Ceftazidime with avibactam for treating severe drug-resistant gram-negative bacterial infections

Antimicrobial resistance develops when bacteria that cause infection develop genetic mutations that make the antimicrobials less effective. Multi-drug-resistant bacteria can spread rapidly in hospitals and residential or care homes. This increases mortality and morbidity when infections can no longer be treated effectively, and when life-saving procedures such as chemotherapy or organ transplantation that rely on antimicrobials to prevent and treat infections cannot be done in people colonised with multi-drug-resistant bacteria. Although drugs in the carbapenem class have historically been reliably active against most common gram-negative bacterial infections, resistance to carbapenems is now increasing. This results in fewer treatment options. Carbapenem resistance is classified based on whether or not the bacteria produce carbapenemase enzymes, which hydrolyse carbapenem antimicrobials, and make them ineffective. There are several treatments for infections with non-carbapenemase resistance mechanisms, but few treatment options for carbapenemase-mediated resistance. Carbapenemase enzymes are grouped into 2 main classes: serine carbapenemases and metallo-beta-lactamases. Ceftazidime–avibactam is active against serine carbapenemases, but not active against metallo-beta-lactamases. The main serine carbapenemases in the UK are Klebsiella pneumoniae carbapenemase and oxacillinases (in particular OXA‑48). Many strains of Enterobacterales such as Escherichia coli and K. pneumoniae producecarbapenemases.

The patient experts at the committee meeting explained that multi-drug-resistant infections are a potential 'death sentence', and people live with 'feelings of fear and hopelessness' knowing that they have few treatment options. They highlighted the negative impact that infections have on people's psychological wellbeing because they may be hospitalised. Multi-drug-resistant infections negatively impact carers and families who may provide financial support. The patient experts explained that there was a high unmet need, particularly for people who are immunosuppressed and likely to develop severe multi-drug-resistant infections. The patient experts emphasised that the adverse effects of existing antimicrobials can significantly affect quality of life. The committee concluded that there was an unmet need, and that patients and their families would welcome new effective treatments with reduced toxicity.

Antimicrobial resistance is a major global health challenge. New antimicrobials, especially those active against multi-drug-resistant pathogens, are subject to strict stewardship to slow the development of resistance. NICE defines antimicrobial stewardship as 'an organisational or healthcare system-wide approach to promoting and monitoring judicious use of antimicrobials to preserve their future effectiveness.' For many antimicrobials, there are few replacements or alternative products in development and even fewer that target multi-resistant pathogens. For many reasons, the pharmaceutical sector sees investment in novel antimicrobials as commercially unattractive. Companies cite as problems the high costs of research and development, post-marketing surveillance, and the logistics of maintaining supply chains. It is difficult for companies to recover these costs because of the strict antimicrobial stewardship, coupled with a limited period of market exclusivity, during which companies expect to generate the most revenue. When generic antimicrobials enter the market at a lower price, this usually results in a substantial drop in sales of the original product. Sales of new antimicrobials may be low if there are few outbreaks of drug-resistant infections during the period of market exclusivity. New antimicrobials have failed in the market. In 2020, only 41 new antimicrobials were in phase 1 to 3 clinical trials, compared with some 1,800 immuno-oncology agents. The committee concluded that there is an urgent need to increase investment for new antimicrobials.

In 2018, EEPRU published a framework for value assessment of new antimicrobials. In 2019, the UK agreed its 5-year action plan for antimicrobial resistance, in which it committed to testing a new way of reimbursing antimicrobials to incentivise research and development. This evaluation was part of a project to test a new reimbursement model in which the payments made by the NHS to the company manufacturing the antimicrobial do not depend on the volume of drugs supplied (also referred as 'delinked' payment or a subscription-based contract). Instead, the payments are based on the benefits that the antimicrobial offers to patients and to the NHS over time, which this NICE evaluation estimated (see section 4.25). This estimate informed commercial discussions between NHS England and the company that manufactures ceftazidime–avibactam. The subscription-based contract between the company and NHS England will last for 3 years with an option to extend it up to 10 years. The committee's first objective was to estimate the incremental population net health benefits of ceftazidime–avibactam against the standard of care, as measured in quality-adjusted life years (QALYs) for the expected eligible population in England. This estimate was based on a model developed by EEPRU using a 20-year time horizon (see section 4.9), and additional evidence submitted by the company and other stakeholders. The committee's second objective was to decide what proportion of the total incremental population net health benefits NHS England should assign to a 10 year contract period.

