactam and amikacin) for 3 days [5]. The combination with aminoglycoside is not recommended by the European Conference on Infection in Leukemia unless septic shock or pneumonia occurs [6]. In the case of persistent fever, antibacterial therapy relies on ceftazidime and vancomycin and amikacin. This first-line treatment regimen is maintained if effective and/or if the isolated bacterial species is sensitive to this combination. If a resistant GNB is isolated, the spectrum is enlarged to include third-generation cephalosporins (3GC) or even imipenem in the case of ESBL (extended-spectrum betalactamase-producing enterobacteria), which hydrolyze 3GC. If a GPC is isolated, vancomycin is given. This antibiotic can be added at an early stage in the case of an MRSA-carrier patient. Finally, the addition of an antifungal drug (amphotericin B) can become necessary if a fungal infection is isolated or suspected.


Patient profiles, frequency and distribution of species

Over an 8-year period, 1413 patients were hospitalized in our haematology department. Of these, 829 had neutropenia with <1000 neutrophils/mm3 (grade 3 according to the WHO classification [7]). Among these 1413 patients, 737 had at least one grade 4 neutropenic episode (<500 neutrophils/mm3). During this period, 6412 blood culture samples (an anaerobic bottle followed by an aerobic bottle) were performed and 723 positive blood culture isolates were identified (i.e. 11.3% of the blood culture samples were positive). During this period, 89 patients underwent an allogeneic BMT. The frequency and distribution of species are presented in Table 1. The most important result of this study is the variation of the epidemiology of bacteraemia: the GPC represent 18.7% (n = 135) of the isolates, and the GNB represent 70.8% (n = 512). The dynamics of the results (Fig. 1) provide a perfect illustration of the variability of these data. The four most commonly isolated species were Escherichia coli (18.5%, n = 134), Pseudomonas aeruginosa (14.8%, n = 107), Stenotrophomonas maltophilia (6.2%, n = 45) and Staphylococcus epidermidis (5.4%, n = 39). The GPC isolates consisted mainly of S. epidermidis (28.9%), Sthaphylococcus aureus (23%, n = 31) and Enterococcus sp. (17.8%, n = 24), of which 55% were E. faecalis (n = 13). The enterobacteriaceae isolates consisted mainly of E. coli (51.7%) and bacteria from the Klebsiella-Enterobacter-Serratia (KES) group (40.9%, n = 106). The non-fermenting GNB (NF-GNB) were dominated by P. aeruginosa (42.3%, of which 35% were of serotype 6), S. maltophilia (17.8%) and Acinetobacter sp. (16.6%, n = 42), of which 64.3% were Acinetobacter baumannii (n = 27). The NF-GNB represented 49.4% of the GNB (n = 253/512 GNB). Among the other NF-GNB, mainly Pseudomonas non-aeruginosa (7.5%, n = 19), in particular Pseudomonas fluorescens and Pseudomonas putida, were found. These were followed by Sphingomonas sp. (4.3%, n = 11).

Table 1. Distribution by species
Streptococcus mitis 12320310121.7Streptococcaceae
Oral streptococcus11111304121.7
Streptococcus pneumoniae 0010000010.1
Enterococcus sp.32730522243.3
Streptococcus (A-G)1020200160.8
Staphylococcus epidermidis 37556733395.4Staphylococcaceae
Staphylococcus hominis 0000000000.0
Staphylococcus haemolyticus 0000001010.1
Other coagulase-negative Staphylococcus0040111291.2
Staphylococcus aureus 36426424314.3
All GPC121827131623101613518.7GPC
Escherichia coli 121522161024171813418.5Enterobacteriaceae
Salmonella sp.1000000010.1
Total enterobacteriaceae212133272845453925935.8
Pseudomonas aeruginosa 1813131398171610714.8Non-fermenting
Acinetobacter sp.41467587425.8
Stenotrophomonas maltophilia 006106788456.2
Other NF-GNB246877169598.2
All NF-GNB241829372927494025335.0
All GNB453962645772947951270.8GNB
Other species974119856598.2
Figure 1.

