Neratinib for extended adjuvant treatment of hormone receptor-positive, HER2-positive early stage breast cancer after adjuvant trastuzumab

3.1 NICE's guideline on early breast cancer recommends offering adjuvant trastuzumab for 1 year in combination with surgery, chemotherapy and radiotherapy as appropriate. In March 2019, NICE's technology appraisal guidance on pertuzumab for adjuvant treatment of HER2-positive early stage breast cancer recommended it, with trastuzumab and chemotherapy, for treating human epidermal growth factor receptor 2 (HER2)‑positive breast cancer in adults who have lymph node-positive disease. Neoadjuvant treatment is also an option, to reduce tumour size before surgery. NICE's technology appraisal guidance on pertuzumab for the neoadjuvant treatment of HER2-positive breast cancer recommended it, with trastuzumab and chemotherapy, for HER2‑positive, locally advanced, inflammatory or early breast cancer at high risk of recurrence. Neratinib is an oral treatment with a marketing authorisation for the extended adjuvant treatment of adults with hormone receptor-positive, HER2‑positive early breast cancer who are less than 1 year from the completion of previous adjuvant trastuzumab-based treatment.

3.2 Since the clinical study for neratinib was done, the treatment pathway has changed. People can now have pertuzumab alongside trastuzumab for adjuvant treatment in HER2‑positive early stage breast cancer if they have lymph node-positive disease. There are no data for neratinib after combination treatment with pertuzumab for early breast cancer. The clinical experts noted that adjuvant pertuzumab treatment is not available for people with node-negative disease and that neratinib could be considered in this population. The clinical experts stated that the decision about the most suitable treatment for patients with node-positive disease would be based on clinical judgement, based on the efficacy of the appropriate treatment and the patient's preferences. The clinical experts explained that some people may prefer neratinib because it's an oral treatment. They may prefer to have adjuvant treatment with just trastuzumab as the HER2‑directed treatment administered subcutaneously, followed by extended adjuvant treatment with neratinib, because both treatments can be given at home, rather than intravenous adjuvant combination treatment with pertuzumab. However, it was noted that neratinib's high level of toxicity (see section 3.7) may mean some people have to go to hospital for diarrhoea treatment. The committee recognised that some people who did not have adjuvant pertuzumab would benefit from extended adjuvant treatment with neratinib. However, it is unclear which people would have neratinib in clinical practice.

3.3 ExteNET (n=2,840) is a phase 3 randomised controlled trial that compared neratinib treatment with placebo in women with HER2-positive breast cancer who had completed adjuvant trastuzumab treatment within 2 years. The company based its submission on a subgroup of women with early hormone receptor-positive cancer who were less than 1 year from completing adjuvant trastuzumab-based treatment (n=1,334). This subgroup, referred to as the 'label' population, is in line with the neratinib marketing authorisation. It showed better clinical effectiveness and fewer side effects than the whole trial. The ERG noted that ExteNET was not designed to have statistical power to detect differences between treatments within subgroups. In addition, it noted that only 80 patients were recruited in the UK and that differences in clinical effectiveness by geographical region were reported. The clinical experts suggested that the trial is generalisable to the UK population. However, they noted that clinical trials tend to recruit fitter participants, and that there may be some limitation to the trial generalisability. The committee concluded that ExteNET is the key trial relevant to this appraisal and suitable for estimating the clinical effectiveness of neratinib.

3.4 ExteNET did not include people who had a pathological complete response after neoadjuvant treatment. A pathological complete response means that there was no residual invasive cancer in the breast or axilla after completing neoadjuvant treatment. The committee concluded that there are no data on the clinical effectiveness of neratinib in people who had a pathological complete response after neoadjuvant treatment.

3.5 ExteNET did not include people who had pertuzumab and trastuzumab as adjuvant treatment. The committee concluded that there are no data on the clinical effectiveness of extended adjuvant neratinib after adjuvant pertuzumab treatment.

3.6 Invasive disease-free survival was the primary outcome in ExteNET. The 5‑year hazard ratio for invasive disease-free survival for the whole trial was 0.73 (95% confidence interval [CI] 0.57 to 0.92). The 5‑year hazard ratio for invasive disease-free survival for the label population was 0.58 (95% CI 0.41 to 0.82). This suggests a statistically significant improvement in invasive disease-free survival for neratinib compared with placebo. Overall-survival data were collected in the trial but had not been fully analysed at the time of this appraisal, when 121 deaths had been recorded in the whole trial. The final analysis is expected after 248 deaths. Overall-survival results for the label population were not available at the time of this appraisal. The committee concluded that neratinib improves invasive disease-free survival compared with placebo in the label population. However, it is not known whether neratinib increases the length of time people live because the final trial results are not available.

