Evidence statements for primary prophylaxis with G(M)-CSF versus no primary prophylaxis with G(M)-CSF

The evidence for primary prophylaxis with colony stimulating factors comes from systematic reviews of randomised trials by Sung, et al., (2007), Bohlius, et al., (2008) and Cooper, et al., (2011). This evidence is summarised in Table 5.1.

Table 5.1

GRADE profile: Is primary prophylaxis with G(M)-CSF (with or without antibiotics) more effective than no primary prophylaxis with G(M)-CSF (with or without antibiotics) at improving outcomes in patients at risk of neutropenic sepsis.

Evidence statements for primary prophylaxis with G(M)-CSF plus antibiotic (quinolone or cotrimoxazole) versus primary prophylaxis with antibiotic

The trials were identified from the systematic review by Sung, et al., (2007) and from the list of excluded studies in a Cochrane review of prophylactic antibiotics versus G-CSF for the prevention of infections and improvement of survival in cancer patients undergoing chemotherapy (Herbst, et al., 2009 ). Most (18/27) of the trials used cotrimoxazole only (specifically for Pneumocystis pneumonia prophylaxis) – these were analysed separately from the nine trials that used quinolones. Three trials that used both quinolones and cotrimoxazole were included in the quinolone group for analysis. The trials were not designed to test the interaction of G(M)-CSF with antibiotics – rather prophylactic antibiotics were part of standard care (many of the these trials also used antiviral and antifungal prophylaxis). This evidence is summarised in Table 5.2.

Table 5.2

GRADE profile: Is primary prophylaxis with G(M)-CSF plus antibiotics more effective than primary prophylaxis with antibiotics at improving outcomes for patients at risk of neutropenic sepsis.

Evidence statements for primary prophylaxis with antibiotic (ciprofloxacin, levofloxacin, ofloxacin or cotrimoxazole) versus no primary prophylaxis

The evidence came from a Cochrane review of antibiotic prophylaxis for bacterial infections in afebrile neutropenic patients following anticancer treatment by Gafter-Gvili, et al., (2005). Data from trials of ciprofloxacin, levofloxacin, ofloxacin or cotrimoxazole were extracted from this review and analysed. Evidence about colonisation with resistant bacteria came from a second systematic review by the same authors (Gafter-Gvili, et al., 2007). An additional trial (Rahman and Khan, 2009) of levofloxacin prophylaxis was identified in our literature search. The evidence is summarised in Table 5.3.

Table 5.3

GRADE profile: Is primary prophylaxis with antibiotics more effective than no primary prophylaxis at improving outcomes in patients at risk of neutropenic sepsis.

Evidence statements for primary prophylaxis with quinolone (ciprofloxacin, levofloxacin or ofloxacin) versus primary prophylaxis with cotrimoxazole

Evidence came from a Cochrane review of antibiotic prophylaxis for bacterial infections in afebrile neutropenic patients following anticancer treatment by Gafter-Gvili, et al., (2005). Evidence about colonisation with resistant bacteria came from a second systematic review by the same authors (Gafter-Gvili, et al., 2007). Data from trials comparing ciprofloxacin, Levofloxacin and ofloxacin to cotrimoxazole was extracted and analysed. The evidence is summarised in Table 5.4.

Table 5.4

GRADE profile: Is primary prophylaxis with quinolone more effective than primary prophylaxis with cotrimoxazole at improving outcomes in patients at risk of neutropenic sepsis.

Evidence statements for primary prophylaxis with G(M)-CSF versus antibiotics

Evidence came from a Cochrane review of prophylactic antibiotics or G-CSF for the prevention of infections and improvement of survival in cancer patients undergoing chemotherapy (Herbst, et al., 2009). This review included two randomised trials directly comparing G(M)-CSF with antibiotics, remarkably few given the large number of trials comparing primary prophylaxis with G(M)-CSF or antibiotics to no primary prophylaxis. Schroeder. et al., (1999) compared G-CSF to ciprofloxacin plus amphotericin-B, Sculier, et al., (2001) compared GM-CSF to cotrimoxazole. The evidence is summarised in Table 5.5.

Table 5.5

GRADE profile: Is primary prophylaxis with G(M)-CSF more effective than primary prophylaxis with antibiotics at improving outcomes for patients at risk of neutropenic sepsis.