The 4 clinical trials identified by EEPRU's literature review are randomised non-inferiority trials of ceftazidime–avibactam compared with meropenem, doripenem and best available therapy in people with infections susceptible to carbapenem antimicrobials. This evaluation focusses instead on using ceftazidime–avibactam to treat severe, multi-drug-resistant infections. In addition, the trials investigated activity against infections with a range of resistance mechanisms that were often unknown at the time of recruitment, and were not limited to OXA‑48-producing Enterobacterales, the only resistance mechanism and pathogen included in EEPRU's economic model (see section 4.10). Infections caused by carbapenem-susceptible OXA-48-producing Enterobacterales would be expected to be uncommon. EEPRU identified 6 observational studies in people with OXA‑48-producing drug-resistant infections treated with ceftazidime–avibactam, but 5 did not report data for comparison treatments. Also, these real-world studies had small sample sizes because of the low prevalence of OXA-48-producing bacteria, and included people with a diverse range of characteristics that likely would have affected their prognosis and how well their infections responded to treatment. The committee concluded that the available clinical trials and observational studies have limited generalisability when evaluating ceftazidime–avibactam in multi-drug-resistant infections caused by OXA‑48-producing Enterobacterales.

Because of the lack of generalisable clinical trials or observational data specifically for OXA-48-producing drug-resistant Enterobacterales, EEPRU assessed the relative clinical effectiveness of ceftazidime–avibactam compared with other antibiotics using the laboratory-assessed susceptibility of a pathogen to antimicrobial treatment instead of using direct evidence on patient outcomes (see section 4.8 for the comparator treatments in EEPRU's model). Susceptibility is assessed in vitro, by culturing a bacterial sample from a patient along with increasing concentrations of the antimicrobial, to determine how well the antimicrobial slows growth. The 'clinical breakpoint' is a threshold of the antimicrobial concentration used to assess the likelihood of treatment success or failure. If the lowest concentration needed to stop bacterial growth is below the breakpoint, the infection is deemed susceptible, and treatment is likely to succeed. The committee was aware that different organisations use different laboratory methods to assess susceptibility and different methodologies to set clinical breakpoints. These organisations include the European Committee on Antimicrobial Susceptibility Testing (EUCAST) and the Clinical and Laboratory Standards Institute (CLSI). EEPRU identified reports linking in vitro susceptibility data to clinical outcomes, but the evidence did not relate to the pathogens and resistance mechanisms of interest in this evaluation. EEPRU used the results of 2 published studies it identified in its literature review, which reported mortality and length of hospital stay conditional on susceptibility to treatment, to model clinical outcomes in the 'empiric treatment setting' of its model (see section 4.10). To model outcomes in the 'microbiology-directed treatment setting', EEPRU used established methods to elicit information from experts to characterise the relationship between susceptibility data and clinical outcomes. These outcomes included mortality, length of hospital stay, and type of hospital ward. EEPRU assumed that outcomes were conditional only upon a pathogen's in vitro susceptibility to the antimicrobial, and that outcomes did not depend on the resistance mechanism causing the infection. Results were available from between 5 and 7 experts, depending on the question. Consultation comments on EEPRU's report suggested that these assumptions were not plausible and introduced uncertainty into the modelling. Consultees commented that in vitro data would not reflect a patient's clinical factors affecting response to treatment or whether the tissue penetration of the antimicrobial differs by infection site. The clinical experts at the committee meeting confirmed that there are many factors other than susceptibility and those identified by EEPRU that affect treatment efficacy and outcomes. The consultees highlighted the small sample size of the expert elicitation. The clinical experts explained that, in the absence of alternative evidence and better estimates, using susceptibility as a predictor of clinical outcomes in EEPRU's model was reasonable. The committee concluded that susceptibility was a reasonable surrogate for clinical outcomes but recognised that it introduced uncertainty into the model.

EEPRU compared pathogen susceptibility to ceftazidime–avibactam with susceptibility to other antimicrobials. It used a network-meta-analysis that combined data from susceptibility studies identified through a systematic literature review and hospital laboratory data provided by the UK Health Security Agency (UKHSA; formerly Public Health England [PHE]). The studies reported the proportion of samples that were susceptible to ceftazidime–avibactam and to comparators. Some studies used EUCAST laboratory methods and breakpoints to assess susceptibility, and some used CLSI methods and breakpoints. EEPRU therefore chose to group data by laboratory methods and breakpoints in different networks because they were not directly comparable. In its base-case economic model, EEPRU used a network meta-analysis of studies that applied EUCAST laboratory methods and breakpoints, supplemented by the data provided by the UKHSA. EEPRU chose this approach because it considered EUCAST methods and breakpoints to be the most applicable to England, because the British Society for Antimicrobial Chemotherapy recommends using EUCAST methods and breakpoints in clinical practice. EEPRU assumed that data reported to the UKHSA, having been collected in the UK, used EUCAST methods. Consultation comments on EEPRU's report noted that EUCAST and CLSI laboratory methods and breakpoints had recently been standardised and CLSI laboratory methods and breakpoints were an acceptable alternative to EUCAST. One consultee cautioned that EUCAST breakpoints should not be applied to data generated by CLSI laboratory methods, and vice versa, although another consultee disagreed. EEPRU did 3 scenario network meta-analyses to test the impact of using different studies:

