Empirical therapy of febrile neutropenic patients with mucositis: challenge of risk-based therapy


N.M.A.Blijlevens University Medical Center St. Radboud Nijmegen, Department of Haematology, P.O. Box 9101, NL-6500 HB Nijmegen, The Netherlands. Tel: +31 24 361 4762. Fax: +31 24 354 2080. E-mail: N.Blijlevens@hemat.azn.nl


Nowadays Gram-positive cocci, especially oral viridans streptococci (OVS) and coagulase-negative staphylococci (CoNS), are the most common bloodstream isolates in febrile neutropenic patients. Although in general these cocci are quite indolent, Streptococcus mitis is associated with serious complications such as sepsis and/or adult respiratory distress syndrome. Neutropenia is the most significant predisposing factor but the impact of mucositis, i.e. damage to the mucosal barrier of mouth and intestines (mucosal barrier injury, MBI), is very much greatly underestimated. Oral mucositis is a strong predictor of OVS bacteremia and simultaneously CoNS bacteremia is clearly associated with mucositis. Treatment with especially high dose cytarabine, cyclophosphamide and idarubicin, when given to allogeneic hematopoietic stem cell transplant recipients, predictably results in mucositis. Hence, the occurrence of mucositis should have implications for complementing empirical therapy with specific drugs such as glycopeptides, because risk patients can be selected based upon the chemotherapeutic therapy administered. An algorithm is presented for dealing with patients at high risk of mucositis and bacteremia due to Gram-positive cocci.


Infection remains the most significant complication of chemotherapeutic-treated cancer patients. Neutropenia is still the main risk factor and there is an inverse correlation between the number of circulating neutrophils and the frequency of infections [1]. Although infection can originate from exogenous sources, the majority of the organisms responsible for infection arise from the patient’s own endogenous microbial flora, particularly from those residing on the mucosal surfaces of the mouth and the gastrointestinal tract. The composition of the endogenous flora is altered by therapeutic interventions, particularly antibiotics, and changes in colonization can be followed by infection [2]. Over the past several years there has been a shift in pathogens from Gram-negative bacilli towards Gram-positive cocci, which now account for approximately 70% of the bacteremic isolates [3]. The widespread use of in-dwelling central venous catheters (CVC) is a significant risk factor for infections due to the coagulase-negative staphylococci (CoNS) and Staphylococcus aureus, whereas overt mucositis predisposes to bacteremia due to oral viridans streptococci (OVS). Whether or not the common use of flouroquinolones for selective oral antimicrobial prophylaxis has contributed to this is still a matter of debate, but such drugs are clearly less effective against Gram-positive bacteria and may lead to their selection [4]. Infections due to OVS are generally quite indolent, but adult respiratory distress syndrome with or without sepsis affects approximately 1 in 10 patients who develop OVS bacteremia after high-dose cytarabine, despite broad-spectrum antibiotic therapy, and carries a mortality rate of around 60%. Emerging multiresistant bacteria such as vancomycin-resistant enterococci and methicillin-resistant staphylococci are also a current concern. Undoubtedly, the prompt institution of empirical therapy at the onset of fever during neutropenia targeting mainly Gram-negative bacilli has improved survival dramatically, but these regimens are not optimal for dealing with infections due to Gram-positive bacteria. Given these considerations it seems timely to evaluate our current strategy for managing fever during neutropenia and pose the questions: should we be using antibiotics aimed at Gram-positive bacteria to complement standard empirical therapy and, if so, under what circumstances and when? To help answer this question we set out an attempt to identify criteria for distinguishing patients at risk for Gram-positive infections specifically in the context of overt mucositis.

Empirical therapy

Most patients with solid cancer treated with chemotherapy develop severe neutropenia for less than 5 days, whereas patients treated with intensive chemotherapy or an hematopoietic stem cell transplant (HSCT) for a hematological malignancy are at risk of developing a profound and prolonged neutropenia lasting at least 10 days. As neutropenic patients are unable to mount an adequate inflammatory or immune response, any fever that develops should be considered to be due to infection until and unless proven otherwise. The strategy of rapid empirical institution of broad-spectrum antibacterial therapy has reduced the mortality rate arising from Gram-negative infection to approximately 10%[5]. Many assorted regimens are used for empirical antibiotic therapy but all can be reduced to three categories: (1) combination therapy with either a β-lactam agent together with an aminoglycoside or two β-lactams drugs, (2) monotherapy with an extended spectrum β-lactam antibiotic, and (3) either of these categories supplemented up front by a glycopeptide, i.e. from the onset of fever. To date no major randomized trials comparing the different antibacterial regimens have been able to show any substantial differences in overall efficacy; however, monotherapy is associated with fewer adverse events and may prove more cost-effective [6]. However, initial therapy is frequently modified prematurely, i.e. within 2–3 days of the onset of fever, even though the median time to defervescence is between 4 and 5 days. In fact, only 15%–20% of patients with a persisting unexplained fever actually require any change to therapy after 72 h, and treatment need only be changed when fever persists and there is evidence of clinical deterioration or of a clinically or microbiologically defined infection [7].

