Clinical evidence (see also full evidence review)

Evidence statements

Short term mortality (febrile neutropenia studies)

A multivariate analysis by Larche, et al., (2003) found that 30 day mortality was higher when time to antibiotic therapy was more than two hours (odds ratio (OR) = 7.05 (95% CI, 1.17 to 42.21 (P = 0.03)). (Table 6.1).

Table 6.1

GRADE profile: Does the length of time before empiric antibiotics are given influence patient outcome.

A multivariate analysis by Lin, et al., found that mortality was higher in patients with an absolute neutrophil count (ANC) of <0.1 × 10⁹/L when time to antibiotic therapy was > 24 hours in a non-ICU setting (OR = 18.0; 95% CI, 2.84 - 114.5; P < 0.01); and in an ICU setting (OR, 5.56; 95% CI, 0.85 - 36.3; P = 0.07). However, for patients who were non-neutropenic (ANC, >0.5 × 10⁹/L) or had ANCs of 0.1 - 0.5 × 10⁹/L, delay was not associated with increased mortality in ICU (OR (ANC 0.1 - 0.5 × 10⁹/L) = 0.59; 95% CI, 0.06 - 6.22; P = 0.66; OR (ANC > 0.5 × 10-⁹/L ) = 0.55; 95% CI 0.29 - 1.02) or non-ICU (OR (ANC 0.1 to 0.5 × 10-⁹/L) = 1.92; 95% CI, 0.17 to 21.3; P = 0.60; OR (ANC > 500) = 1.78; 95% CI 0.89 to 3.44).

This evidence is of very low quality and is indirect on the basis that patients had bacteraemia or septic shock.

Overtreatment, severe sepsis, length of stay, duration of fever and quality of life

These outcomes were not reported by the identified studies. The outcome of severe sepsis was not relevant to the included studies, which included only participants who had bacteraemia or severe sepsis at study entry.

Cost-effectiveness evidence

A literature review of published cost-effectiveness analyses did not identify any relevant papers. This topic focused on the optimal timing of a change in management strategy. The difference between strategies were considered unlikely to lead to large differences in cost, but rather be guided by differences in patient outcomes and other considerations such as service configuration that. It was agreed that these considerations would probably be difficult to accurately capture using economic modelling and therefore further health economic analysis was not undertaken

Linking Evidence to Recommendations

The aim of this topic was to see if the length of time before empiric antibiotics are given influences a patient's outcome. For this topic the GDG considered the outcomes of overtreatment, mortality, severe sepsis, length of stay, duration of fever and quality of life to be the most relevant to this patient population as these are the adverse consequences of unnecessarily being given antibiotics and staying in hospital. No evidence was reported for any of these outcomes.

The search was therefore widened to include patients with general suspected bacterial infections. Evidence was found for short term mortality but this was not directly relevant to the patient population and the study reported patients who had bacteraemia or septic shock. The GDG noted that the evidence was classified by GRADE as being of ‘very low’ quality and no studies defined the optimal time for administering antibiotics. The GDG agreed that data for short term mortality could be used as it was the only data available.

The GDG also acknowledged the one hour to antibiotic pathway from the National Cancer Peer Review Programme, Manual for Cancer Services. The GDG felt that there was insufficient evidence to support recommending a specific time period for administering antibiotics. However the GDG recognised that benefits such as increased patient survival and a reduction in complications could be gained from administering antibiotics as soon as possible.

The GDG noted that no relevant, published economic evaluations had been identified and no additional economic analysis had been undertaken in this area. The opinion of the GDG was that there may be potential cost implications of unnecessary treatment. However they felt that improvements in patients' survival outweigh any potential costs. The GDG also noted that adverse events for the patient, and the costs associated with dealing with these would be avoided as a result of urgent antibiotic intervention.

Therefore the GDG decided to recommend that patients with suspected neutropenic sepsis should be treated as an acute medical emergency and receive empiric antibiotic therapy without delay.