• One scenario used all the studies regardless of whether they used EUCAST or CLSI laboratory methods or breakpoints.

• Another scenario used only the UKHSA susceptibility data.

• The final scenario used the only study that focused solely on the resistance mechanism of interest, OXA-48 (that is, the study excluded pathogens with co-existing OXA-48 and metallo-beta-lactamase resistance mechanisms), and used EUCAST methods.
  The committee was aware that other susceptibility studies, and data from UKHSA, did not explicitly exclude pathogens with co-existing metallo-beta-lactamases.
  
  EEPRU explained the limitations of the scenario analyses. The committee heard that it might not be appropriate to combine data based on EUCAST breakpoints and laboratory methods with data based on CLSI breakpoints and laboratory methods. The committee preferred EPRUU's base-case network meta-analysis over EEPRU's first scenario. The committee agreed that the second scenario was inappropriate because the UKHSA data were not generalisable to current practice in England because susceptibility test results have not always been routinely submitted to UKHSA, and some relevant comparators were missing. The committee felt that the final scenario was inappropriate because the results seemed clinically implausible (for example, the estimated odds ratio for being susceptible to ceftazidime–avibactam compared with colistin was substantially higher than in the other scenarios) and because it was unlikely to be generalisable to England; this is because the study was done in Spain, which has different rates and mechanisms of carbapenemase-mediated resistance. The committee concluded that the network meta-analysis EEPRU used in its base-case economic model was an appropriate source of susceptibility evidence.

Current standard care for treating infections suspected or confirmed to be caused by OXA-48-producing Enterobacterales includes a range of antimicrobials. Treatment choice depends on the infection site, whether the pathogen and resistance mechanism has been confirmed by microbiology testing, and whether the pathogen has additional mechanisms of resistance. EEPRU included a range of comparators, such as meropenem with colistin; fluoroquinolones with meropenem; aminoglycosides; cephalosporins; aztreonam with fosfomycin or colistin; tigecycline with meropenem, colistin or both; and other combinations. When more than 1 formulation was available, EEPRU assumed all comparators were given intravenously. The clinical experts explained that treatment is usually a combination of 2 or 3 agents, and confirmed that EEPRU's comparators were appropriate. To simplify its approach to modelling, EEPRU classified patients into 2 groups of people with an infection that was:

• susceptible to colistin-based therapy or aminoglycoside-based therapy

• not susceptible to either colistin-based therapy or aminoglycoside-based therapy.
  
  The clinical experts on the committee and at the meeting agreed that it was appropriate to consider colistin and aminoglycosides separately from other antimicrobials because they are associated with a risk of renal toxicity, which is higher with colistin than aminoglycosides. The clinical experts explained that a proportion of people at risk of severe and potentially irreversible renal damage would not be offered colistin or aminoglycosides in practice, even if no other effective antimicrobials were available (see section 4.16). The committee concluded that the comparators and classification of comparators in EEPRU's analyses were appropriate.

EEPRU quantified the benefits of ceftazidime–avibactam in 2 stages. First, it developed a new decision analytic model to estimate the costs and benefits of ceftazidime–avibactam over a patient's lifetime (the 'patient-level model'). It modelled the clinical effectiveness, safety, quality of life, costs and resource use associated with ceftazidime–avibactam and its comparators. To inform a 'value-based' delinked payment contract between NHS England and the company, the output of the model is incremental net health benefit, expressed in QALYs at a population level. This differs from NICE's usual approach in health technology assessment of estimating the incremental cost-effectiveness ratio (ICER) at a patient level. EEPRU set the price of the drug to zero, and modelled costs of ceftazidime–avibactam related only to the use of healthcare resources. To convert any cost savings (or losses) associated with ceftazidime–avibactam (for example, reduced or increased time spent in hospital) into health benefits measured in QALYs, EEPRU used an estimate of health opportunity cost. As per the NICE scope for this evaluation, EEPRU used £20,000 per QALY as the estimate of health opportunity cost. This means that for every £20,000 saved, 1 QALY of health can be generated in the NHS. In the second stage, after estimating the per-patient benefits of ceftazidime–avibactam, EEPRU considered the size of the population currently eligible for treatment and how this would change over time to account for a growing number of people with infections and emerging resistance to ceftazidime–avibactam and other antimicrobials. EEPRU modelled the benefits of ceftazidime–avibactam over a 20-year time horizon. This allowed EEPRU to estimate the long-term costs and benefits of ceftazidime–avibactam at the population level.