3.7 Neratinib is associated with high rates of diarrhoea, nausea and fatigue and may require hospital visits to treat diarrhoea. In ExteNET, diarrhoea prophylaxis was not used, although people were treated for diarrhoea as needed. The company explained that if diarrhoea prophylaxis is used, diarrhoea lasts approximately 5 days. The clinical experts confirmed that diarrhoea is not an issue for all people and that it can usually be managed with prophylaxis. One clinical expert reported feedback from 2 people who have taken neratinib. One person did not have any diarrhoea, and the other had severe symptoms. It took some time to adjust the diarrhoeal treatment for this person before the symptoms diminished, but the person wanted to continue taking neratinib. The committee agreed that diarrhoea toxicity is high in some people but could be managed with prophylaxis and diarrhoeal treatments.

3.8 The company developed a 5‑state Markov model to evaluate the cost effectiveness of neratinib (the states were: invasive disease-free, local recurrence, remission, distant recurrence and dead). In the absence of overall-survival data, invasive disease-free survival in the label population of ExteNET and a general population mortality rate was used to estimate overall survival. The ERG noted that death from breast cancer is only possible from the distant recurrence health state. Mortality risk for all other health states (invasive disease-free, local recurrence and remission) is based on general population mortality. The ERG said that this assumption could underestimate the cost-effectiveness results. The company explained that patients are most likely to move into the distant recurrence health state first and that no patients in ExteNET died from breast cancer without first experiencing a distant recurrence. The clinical experts and the ERG agreed that this was a reasonable assumption. The committee was satisfied that the use of the general mortality in the model was appropriate.

3.9 The company investigated whether neratinib's treatment effect could be modelled assuming proportional hazards between the neratinib and placebo arms of the trial. It concluded that it could, and invasive disease-free survival data in the label population were pooled and modelled together with a treatment effect as a covariate. The company chose a flexible-spline Weibull with 1 knot to model invasive disease-free survival. The ERG explained that the assumption of proportional hazards is uncertain because some of the analyses provided by the company suggested that it is not valid. The ERG assessed an overall-goodness-of-fit of the models considered by the company, and stratified models for which the proportional hazard assumption is not needed. It found that the stratified generalised gamma model provided the best overall fit for the invasive disease-free survival data and included this model in its preferred base case. In response to technical engagement the company considered the ERG's model to be a conservative approach resulting in cost-effectiveness results that are at the high end of the most plausible range. The clinical experts considered both extrapolations plausible. The committee concluded that it is unclear which approach to invasive disease-free survival modelling is the most appropriate and that both approaches could be plausible.

3.10 The company assumed that the treatment effect observed in ExteNET would last beyond the trial time horizon until patients had a risk of invasive disease-free survival equal to the mortality rate in the general population. The duration of the treatment effect depends on the curve used to model invasive disease-free survival and general population mortality. The company assumed a continued effect and applied the invasive disease-free survival 5‑year hazard ratio from the ExteNET label population of 0.58 (95% CI 0.41 to 0.82) from month 62.98 (as observed 4 years after 1 year of treatment with neratinib) to month 129 (when neratinib and general population mortality hazards are the same). This continued effect is followed with an implicit taper period until month 176 (when placebo and general population mortality hazards are the same). Using the ERG's preferred method to extrapolate invasive disease-free survival, the treatment effect stops at month 140. However, the ERG assumed that the tapering starts at the end of the trial and lasts until month 140. In response to the technical engagement, the company considered the ERG's approach to modelling neratinib treatment effect and duration to be plausible, although conservative. The clinical experts considered the ERG's approach to be appropriate. The committee concluded that the ERG's approach to modelling neratinib treatment effect and duration is appropriate for decision making.

3.11 Both the company's and ERG's preferred incremental cost-effectiveness ratios (ICERs) for people for whom trastuzumab is the only HER2-directed adjuvant treatment they have had, and who did not have a pathological complete response to neoadjuvant treatment (see section 3.4 and section 3.5) are within the range NICE normally considers an acceptable use of NHS resources. ICERs were presented as commercial in confidence to maintain the confidentiality of the proposed commercial agreement for neratinib, and therefore cannot be reported here. The committee was aware of the uncertainty associated with some of the inputs and assumptions in the model (for example, invasive disease-free survival modelling as discussed in section 3.8 and section 3.9), and the impact of the additional uncertainties as summarised in the technical report (table 2 on pages 26 to 29, and table 3 on pages 30 and 31). It considered that the most plausible ICER is unknown and could be higher or lower than the company's and ERG's preferred ICERs. However, it agreed that the most plausible ICER is unlikely to exceed the range NICE normally considers an acceptable use of NHS resources. It therefore recommended neratinib for hormone receptor-positive, HER2‑positive early breast cancer in adults who completed adjuvant trastuzumab-based treatment less than 1 year ago, only if trastuzumab was the only HER2‑directed adjuvant treatment they have had, and if their cancer did not have a pathological complete response to neoadjuvant treatment (if they had neoadjuvant chemotherapy-based regimens).