None of the commonly used empirical regimens provides optimal treatment of infections due to streptococci, coagulase-negative staphylococci, Staphylococcus aureus or enterococci. Several prospective trials evaluated the addition of vancomycin or a glycopeptide to the empirical therapy either at the onset of fever or as rescue therapy. There was no advantage in adding a glycopeptide to empirical therapy with ceftazidime for treating bacteremia due to streptococci, as the glycopeptide increased toxicity without offering any measurable improvement in efficacy [8]. Others have shown that there is no need to incorporate a glycopeptide into an initial empirical regimen unless there is a risk that methicillin-resistant Staphylococcus aureus is involved [9]. This seems to have been contradicted by the results of a large randomized trial that showed a better response for Gram-positive infections in the group treated with the empirical regimen containing vancomycin, but there was no advantage in terms of reducing either the incidence or the duration of fever and patients had similar fatality rates whether or not they received vancomycin initially or later on [10]. Moreover the costs were higher and the adverse reactions, particularly nephrotoxicity, increased after adding vancomycin.

Infections caused by Gram-positive cocci are generally indolent and early attributable mortality is low, so the outcome is seldom affected adversely by delaying therapy directed specifically towards these bacteria. This is illustrated by a large single-centre report of 7 years follow-up of 409 HSCT recipients. Whilst 319 (78%) developed fever, infection was only documented in 24.6% of cases, of which 57.9% involved Gram-positive bacteria and none of the patients died as a result [11]. A recently published prospective, double-blinded, placebo-controlled single-center study showed that, with the exception of infections caused by methicillin-resistant staphylococci, Gram-positive infections in neutropenic patients who remained febrile after 72–96 h of monotherapy with imipenem responded well without the addition of teicoplanin, although the response was slower [12]. In another report, the clinical outcome was not better when patients with skin and soft tissue infections associated with Gram-positive bacteremia at the onset of fever received empirical vancomycin in addition to either ceftazidime or piperacillin-tobramycin [13]. Although only half of the isolates of the coagulase-negative staphylococci were susceptible for these empirical regimens, the addition of vancomycin only resulted in a higher rate of CoNS bacteremia eradication that was not translated into any improvement in the clinical response.

The routine use of vancomycin is not without risk. Widespread use has undoubtedly contributed to the increased risk of acquiring vancomycin-resistant enterococci and less susceptible Staphylococcus aureus and S. hemolyticus[3]. Resistance is also not confined to vancomycin. Methicillin resistance is common amongst the coagulase-negative staphylococci (80% of 3908 isolates in North America) [3]. Prior exposure to ciprofloxacin (and by implication other fluoroquinolones) also selects S. epidermidis that are resistant to methicillin, erythromycin, gentamicin, sulfonamide, trimethoprim as well as ciprofloxacin from the axilla within a few days of starting treatment, since the drug is excreted in sweat [14] Excretion of β-lactam antibiotics in the sweat may also explain why staphylococci so rapidly becomes resistant to these drugs [15]. Similarly, as many as 60% of viridans streptococci involved in bacteremia in neutropenic patients are resistant to penicillin largely as a result of exposure to β-lactam antibiotics during the previous 2 weeks [16]. Besides selecting resistance, the liberal use of antibiotics also increases the risk of superinfection, organ toxicity and overall costs.

Hence, we take the view that modification of the initial em-pirical regimen therapy is only warranted if a clinically or micro-biologically defined Gram-positive infection is not responding.

