Larotrectinib for treating NTRK fusion-positive solid tumours

3.1 Traditional oncology approaches treat tumours based on their type. More recently, targeted therapies based on the tumour's genetic information have been used for some indications. Larotrectinib is indicated for any solid tumour with a neurotrophic tyrosine receptor kinase (NTRK) gene fusion. Because many tumour types respond to it, the company considers larotrectinib to be 'tumour-agnostic' or 'histology-independent'. NTRK gene fusions may be able to drive tumour growth, so targeting treatment to the cause of the disease could mean higher rates of response to therapy and potentially better outcomes. The committee accepted that it was expected to appraise larotrectinib within its conditional marketing authorisation using NICE's single technology appraisal process. But it recognised the challenges of appraising a histology-independent treatment within this process.

3.2 NTRK gene fusions occur rarely (less than 1%) in common tumours such as lung, colorectal and breast cancers. Some rare tumour types have more than 90% NTRK fusion prevalence (for example, mammary analogue secretory carcinoma and infantile fibrosarcoma). There are many tumour types with known NTRK gene fusions and all solid tumour types are included in larotrectinib's marketing authorisation. NTRK fusions can involve portions of the NTRK1, NTRK2 and NTRK3 genes with another unrelated gene partner (over 80 different partner genes identified). The European public assessment report (EPAR) for larotrectinib notes there is evidence that the most frequent fusions found in high NTRK-prevalent tumours, ETV6‑NTRK3 fusions, drive tumour growth regardless of other characteristics. The EPAR also states that larotrectinib has shown activity in gastrointestinal stromal tumours with NTRK gene fusions and this likely reflects a similar role of NTRK gene fusions in driving tumour growth. For all other NTRK fusions, their role in driving cancer growth has not been properly studied. It is not known if there are tissue-specific mechanisms for bypassing response to drugs for NTRK fusions or effects from other drivers of tumour growth. The committee concluded that better characterisation of NTRK gene fusions was needed to fully support the histology-independent approach.

3.3 It is not known whether patients with tumours that have NTRK gene fusions have a different prognosis to those who do not have them. Evidence of an association between NTRK gene fusions and different disease presentation is weak and based on data from very few patients. The company assumed there was no prognostic effect of NTRK gene fusions in its base-case analysis. The ERG considered that it was unclear whether NTRK fusions affect prognosis directly or whether they are associated with other factors that affect prognosis such as age and Eastern Cooperative Oncology Group (ECOG) status. Prognosis could also vary by tumour type and NTRK gene fusion type. The committee concluded that further data would be needed to establish whether NTRK gene fusions affect prognosis.

3.4 There is no defined clinical pathway for people with solid tumours with NTRK gene fusions. Treatment currently follows care guidelines for specific tumour types, with surgery, targeted therapy, immunotherapy and chemotherapy for the more common cancers. Treatment for rarer cancers is generally limited to surgery, radiotherapy and chemotherapy. The patient experts explained that people who have a solid tumour with a gene alteration would want treatment with a targeted therapy because longer survival and a better side effect profile are likely. The aim of treatment for some inoperable tumours is to shrink the solid tumour so that surgery might be a treatment option. The committee concluded that people with NTRK fusion-positive solid tumours would value new treatment options.

3.5 Larotrectinib is indicated for 'the treatment of adult and paediatric patients with tumours that display a NTRK gene fusion and who have disease that is locally advanced, metastatic or where surgical resection is likely to result in severe morbidity, and no satisfactory treatment options'. The ERG considered that the positioning of larotrectinib would depend on which treatments clinicians consider unsatisfactory. This would vary because it involves assessing response rates and adverse events and discussing options with patients. The clinical evidence for larotrectinib comes from a population with different tumour types who have had different previous treatments. Most patients had previous systemic therapies but a proportion had no previous treatments. The patient experts considered that any targeted therapy should be used as early in the treatment pathway as possible to maximise patient access. The clinical expert stated that for some sarcomas there are very few treatment options. In larotrectinib's summary of product characteristics, it states that it should only be used if there are no treatment options with established clinical benefit, or when such treatment options have been exhausted. This is because the regulatory authority considered that larotrectinib's benefit had been established in single-arm trials in a relatively small sample of patients. But it also considered that its effect may differ depending on tumour type and other possible gene alterations, so it should not displace any effective therapies. The company therefore positioned larotrectinib as a last-line treatment, after all other treatments have been tried, by choosing the comparators: best supportive care for common cancers and chemotherapy for rarer cancers. The committee considered this positioning appropriate and in line with the marketing authorisation. However, it recognised that if larotrectinib's efficacy was further established, it was possible that clinicians would want to use it earlier in the pathway. This would displace other potentially effective therapies and is outside larotrectinib's current marketing authorisation. Further evidence about this is being collected as part of larotrectinib's conditional marketing authorisation. The committee concluded that larotrectinib's positioning was a major uncertainty and collecting further data would determine how larotrectinib would be used in clinical practice.