Evidence statements for primary prophylaxis with pegfilgrastim versus filgrastim

Evidence came from a systematic review and meta-analysis of prophylactic G-CSFs which included a comparison of pegfilgrastim versus filgrastim for the prevention of neutropenia in adult cancer patients with solid tumours or lymphoma undergoing chemotherapy (Cooper, et al., 2011). This review included five randomised trials. The literature search identified an additional phase II randomised trial comparing pegfilgrastim to filgrastim for prophylaxis in children with sarcoma receiving chemotherapy (Spunt, et al., 2010). The evidence is summarised in Table 5.6.

Table 5.6

GRADE profile: Is primary prophylaxis with pegfilgrastim more effective than primary prophylaxis with filgrastim at improving outcomes for patients at risk of neutropenic sepsis.

Evidence statements for primary prophylaxis with granulocyte infusion versus no prophylaxis with granulocyte infusion

Evidence came from a Cochrane review of granulocyte transfusions for preventing infections in patients with neutropenia or neutrophil dysfunction (Massey, et al., 2009). This review included ten trials, all but one of which were carried out before 1988. The evidence is summarised in Table 5.7.

Table 5.7

GRADE profile: Is primary prophylaxis with granulocyte infusion more effective than no such prophylaxis at improving outcomes in patients at risk of neutropenic sepsis.

Evidence statements for secondary prophylaxis with G(M)-CSF versus placebo or nothing (with or without antibiotics)

The literature search identified one randomised trial (Leonard, et al., 2009) published in abstract form only. This trial compared secondary prophylaxis using G-CSF with standard management (dose delay or reduction) in patients with early stage breast cancer receiving anthracyline or anthracycline-taxane sequential regimes. The evidence is summarised in Table 5.8.

Table 5.8

GRADE profile: Is secondary prophylaxis with G(M)-CSF more effective than no secondary prophylaxis at improving outcomes in patients with a prior episode of neutropenic sepsis.

Evidence statements for secondary prophylaxis with antibiotics versus no secondary prophylaxis (with or without G(M)-CSF)

No trials of antibiotics for secondary prophylaxis were identified. One low quality randomised trial compared G-CSF plus ciprofloxacin or ofloxacin to G-CSF alone for secondary prophylaxis (Maiche and Muhonen, 1993). The evidence is summarised in Table 5.9.

Table 5.9

GRADE profile: Is secondary prophylaxis with quinolone plus G-CSF more effective than secondary prophylaxis with G-CSF alone at improving outcomes in patients with a prior episode of neutropenic sepsis.

Evidence statements for secondary prophylaxis with G-CSF versus antibiotics for secondary prophylaxis

No trials were identified.

Study quality and results

All included papers were deemed partially applicable to this guideline. The most common reason for partial applicability was that the analyses did not include all options considered relevant for the topic. For example, most economic studies about G(M)-CSF omit quinolones. Other reasons for partial applicability included: analysis conducted in countries other than the UK, health effects not expressed in QALYs.

Seven papers were deemed to have very serious limitations. The most common reason for serious limitation was that the analyses considered the combined effectiveness of chemotherapy and G(M)-CSF, but did not count the cost of chemotherapy (Borget, et al., (2009); Danova, et al., (2008); Liu, et al., (2009); Lyman, (2009 a); Lyman, (2009 b); Ramsey, (2009). except Whyte, et al., (2011) is the only study that counted the cost of chemotherapy. However this study was also deemed to have very serious limitations as it did not use data from the best available source (ideally data should come from a recently conducted systematic review); and the estimates of clinical data used in Whyte, et al., (2011) differ substantively from the data reported by best available evidence (Note 28, Table 5.10)

The other three papers were deemed to have potentially serious limitations (Lathia, et al., 2009; Timmer-Bonte, et al., 2008; Timmer-Bonte, et al., 2006). The most common reason for potentially serious limitation was that the analyses did not use data from the best available source (ideally data should come from a recently conducted systematic review), but are similar in magnitude to the best available estimates.

Evidence statements

Eight studies were identified for patients with a solid tumour and two studies for patients with non-Hodgkin lymphoma. No economic evidence has been identified for patients with other types of cancer.