Clinical evidence (see also full evidence review)

Evidence statements

Evidence from trials directly comparing single agent with combined treatment

There was moderate quality evidence from 44 studies extracted from a systematic review by Paul et al (2007) with over seven thousand episodes of neutropenia and fever which did not show a significant difference in the risk of all cause mortality between monotherapy and combined therapy. This evidence is summarised in Table 6.2.

Table 6.2

GRADE profile: Is empiric IV antibiotic monotherapy more effective than empiric IV antibiotic combined therapy in the treatment of patients with neutropenic sepsis.

Moderate quality evidence from 55 studies showed that treatment failure was less likely with monotherapy than combined therapy, when combined therapy used a narrower spectrum antibiotic than was used for monotherapy (52 studies from Paul et al, 2007; Pereira et al., 2009; Yildirim et al., 2008 and Zengin et al., 2011). Fifteen studies where the same beta- lactam was used for both monotherapy and combined therapy, however, found treatment failure more likely with monotherapy.

Moderate quality evidence showed that monotherapy was associated with fewer adverse events, including nephrotoxicity (Paul et al, 2007).

Moderate quality evidence showed that monotherapy and combined therapy had similar rates of bacterial secondary infection.

Low quality evidence showed fungal secondary infection was more likely with combined therapy.

Very low quality evidence from two studies with 152 patients suggested that colonisation of resistant Gram-negative bacteria was more likely with monotherapy, but such bacteria were only detected in six patients overall.

There was no evidence about quality of life and no useful evidence about the duration of hospital stay.

Evidence from mixed treatment comparison

A mixed treatment comparison was done for this guideline using 108 trials identified in two Cochrane reviews by Paul, et al., (2007 and 2010). These trials were either comparing single agent beta-lactams with each other (Paul, et al., 2010) or comparing single agent beta-lactams with combined beta-lactam/aminoglycoside treatment (Paul, et al., 2007).

The summary estimates from the mixed treatment comparisons showed good model fit (residual deviance ∼ 126, compared with 148 data points). The Deviance Information Criterion was minimised when covariates indicating year of publication, age of patients, and proportion of haematological malignancy were not entered into the model. Additionally, none of these covariates were significant (i.e. their 95% credible intervals all crossed log-zero; no effect).

The treatment most likely to be best at reducing overall mortality was the use of a single agent ureidopenicillin. This was reflected in direct and indirect estimates (Tables 6.3 to 6.5). Carbapenems alone compared with ureidopenicillin had higher overall mortality, equivalent infectious mortality and marginally less risk of ‘treatment failure’.

Table 6.3

Results of mixed treatment comparison of empiric antibiotic monotherapies and empiric combined therapies.

Table 6.4

Comparison of results from pairwise and mixed treatment comparisons of empiric antibiotic monotherapies and empiric combined therapies for mortality.

Cost-effectiveness evidence

A literature review of published cost effectiveness analyses identified two relevant papers, Corapcioglu and Sarper (2005) and Paladino, (2000). The results of both studies are summarised in Table 6.5. No further health economic analysis was undertaken as the cost difference between monotherapy and dual therapy was relatively low.

Table 6.5

GRADE profile: Cost effectiveness of antibiotic monotherapy compared with antibiotic dual therapy.

Linking Evidence to Recommendations

The aim of this topic was to consider what was the most effective empiric intravenous antibiotic treatment of patients with neutropenic sepsis.

The GDG considered the outcomes of overall mortality, adverse effects and allocated treatment failure to be the most clinically relevant to the question. The adverse effects that the GDG considered included nephrotoxicity, the development of antibiotic resistance and development of Clostridium difficile infection. The GDG decided that overall mortality was more important than allocated treatment failure, based on available evidence from studies and current clinical practice. The overall quality of the evidence as classified by GRADE was ‘moderate’ in addressing mortality and treatment failure, and low or very low in relation to adverse effects.