3.6 All solid tumour types could potentially have NTRK gene fusions although they are rare in common tumour types (see section 3.2). Therefore, many people would need screening to identify who would benefit from larotrectinib. Currently, NTRK testing is not routinely done in the NHS for all solid tumours. However, it is available for mammary analogue secretory carcinoma and secretory breast carcinoma with immunohistochemistry techniques (a method that uses antibodies to detect the gene fusion protein). Whole genome sequencing (a method of determining the whole DNA sequence of a cancer, used for discovering mutations) can also identify NTRK gene fusions. It is available for children's cancers and sarcomas, although confirmation of the results with another targeted DNA or RNA test is needed (for example next generation sequencing, which is a faster method of sequencing targeted regions of the cancer's DNA). The committee concluded that the diagnostic pathway for NTRK gene fusions was important, with implications for identifying patients and on costs of diagnosis.

3.7 The Cancer Drugs Fund clinical lead explained that NHS England is currently establishing a national service for cancer genomic testing, to replace all local testing. It involves setting up 7 laboratory hubs across England to do genomic testing by next generation sequencing and interpret all results. The committee understood that until the laboratory hubs are fully established, next generation sequencing will be done after all NHS-commissioned treatment options have been tried. When the hubs are fully established, next generation sequencing to identify gene alterations, including NTRK gene fusions, will be done when locally advanced or metastatic solid tumours are first diagnosed. The Cancer Drugs Fund clinical lead estimated that 100,000 solid tumours will be tested per year once the service is fully established. He noted that other targeted therapies will likely become available soon for different diseases and genomic testing will also be needed before these treatments are used. The committee acknowledged the ongoing developments in genomic testing practice to identify NTRK fusion-positive solid tumours. It considered that the rapid change to the diagnostic testing pathway being led by NHS England was a unique situation. The committee concluded that the diagnostic testing pathway was uncertain until NHS England establishes a national service for cancer genomic testing.

3.8 The diagnostic specificity of any test needs to be very high to exclude patients who do not have an NTRK gene fusion and so would not benefit from larotrectinib. This particularly affects common tumour types with low NTRK prevalence (for example, lung, colorectal and breast tumours) in which the number of false-positive results could outnumber true-positive results if the test is not specific enough. The committee was aware that currently, DNA-based next generation sequencing and whole genome sequencing were not sensitive enough to screen for NTRK gene fusions. Clinical experts considered that DNA- and RNA-based next generation sequencing with a confirmatory targeted DNA, RNA or immunohistochemistry test when a positive result is obtained would be appropriate and minimise the number of false-positive results. Some factors may reduce the specificity of these tests, such as tissue degradation and human error. The committee noted that the last-line positioning of larotrectinib reduced the risk of displacing active treatments for patients with false-positive results for NTRK gene fusions, whose tumours would not respond to larotrectinib. However, there were concerns about the ethics of treatment for any patients with false-positive results because of adverse events. The committee concluded that diagnostic techniques would improve as the genomic laboratory hubs validate their techniques and NTRK gene fusions are better characterised (see section 3.2).