The marketing authorisation of ceftazidime–avibactam is broad. EEPRU's analysis was narrower than the marketing authorisation and focused on populations for which it expected ceftazidime–avibactam to have the greatest clinical benefit, and referred to these as 'high-value clinical scenarios'. EEPRU divided the clinical scenarios into 2 treatment settings: 'empiric' and 'microbiology-directed'. 'Empiric' reflects clinically urgent infections requiring empiric treatment, when clinicians strongly suspect a particular resistant organism and its mechanism of resistance. EEPRU defined the empiric treatment setting as fulfilling one of the following criteria: a person previously admitted to a hospital with a high prevalence of the suspected pathogen, a ward outbreak, or cultures taken during the current or previous hospital stay showing the person had an infection or bacterial colonisation. The second setting was 'microbiology directed' and referred to an identified organism with tested and confirmed microbiological susceptibility. EEPRU included several high-value clinical scenarios in its patient-level analyses: hospital-acquired pneumonia and ventilator-associated pneumonia treated empirically; and complicated urinary tract infection, hospital-acquired pneumonia and ventilator-associated pneumonia treated in the microbiology-directed setting. EEPRU focused on infections with Enterobacterales with OXA‑48 mechanisms of resistance. In its population-level model, EEPRU included additional groups of patients for which ceftazidime–avibactam is expected to have clinical benefit and be used in practice: people with bloodstream and intrabdominal infections. Consultation comments on EEPRU's report suggested that ceftazidime–avibactam is effective against, and would be used to treat, pathogens and resistance mechanisms that EEPRU had not included in either its patient- or population-level analysis. For example, serine carbapenemase-producing Pseudomonas aeruginosa against which other treatment options are not available or appropriate. Consultees highlighted the importance of ceftazidime–avibactam for people with compromised immune systems (for example, pre- or post-transplantation, or during cancer treatment), and other scenarios including, but not limited to, renal complications, cystic fibrosis, and burns. The committee agreed that EEPRU's analysis excluded populations that would benefit from ceftazidime–avibactam. EEPRU's estimates of the number of people eligible for ceftazidime–avibactam ranged between 300 and 500 people per year in England, while the company estimated that 1,400 people in the UK receive ceftazidime–avibactam per year. The committee was aware that its guidance applies only to England. A committee member with specialist expertise in infectious disease noted that recent data from NHS England suggests that approximately 1,200 people per year in England receive ceftazidime–avibactam. The committee noted that current usage of ceftazidime–avibactam was limited and guided by principles of good antimicrobial stewardship, so agreed that the annual estimates are likely to reflect appropriate use. The committee concluded that the current population size is likely to be approximately 2 to 3 times bigger than EEPRU's estimate.

When modelling benefits of ceftazidime–avibactam at the population level, EEPRU included additional groups of patients for which ceftazidime–avibactam is expected to have clinical benefit and be used in practice (see section 4.10). Because EEPRU did not include these groups in its patient-level model, it was unable to estimate the patient-level QALY gains. EEPRU assumed that QALY gains in people with bloodstream infections were the same as those in people with hospital-acquired pneumonia and ventilator-associated pneumonia. EEPRU further assumed that QALYs gains in people with intra-abdominal infections were the same as those in people with complicated urinary tract infections. The committee noted there was no evidence to show that QALY gains would differ between high-value clinical scenarios and these other infection sites in a wider-use population. In the absence of evidence, the committee recognised that EEPRU's assumptions introduced further uncertainty in the model, but concluded that it is reasonable to generalise incremental benefits of ceftazidime–avibactam to a wider population using results from the high-value clinical scenarios.