Illustration of the variations in isolated species over a period of 15 years (ordonate, positive bloodstream (%); GPC, Gram-positive cocci; NF-GNB, non-fermenting Gram negative bacilli (Pseudomonas aeruginosa, Acinetobacter, Stenotrophomonas maltophilia…)); GNB, Gram-negative bacilli).

Resistance distribution

Antibiotic susceptibilities were studied by class of antibiotic. Only two of the S. aureus isolated from blood cultures between 2003 and 2010 were methicillin resistant (MRSA = 6.5%). However, 61.2% (n = 30) of coagulase-negative staphylococci harboured this resistance. Considering glycopeptides, one S. epidemidis strain showed an intermediate resistance to teicoplanin, but was sensitive to vancomycin. No strain of S. aureus was isolated with an intermediate resistance to glycopeptide antibiotics. Sixty per cent (n = 81) of acquired penicillinase and 27% (n = 36) of high-level penicillinase were observed in E. coli. Almost 4% (n = 10) of the enterobacteriaceae were resistant to 3GC, the phenotypes were hyperproduction of the chromosomal cephalosporinase in five Enterobacter sp. and ESBL in five E. coli. In E. coli, as in the KES group, 27% were resistant to fluoroquinolones (this rate was 10% in the previous study from 1996 to 2002). Concerning the NF-GNB, especially for P. aeruginosa, the phenotypes were distributed over penicillinase (23.4%, n = 25) and cephalosporinase (13.1% were resistant to ceftazidime, n = 14). The impermeability rate of imipenem was 9.3% (n = 10). In the case of Acinetobacter sp., the resistance rates were similar: 16.7% (n = 7) were resistant to ticarcillin. None of the isolated Acinetobacter sp. were resistant to imipenem and to the piperacillin-tazobactam combination, and 31% (n = 13) were resistant to fluoroquinolones. For the 45 isolated strains of S. maltophilia, the in vitro resistance rates were as follows (R%): ticarcillin-clavulanic acid (31%, n = 14), piperacillin-tazabactam (62%, n = 28), ciproflaxin (27%, n = 12), trimethoprim-sulphamethoxazole (4%, n = 2). Forty-nine per cent (n = 22) of the strains were resistant to amikacin and 49% (n = 22) were resistant to ceftazidime.



Patients with onco-haematological diseases combine multiple infection risk factors: immunodeficiency, modification of the endogenous flora, and alteration of the skin and intestinal barriers by catheters and chemotherapy. As the latter phenomenon can lead to life threatening bacteremia, SDD can be carried out in patients with grade 4 neutropenia. This SDD is prescribed during aplasias following allogeneic and autologous BMT and induction chemotherapy for acute leukaemias, in order to reduce the digestive bacterial concentration of GNB species [8]. The investigation of cases of fever in neutropenic patients will include carrying out a stool culture, an ECBU, a chest X-ray and blood cultures taken at the same time (if needed) from central and peripheral catheters, to elucidate the involvement of the central catheter in the sepsis [9]. Nevertheless, from the microbiological point of view, nearly 70% of febrile neutropenia cases remain undocumented [10, 11]. The increasing use of highly active cytotoxic treatments and allografts has led to an increase in the number of neutropenic patients treated in onco-haematology. Over a period of 8 years, the neutropenic population represented almost 60% of patients hospitalized in our department. These patients are at risk of infection: during the period 2003–2010, 46% of patients had one or more bacteraemias during a grade 3 neutropenic episode.


The main feature of these trends is the constant increase in GNB, in particular NF-GNB, in absolute numbers and incidence rate. Table 2 shows the variation in absolute terms of the various bacterial families or species between 1996–2002 and 2003–2010. With essentially equal numbers of blood cultures included in the two study periods, 2 times as many E. coli, 2.5 times as many NF-GNB and 3 times more KES, were isolated in the second period than in the first. The NF-GNB have followed the evolution of GNB; 15 years ago they already represented half of all isolated GNB. The first isolated species in our study is E. coli. This result is in line with various studies dealing with large series of immunocompetent and immunosuppressed patients [12-15]. Other studies performed in patients with well-established cancer showed the opposite trend (i.e. a predominance of GPC, in particular negative coagulase staphylococcus). However, very few studies of the microbiology of bacteraemia in onco-haematology patients have been published. A study performed in India has shown that in hospitalized patients with acute myeloid leukaemia and presenting with bacteraemia, the most commonly encountered species was P. aeruginosa, followed by enterobacteriaceae such as Klebsiella pneumoniae and E. coli [16]. In 2008, Cattaneo et al. [17] revealed a change in the epidemiology of bacteraemia in patients with haematological malignancies and an increase of E. coli, but also of P. aeruginosa. Finally, in a recent study of haematology patients by the same authors between June 2004 and January 2010, from 441 positive blood cultures nearly 60% were identified as GNB, of which 66 were P. aeruginosa (15%) [18]. Another study, carried out over a 1-year period, showed that in neutropenic patients coagulase-negative staphylococci represented more than 50% of blood culture isolates [3]. Similar results were reported in 2009 by a Moroccan team [19].