Aim

The aim of this economic analysis was to examine which of the following prophylactic strategies is the most cost-effective for cancer patients who are receiving chemotherapy

• Nothing/placebo
• Primary prophylaxis with quinolones
• Primary prophylaxis with G-CSF
• Primary prophylaxis with G-CSF and quinolones
• Primary prophylaxis with PEG-G-CSF
• Secondary prophylaxis with quinolones
• Secondary prophylaxis with G-CSF
• Secondary prophylaxis with G-CSF and quinolones
• Secondary prophylaxis with PEG-G-CSF

Evidence was reported for both quinolones and cotrimoxazole as antibiotic prophylaxis. However the GDG chose to focus on the evidence related to quinolones because of concerns that changing anti-microbial resistance patterns meant the cotrimoxazole trials may no longer be applicable (Gafter-Gvili, et al., 2005).

A subgroup analysis was conducted for the following three patient groups:

• Patients with a Solid tumour (aged over 18 years)
• Patients with Non-Hodgkin lymphoma (aged over 18 years)
• Patients with Hodgkin lymphoma (aged over 18 years)

The economic analysis does not cover:

• Cancer patients whose chemotherapy regimen includes G-CSF.
• Cancer patients with planned inpatient treatment of greater than 10-days post-chemotherapy. It is acknowledged that the costs of prophylaxis and treatment of neutropenic sepsis for inpatient-only management are lower than outpatient management.
• Paediatric cancer patients (aged less than 18 years). Due to considerable clinical heterogeneity in the treatment regimens for this patient group, and a paucity of direct evidence, a representative model for economic analysis could not be built
• The impact of different prophylactic strategies on subsequent courses of chemotherapy. The consequence of this bias is discussed in detail in section A9.2.3.
• Antibiotic resistance. The best available evidences identified to address the issue of antibiotic resistance caused by use of quinolones was derived from two systematic reviews: one was a review conducted for this guideline (see Page 79-83); and the other was a Cochrane review undertaken by Gafter-Gvili, et al., (2005). The conclusions of these two reviews were very similar. After use of quinolones, although there is an increase in colonisation with bacteria resistant to quinolones, there was no statistically significant increase in the number of infections caused by pathogens resistant to quinolones. The GDG were aware of the potential limitations of these two reviews but could not find any better evidence to answer the clinical question. Therefore the GDG decided to qualitatively consider the potential increase in antibiotic resistance and its impact on cancer patients when agreeing their recommendations, instead of quantitatively model it in the economic analysis.

Results

Adult/elderly patients with a solid tumour who can take fluoroquinolone

For adult patients with a solid tumour and who can take quinolone clinical evidence was available for all nine strategies of interest (Section A3.1.2). Compared to quinolone alone, G(M)-CSF and G(M)-CSF + quinolone are more expensive and less effective in terms of preventing neutropenic sepsis. Therefore all primary and secondary prophylactic strategies involving G(M)-CSF and G(M)-CSF + quinolone were excluded from the analysis. As a result cost-effectiveness was only formally examined for the following five strategies:

• Nothing/placebo
• Primary prophylaxis with quinolone
• Secondary prophylaxis with quinolone
• Primary prophylaxis with PEG-G-CSF
• Secondary prophylaxis with PEG-G-CSF

The incremental costs and incremental QALYs in the base case analysis for each of the five strategies are summarised in Table 5.11. Taking primary prophylaxis with quinolone as the reference (least expensive) strategy, all other strategies were shown to be less effective and also more costly except primary prophylaxis with PEG-G-CSF. Compared to the reference strategy, use of primary PEG-G-CSF produces 3.3×10^–4 more QALYs and incurs £1,899.8 in additional costs. This yields an incremental cost-effectiveness ratio (ICER) of £5.7 million/QALY, which exceeds the NICE willingness to pay (WTP) threshold of £20,000/QALY. Therefore primary prophylaxis with PEG-G-CSF was considered not to be cost effective. At a willingness to pay (WTP) threshold of £20,000/QALY, primary prophylaxis with a quinolone is the most cost-effective strategy. This conclusion was robust to structural sensitivity analysis and all one-way sensitivity analysis tested (Section A4.2) except for relative risk of neutropenic sepsis (quinolones versus nothing/placebo).

Table 5.11

Incremental costs and effectiveness by treatment strategy for solid tumour patients who can take quinolone (baseline risk of neutropenic sepsis of one course of chemotherapy: 34.41%).

