Which children with fever and neutropenia can be safely treated as outpatients?



Craig A. Mullen, M.D., Ph.D., Department of Paediatrics, University of Texas M. D. Anderson Cancer Center, Box 88, Room B7·4518, 1515 Holcombe Boulevard, Houston, TX 77030, USA. E-mail: mullen@mdacc.tmc.edu

Myelosuppression is the dose-limiting toxicity for most drug regimens in paediatric oncology. Consequently, fever and neutropenia commonly complicate cancer therapy. Management of fever and neutropenia consumes a substantial portion of the effort of most paediatric oncologists and of health care system resources. The purpose of this brief annotation is to offer an opinion about what role outpatient care can play in the management of this condition in children with cancer. A group from St. Jude Children's Hospital has recently reviewed outpatient management of fever and neutropenia in children (Patrick & Shenep, 1999). The published work on outpatient management of adult cancer patients will not be reviewed (Rolston et al, 1996; Sundararajan et al, 1997) owing to space limitations, although it should be noted that successes in adult oncology have encouraged trials in paediatrics.

Biological factors

Chemotherapy damages normal host integument and depresses the number of polymorphonuclear cells and monocytes. Together, these deficits place patients at high risk of bacterial and fungal infection. Bodey et al (1966) demonstrated the relationship between neutropenia and the incidence of fever. Numerous studies since then have clearly shown that both the depth and duration of neutropenia are risk factors for infection (Pizzo, 1993; Patrick & Shenep, 1999). More serious complex infections occur in patients with protracted, profound neutropenia. Two factors influence the depth and duration of neutropenia: the status of the marrow at the time of chemotherapy and the dose intensity of the chemotherapy. Patients with substantial replacement of normal marrow elements with malignant cells or those with empty marrows after myeloablative therapies suffer more prolonged neutropenia. The second important risk factor for infection is damage to normal skin and mucous membrane barriers. A number of commonly used drugs (e.g. anthracyclines, methotrexate, alkylating agents) produce severe mucositis.

Conventional care

Over the past 30 years, the standard of care for children with fever and neutropenia has been the empirical administration of broad-spectrum intravenous (i.v.) antibiotics in hospital within hours of the manifestation of fever and the continuation of this therapy in hospital until resolution of both fever and neutropenia (Pizzo, 1993; Hughes et al, 1997). This approach is extremely effective in preventing death or serious morbidity. The death rate for fever and neutropenia in children is 1% (Hann et al, 1997). No alternative approach will be able to improve upon these results. Therefore, proposals for alternative approaches will have to demonstrate that they do not make matters worse in terms of mortality and morbidity, and justify change on the basis of quality of life and/or economics.

Treatment failure in fever and neutropenia

In most reported clinical trials of management of fever and neutropenia, the end-points reported are either duration of fever or the fraction of cases in which the initial antibiotic regimen needed to be modified because of persistent fever. This is understandable as these are quantifiable variables that lend themselves to statistical comparisons. In most cases, fever and neutropenia will resolve at the time of recovery of bone marrow production of monocytes and polymorphonuclear cells regardless of the particular broad-spectrum antibiotic regimen administered. The average number of days of fever (three) is quite close to the average number of days from presentation to recovery from neutropenia (four to five) (Aquino et al, 1997; Mullen et al, 1999a).

In 1% of patients, fever and neutropenia result in death (Hann et al, 1997). Treatment can fail in two common ways. The first is septic shock. The other is secondary infection in the setting of protracted neutropenia. When septic shock occurs, it often occurs early, i.e. in the first 2 d. Baorto and colleagues recently reviewed more than 1100 episodes of paediatric fever and neutropenia at three USA cancer centres and observed that 7 out of the 11 intensive care unit (ICU) transfers occurred within 1 d of presentation (Baorto, Washington University, St. Louis, MO, USA, personal communication), and none occurred after the third day of treatment. Exceptions do occur and an important exception to this observation is the patient with pneumonia whose pulmonary function may worsen over several days. The second pattern of treatment failure is secondary infection in a patient with protracted neutropenia, i.e. usually more than 7 d of neutropenia after manifestation of fever. Secondary bacteraemias with organisms resistant to the original drugs or fungal infections are seen (Pizzo et al, 1979, 1982; Wingard et al, 1987).