3.9 The company presented a pooled analysis of 102 patients from 3 trials:

3.10 The company considered the results to be generalisable to NHS clinical practice and did not do any adjustments for baseline characteristics. The ERG noted that patients in each trial were recruited by convenience sampling and there was no systematic attempt to represent the distribution of tumour types in NHS clinical practice. Therefore, a substantial number of patients in the trials had high NTRK-prevalent tumour types because these patients were easier to recruit. The over-representation of high NTRK-prevalent tumour types meant that:

3.11 The primary outcome measure of the 2 larger trials was overall response rate. The committee considered that the evidence showed a clinically relevant overall response rate of 72% across multiple tumour types. But it noted that this may not be generalisable to the broader range of tumour types expected in NHS clinical practice (see section 3.10). Also, it was aware of considerable uncertainty about the extent to which the response translated into clinically meaningful survival benefits. The progression-free survival data were immature, with only 37% of patients having progressed disease (excluding primary CNS tumours). Also, overall survival data were very immature with only 14 of 102 patients dying in the trials. The immaturity of the data meant that the extrapolation of survival estimates was very uncertain (see section 3.21). The committee concluded that the immaturity of the survival data meant that the size of larotrectinib's benefit on long-term survival could not be reliably estimated.

3.12 The committee was aware of several biological reasons why heterogeneity, or a difference, in response to larotrectinib might be seen. For example, response might be different by histology, by NTRK gene fusion or fusion partner, by the presence of co‑drivers of the disease and by age (for example for children's indications). None of the statistical protocols of the trials included in the pooled analysis were designed to test heterogeneity in response by any factor. NAVIGATE was a 'basket' trial (that is, a trial that included patients who had different types of cancer but the same gene mutation) which involved many small groups of patients stratified by tumour type. If a response was seen in one of these groups, a response to larotrectinib for that tumour type was assumed and the basket was further expanded to increase the sample size. However, the company later pooled all the analyses from the 3 studies of all patients in whom efficacy could be evaluated (see section 3.9). The EPAR states that selection bias on pooling the data is possible and there may be some tumour types that do not respond (type 1 error). The ERG explained that the company assumed an equal response to larotrectinib independent of tumour type and response was not formally assessed by tumour type. The committee concluded that the trials were not designed to assess heterogeneity in response.

3.13 The committee noted the challenges of assessing response because the individual subgroups were too small for meaningful analysis. For example, some tumour types with few patients in the trials, such as pancreatic tumours (n=1) and congenital mesoblastic nephroma (n=1), had 0% response or 100% response respectively. This may be because of chance findings or because of biological differences in the tumour type. For example, as noted in the EPAR, tissue-specific mechanisms for bypassing response to drugs have been seen when targeting other gene mutations (such as colorectal cancer not responding to BRAF mutation inhibitors). Tumour tissue can also affect type of NTRK gene fusion, gene fusion partners and presence of other mutations. The response to larotrectinib by NTRK gene fusion showed heterogeneity, with low response in NTRK2 and high response in NTRK3 (82%). There was also heterogeneity in response to larotrectinib by gene fusion partner. The most common fusion partner, ETV6‑NTRK3, had an overall response of 84%, higher than all other fusion partners combined. The Cancer Drugs Fund clinical lead commented that it is biologically plausible that for people with high NTRK-prevalent tumours, such as those with ETV6‑NTRK3 (see section 3.2), there would be a higher response to larotrectinib and greater benefits. The clinical experts could not comment on whether heterogeneity in the response to larotrectinib would be expected across different tumour types. The company's model assumed equal response for all tumour types and fusion types, based on the overall response from the trial population (see section 3.9) and did not explore any other assumption. The committee concluded that given the observed differences in response and the poor characterisation of NTRK gene fusions and fusion partners, assuming equal response was inappropriate. Adjustment should be made for potential differences by subgroup.

3.14 The ERG presented the Bayesian hierarchical model (BHM) framework as an approach to characterise heterogeneity. The BHM framework was developed specifically for basket trials and is useful when there is limited information. This method allows for analysis of a pooled overall response rate, adjusting for the observed heterogeneity and borrowing strength across different tumour types to avoid extreme results. Extreme results may be seen because of limited patient numbers in trials; even small changes in the absolute number of people whose disease responds can considerably affect the response rate (see section 3.13). The ERG noted that because the tumour types with more response events also had high response rates, the low response rates seen in tumour types with fewer response events reduced the overall response in the pooled analysis from 72% to 57% (including primary CNS tumours, see section 3.15). The committee acknowledged that the results of the ERG's BHM approach were substantially different to the company's approach, which assumed that the response was the same across the different tumour types. The committee considered the reduced overall response output of the BHM, with wide credibility intervals, to be more appropriate for decision making because it incorporated some adjustment for heterogeneity. The ERG noted that the BHM approach could also be used for survival outcomes because response was not explicitly modelled in the company's base case. However, the company did not provide the data needed to do this analysis and the survival results were likely to be too immature for meaningful interpretation of the results. The committee concluded that the BHM was a useful tool for exploring heterogeneity in response to larotrectinib and, based on the current limited dataset, should be considered as part of its decision making.