Table 2. Epidemiology between 1996–2002 and 2003–2010
 Positive blood cultureGPCGNBEnterobacteriacaeNF GNBAnaerobicYeastOther speciesa
Escherichia coliKESOther
  1. a

    Haemophilus sp., Campylobacter sp., Branhamella sp., Bacillus sp.

  2. Statistical analysis was performed using the chi-square test; GPC, Gram-positive cocci; NF-GNB, non-fermenting Gram-negative bacilli (Pseudomonas aeruginosa, Acinetobacter, Stenotrophomonas maltophilia…); GNB, Gram-negative bacilli.

Study 1996–2002690432 (62.6%)216 (31.3%)60381410491023
Study 2003–2010723135 (18.7%)512 (70.8%)1341061925341359
  p 0.001p 0.001       

This re-emergence of GNB can be partially explained by a decrease in the prophylactic use of fluoroquinolones and the treatment of cases of febrile neutropenia [20]. It can also be explained by the lack of active antibacterial agents for naturally multi-resistant species (Acinetobacter, P. aeruginosa and S. maltophilia). In our department, fluoroquinolones are only used as a prophylaxis for allogeneic BMT, and on a case-by-case basis, when the patient's renal function contraindicates the use of aminoglycosides and for the induction of acute leukaemia.

The influence of SDD is more difficult to evaluate: theoretically, it targets the Gram-negative digestive reservoir, in particular emerging P. aeruginosa. Although they are naturally resistant to colistin, Proteus, Serratia and Providencia do not emerge. SSD thus remains controversial as a daily practice, and if using SDD might reduce bacteraemia rates, infection-related fatality rates are not reduced [21]. Selective pressure related to empirical first-line treatment of febrile neutropenia could also partially explain the growing incidence of P. aeruginosa. Among the various GNBs, S. maltophilia occupies a central position. This trend was also observed by another team [22]. This increase in incidence rate raises difficulties, especially in terms of treatment, as a consequence of the natural resistance of this species, in particular to carbapenems. The respiratory and gastrointestinal tracts are reservoirs for S. maltophilia. The increase in S. maltophilia bacteraemia, which reveals antibiotic selective pressure, needs to be carefully monitored as a consequence of the risk of a therapeutic impasse. For the forty-five patients with S. maltophilia bacteraemia, it would be interesting to know the number of treatments including imipenem prior to the bacteraemia; unfortunately these data are difficult to record in a retrospective study.

In parallel with this phenomenon, the incidence of GPC has decreased. This decline has multifactorial origins: a decrease through improved control of the implementation and monitoring of central venous lines, the controlled prescription of fluoroquinolones [23] and the appropriate use of glycopeptides. The rate at which anaerobic bacteria are isolated remains low, and similar to the rates found in the literature [24]. The proportion of candidaemia also remains low, with no change with respect to the last decade. The 2012 data from the ‘Interscience Conference on Antimicrobial Agents and Chemotherapy’ reveals a 2% incidence in grafted patients [25]. In the present study, only three cases of candidaemia (i.e. an incidence rate of 0.4%), all of which occurred in grafted patients, were observed. The study of patients’ stool flora in onco-haematology, allowing on one hand the SDD to be confirmed, and on the other hand a probabilistic antibiotherapy to be adapted in the case of febrile episodes, reveals discrepancies in the data [26]. Indeed, in the same 2003–2010 study period, quantitative stool cultures revealed a clear predominance of GPC, enterococci in particular. GNBs were rarely isolated from quantitative stool cultures. Nevertheless, the study of this stool flora does not allow digestive translocations to be predicted or an antibiotherapy to be adapted. These data partially explain why the systematic monitoring of digestive flora in neutropenic patients remains controversial [27] and is no longer performed in many hospitals.