Requirements for successful outpatient management of fever and neutropenia

A satisfactory approach to outpatient management of fever and neutropenia must have a failure rate similar to that seen with conventional inpatient care. Mortality or serious morbidity owing to suboptimal management in the outpatient setting will outweigh the economic and social benefits of outpatient care. Some thoughtful clinicians may conclude that this standard is unattainable and therefore no patients with fever and neutropenia should be treated as outpatients. The experience at our centre and several others indicates that safe, satisfactory outpatient care can be given in selected settings. This depends upon selection of patients at low biological and sociological risk and a health care delivery system that can efficiently deliver outpatient care.

Biological risk Over the past decade, a considerable body of work has shown that patients with fever and neutropenia are a clinically heterogeneous group (Talcott et al, 1988; Buchanan, 1993; Rolston et al, 1996). At one end of the spectrum one sees the desperately ill leukaemia patient undergoing induction chemotherapy and, at the other, the leukaemic in maintenance phase with an absolute neutrophil count (ANC) of 0·4 × 109/l who has a fever but is otherwise well. While most clinicians would acknowledge that the former is high risk and the latter low risk, the clinical task is to devise objective criteria that can be used prospectively to identify low-risk patients whose clinical features fall between the extremes above.

The majority of studies trying to establish risk criteria have used bacteraemia as the defining feature of a patient at high risk (Jones et al, 1996; Lucas et al, 1996; Rackoff et al, 1996). In 15–20% of episodes, bacteraemia will be documented (Lucas et al, 1996; Rackoff et al, 1996), although the rate is substantially higher (10–49%) in patients receiving myeloablative therapy (Mullen et al, 2000). Some studies have demonstrated that both profound neutropenia (e.g. ANC < 0·2 × 109/l) and severe mucositis are associated with a higher probability of bacteraemia (Lucas et al, 1996; Rackoff et al, 1996). Some studies have suggested that extremely high fevers (> 39°C) may also have a statistically significant association with bacteraemia (Rackoff et al, 1996). However, none of these factors have proved to be efficient at prospectively identifying a low-risk group in which the risk of bacteraemia is virtually nil or in identifying patients certain to have bacteraemia. Even if such criteria for identifying bacteraemic patients were developed, it is not clear that they would be useful in identifying patients suitable for outpatient care. Bacteraemia is an imperfect surrogate measure for a severe infection. Not every fatal or severe infection is associated with bacteraemia. Moreover, the majority of bacteraemic patients have a satisfactory, uncomplicated response to antibiotics. In our recently reported trial, four patients with bacteraemia were treated exclusively as outpatients (three with oral antibiotics); all had a good clinical response to therapy at the time the cultures were reported as positive and, thus, continued outpatient care (Mullen et al, 1999a).

Most trials of outpatient paediatric fever and neutropenia have thus used relatively simple clinical criteria to define low-risk patients by exclusion of ‘ill’ patients (Table I describes eligibility criteria). Patients undergoing leukaemia induction therapy or bone marrow transplantation have been excluded as they predictably have profound, protracted neutropenia. Clinically unstable patients, and those with hypotension, respiratory distress or haemorrhage, have also routinely been placed in the high-risk groups. Patients with defined soft tissue infections or evidence of pneumonia have also been excluded. Most trials have also excluded patients with severe mucositis. With severe mucositis, an increased risk of bacteraemia (especially with gram-positive organisms) has been observed (Wingard, 1990; Ruescher et al, 1998; Peterson, 1999). An additional practical consideration in patients with mucositis is that inpatient care is often necessary for administration of i.v. narcotics for pain and i.v. fluids for treatment and/or prevention of dehydration.