EEPRU estimated the number of people currently eligible for ceftazidime–avibactam using data from the UKHSA Second Generation Surveillance System (SGSS), a national database of microbiology test results from 98% of hospital laboratories in England. It includes information on the mechanism of resistance and susceptibility to different antimicrobials for each isolate tested and submitted. It does not include direct information on the site of infection, which must be inferred from the specimen type submitted so is uncertain, as confirmed by the clinical experts at the committee meeting. The clinical experts explained that the UKHSA SGSS data represent isolates classified as susceptible to ceftazidime–avibactam through laboratory testing, rather than infections treated by ceftazidime–avibactam in practice. Therefore, the UKHSA SGSS data may overestimate the eligible population because it includes isolates that may not cause significant clinical illness needing antimicrobial treatment. The committee also heard that the UKHSA SGSS data might underestimate the eligible population because not all hospitals have a microbiology laboratory, and the data submitted to the SGSS from other hospitals may be incomplete. The clinical experts did not know whether the overall effect of these factors resulted in EEPRU overestimating or underestimating the eligible population size. EEPRU explored 2 ways of establishing the infection site from the SGSS data: based on the UKHSA's classification of the specimens or based on classification by EEPRU's clinical advisers. EEPRU's clinical advisers considered that the UKHSA's classification system would underestimate the number of people eligible for ceftazidime–avibactam, because it excluded several specimen types. For example, the UKHSA's classification excluded sputum samples from estimates of pneumonia, and excluded urine specimens from women from estimates of complicated urinary tract infections. The committee noted that EEPRU estimated an eligible population size of 300 people when using the UKHSA's classification, and 500 people when using the clinical advisers' classification. On balance, while acknowledging uncertainty, the committee concluded that it preferred the clinical advisers' infection site classification.

To forecast how the population eligible for ceftazidime–avibactam might change over the 20-year modelled time horizon, EEPRU used historical data on population size for people infected with the pathogen and resistance mechanism of OXA‑48-producing Enterobacterales, provided by the Antimicrobial Resistance and Healthcare Associated Infections national reference laboratory. EEPRU excluded data from before October 2012 because of small patient numbers, and excluded data after March 2018 because of an anomalous decrease in reporting caused by changes in guidelines. EEPRU applied 2 alternative methods to extrapolate data to forecast growth in the patient population: a 'persistent growth' model in which the growth persists over time, and a 'damped trend' model in which the population grows in the short term and stabilises in the long term. The committee appreciated that the choice of model had a significant effect on the long-term estimates. EEPRU provided base-case economic analyses including both approaches. The committee recognised that there was little evidence to support one over the other, and there was considerable uncertainty in both. However, the committee noted that the persistent growth model best fitted the data and was more clinically plausible. The committee concluded that it was more appropriate to assume that the population size of people with OXA-48-producing Enterobacterales infections would continue to grow over the modelled time horizon rather than stabilise.

Based on evidence that resistance develops to a new antimicrobial as its usage increases, EEPRU assumed that resistance to ceftazidime–avibactam would also increase over the model's 20-year time horizon. EEPRU used data from the European Antimicrobial Resistance Surveillance Network to model the relationship between antimicrobial use and resistance, which predicted a small increase in resistance of 0.03% over 20 years. EEPRU believed that this value underestimated true resistance and explored 4 alternative assumptions in its base-case model: resistance to ceftazidime–avibactam reaching 1%, 5%, 10% or 30% after 20 years. EEPRU and the company agreed that 30% was an extreme estimate. The company noted that current resistance to ceftazidime–avibactam is not mediated by OXA-48 carbapenemases, but other carbapenemases. The clinical experts explained that it is not unusual to see a 20% increase in resistance to antimicrobials over a 20‑year period. They explained that indiscriminate use of ceftazidime–avibactam would result in at least a 10% increase in resistance over the 20-year modelled time horizon, but if principles of good antimicrobial stewardship were followed, then the increase in resistance would be between 3% and 5%. The committee concluded that it was reasonable to assume a 5% increase in resistance to ceftazidime–avibactam over the 20-year modelled time horizon.

In its base-case model, EEPRU assumed that resistance to the comparators remains constant over time, because it found little evidence to inform extrapolations of current resistance rates. However, EEPRU acknowledged that resistance to comparators would likely increase over time, either because new multi-drug-resistant pathogens would emerge, or because currently susceptible pathogens would become resistant to existing drugs. This would increase the incremental benefits of ceftazidime–avibactam. The committee noted that in modelling the emergence of resistance to existing antimicrobials, it was important to account for the benefits of being prepared for a catastrophic emergence of widespread multi-drug-resistant infections (sometimes referred to as 'insurance value', see section 4.22). To reflect this, EEPRU provided additional exploratory scenario analyses to reflect a situation in which a new multi-drug-resistant pathogen emerges, against which ceftazidime–avibactam is the only effective treatment. In the absence of evidence to inform the probability, timing and impact of such an event, EEPRU used the following estimates suggested by a committee member with specialist expertise in infectious disease:

• probability of the emergence of highly resistant pathogen(s): 1%

• time to event: 10 years

• number of people affected in the first year: 25

• annual growth in number of infections: 20%.
  