To aid the GDG in making a recommendation they undertook a meta-analysis derived from data from published systematic reviews and using a mixed treatment comparison analysis. This demonstrated reduced mortality with empiric ureidopenicillin monotherapy, compared to carbapenem therapy or treatment with the addition of aminoglycosides, with reduced nephrotoxicity in this group. This was despite an increased chance of needing to alter therapy during the episode. Subgroups relating to age, cancer type and methodology of the studies included did not show significant differences in outcomes, and so were considered to support a universal recommendation. Additionally, concerns about the use of cephalosporins and their effect in promoting Clostridium difficile infection limited the recommended monotherapy to piperacillin with tazobactam. Local microbiological resistance patterns were also felt to be very important, as high rates of resistance to the chosen empiric agent could lead to treatment failure and avoidable mortality.

The GDG noted that patients with penicillin allergy would not be able to receive piperacillin with tazobactam. However the evidence appraised did not support recommending a specific alternative for this group of patients. The opinion of the GDG was that in this situation, clinicians would need to be able to use their clinical judgement – taking into account whether the allergy was severe (anaphylaxis, angio-oedema and bullous skin eruptions) in which case no beta-lactams should be given, or the more common adverse events of skin rash or nausea where an alternative beta-lactam may be considered. The GDG therefore did not make a specific recommendation on what empiric antibiotic therapy to give patients with a penicillin allergy.

The GDG noted that no relevant, published economic evaluations had been identified and no additional economic analysis had been undertaken in this area. Both of these papers were partially applicable to the question, but both had serious limitations. The conclusion derived from these papers was that monotherapy can be cost effective compared to dual therapy.

The GDG considered the possible clinical scenarios for resource usage, potential costs of delivering excess drug, with intensive monitoring of aminoglycoside levels and subsequent costs of toxicity, against the potential reduced likelihood of resistance to both chosen empiric agents being present.

Therefore the GDG decided to recommend that patients with suspected neutropenic sepsis should be offered beta lactam antibiotic monotherapy with piperacillin with tazobactam as initial empiric treatment, unless there are local microbiological contraindications. They also agreed that aminoglycoside, either in mono or dual antibiotic therapy should not be used for the initial empiric treatment of patients with suspected neutropenic sepsis unless there are local microbiological indications.

Clinical evidence (see also full evidence review)

Evidence statements

The evidence for all outcomes is summarised in Table 6.6.

Table 6.6

GRADE profile: What is the role of empiric glycopeptide antibiotics (antibiotics chosen in the absence of an identified bacterium) in patients with central lines and suspected neutropenia or neutropenic sepsis.

Short term mortality

Five studies reported short term mortality (de Pauw, et al., 1990; EORTC, 1991; Ramphal, et al., 1992; Molina, et al., 1993; Novakova, et al., 1991). There was very low quality evidence of uncertainty about the difference between antibiotics administered alone, and the same empiric antibiotics administered with the addition of glycopeptides, RR = 0.97 (95% CI 0.63 – 1.50) in four studies with1083 participants.

Critical care, length of stay and line preservation

These outcomes were not reported by any of the included studies.

Antibiotic resistance

Only one study reported antibiotic resistance (Novakova, et al., 1991). Rates of resistance were very low in both groups (2/51 (4%) in the group who received empiric antibiotics alone and 0/52 (0%) in the group who received empiric antibiotics plus glycopeptides).

Proven Bacteraemia

Two studies with 150 participants reported proven bacteremia as an outcome (Del Favero, et al., 1987; Novakova, et al., 1991). There was very low quality evidence of uncertainty about whether antibiotics administered alone or empiric antibiotics administered with glycopeptides was more effective in terms of proven bacteraemia, RR = 0.80 (95% CI 0.42 – 1.53).

Nephrotoxicity

In five studies with 1160 participants, there was very low quality evidence of a significant difference between antibiotics administered alone, and the same empiric antibiotics administered with glycopeptides, with a greater number of individuals receiving the latter regimen experiencing nephrotoxicity, RR = 0.57 (95% CI 0.33 – 0.99).

Hepatic toxicity

Two studies with 856 participants reported hepatic toxicity as an outcome. There was very low quality evidence of a significant difference between empiric antibiotics administered alone, and antibiotics administered with the addition of glycopeptides. A greater number of individuals in the latter group experienced hepatic toxicity, RR = 0.53 (95% CI 0.33 – 0.99).