3.15 There were 9 out of 102 patients with primary CNS tumours and these tumours had a much lower response to larotrectinib than other types. Although the results for this group were analysed separately from the main efficacy analysis, they were included in the economic analysis. The company explained that tumour response to larotrectinib in patients with primary CNS tumours was assessed by investigators using the response assessment in neuro-oncology (RANO) criteria, instead of the independent review committee assessed response evaluation criteria in solid tumours (RECIST) v1.1. Surgery and radiotherapy for CNS tumours can lead to varying amounts of scarring and inflammation, which makes it difficult to assess response using RANO criteria. The company considered that this explained the lower response compared with other tumour types. However, in the EPAR it states that larotrectinib is a substrate of P‑glycoprotein, a key constituent of the blood–brain barrier. This may mean that the dose of larotrectinib reaching the brain is reduced. The ERG considered that primary CNS tumours may have a lower response to larotrectinib or that the high proportion of NTRK2 gene fusions within primary CNS tumours may explain the low response. The ERG noted that including the primary CNS data in the BHM reduced the estimated overall response rate. But the ERG did not state a preference for including or excluding primary CNS data from the BHM. The committee concluded that there was uncertainty about whether primary CNS tumours respond to larotrectinib. However, it considered that the primary CNS data should be included in the BHM until more is known about larotrectinib's efficacy in the brain and NTRK2 gene fusions are better characterised. This is so the results are more generalisable to the population covered in the marketing authorisation.

3.16 To establish relative clinical effectiveness compared with current clinical practice, larotrectinib was compared with multiple comparators by tumour site-specific pathway. The company's base-case analysis created a pooled comparator arm by using data from last-line treatment arms in published NICE appraisals and in the literature to represent best supportive care. Survival estimates were weighted by the distribution of the efficacy population to allow a naive indirect comparison with an estimate of survival expectancy for this population. The ERG considered this method could introduce bias in many ways, including generalisability of the trial population (see section 3.10) and uncertainty over the potential prognostic importance of NTRK (see section 3.3). This method also limited the ability to adjust for any heterogeneity (see section 3.13). This was because it assumed that survival was independent of tumour type and other factors, therefore assuming a common natural history for all tumour types in the analysis. The ERG explained that the size and direction of bias was impossible to establish. The committee appreciated the difficulty in estimating a comparator population for rare tumours across multiple tumour types. But it concluded that the company's method had unadjusted bias that did not account for significant heterogeneity and could not adjust for important prognostic factors and baseline characteristics.

3.17 The company presented 2 further indirect treatment comparisons, as confirmatory analyses, to approximate the comparator arm. A response-based analysis assumed patients whose tumours did not respond to larotrectinib were equivalent to those who had best supportive care and were assumed to represent the comparator arm. The advantages of the response-based analysis were that the eligibility criteria of the trial were considered, and the ERG's adaptation of this model allowed for exploration of heterogeneity (see section 3.13). However, this method assumed that response was a surrogate outcome for survival, which may differ significantly by tumour type and have known biases. A previous line of treatment analysis compared the time to progression on the previous line of treatment with the time in progression-free survival on larotrectinib for each patient. This ratio was used as a hazard ratio to approximate a comparator arm for progression-free survival and it was assumed that overall survival behaved the same. The advantage of this analysis was that each patient acted as their own control, which was particularly beneficial for a histology-independent population. However, the ERG had concerns with how this analysis was implemented. It considered that a patient's previous unsuccessful line of therapy may not represent best supportive care and it noted that the method was uninformative for overall survival, which was a major uncertainty of the base-case analysis. The committee noted that both confirmatory analyses also had substantial biases and that all the indirect treatment comparisons had structural uncertainty. Therefore, it concluded that it would consider the outputs of all indirect treatment comparisons in its decision making.