The practical definition of a low-risk patient based on these clinical trials is one who has not received near-myeloablative therapy and who, at presentation, has stable vital signs, no evidence of pneumonia, no focal tissue-based infection and no significant mucositis, regardless of actual temperature or absolute neutrophil count. Approximately 5% of a population defined in this manner will prove to be bacteraemic (Mullen et al, 1999a).

Sociological risk With outpatient care, administration of antibiotics and monitoring of condition become the responsibility of the family. These tasks can be reliably done by some families but not by others. The assumption that compliance will occur because the condition being treated is potentially life-threatening is not valid. Studies of compliance with outpatient oral chemotherapy in children with cancer indicate that more than 30% of patients will not be fully compliant (Tebbi et al, 1986; Lau et al, 1998). The carers must have sufficient knowledge to be able to recognize adverse changes in condition, have complete commitment to regular administration of antibiotics and follow-up clinic visits, and have transportation and communication resources in the home in order for them to be able to speak with staff and/or bring the child to clinic should there be a change in condition. From a practical standpoint, at our institution the following factors were used as exclusion criteria for outpatient care: any history of non-compliance, lack of language skills which prevents adequate teaching of the carers and which may prevent effective reporting by the carer to the staff, lack of reliable transportation or lack of a telephone in the residence. In one trial, such criteria excluded 8% of episodes from eligibility for outpatient care (Mullen et al, 1999a).

Health care system competence Outpatient management of fever and neutropenia requires careful follow-up by the medical team. Providing outpatient care requires a significant commitment of availability to the patient. If outpatient management of fever and neutropenia is to be offered, several criteria of health care system competence should be met. First, the treating physician should be experienced in treating children with cancer. Second, the child should be monitored daily by the staff. Third, 24-h telephone support and clinic/emergency room access should be available should the child's condition change. Fourth, the outpatient clinic and laboratory should run efficiently, so as not to turn check-ups into an all-day affair for the family that would largely erode the benefits of outpatient care. Most paediatric oncology services should be able to provide this level of service.

One potential problem with advocating outpatient care for some patients with fever and neutropenia is that both medical carers and families may lose some of the appropriate respect for this potentially fatal condition. For example, it is conceivable that in a setting in which outpatient care is considered a component of standard care, some practitioners may simply see the patient once at presentation, dispense antibiotics and schedule follow-up in 1 week. No clinical trial of outpatient care for fever and neutropenia has used this approach. Similarly, families that experience outpatient care for fever and neutropenia may fail to appreciate the potential gravity of the condition. [Indeed, in our first trial we encountered two episodes of non-compliance (2·7% of episodes). One patient on oral antibiotics failed to return for daily clinic visits. In another, an adolescent with Ewing's sarcoma who was treated successfully for fever and neutropenia as an outpatient developed fever after her next cycle of chemotherapy. She did not promptly report her fever to her physician, but was simply treated by her family with some unused oral antibiotics remaining from the prior episode.] The very low mortality rate for conventionally treated fever and neutropenia is largely owing to the immediate reporting of fever and rapid initiation of empirical antibiotic therapy. If a paediatric oncology service chooses to treat some episodes of fever and neutropenia in the outpatient clinic, it should also provide a high level of patient education about the nature of the problem.