Cost-effectiveness evidence

A literature review of published cost-effectiveness analyses did not identify any relevant papers. Although this topic was identified as important for further health economic analysis, due to the large patient subgroup and potentially significant differences in costs of the interventions; the lack of directly relevant clinical and economic evidence meant that it was not possible to undertake further health economic analysis.

Linking Evidence to Recommendations

The aim of this topic was to identify the benefits and risks of adding vancomycin, teicoplanin or linezolid to first line antibiotics in patients with a central venous access device with no external signs of line infection but with suspected neutropenia or neutropenic sepsis.

The GDG considered the outcomes of death, critical care, length of stay, line preservation (device remains in situ), antibiotic resistance, proven bacteraemia and toxicity to be the most clinically relevant to the question. No evidence was reported for critical care, length of stay or line preservation. Evidence was available for proven bacteraemia, toxicity, antibiotic resistance and death which was reported as short term mortality. They also considered an additional outcome reported by the evidence of the presence of a super-infection, as this was also relevant to the question.

The GDG noted that there was very little evidence available for this topic. The evidence that was available was assessed by GRADE as being of ‘low’ quality for all outcomes due to imprecision (low number of events) and indirectness (only one study reported on patients with a central line, and the standard empiric drugs used in the available studies are no longer recommended in clinical practice).

The GDG noted that the evidence had shown no significant difference in the incidence of death or proven bacteraemia between antibiotics administered alone or antibiotics administered with the addition of a glycopeptide. In addition, the GDG were aware that the evidence had shown increased harms such as kidney and liver toxicity from the empiric use of glycopeptide antibiotics. They also noted that there is no available evidence to show that not using glycopeptide antibiotics has any detrimental effect on line preservation.

The GDG noted that no relevant, published economic evaluations had been identified and no additional economic analysis had been undertaken in this area. The GDG based on their clinical experience considered that there may be potential cost savings from stopping the use of empiric glycopeptide antibiotics in this setting along with a reduction in therapeutic drug monitoring costs.

Given the lack of evidence of clinical benefit and the evidence of increased harms, the GDG recommended that empiric glycopeptide antibiotics should not be used in patients with a central venous access device.

Clinical evidence (see also full evidence review)

Cost-effectiveness evidence

A literature review of published cost-effectiveness analyses did not identify any relevant papers. Further health economic analysis was not undertaken as the topic did not lend itself to economic evaluation as there was no comparative analysis of cost and outcomes

Linking Evidence to Recommendations

The aim of this topic was to identify if patients with central venous access devices and neutropenic sepsis would benefit from the immediate removal of their central line.

The GDG considered the outcomes of mortality, severe sepsis, length of stay, duration of fever, line preservation and complications to be the most clinically relevant to the question. No evidence was reported for overall mortality, severe sepsis, length of stay or complications. Evidence was available for duration of fever, line preservation and duration of antibiotics. The GDG considered the additional outcome of infectious mortality as a surrogate marker for overall mortality. The reported evidence for duration of fever and duration of antibiotics was not considered useful by the GDG as it did not relate to empiric management.

The evidence for all outcomes was ‘low’ quality. The GDG acknowledged that the available evidence was indirect as it focused on targeted rather than empiric management and would therefore need to be extrapolated backwards. The GDG also noted that the number of events reported in the data was low, and the studies investigated disparate practice, making it difficult to compare and draw meaningful conclusions.

From the available evidence, the GDG were unable to identify a group of patients that would benefit from the removal of their central lines during the empiric phase of treatment. They considered that not removing a line would have the benefit of maintaining venous access during a period of acute illness, together with a reduction in possible traumatic or invasive interventions.

The GDG noted that not removing a line might be associated with an increased risk of complications from a central line infection such as severe sepsis. The GDG recognised that there will be some patients with uncontrolled infection where there is a strong clinical suspicion of a central line infection, who may require central line removal. The management of specific infections is outside the scope of the guideline. The GDG emphasised that this would only relate to a small proportion of episodes, and have used the phrased “initial” empiric management to indicate that line removal should not be the first response in every patient with a central line. The opinion of the GDG was that the benefits from not removing the line outweigh any risks associated with removing the central line empirically.