3.18 The company's economic models were based on the indirect treatment comparisons (see section 3.16 and section 3.17). The company's base-case structure was a 3‑state partitioned survival model (progression-free, progressed and death). Survival estimates were extrapolated from the larotrectinib efficacy population and comparator arm survival estimates calculated using the method in section 3.16. Other model parameters such as utility, time on treatment and adverse events were also included to create part-models (termed 'engines' in the company submission) for each tumour type comparator. The company also provided economic models based on the confirmatory analyses in section 3.17. The ERG adapted the response-based model to a dual-partitioned survival model with the larotrectinib arm response defined by output from the BHM (see section 3.14). This allowed for some exploration of uncertainty from heterogeneity in response. The ERG also considered that this would account for issues with post-progression treatments (see section 3.22) because these would be the same in both treatment arms. The ERG considered the previous line of treatment analysis was not implemented appropriately and was uninformative for overall survival, so did not consider this model structure further. The committee recognised the difficulty in modelling treatments for single-arm data, and that the unique challenges of histology-independent treatments further complicated the modelling issues. The committee understood that there were limitations and uncertainties with each of the modelling approaches. It considered that when more data were available, the different model structures could be explored more fully. The committee concluded that the most appropriate model structure for decision making was uncertain.

3.19 The company's base-case model used the full efficacy evaluable population from the pooled analysis, which the committee considered was not generalisable to NHS clinical practice (see section 3.10). The company did not provide any information for the effectiveness of larotrectinib by individual tumour type, so it was not possible to remove patients who were over-represented from the economic model or to adjust the population to make it more generalisable. The over-representation of rare and high NTRK-prevalent tumour types introduced several issues, including:

3.20 The committee was aware that only a small number of patients in clinical practice would have tumour types that could potentially be cured because locally advanced and metastatic cancer is generally incurable. However, children with tumour types that could potentially be cured were included in the SCOUT trial in much greater proportions than would be seen in clinical practice (see section 3.10 and section 3.19). Most children in the evaluable population from SCOUT had no other curative options besides amputation or disfiguring surgery. The company considered that some of these patients would be cured without larotrectinib, with lifelong morbidity from amputation or disfigurement. Therefore, it considered that a cure model could be appropriate. But cure could not be determined because of short follow up, limited numbers of patients and high censoring of data. The committee noted considerable uncertainty around the potential for a cure and that the curative role of surgery had not been explored in either treatment arm. It also noted that the comparator part-model for these patients (grouped as paediatric sarcomas) did not model lifelong survival over the full-time horizon. Therefore, the current model structure captured the benefit of a cure in the Kaplan–Meier survival analysis in the larotrectinib arm but did not model the possibility of cure in the comparator arm. The committee concluded that this issue strongly biased the modelled cost-effectiveness results in favour of larotrectinib. The committee was aware that the modelled cure rates were speculative because the cure rate in the trial population was unknown and likely to be negligible for a generalisable population. The potential for a cure in the trial population supported why the model structures proposed were not appropriate for a heterogeneous non-generalisable population. The committee concluded that a different model structure should be used to explore the effect of a cure. Also, further information was needed on how larotrectinib would be used in clinical practice.

3.21 Data on overall survival and progression-free survival were incomplete and the clinical trials are ongoing. To extrapolate progression-free survival and overall survival for larotrectinib, the company fitted standard distributions to the data (the exact distributions chosen are confidential and cannot be reported here). For the pooled comparator arm, the curve accepted by the committee in previous NICE guidance was used when available and assumptions used when no NICE guidance was available. The ERG noted the considerable uncertainty in extrapolating data that are immature (see section 3.11). The overall survival extrapolation showed substantial separation from the progression-free survival extrapolation, which contributed to an implausible post-progression survival estimate (see section 3.22). During the technical engagement stage, the company provided overall survival data from an updated data-cut. The ERG considered that there were no clear differences in survival characteristics in these updated data. But it noted that a few events at the extreme end of the curve dramatically affected the survival extrapolation and modelled survival gain. The committee considered that this could be a result of uncertain extrapolation and may not suggest a true decline in survival. This was compounded by uncertainty about the possibility of including patients whose disease was cured in the larotrectinib Kaplan–Meier curves (see section 3.20). The committee concluded that the extreme sensitivity of the model output to the survival extrapolations meant that extrapolation did not provide results that the committee could trust, but that data on longer-term extrapolation could be collected in the Cancer Drugs Fund.