Clinical trials

Several centres have reported relatively small trials of outpatient management of fever and neutropenia in paediatric oncology. [One large series of 1300 episodes in 135 patients has also been reported (Sahu et al, 1997), but is not discussed here as the report lacks detail on outcomes.]Table I summarizes some important aspects of these studies. Most have used the same eligibility criteria for a low-risk patient described above, i.e. excluded patients who at presentation were haemodynamically unstable, had pneumonia, serious focal infections or had advanced malignant disease. Many specifically excluded patients with severe mucositis, vomiting or dehydration. In all, patients were treated until both fever and neutropenia resolved; typical durations of fever were 2–4 d and of treatment 3–7 d. A variety of oral and i.v. antibiotic regimens were used. Treatment protocols called for hospitalization for protracted fever (e.g. > 4–5 d), worsening condition or (in some studies) positive blood cultures. Preis et al (1997) reported on 64 episodes in Germany treated with i.v. ceftriaxone which, in some cases, was supplemented with teicoplanin. Twelve patients (19%) required hospitalization. Lau et al (1994) reported 23 episodes in Canada in which patients initially received 72 h of in-hospital treatment with i.v. antibiotics and then were switched to oral antibiotics; two patients (9%) required resumption of i.v. antibiotics. Kaplinsky et al (1994) reported 50 episodes in Israel treated with outpatient i.v. ceftriaxone. The first dose was given in clinic and subsequent doses at home by parents or a home health nurse. Nine patients (18%) were subsequently admitted to hospital for fever persisting > 96 h or positive blood cultures. Shemesh et al (1998) reported 60 episodes in Israel treated in a similar manner but with a variety of i.v. antibiotic regimens. Thirteen patients (22%) were ultimately hospitalized, the majority for blood cultures positive for Pseudomonas. Mustafa et al (1996) described 19 episodes in Dallas treated with once-daily i.v. ceftriaxone; one patient was hospitalized after 48 h because of a deteriorating condition. Malik (1997) reported 91 episodes in Pakistan treated with oral ofloxacin. Patients were selected for this treatment if they lived more than 12 h from the medical centre and treated themselves after contacting the oncologist. Eight patients (9%) required hospitalization for persistent fever or declining condition; the majority of these hospitalized patients had protracted neutropenia. We recently reported 73 episodes in Houston in which patients received one dose of i.v. ceftazidime during a 3–16-h period of extended observation in clinic at presentation (Mullen et al, 1999a). Patients who were stable were then subsequently randomized to treatment with i.v. ceftazidime via portable infusion pump or oral ciprofloxacin. Ten patients (14%) required subsequent hospitalization for fever lasting > 5 d (four), emesis (three), non-compliance (one), deterioration (one child with a worsening viral respiratory tract infection) or a protocol violation (one). Regression analysis identified duration of neutropenia as the only statistically significant variable associated with the need for hospitalization. Patients with leukaemia often had longer periods of neutropenia. There was no statistically significant difference in outcome between the oral and i.v. treatment groups. Petrilli et al (2000) reported 138 episodes in Brazil that included only patients with solid tumours or non-Hodgkin's lymphoma and in which patients were randomized to receive oral ciprofloxacin or i.v. ceftriaxone. They did not observe a statistically significant difference in success without modification of the treatment regimen in the two treatment groups (83% in the ciprofloxacin group compared with 75% in the ceftriaxone group). However, 5% of episodes in the oral antibiotic group experienced adverse gastrointestinal symptoms compared with none in the i.v. therapy group.

Table I.  Published reports of outpatient management of fever and neutropenia in paediatric cancer patients.
ReferenceInclusion criteriaExclusion criteriaTreatmentNo.DeathsICUHosp.% + BC%FUO
  1. Inclusion criteria, those in addition to fever and neutropenia; No., the number of episodes treated; ICU, the number of patients requiring ICU care; Hosp., the number of episodes in which outpatients required hospitalization; % + BC, the percentage of episodes in which blood cultures were positive; %FUO, the percentage of episodes of fever of unknown origin; NR, not reported; NA, not applicable; CVC, central venous catheter.

Preis et al (1997)Clinically stableSeptic shocki.v. ceftriaxone640012NRNR
Lau et al (1994)Negative blood culturesSepsisi.v. gentamicin, ticarcillin
and cloxacillin × 72 h,
followed by p.o. cefixime
Kaplinsky et al (1994) Hypotension, dehydration,
rigour, bleeding, pneumonia
Shemesh et al (1998)< 1 h from medical centre,
competent carers, CVC present
Any serious co-morbidity,
CNS or pulmonary infection
Mustafa et al (1996)ANC > 0·1 × 109/l or platelet
count ≥ 75 × 109/l, reliable parents,
close to medical centre, > 1 year of age
Any serious co-morbidity, leukaemia
induction, serious mucositis, unable
to take p.o., abdominal tenderness
i.v. ceftriaxone190010%74%
Malik (1997)Capable of taking oral medicationsSevere mucositis, vomiting, hepatic
insufficiency, serious co-morbid conditions
p.o. ofloxacin850082·3%84%
Mullen et al (1999a)
Taking oral fluids, close to medical
centre, > 2 years of age
Haemodynamically unstable, pneumonia,
severe mucositis, leukaemia induction,
i.v. ceftazidime vs. p.o.
ciprofloxacin (randomized)
Petrilli et al (2000)Solid tumours, non-Hodgkin's
lymphoma, > 3 years of age
High-dose chemotherapy, leukaemia,
haemodynamically unstable
i.v. ceftazidime vs. p.o.
ciprofloxacin (randomized)