The main purpose of a surveillance of a specific disease such as bacteraemia is to detect shifts in antimicrobial susceptibility of the involved bacteria and should guide the choice of an empirical therapy. The rate of penicillinase was 60%, as opposed to 51% to 60% reported in various studies [26]. Furthermore, the 1998–2009 ONERBA study of the monitoring of bacteraemia reveals a 12% fall in the sensitivity of enterobacteria to fluoroquinolones, with a sensitivity of 80% to 85%. In our department, the rate of ciprofloxacin resistance is higher (25%), despite the absence of fluoroquinolones in the treatment protocols. This significant increase of ciprofloxacin resistance was described by European Centre for Disease Prevention and Control (ECDC) in invasive French E. coli between 2001 (9%) and 2010 (between 10% and 25%) [28]. In the literature, 5% of bacteraemia are related to ESBL [29]. In our work, a slight (4%) but not significant increase of ESBL was identified. In NF-GNB, the study of P. aeruginosa resistance to ‘anti-pseudomonas’ molecules reported similar or even lower resistance rates than those reported by the ONERBA and ECDC: 60% to 65% sensitivity to ticarcillin vs. 75% in our study, and 75% to 85% of strains sensitive to ceftazidime vs. 87%. The resistance of P. aeruginosa to imipenem is mainly related to the loss of D2 porin: the resistance rates found by ONERBA and in our study are close to 10% [26], whereas ECDC found a resistance rate of 17.8% [28]. Fluoroquinolones and aminoglycosides remain highly efficient and these must be used in combination. According to the studies, the resistances to ciprofloxacin and amikacin vary respectively from 60% to 80% and from 80% to 90% [28-31]. Furthermore, no carbapenemase or methylase producing bacteria were isolated. In our work, a small proportion (6.5%) of oxacillin-resistant S. aureus was identified. This feature is discordant with ECDC, which reported a rate of oxacillin resistance among S. aureus of 21.6% [28]. This feature may be explained in part by the low number of strains in our work. No vancomycin-resistant Enterococcus sp. was detected, despite the use of this glycopeptide as a second-line treatment for febrile neutropenia. Table 3 shows the change in resistance of the main bacterial species, between 1996–2002 and 2003–2010.

Table 3. Antibiotic resistance between 1996–2002 and 2003–2010
 MRSACoagulase negative Staphylo-coccus méthicillin-resistantEscherichia coli - Pénicilli-naseE coli -ESBLE  coli –fluoroquinolones resistancePseudomonas aeruginosa – resistance ticarcillin, ceftazidime, imipenemStenotrophomonas maltophilia – Ticarcillin-clavulanic acid résistance
TicCazImi (D2)
  1. Statistical analysis was performed using the chi-square test; NS, not significant; MRSA, methicillin resistant Sthaphylococcus aureus; ESBL, extended spectrum beta lactamase; Tic, ticarcillin; Caz, ceftazidime; Imi, imipenem.

Study 1996–200211.8%65%60%0%10%25.5%8.5%6.4%21.4%
Study 2003–20106.45%61.2%60%4%25%23.4%13.1%9.3%31%
    NSp 0.001   NS


The aggressiveness and duration of antineoplastic therapy explains why infections remain one of the main complications of neutropenia. In this context, the emergence of new or unusual bacteria is highly likely. Antibiotic selective pressure and long periods of hospitalization could explain the emergence of multiresistant bacteria. As a consequence, epidemiological surveillance is indispensable. The evolution of flora towards GNB should encourage the highest level of caution. Moreover, the results of this study lead us to question the correct observance of SDD in neutropenic patients. A prospective study of bacteraemia in neutropenic patients would allow this crucial debate to be taken further.


We are grateful to bacteriology technicians from the Percy Military Medical Centre, Clamart, France.

Transparency Declaration

The authors declare no conflicts of interest.