3.22 The ERG noted that the life years gained after progression were greater than both the progression-free life years gained and overall survival in the comparator arm, which it considered implausible. The ERG considered this could be a result of the highly uncertain extrapolation or because of the high proportion of patients who had post-progression treatments, such as further larotrectinib or an experimental treatment (LOXO‑195) for people whose disease was resistant to TRK-inhibitors. The ERG considered that LOXO‑195 would not be used in clinical practice because the tumour would need to be TRK-inhibitor resistant before people could have this treatment. The committee considered it appropriate to adjust for the benefits of these treatments. However, it noted the difficulty of using adjustment techniques for an unknown treatment effect in immature survival data. The clinical expert stated that a high depth of response (prolonged benefit from shrinking the tumour) to larotrectinib might explain why post-progression survival could be higher than progression-free survival. Having a smaller tumour could mean longer survival even after developing resistance to larotrectinib. The committee considered this concept to be possible but highly speculative because the current evidence base was very immature. It also noted that TRK-inhibitor resistance mechanisms were not well characterised and would not explain the size of the discrepancy in life years gained. The ERG provided 2 scenarios based on crude adjustment of larotrectinib post-progression survival to match post-progression survival in the comparator arm (no comparative effect of larotrectinib after progression) or overall survival in the comparator arm. The committee considered these scenarios to be more plausible than the company's base case and appropriate for showing the upper limit of plausible cost-effectiveness estimates. However, it considered that the survival estimates, both progression-free and post-progression, were likely to be affected by immature extrapolation (see section 3.21) and including survival data from patients whose disease was cured (see section 3.19). The committee concluded that the post-progression survival estimates were implausible and that the ERG scenarios did not fully capture the issues with modelling survival.

3.23 The company provided utility values derived from health-related quality-of-life data collected in the SCOUT and NAVIGATE trials, mapped to EQ‑5D‑3L utility values for the pre-progression and progressed health states. For the comparator arm utility values, a similar approach to that used for estimating survival was used (see section 3.16). Utility values from published NICE guidance were pooled and weighted by the distribution of the efficacy population. The ERG considered that there was considerable uncertainty in the utility value estimate of the post-progression health state for larotrectinib because it was based on few assessments of patients, most of whom were children. The committee noted that some patients could potentially be cured (see section 3.19). It considered that the evidence for the post-progression utility values for larotrectinib was weak and there was no plausible reason why post-progression utility would be so much higher for larotrectinib than for the comparator arm for the entire population. The ERG provided a scenario with equal post-progression utility values for larotrectinib and the pooled comparator. The committee agreed that this scenario was more appropriate.

3.24 The company's utility values suggested a difference in pre-progression utility between larotrectinib and the pooled comparator. The Cancer Drugs Fund clinical lead considered this to be implausible because the modelled comparator in NHS clinical practice was likely to be best supportive care. Therefore, any difference in utility values between arms would represent positive effects from reduced tumour size in the pre-progression state for larotrectinib and the difference in adverse effects between treatments not captured by the adverse event modelling. The committee considered that the high number of patients having chemotherapy rather than best supportive care (see section 3.10) would bias this difference in favour of larotrectinib. The committee concluded that it was appropriate to do a sensitivity analysis to see the effect of reducing larotrectinib pre-progression utility values on this bias. The committee considered this scenario to be more plausible for a population that was generalisable to NHS clinical practice.