Together, these eight papers describe 494 episodes of low-risk fever and neutropenia in paediatric cancer patients treated in one manner or another with outpatient care. In none of the eight studies were any deaths or ICU transfers attributed to outpatient treatment of fever and neutropenia. The percentage of patients ultimately hospitalized ranged between 4% and 22%. While these reports have concluded that outpatient care for fever and neutropenia is feasible and safe, one cannot, with complete certainty, conclude that outpatient and inpatient care for low-risk patients are truly equivalent in efficacy. As the death and ICU transfer rate is on the order of 1% in conventionally treated patients of all risk groups, a comparative trial of low-risk patients involving hundreds of episodes would be required to establish equivalence. It is improbable that such a trial will ever be conducted.


Several reports have demonstrated that outpatient care for fever and neutropenia is less costly (Mustafa et al, 1996; Mullen et al, 1999b). Charges for outpatient care have been estimated at $900–$1900 per episode compared with estimated charges of $4500–$6500 for inpatient care of identical duration. Some of the savings result from shifting responsibilities from the health care system to families. Despite this, in two studies in which patient and family satisfaction was reported, nearly all families were quite pleased with outpatient care (Shemesh et al, 1998; Mullen et al, 1999b).

Oral vs. intravenous antibiotics

Over the past decade, published trials of outpatient fever and neutropenia in paediatric and adult oncology have employed both oral and intravenous antibiotics, with neither being substantially superior to the other (Sundararajan et al, 1997). Recently, two large comparative trials of oral (ciprofloxacin plus amoxicillin/clavulanate) and intravenous (e.g. ceftazidime or ceftriaxone plus amikacin) antibiotics in low-risk hospitalized patients have failed to show significant differences in outcome (Freifeld et al, 1999; Kern et al, 1999). In our outpatient trial, many patients and families expressed a strong preference for oral antibiotics because of the ease of administration and reduced time needed in clinic for replacing antibiotic solutions in portable infusion pumps. In addition, oral regimens are significantly less expensive than intravenous regimens (the median charges are ≈$500 less per episode) (Mullen et al, 1999b). However, there are circumstances in which i.v. regimens are preferable. First, compliance is more easily enforced with i.v. antibiotics. This may be an important issue with some adolescent patients and may also be important for some very young children who have difficulty taking oral medications. Second, nausea and vomiting complicate about 10% of episodes treated with oral antibiotics (Freifeld et al, 1999; Kern et al, 1999; Mullen et al, 1999a). This may be as a result of the antibiotics themselves and/or underlying mucositis from the chemotherapy.


Children with fever and neutropenia at low risk of complex infections can be prospectively identified by clinical criteria mentioned above and in Table I, and managed safely in the outpatient setting with either oral or intravenous broad-spectrum antibiotics. No single antibiotic regimen is clearly better than others. For parenteral use, third- or fourth-generation cephalosporins (ceftazidime or cefepime) have been effective as monotherapy, while oral quinolones supplemented with another oral gram-positive agent have been satisfactory (e.g. ciprofloxacin plus amoxicillin/clavulanate or clindamycin). Choice of antibiotics should be based on local patterns of infection and drug resistance. An extended period of observation (several hours to overnight) in clinic during which antibiotic treatment is initiated may increase the clinical confidence with which a particular patient is deemed suitable for outpatient care. Closely daily follow-up is necessary as approximately 15% of such patients may ultimately require hospitalization for protracted fever, vomiting or a worsening infection.