When the relative risk of neutropenic sepsis (quinolones versus nothing/placebo) was above 0.787, nothing/placebo became the most cost-effective strategy, at a WTP threshold of £20,000/QALY. Primary or secondary prophylaxis with PEG-G-CSF is never the most cost-effective strategy, even when extreme scenarios were considered, for example: 100% risk of neutropenic sepsis per cycle of chemotherapy, 90% drug discount of PEG-G-CSF, extended length of hospital stay for neutropenic sepsis patients (6-day for low-risk patients and 14-day for high-risk patients) etc.

The result of the probabilistic sensitivity analysis shows that the probability of primary prophylaxis with quinolone becoming cost-effective is always 100%, at a willingness to pay between £10,000 to £40,000 per QALY.

Adult/elderly patients with a solid tumour who cannot take fluoroquinolone

For adult patients with a solid tumour who cannot take quinolone, cost-effectiveness was only formally examined for the following strategies (all strategies containing quinolone were excluded):

• Nothing/placebo
• Primary prophylaxis with G(M)-CSF
• Secondary prophylaxis with G(M)-CSF
• Primary prophylaxis with PEG-G-CSF
• Secondary prophylaxis with PEG-G-CSF.

The incremental costs and incremental QALYs in the base case analysis for each of the five strategies are summarised in Table 5.12. Taking nothing/placebo as the reference (least expensive) strategy, the other four strategies were shown to be more effective but were each associated with a very high ICER (all > £0.6 million/QALY) and were not considered to be cost-effective. Therefore at a willingness to pay (WTP) threshold of £20,000/QALY, nothing/placebo is the most cost-effective strategy. This conclusion was robust to structural sensitivity analysis and all one-way sensitivity analysis tested (Section A4.2),except for discounting the cost of PEG-G-CSF. At a WTP threshold of £20,000/QALY:

Table 5.12

Incremental costs and effectiveness by treatment strategy for solid tumour patients who can not take quinolone (baseline risk of neutropenic sepsis of one course of chemotherapy: 34.41%).

• When the discount to the cost of PEG-G-CSF was over 73.85% (corresponding price: £179.5 per single subcutaneous injection (6mg), secondary prophylaxis with PEG-G-CSF became the most cost-effective strategy.
• When the discount to the cost of PEG-G-CSF was over 84.13% (corresponding price: £108.9 per single subcutaneous injection (6mg), primary prophylaxis with PEG-G-CSF became the most cost-effective strategy.

Primary or secondary prophylaxis with G(M)-CSF is never the most cost-effective strategy for any of the three patient groups of interest, even when extreme scenarios were considered, for example: 100% risk of neutropenic sepsis per cycle of chemotherapy, 90% drug discount, reduced days of using G(M)-CSF (5-day per cycle of chemotherapy), reduced daily dose (one vial of G(M)-CSF for all adult patients regardless of weight) etc.

The result of the probabilistic sensitivity analysis shows that the probability of nothing/placebo becoming cost-effective is always 100%, at a willingness to pay between £10,000 to £40,000 per QALY.

Adult/elderly patients with non-Hodgkin lymphoma

For adult patients with non-Hodgkin lymphoma, no clinical evidence was identified for the use of quinolone alone for either primary or secondary prophylaxis therefore neither strategy was included in this analysis.

Compared to G(M)-CSF alone, G(M)-CSF + quinolone is more expensive and less effective in terms of preventing neutropenic sepsis so both primary and secondary prophylactic G(M)-CSF + quinolone strategies were excluded. As a result cost-effectiveness was only formally examined for the following five strategies:

• Nothing/placebo
• Primary prophylaxis with G(M)-CSF
• Secondary prophylaxis with G(M)-CSF
• Primary prophylaxis with PEG-G-CSF
• Secondary prophylaxis with PEG-G-CSF

The incremental costs and incremental QALYs in the base case analysis for each of the five strategies are summarised in Table 5.13. Taking nothing/placebo as the reference (least expensive) strategy, the other four strategies were shown to be more effective, but were each associated with a very high ICER (all > £1.2 million/QALY) and were not considered to be cost effective. Therefore at a WTP threshold of £20,000/QALY, nothing/placebo is the most cost-effective strategy. This conclusion was robust to structural sensitivity analysis and all one-way sensitivity analysis tested (Section A4.2), except for discounting the cost of PEG-G-CSF. At a WTP threshold of £20,000/QALY:

Table 5.13

Incremental costs and effectiveness by treatment strategy for non-Hodgkin lymphoma patients (baseline risk of neutropenic sepsis of one course of chemotherapy: 44.22%).