3.25 The committee understood that the NICE methods guide was not designed to address a system-wide change in diagnostic techniques and the cost of testing would depend on NHS England's testing strategy. The company considered that the planned changes in genomic testing (see section 3.7) would mean that none of the testing costs would be borne by larotrectinib in the economic model. This was because the system would be independent of testing for NTRK gene fusions and would not be used solely for identifying NTRK gene fusions for a particular drug. The ERG's interpretation of the methods guide was that associated costs of the diagnostic test should be incorporated into the clinical- and cost-effectiveness assessments. The committee agreed that the ERG's interpretation, which included assessment of the cost of larotrectinib in current NHS clinical practice, most closely matched the methods guide. The ERG proposed a pragmatic screening pathway for each tumour type, based on adapting current testing provisions for some tumour types that already have some genomic testing. For tumour types that are not currently tested for any genetic mutations, the ERG assumed that immunohistochemistry would be followed by confirmatory next generation sequencing if there was a positive result. This technique provided an average cost of screening for a single patient with an NTRK gene fusion, which was included in the cost-effectiveness estimates. The committee considered this analysis was reasonable because it reflected current clinical practice, but recognised that NHS England is rapidly moving towards a national service for cancer genomic testing. Therefore, the proposal from NHS England to implement next generation sequencing at diagnosis of locally advanced or metastatic disease was likely to reflect the near future once the changes to the diagnostic pathway have been established. NHS England proposed a cost per patient with an identified NTRK fusion-positive tumour to be included in the model. The committee noted that this cost was substantially less than in the ERG's scenario. The committee concluded that it was appropriate to include diagnostic testing costs in the model.

3.26 The company used the average adult and children's doses used for the efficacy population to calculate the dose of larotrectinib applied in the model. The ERG considered this to be inappropriate because some children had a lower dose in SCOUT because it was a phase 1 dose-finding study. The ERG provided a scenario in which the full children's dose was adjusted using the percentage of adults adhering to treatment. The committee agreed that the ERG's adjustments were appropriate to better generalise the dose in SCOUT to NHS clinical practice, but it noted that this did not account for potential differences in clinical effect. Additionally, the company supplied scenarios considering larotrectinib wastage after stopping treatment for 2- and 4‑week treatment supplies. The committee noted that these scenarios had little effect on the cost-effectiveness estimates but considered it appropriate to include the 4‑week treatment drug wastage scenario in the model. The committee also noted that these assumptions relied on using hard capsules, which are not yet available, so an additional scenario using the oral solution should have been provided.

3.27 The company considered that administration costs and resource use should be applied in the model by the distribution of patients in the efficacy analysis. The committee considered this inappropriate because the over-representation of rare tumours that were modelled as having chemotherapy (see section 3.10) may have resulted in bias in favour of larotrectinib. However, it considered that the true distribution could not be calculated. The Cancer Drugs Fund clinical lead commented that the administration costs for oral chemotherapy were not included for larotrectinib. At the technical engagement stage, the company provided a scenario with these administration costs. The committee concluded that this scenario was appropriate for including in the economic model.

3.28 The company modelled the costs of larotrectinib treatment until progression, but some patients continued to have larotrectinib after progression (see section 3.22). The company provided a scenario using the time to treatment discontinuation seen in the trial to model larotrectinib costs. The committee considered this to be more appropriate because the modelled costs of larotrectinib should match the modelled benefits. However, it noted there would be further uncertainty with extrapolation of time to treatment discontinuation, and that other methods of including these costs may be more appropriate because of the number of patients who had post-progression treatment with larotrectinib and the immaturity of the data. The committee did not see evidence for how long patients had treatment with larotrectinib after disease progression and could not comment on whether this was generalisable to NHS clinical practice. The committee concluded that the costs of post-progression larotrectinib should be included, but this issue had not been fully explored. Further data collection within the Cancer Drugs Fund could reduce this uncertainty.

3.29 The committee considered the advice about life-extending treatments for people with a short life expectancy in NICE's guide to the methods of technology appraisal. The company proposed that larotrectinib met the criteria for life-extending treatments for people with a short life expectancy (normally less than 24 months). The committee understood that the end-of-life criteria were not designed for histology-independent treatments and it was not presented with the data needed to assess the criteria for people with NTRK gene fusions specifically. Instead, it was presented with life expectancy data for people with the relevant tumour type irrespective of NTRK gene fusion status and life extension data estimated from the model. It acknowledged the challenges with the data available, for example:

3.30 The company's base-case model gave a deterministic incremental cost-effectiveness ratio (ICER) of £16,155 per quality-adjusted life year (QALY) gained for larotrectinib compared with current clinical management. This included an updated simple discount patient access scheme for larotrectinib after consultation. The committee considered that the base case should also include the following committee preferences:

3.31 Having concluded that larotrectinib could not be recommended for routine use, the committee then considered if it could be recommended for treating NTRK fusion-positive solid tumours within the Cancer Drugs Fund. The committee discussed the arrangements for the Cancer Drugs Fund agreed by NICE and NHS England in 2016, noting NICE's Cancer Drugs Fund methods guide (addendum). It considered that some of the inherent uncertainty in appraising a histology-independent treatment could be resolved through further data collection. The committee was aware that the company expressed a preference for larotrectinib being available in the Cancer Drugs Fund because of the uncertainty in the appraisal and while data mature. The committee was aware that more larotrectinib clinical trial data are expected. Also, the company has some data collection ongoing as part of the regulatory commitments required by the conditions of the marketing authorisation. The committee concluded that some of the uncertainty associated with larotrectinib's use could be addressed through collecting more data. It noted that:

3.32 The committee noted that it had not seen evidence that was likely to be generalisable to clinical practice, or about how larotrectinib was likely to be used in clinical practice. It acknowledged that there was substantial clinical uncertainty about the population, modelling of comparator treatments, survival estimates and utility values. However, it considered that the data from larotrectinib trials were promising, because tumour response rates were good, and it showed that larotrectinib was likely to improve overall and progression-free survival. The committee noted that many of the key clinical uncertainties could be addressed by collecting data in the Cancer Drugs Fund (see section 3.31). The committee then considered if larotrectinib showed plausible potential to be cost effective at the end of the managed access agreement. It noted that the range of ICERs from the available analyses, which included most of the committee's preferred assumptions for the population from the trials compared with established practice, were within what is usually considered a cost-effective use of NHS resources, if larotrectinib meets the end-of-life criteria. However, the committee considered that, because of the many underlying clinical uncertainties, the ICERs were not reliable, and that cost-effectiveness estimates could be improved by collecting data in the Cancer Drugs Fund. The committee recalled, from NHS England's Appraisal and funding of cancer drugs from July 2016 (including the new Cancer Drugs Fund) – a new deal for patients, taxpayers and industry, that the Cancer Drugs Fund is designed 'to offer pharmaceutical companies that are willing to price their products responsibly with a new fast-track route to NHS funding for the best and most promising drugs'. The committee concluded that larotrectinib met the criteria to be included in the Cancer Drugs Fund as an option for people with NTRK fusion-positive solid tumours, if the conditions in the managed access agreement are followed.

3.33 The company considered larotrectinib to be innovative because it targets a gene fusion instead of a tumour type. The patient and clinical experts agreed. The committee considered larotrectinib to be innovative because it represents a major change in the treatment of NTRK fusion-positive solid tumours. The committee understood that important innovations were already underway as part of the NHS long-term plan to improve genomic testing in clinical practice. These advances will likely help the uptake of treatments targeted to a gene alteration. The Cancer Drugs Fund clinical lead explained that histology-independent treatments entering the market are accelerating the advances in genomic testing in the NHS. It is estimated that 100,000 solid tumours will be tested per year once the genomic medicine service is fully established, thought to be in the next 2 years. The committee acknowledged that the improvements in genomic testing would bring wider benefits to the NHS and that these benefits have not been captured in the QALY calculation. The committee concluded that larotrectinib would be beneficial for patients, but it had not been presented with evidence of any additional benefits specific to larotrectinib that were not captured in the measurement of the QALY.

3.34 The company did not consider there to be any equality issues. However, it considered that the uncertainty inherent in this appraisal may pose an equity issue. There is no precedent for appraising technologies when their clinical trials have a basket trial design and a high number of comparators across multiple tumour types. The company considered that patients should have equity of access while health technology assessment methods adapt to these challenges. The committee considered that NICE's single technology appraisal process was appropriate for appraising larotrectinib. It concluded that the uncertainties associated with the trial design (see section 3.14) and multiple comparators (see section 3.17) had been appropriately accounted for in its decision making. The Cancer Drugs Fund clinical lead also noted that there may be issues with accessing larotrectinib because the genomic testing needed to identify NTRK fusion-positive solid tumours is still being established as a national service (see section 3.7). The committee understood that any variation in access to genomic testing will be resolved in the next 1 to 2 years.