Oral viridans streptococcal (OVS) bacteremia

The incidence of bacteremia in the first 2 weeks after transplant is higher in allogeneic HSCT than in autologous HSCT [11]. In a retrospective analysis of 200 autologous transplants OVS were isolated from the blood at a median of 6 days (range: 2–8 days) after transplant [17]. This coincides with the period when mucositis reaches its peak and MBI is in the ulcerative phase (Figure 1) [18]. Oral mucositis is an important risk factor for streptococci bacteremia in autologous HSCT, as recipients with oral ulcerative mucositis were found to be three times as likely to develop OVS bacteremia as those without mucositis [19]. Similarly, Donnelly reported a higher incidence of OVS more due to the marked mucositis associated with treatment intensification than the use of antimicrobial prophylaxis [20]. Cytarabine (cytosine arabinoside or Ara-C) has been identified as the major risk factor for nosocomial OVS bacteremia with a rate of 11% in 209 allogeneic HSCT recipients [21]. The rate is even higher (30%) when the dose of cytarabine exceeds 1 g/M2[22]. Conditioning intensification with anthracyclines, and in particular idarubicin, results in more pronounced mucositis [23] and higher rates of OVS bacteremia (15.5%). Hence, mucositis plays a pivotal role in the occurrence of OVS during neutropenia early after transplant. Normally, OVS contribute to the ‘colonization resistance’, helping maintain the integrity of the integument and prohibiting pathogenic microbes from gaining a foothold. It is common practice to administer antimicrobial agents, particularly the fluoroquinolones, prophylactically to HSCT recipients and to offer them antiseptics such as chlorhexidine to prevent mucositis. Whatever the merits of this practice, it inevitably leads to marked shifts in the resident oral flora towards the more resistant species, particularly the viridans (α-hemolytic) streptococci. This shift is more profound in patients with overt oral mucositis [24]. MBI is almost certainly not restricted to the oral cavity and probably extends throughout the entire alimentary tract. Hence, the stomach and small intestine could also be a portal of entry assuming colonization with these streptococci occurs as a result of the achlorhydria induced by H2 histamine-antagonists and proton pump inhibitors, since the use of these agents has been noted as a risk factor for the so-called ‘alpha–strep syndrome’[25]. Obviously, MBI is itself a risk factor for OVS bacteremia, although detection of the bacteria in blood does not necessarily always indicate systemic infection since transient bacteremia also occurs in healthy persons after dental manipulation [26]. Moreover, these bacteria do not elaborate exotoxins nor are they professional pathogens. Hence, OVS bacteremia might simply signal the presence of mucosal barrier injury rather than infection. Nevertheless, one particular species, Streptococcus mitis, is apparently associated with more serious complications like sepsis and/or adult respiratory distress syndrome (ARDS), mainly after treatment with high dose cytarabine [25,27]. Recently, OVS bacteremia was documented in 88 out of 485 episodes in neutropenic patients with cancer, ARDS and/or sepsis developed in 10 cases (11%), of which seven involved Streptococcus mitis. These complications were associated with a high mortality (80%) and occurred within 24 h of the onset of bacteremia between 5 and 13 days after transplant. When patients with serious complications were compared with those without complications, severe mucositis, high-dose chemotherapy with cyclophosphamide and allogeneic HSCT were identified as significant variables [28]. Resistance to penicillin was not associated with the occurrence of serious complications, although 30% of the strains showed diminished susceptibility to penicillin and approximately half of these strains were resistant to ceftazidime. The ARDS syndrome could be provoked by changes in the pulmonary endothelium and lung macrophages induced by cytotoxic chemotherapy, which, in turn, induces cytokine production perhaps triggered by infection with Streptococcus mitis[29]. Hence, although many have tried broadening the prophylactic or empirical treatment regimen with penicillin-G, vancomycin [30], roxithromycin [31] or switching to imipenem or meropenem, administering corticosteroids (2–3 mg/kg IV 3–5 days) pre-emptively may prove more effective in preventing the development of ARDS and sepsis associated with Streptococcus mitis after high-dose chemotherapy (Figure 2) [29]. The rate of OVS bacteremia is indeed reduced after addition of other antibiotics but respiratory morbidity and mortality remains unchanged. The preventive effect of corticosteroids also serves to highlight the fact that serious pulmonary complications after Streptococcus mitis bacteremia represent more of an immunologically mediated phenomenon than a microbiological problem. This approach will also reduce the use of antibiotics, limit the development of resistance and overgrowth of other micro-organisms, while reducing significantly the morbidity and mortality attributable to this form of sepsis syndrome.

Figure 1.

Least-squares regression curves of the daily mucositis score and granulocytes of 28 allogenic hematopoietic stem cell transplant recipients who had received idarubicin, cyclophosphamide and total body irradiation as conditioning therapy. The course of oral mucositis parallels that of neutropenia [1 8].

Figure 2.

Algorithm for the management of the febrile neutropenic patient with mucositis. Patients who have developed mucositis following treatment with high-dose cytarabine, idarubicin or cyclophosphamide are considered to be at a high risk of developing bacteremia. This will be known within 72 h of starting empirical monotherapy, allowing other drugs to be added as and when necessary.