• When the discount to the cost of PEG-G-CSF was over 83.49% (corresponding price: £113.3 per single subcutaneous injection (6mg), secondary prophylaxis with PEG-G-CSF became the most cost-effective strategy.
• When the discount to the cost of PEG-G-CSF was over 89.12% (corresponding price: £74.7 per single subcutaneous injection (6mg), primary prophylaxis with PEG-G-CSF became the most cost-effective strategy.

Primary or secondary prophylaxis with G(M)-CSF is never the most cost-effective strategy, even when extreme scenarios were considered, for example: 100% risk of neutropenic sepsis per cycle of chemotherapy, 90% drug discount of G(M)-CSF, reduced days of using G(M)-CSF (5-day per cycle of chemotherapy), reduced daily dose (one vial of G(M)-CSF for all adult patients regardless of weight) etc.

The result of probabilistic sensitivity analysis shows that the probability for nothing/placebo becoming cost-effective is always 100%, at a willingness to pay between £10,000 to £40,000 per QALY.

Adult/elderly patients with Hodgkin lymphoma

For adult/elderly patients with Hodgkin lymphoma, clinical evidence was only available for the use of G(M)-CSF for either primary or secondary prophylaxis. Therefore cost-effectiveness was only formally examined for the following three strategies:

• Nothing/placebo
• Primary prophylaxis with G(M)-CSF
• Secondary prophylaxis with G(M)-CSF

The incremental costs and incremental QALYs in the base case analysis for each of the three strategies are summarised in Table 5.14. Taking nothing/placebo as the reference (least expensive) strategy, the other two strategies were shown to be more effective, but were each associated with a very high ICER (both > £18.2 million/QALY) and were therefore not considered to be cost effective. Therefore at a WTP threshold of £20,000/QALY, nothing/placebo is the most cost-effective strategy. This conclusion was robust to all one-way sensitivity analysis tested (Section A4.2).

Table 5.14

Incremental costs and effectiveness by treatment strategy for Hodgkin lymphoma patients (baseline risk of neutropenic sepsis of one course of chemotherapy: 20.27%).

Primary or secondary prophylaxis with G(M)-CSF is never the most cost-effective strategy, even when extreme scenarios were considered, for example: 100% risk of neutropenic sepsis per cycle of chemotherapy, 90% drug discount of G(M)-CSF, reduced days of using G(M)-CSF (5-day per cycle of chemotherapy), reduced daily dose (one vial of G(M)-CSF for all adult patients regardless of weight) etc.

The result of the probabilistic sensitivity analysis shows that the probability of nothing/placebo becoming cost-effective is always 100%, at a willingness to pay between £10,000 to £40,000 per QALY.

Recommendations

• For adult patients (aged 18 years and older) with acute leukaemias, stem cell transplants or solid tumours in whom significant neutropenia (neutrophil count 0.5 × 10⁹ per litre or lower) is an anticipated consequence of chemotherapy, offer prophylaxis with a fluoroquinolone during the expected period of neutropenia only
• Rates of antibiotic resistance and infection patterns should be monitored in treatment facilities where patients are having fluoroquinolones for the prophylaxis of neutropenic sepsis¹⁶.
• Do not routinely offer granulocyte-colony stimulating factor (G-CSF) for the prevention of neutropenic sepsis in adults receiving chemotherapy unless they are receiving G-CSF as an integral part of the chemotherapy regimen or in order to maintain dose intensity.

16

For more information see the Department of Health's Updated guidance on the diagnosis and reporting of Clostridium difficile and guidance from the Health Protection Agency and the Department of Health on Clostridium difficile infection: how to deal with the problem.

Linking Evidence to Recommendations

The aim of this topic was to identify if prophylactic treatment with antibiotics, growth factors and/or granulocyte infusion could improve short term outcomes in patients receiving anticancer treatment. This topic did not investigate the effect of G-CSF as an integral part of a chemotherapy regimen (for example, CHOP-14) or on dose intensity of chemotherapy.