  EEPRU explored the impact of varying these parameter estimates using plausible ranges provided by the same committee member. EEPRU maintained the susceptibility to ceftazidime–avibactam at 90% over the long term. For the scenario in which a new multi-drug-resistant organism emerged, EEPRU presented incremental net health benefit results for infection sites separately. It was unable to present the overall population-level results across all infection sites because it lacked evidence for the proportion of patients for each site. The committee would have preferred to see results for the total population. It was also concerned that the scenario did not include the pathogens modelled in the base-case analysis. The committee considered that resistance to comparators was likely to increase, but that EEPRU's scenario analysis was highly uncertain, and was not entirely relevant to the population under consideration. The committee recognised EEPRU's challenges when modelling the unknown. It concluded that the model underestimates the benefits of ceftazidime–avibactam by not accounting for increased resistance to comparators.

In its base-case model, EEPRU assumed that a proportion of patients would have infections resistant to all existing antimicrobials other than colistin- or aminoglycoside-based regimens. However, consultation comments on EEPRU's report highlighted that some people cannot tolerate the renal toxicity associated with colistin and aminoglycosides, or tolerate its treatment – renal replacement therapy. The comments noted that these people would not be offered these treatments, even if no other therapy were available. Instead, they would be offered 'multi-drug salvage therapy', a regimen combining multiple agents: no single drug would be expected to be effective in isolation, but there could be some benefit when used in combination. EEPRU did not account for this in its base-case model. In response to the consultation comments, EEPRU did a scenario analysis to estimate the incremental benefit of ceftazidime–avibactam in this subgroup of patients. Rather than modelling this population separately, EEPRU derived a weighted average incremental benefit that accounted for the proportion of the total treated population whose infection would be susceptible to colistin or aminoglycosides but who would not be offered these treatments because of the high risk of renal toxicity. In the absence of empiric evidence, EEPRU based its analysis on advice from the committee, which suggested that 20% to 40% of patients would be unable to take colistin or aminoglycoside-based regimens. The committee understood that the risk of renal toxicity is lower with aminoglycosides than with colistin (see section 4.8). A committee member with specialist expertise in infectious diseases stated that the proportion of people unable to take colistin would be close to 40%, but recognised that renal dosing (adjusting the dose based on renal function, to reduce the risk of renal toxicity) would allow colistin to be offered to some of these people. The committee heard from a clinical expert that approximately 5% to 10% of people would be unable to take aminoglycosides. On balance, the committee concluded that the most plausible scenario was the one in which EEPRU assumed that 20% of people cannot have colistin or aminoglycosides, even if no other effective antimicrobial were available. In the empiric treatment setting, this represented 20% of the total treated population. In the microbiology-directed setting, EEPRU assumed that clinicians would consider colistin or aminoglycosides as a treatment option for the 35% of people whose infections would be resistant to non-colistin-based or non-aminoglycosides-based regimens. This means that the proportion of people in the overall microbiology-directed setting who would not be offered colistin or aminoglycosides was 7%.

Several benefits that are important for antimicrobials (see sections 4.18 to 4.22) were not fully captured in EEPRU's analysis. Some of these would increase the estimated incremental benefits of ceftazidime–avibactam. The committee considered the extent to which each element of value was captured in EEPRU's model.

Spectrum value refers to the benefits of a new, effective, narrow-spectrum antimicrobial replacing broad spectrum antimicrobials, reducing problems of antimicrobial resistance associated with their use. EEPRU did not model spectrum value for ceftazidime–avibactam because it considered that ceftazidime–avibactam has a broad spectrum of activity. The clinical experts agreed with EEPRU's assumption that spectrum value was unlikely to be relevant for ceftazidime–avibactam because under a policy of responsible antimicrobial stewardship, it would replace treatments with a similar spectrum of activity. The committee concluded spectrum value was not a source of benefit in this evaluation.

Transmission value refers to the benefits of a new antimicrobial reducing transmission of a given pathogen from treated people to other people; the value is in reducing the incidence of resistant infection. EEPRU did not include transmission value in its analysis, because changes impacting transmission are broad and can have opposite effects. For example, if ceftazidime–avibactam reduced the length of hospital stay, treatment could reduce transmission, but if treatment lengthened life this could increase the length of hospital stay and increase transmission. EEPRU was advised by its clinical experts that even after successful treatment pathogens may remain in the gut which risks transmission; a committee member with specialist expertise in infectious disease agreed. The committee considered that the overall direction of effect is unclear and there is a lack of evidence to support one direction or the other. The committee concluded that transmission value was unlikely to be a source of benefit but acknowledged that this was an area of uncertainty.