Coagulase-negative staphylococcal (CoNS) bacteremia

Blood cultures of 409 HSCT recipients showed that CoNS bacteremia exceeds OVS bacteremia rates 3-fold (45% vs. 16%) [11]. CoNS bacteremia, especially Staphylococcus epidermidis is frequently related to inserted CVCs, as CoNS are predominant members of the skin flora. However, CoNS are also present in the endogenous flora of the mucous membranes of mouth and intestines of neutropenic patients [32]. Moreover, serial surveillance cultures of patients with acute leukemia showed a significantly higher gingival and rectal S. epidermidis colonization index for patients given oral prophylaxis consisting of gentamicin plus nystatin. When vancomycin was reincorporated into the oral prophylactic regimen, the incidence of S. epidermidis bacteremia declined from 7.6 per 1000 days to 2.0 per 1000 days. A prospective study reported a significant increase of CoNS bacteremia from 3 to 19 episodes per 1000 admissions from 1986 to 1993 [33]. Almost half of the CoNS were related to the presence of central venous catheter, whilst in 37% the source remained unknown. Unfortunately, the results of the surveillance cultures were not reported, but the alimentary tract might have been the origin of the CoNS. In 44 hematology patients plasmid pattern analysis was performed on 340 surveillance isolates of nares, throat, skin, rectum and urine and on 201 bloodstream isolates [34]. Patients were colonized with numerous different CoNS strains (median 5 per patient) and strains isolated from blood cultures matched those isolated from mucosal sites in 70% of cases, compared with only 30% of those recovered from the skin. More recently, chromosomal DNA analysis was used to demonstrate that the first episode of bacteremia due to S. epidermidis and S. oralis originated from the mouth, whereas the CVC was the origin for a second episode involving another strain of S. epidermidis[35]. This or a similar technique could be used routinely for surveillance cultures to undertake a prospectively randomized study in HSCT transplant recipients to investigate whether oral decontamination of the mucosa in high-risk patients actually leads to a reduction of CoNS bacteremia as well as CVC colonization and CVC-related infections. Indeed, if the CoNS causing bacteremia originate from a colonized CVC the rational approach would be to remove the catheter, whereas if the mouth or mucous membranes in general are the source, then oral decontamination would be a more logical and preferable way to prevent CoNS bacteremia. A simpler, pragmatic approach to determine the possible origin of the CoNS bacteremia would be to note the day of onset of bacteremia and the number of lumen blood cultures yielding growth. CoNS bacteremia occurring in the second or third week after starting chemo-(radio-) therapy, in which each lumen becomes positive simultaneously, would indicate the mucosa to be the source, especially if a similar strain is isolated from contemporaneous oral and fecal cultures since this suggests mucosal colonization just when MBI is at its most severe. In this case the CVC would not need to be removed. Nor would giving a glycopeptide parenterally be of any use since these drugs are not excreted into saliva or the gut to any appreciable extent. By contrast, recovery of a strain from only one lumen would suggest true colonization of the CVC, which should be removed as soon as possible. Only when CoNS bacteremia persists (i.e. at least two consecutive sets of blood cultures from a peripheral vein taken on different days and yielding a similar strain) or there is evidence of progressive skin/soft tissue infection should empirical therapy be modified by adding a glycopeptide parenterally (Figure 2).


Much effort has been made in establishing optimal antibiotic therapy for Gram-positive infections in febrile neutropenic patients. Apart from neutropenia, a crucial factor in the emergence of Gram-positive infections is the damage caused by cytostatic chemotherapy to the mucosal membranes of mouth and intestines, as these constitute the frontline of defence against invading micro-organisms. The pattern of mucositis is predictable and high-risk patients can be selected based upon the type of cytostatic therapy given. Cytarabine given in high doses, cyclophosphamide and idarubicin are associated with a high incidence of OVS and CoNS bacteremia in neutropenic allogeneic HSCT recipients. Hence, rather than adding more antibiotics to empirical therapy in a vain attempt to overcome these infections, which only increases the risk of resistance and toxicity, more efforts should be put into developing the means of protecting the mucosa from damage. Until that time, a wait-and-see approach is the only sensible option. Similarly, the management of ARDS and sepsis syndrome associated with OVS bacteremia should be managed by adding a corticosteroid and not an antibiotic to empirical therapy. The addition of antibiotics targeted against Gram-positive bacteria to complement standard empirical regimens should be reserved for those cases in which bacteremia persists or clinically defined infections progress.