The GDG noted that no relevant, published economic evaluations had been identified and no additional economic analysis had been undertaken in this area. The opinion of the GDG was that there may be potential additional costs associated with extending treatment in those patients who have a proven line infection. However the GDG also noted that there may be potential cost savings by avoiding the replacement of central lines. The GDG were unable to determine whether the costs and savings would balance but believed that the clinical benefits far outweigh any potential increase in costs.

Therefore the GDG recommend that central venous access devices should not be removed as part of the initial empiric treatment of patients with neutropenic sepsis.

Clinical evidence (see also full evidence review)

The evidence for all outcomes is summarised in Table 6.7 and Table 6.9.

Table 6.7

GRADE profile: Is inpatient management more effective than outpatient management for patients with neutropenic sepsis.

Table 6.8

GRADE profile: Is outpatient oral antibiotic treatment more effective than outpatient intravenous antibiotic treatment.

Table 6.9

Modified GRADE profile: Inpatient versus Ambulatory care (all different forms).

Cost-effectiveness evidence

A literature review of published cost-effectiveness analyses identifed two Canadian studies (Teuffel, et al., 2010; Teuffel, et al., 2011b) comparing the cost-effectiveness of inpatient care with ambulatory management strategies. The results of both studies are summarised in Table 6.10.

As there was no definition of what constitutes a specific inpatient management strategy for this question, costing and evaluating health outcomes using economic modelling was not feasible. The GDG anticipated that the different management strategies were unlikely to result in large differences in patient outcomes and those strategies that minimise or reduce the duration of inpatient care would generally be less costly. Given that there was little uncertainty surrounding the economics of this question, further health economic analysis was not undertaken.

Linking Evidence to Recommendations

The aim of this topic was to see if there is any difference between the outcomes of patients with neutropenic sepsis who are given antibiotics in hospital compared to those given antibiotics at home.

The GDG considered the outcomes of death within 30 days, critical care, clinically documented infection, length of stay, hospital re-admission and quality of life to be the most clinically relevant outcomes that would benefit patient care. No evidence was found for critical care, clinically documented infection or quality of life. Evidence was reported for mortality (30 days), hospital re-admission and length of stay. The GDG also considered an additional outcome reported by the evidence of treatment failure (a composite outcome of readmission and modification of antibiotics), which showed no significant association between outpatient management, drug administration and treatment failure. The GDG noted that the evidence was classified by GRADE as being of ‘low’ to ‘moderate’ quality.

The GDG acknowledged that the available data was limited due to the low event rate, very few patients experiencing adverse outcomes, and also the study design, (few studies used adequate sequence generation and concealment and none of the studies were blinded, few reported intention to treat analysis). The GDG also noted that the risk of treatment failure for this patient population was low, and providing they have been properly risk assessed the risk of death was minimal.

The GDG noted that there was a potential risk of treatment failure and death in the low risk population but this was minimal in the evidence. However it was the clinical opinion of the GDG that the benefits of offering outpatient antibiotic therapy would improve a patient's quality of life. Choice of antibiotics will be influenced by prior prophylaxis. One clinical indication not to offer oral antibiotics may be the use of prophylactic quinolones. For those patients without prior prophylaxis oral antibiotic regimes with a quinolone and/or co-amoxiclav have been most frequently used.

The GDG noted that no additional economic analysis had been undertaken in this area. A literature review of published cost-effectiveness analyses identified two relevant papers, both of which were partially applicable to the question. These studies had minor limitations and concluded that IV antibiotics administered at home was the most cost-effective regimen. However the GDG noted that these studies were based on once daily administration of an antibiotic that is not available in the UK.

The GDG recognised that some patients would not be suitable for out patient therapy due to their clinical and social circumstances, for example those patients who are not thought to recognise there illness or are able to return to hospital. Therefore the GDG decided to recommend that patients at low risk of severe sepsis can be considered for outpatient antibiotic therapy but did not specify a route of administration.