Enablement value refers to the benefits of being able to perform medical procedures because of new antimicrobials for resistant infections with few treatment options. When possible, EEPRU included some aspects of this value in its analysis, including the improved treatment of postoperative infections, and the benefits of freeing up hospital resources, that would otherwise be used for treating infections, to enable healthcare and procedures in other patients. It did not include other aspects of enablement, such as increasing the number of procedures that can go ahead in people whose infections are treated, or keeping wards open during an outbreak. The committee was aware that treating a single multi-drug-resistant infection can be costly because, to reduce the risk of transmission, staff allocated to this person are unable to care for other people. It noted that the reduced renal toxicity of ceftazidime–avibactam compared with antimicrobials that clinicians would otherwise offer would free up hospital resources by reducing the number of people needing dialysis and enabling other procedures to go ahead. The committee noted that enabling procedures to go ahead was a benefit of ceftazidime–avibactam. The committee noted that improvements in medicine meant that the number of procedures and interventions including organ transplantation and new cancer treatments has increased in recent years and will continue to increase in the next 5 to 10 years and beyond. The committee recognised that the magnitude of ceftazidime–avibactam's enablement value depends, in part, on the value of the 'enabled' procedures. The committee was also aware that the model did not capture the value provided by ceftazidime–avibactam of reducing staff time and other hospital resources that are lost because of procedures cancelled because of infection. The committee acknowledged the challenges in modelling enablement value, and concluded that EEPRU's model had not fully captured it.

Diversity value refers to the benefits that new antimicrobials offer by diversifying the range of treatments available, thereby reducing use of individual treatments. EEPRU did not model strategies involving diverse prescribing, which it considered inappropriate in high-value clinical scenarios without effective alternative treatments. EEPRU considered that ceftazidime–avibactam should not be used outside of high-value clinical scenarios to avoid developing resistance associated with other antimicrobials. The clinical experts suggested that EEPRU's model underestimated diversity value, explaining that ceftazidime–avibactam will reduce use of carbapenems, and provide an alternative treatment option when there are supply issues with other antimicrobials. The committee noted that diversity value is important when treating severe infections in intensive care units, because people may have organ failure and have few treatment options. It is therefore important to have a diverse range of antimicrobials available in this setting because relying on a limited range of antimicrobials will drive resistance. The committee concluded that diversity value was an uncaptured value that would increase the net health benefits of ceftazidime–avibactam.

Insurance value refers to the benefits of reserving a new antimicrobial until resistance eliminates current alternatives as options, or the benefits of being prepared for a catastrophic emergence of widespread multi-drug-resistant infections against which only the new antimicrobial is effective. The committee was aware that EEPRU did not model a scenario in which ceftazidime–avibactam is held back (that is, not used at all to preserve its effectiveness). It recalled EEPRU's scenario in which a new multi-drug-resistant pathogen emerges against which ceftazidime–avibactam is the only effective treatment (see section 4.15). The committee noted that these analyses were based on adopting a risk-neutral perspective, but agreed that a risk-averse perspective is likely to be more appropriate for estimating the insurance value of an antimicrobial. Being risk averse means paying more than the expected value of a product (in this case, a new antimicrobial) to insure against unwanted future events. However, the committee acknowledged that it had no basis to determine the additional value that the NHS would be willing to pay to avoid a situation in which an infection emerged that was resistant to all available treatments. The committee concluded that EEPRU's model had not fully captured the potential 'insurance value' of ceftazidime–avibactam.

The committee recalled its preferred assumptions from the options presented by EEPRU:

• The network meta-analysis of susceptibility studies used in EEPRU's base-case economic model was an appropriate source of evidence for clinical outcomes (see section 4.7).

• OXA-48-producing Enterobacterales infections are likely to increase over the modelled 20-year time horizon, that is, follow a persistent growth trend (see section 4.13).

• The clinical advisers' classification of infection site is more appropriate than the UKHSA's classification for estimating the number of people currently eligible for ceftazidime–avibactam (see section 4.12).

• Resistance to ceftazidime–avibactam will increase by 5% over the 20-year modelled time horizon (see section 4.14).

• 20% of patients would not be offered colistin or aminoglycoside-based treatment regimens (see section 4.16).
  