The GDG assessed the clinical effectiveness of antibiotics, growth factors and granulocyte infusions in all patient groups. No evidence was found of a clinical benefit for granulocyte infusions and so cost-effectiveness analysis was undertaken only for growth factors and antibiotics. Because of the heterogeneity and complexity of anticancer treatment, formal cost-effectiveness analysis focused on the group of adult patients receiving outpatient treatment for solid tumours, Hodgkin lymphoma and non-Hodgkin lymphoma.

Studies of patients with stem cell transplants or leukaemia were excluded from the formal cost-effectiveness analysis because the GDG recognised that the costs of prophylaxis for inpatient-only management are very different from outpatient management. Paediatric patients were also excluded from the formal cost-effectiveness analysis because of considerable clinical heterogeneity in the treatment regimens, and a paucity of direct evidence which precluded building a meaningful model for analysis.

The GDG considered that the outcomes of death (short-term mortality), incidence of neutropenic sepsis, bacterial resistance, secondary infection, critical care, length of hospital stay and quality of life were the most clinically relevant. No evidence was reported for secondary infection, critical care or quality of life. Evidence was available for short-term mortality, bacterial resistance, and incidence of neutropenic sepsis. Overall the evidence for all outcomes was of ‘low’ quality with potential bias as assessed by GRADE.

The GDG noted that evidence directly comparing growth factors and antibiotics (ciprofloxacin and cotrimoxazole) was very sparse and of low quality. The GDG were surprised to find that the vast majority of evidence compared growth factors, predominately G-CSF, against no prophylaxis. The GDG also noted that there were very limited data available on the combination of growth factors with fluoroquinolones.

The GDG noted that the clinical evidence comparing antibiotics with placebo was of low quality and showed that antibiotics were effective at reducing overall short term mortality and incidence of neutropenic sepsis. The clinical evidence comparing growth factors against placebo was of low quality and showed no difference in effect on overall short term mortality.

However this evidence did show that growth factors reduce the incidence of neutropenic sepsis and they were also reported to shorten the length of hospital stay. The GDG considered that reduced overall short term mortality was the most important outcome.

Sparse evidence also reported antibiotic resistance with the use of prophylactic antibiotics. This demonstrated that whilst isolation of bacteria resistant to the prophylactic antibiotic may have increased there was still a reduction in overall mortality. The GDG recognised that prophylactic antibiotics contribute to antibiotic resistance but concluded that in patients receiving anticancer treatment the evidence suggests the benefits outweigh the risk.

Evidence was reported for both fluoroquinolones and cotrimoxazole as antibiotic prophylaxis. However the GDG chose to focus on the evidence related to fluoroquinolones because of concerns that changing anti-microbial resistance patterns meant the cotrimoxazole trials may no longer be applicable (Gafter-Gvili, et al., 2005). Consequently the GDG acknowledged that any recommendations made would only be able to focus on fluoroquinolones, because these studies were more recent include a large study undertaken in the UK population and would therefore more accurately reflect the current microbiological environment. The GDG were aware that this approach would exclude all evidence related to antibiotic prophylaxis with cotrimoxazole and that the smaller number of studies would decrease the precision in the estimates of effect, with the potential to increase uncertainty around any recommendation.

The GDG noted that international guidelines such as American Society of Clinical Oncology (Smith et al, 2006), The National Comprehensive Cancer Network (NCCN, 2011) and European Organisation for Research and Treatment of Cancer (Aapro, et al., 2010) recommend the use of G-CSF in selected patients who have a neutropenic sepsis risk of greater than 20%. The GDG also noted that these guidelines were based on the comparison of G-CSF with no prophylaxis, rather than with antibiotics, and did not assess the cost-effectiveness of their recommendations. In addition these guidelines had been developed in non UK healthcare settings.

The GDG considered the issue of paediatric patients carefully, balancing the potential benefits of extrapolating evidence from adult patients against the risks of adverse effects from the medications. Potential similarities between children undergoing stem cell transplantation and treatment for acute leukaemia in adults were considered, as were the documented differences between children and adults in the range of infecting organisms, underlying malignant diagnoses and treatment regimens. The GDG noted a large RCT was in progress by the Children's Oncology Group in North America addressing this question. They also noted the very different treatments used in treating the majority of children and young people with solid tumours compared to the majority of adult solid tumours. The GDG therefore concluded that there was too little evidence to recommend the use of either antibiotics or GCSF in this group, but identified this as an area for research.

The results of the cost-effectiveness analysis showed that for adult patients with solid tumours, primary prophylaxis with fluoroquinolones was more cost-effective than other strategies. This conclusion was robust to sensitivity analysis. For adult patients with solid tumours who cannot receive fluoroquinolones, no prophylaxis was shown to be the most cost-effective strategy. However, this result was shown to be sensitive to adjustments in several of the inputs to the model. As a result of this uncertainty the GDG did not feel able to make a recommendation for adult patients with solid tumours who cannot receive fluoroquinolones.

Little clinical evidence was found comparing fluoroquinolones with no prophylaxis for patients with lymphoma (Hodgkin or non-Hodgkin). Therefore the cost-effectiveness analysis only compared G-CSF or G-CSF + fluoroquinolone with no prophylaxis in these patients. The results showed that although G-CSF or G-CSF + fluoroquinolone could reduce the incidence of neutropenic sepsis; the ICER of both strategies was far above NICE'S £20,000 per QALY threshold and consequently the strategy of no prophylaxis was the most cost effective. However given that data were not available to compare all the strategies of interest the GDG was uncertain whether prophylaxis with antibiotics and/or G-CSF was clinically and cost-effective for lymphoma patients. They therefore decided not to make any recommendations on this issue.

Based on their clinical experience the GDG considered that for patients undergoing stem-cell transplantation and during intensive treatment for acute leukaemia the additional costs of antibiotic prophylaxis would be small and vastly outweighed by the improvement in short term mortality.

A systematic review of published economic evidence for this topic identified 10 papers that were relevant. However all papers had either very serious or potentially serious limitations. Therefore the GDG decided to use the results of the cost-effectiveness analysis conducted as part of this guideline to inform their recommendations.

The GDG considered that the benefits of recommending the use of fluoroquinolones in primary prophylaxis for this subset of patients would be fewer deaths and hospital admissions and potentially improved quality of life. The GDG noted that there are risks associated with recommending primary prophylaxis with fluoroquinolones, such as resistant bacterial infections and super-infection with Clostridium difficile and that monitoring for antimicrobial resistance should be carefully undertaken. However, the GDG noted, based on their clinical experience, that the death rate from such infections in this population is likely to be less than the death rate from neutropenic sepsis. The GDG also noted that the use of fluoroquinolones can have side effects, but agreed that the benefit of saving lives outweighed any potential harms.

The GDG noted the high ICER for G(M)-CSF in the prevention of neutropenic sepsis. Whilst the GDG acknowledged that clinicians in some settings are able to source G(M)-CSF products at substantially reduced prices, it was noted that these arrangements are fluid and regional, and therefore no national recommendations can be based on these costs. Not withstanding this, the GDG noted that the nursing costs of administering G(M)-CSF for preventing neutropenic sepsis result in this intervention not being cost effective, even at reduced prices.

One-way sensitivity analysis has shown that the economic model was sensitive to discounting the cost of PEG-G-CSF. PEG-G_CSF becomes cost-effective for secondary prophylaxis in patients with solid tumours who cannot take fluoroquinolones at less than £179.83 per single subcutaneous injection (6mg). However the GDG considered that is was unlikely that PEG-G-CSF would be available at these levels of discount. It was not possible to calculate similar thresholds for other patient groups because of a lack of clinical evidence (see Appendix A). These elements of uncertainty along with the high ICER described by the economic model led the GDG not to recommend the routine use of G(M)-CSF for the prevention of infectious complications and death from neutropenic sepsis. The GDG were aware that G(M)-CSF is an integral part of some chemotherapy regimens, or is used for maintaining dose intensity. Although this was outside the scope of this guideline and the evidence on this has not been reviewed, the GDG agreed that the use of G(M)-CSF for these indications should be acknowledged in the recommendation.

Based on the clinical evidence and the results of the cost-effectiveness analysis the GDG decided to recommend primary prophylaxis with fluoroquinolones for patients with acute leukaemias, stem cell transplants and adult patients with solid tumours during the period of expected neutropenia. They also recommended further research be undertaken in examining the cost-effectiveness of antibiotics and G-CSF in preventing neutropenic sepsis in children and young people. The GDG noted that in making a recommendation for primary prophylactic treatment a recommendation for secondary prophylactic treatment was no longer relevant. Because of the limited data available on the combination of growth factors with antibiotics, the GDG did not feel able to make any recommendations on this.