  Using these assumptions, the incremental net health benefit of ceftazidime–avibactam was approximately 3,700 QALYs. The committee also recalled its conclusions about the benefits of ceftazidime–avibactam not captured in EEPRU's analysis, specifically:

• The population for which ceftazidime–avibactam is suitable is likely to be 2 to 3 times larger than EEPRU's estimate (see section 4.10). The committee understood that increasing the population size would increase the incremental benefit of ceftazidime–avibactam. On balance, the committee concluded that the increased population size would double the incremental QALYs for ceftazidime–avibactam.

• The model did not capture all elements of value. The committee identified that enablement value, diversity value and insurance value were not fully captured (see sections 4.17 to 4.22). It also identified that the model had underestimated the benefits of ceftazidime–avibactam by not accounting for increased resistance to comparators over time (see section 4.15). The committee concluded that the estimate of incremental QALYs should be increased by a further 20% to account for uncaptured value.
  
  The committee concluded that the incremental net health benefit of ceftazidime–avibactam would be approximately 8,880 QALYs over the 20-year modelled time horizon, when the technology is used within its marketing authorisation and in line with the criteria in section 1.1. It acknowledged that there was a large degree of uncertainty around this estimate because of uncertainties in the model results and in estimating uncaptured benefits (see section 4.24).

EEPRU's probabilistic sensitivity analysis resulted in a broad range of estimates of incremental QALYs. This indicates that uncertainty around the parameter values in the model affects the population-level value of ceftazidime–avibactam. The committee recalled several areas of uncertainty in the evaluation that relate to the model structure and to the assumptions made by EEPRU in the absence of evidence. These included the association between in vitro susceptibility and clinical outcomes, the trends in antimicrobial usage and resistance over time, the limitations of the data from the UKHSA SGSS to estimate the size of the population for which ceftazidime–avibactam is suitable, and the uncaptured benefits. The committee concluded that the QALY estimates were associated with significant uncertainty, and encouraged research to develop best practice in the health economic evaluation of antimicrobials (see sections 5 and 6).

Having concluded that the total benefits over the 20-year time horizon would be approximately 8,880 QALYs (see section 4.23), the committee considered what proportion of this should be assigned to a 10-year contract period. It considered that this should be at least as much as the rewards typically earned by companies during the first 10 years of marketing a non-antimicrobial. Assigning a lower proportion would not address the issues of market failure for new antimicrobials nor create a 'pull incentive' for investment. EEPRU presented the committee with evidence that the proportion of benefits of non-antimicrobial drugs in their first 10 years on the market is about 60%. The committee's view was that the proportion of benefits that should be assigned to a 10-year contract period ranged from 60% to 100%. The committee concluded that the proportion of QALY benefits to assign to each year of the 10-year contract period should be a minimum of 60%, resulting in a minimum of 530 QALYs per year.

The committee agreed that good antimicrobial stewardship is extremely important to preserve the effectiveness of ceftazidime–avibactam and to minimise the risk of developing resistance. It was aware of NICE's guideline on antimicrobial stewardship. The committee agreed that ceftazidime–avibactam should be reserved for people with no suitable alternative treatment options, either because clinicians expect or have confirmed that the infection is resistant to other antimicrobials, or because there are concerns about the toxicity or availability of alternative treatments (see section 4.16). The committee considered that ideally clinicians would offer ceftazidime–avibactam only after tests for microbiology susceptibility and mechanisms of resistance have confirmed that the pathogen is resistant to other suitable treatment options and susceptible to ceftazidime–avibactam. However, it recognised that having these test results before starting treatment was not always possible, for example, if a person's condition is clinically unstable with a fast-progressing infection that is not responding to other antimicrobials. The committee agreed that it would be appropriate to offer ceftazidime–avibactam in the absence of test results, only if clinicians strongly suspect that the infection will be susceptible to ceftazidime–avibactam, and not susceptible to other suitable antibiotics. The committee noted that the estimates of incremental net health benefit for ceftazidime–avibactam were based on using it under these conditions. The committee was aware that the marketing authorisation for ceftazidime–avibactam states that it should be offered 'only after consultation with a physician with appropriate experience in the management of infectious diseases' and agreed that this was essential to limit antimicrobial resistance. The committee concluded that ceftazidime–avibactam should be offered only when there are no suitable alternative treatment options, and only when tests for microbiological susceptibility and mechanisms of resistance have confirmed that the infection is susceptible to ceftazidime–avibactam and resistant to other suitable treatment options, or when there is an urgent need to treat an infection expected to be susceptible to ceftazidime–avibactam and the results of these tests are not yet available.