Draft guidance consultation

Breast cancer has tumour subtypes that guide treatment. Oestrogen receptor-positive (often referred to as hormone receptor-positive), HER2-negative breast cancer is the most common subtype. PIK3CA gene mutations occur in around 30% to 40% of oestrogen receptor-positive, HER2-negative tumours. Breast cancer with a PIK3CA mutation is associated with a poorer prognosis than breast cancer without the mutation. A patient expert emphasised the importance of having targeted treatment options available at earlier points of the treatment pathway. This is because targeted treatment may increase the time that the condition is relatively stable, delay the need for chemotherapy and allow people with the condition to continue their daily lives. A patient expert explained that people with metastatic breast cancer have repeated progression, which affects many aspects of life, with symptoms including fatigue, cognitive impacts, emotional burden, anxiety and cumulative side effects. They also reported patient perspectives on the importance of continuing stable first-line treatment for longer, and that targeted treatment gives people hope of living longer and having a better quality of life. The committee agreed that there is an unmet need for targeted treatments for oestrogen receptor-positive, HER2-negative, PIK3CA-mutated, metastatic breast cancer.

In advanced breast cancer:

• primary endocrine resistance is defined as:

  • relapse during the first 2 years of adjuvant endocrine treatment, or

  • progression within 6 months of starting endocrine treatment, for metastatic cancer.

• secondary endocrine resistance is defined as:

  • relapse after the first 2 years of adjuvant endocrine treatment, or

  • relapse within 12 months of completing adjuvant endocrine treatment, or

  • progression at least 6 months after starting endocrine treatment, for metastatic cancer.

The clinical evidence for inavolisib plus palbociclib and fulvestrant came from INAVO120, a phase 3, multicentre, double-blind, randomised controlled trial. The trial included adults with PIK3CA-mutated, hormone receptor-positive, HER2-negative, locally advanced or metastatic breast cancer that had progressed either on or within 12 months of completing adjuvant endocrine treatment. People were randomised to inavolisib plus palbociclib and fulvestrant (n=161) or placebo plus palbociclib and fulvestrant (n=164). People remained on treatment until disease progression, toxicity or loss of clinical benefit. The primary outcome was progression-free survival (PFS), and secondary outcomes included overall survival (OS), response rate, duration of response, time to off-treatment and adverse effects. At the PFS full-analysis set clinical cut-off date (September 2023), the mean PFS for inavolisib plus palbociclib and fulvestrant was 15.0 months (95% confidence interval [CI] 11.3 to 20.5), compared with 7.3 months (95% CI 6.6 to 9.3) for placebo plus palbociclib and fulvestrant (hazard ratio [HR] 0.43; 95% CI 0.32 to 0.59; p<0.0001). At the OS clinical cut-off date (November 2024), the mean OS for inavolisib plus palbociclib and fulvestrant was 34.0 months (95% CI 28.4 to 44.8) compared with 27.0 months (95% CI 22.8 to 38.7) for placebo plus palbociclib and fulvestrant (HR 0.67; 95% CI 0.48 to 0.94; p=0.019). The committee noted the maturity of the trial data and concluded that the INAVO120 trial results showed that inavolisib plus palbociclib and fulvestrant statistically significantly increased PFS and OS compared with placebo plus palbociclib and fulvestrant.

In INAVO120, only 3 people (2 in the inavolisib plus palbociclib and fulvestrant arm; 1 in the placebo plus palbociclib and fulvestrant arm) had a CDK 4 and 6 inhibitor as part of adjuvant or neoadjuvant treatment. The EAG noted that, in NHS clinical practice, aromatase inhibitor monotherapies were used as adjuvant treatments for hormone receptor-positive, HER2-negative early breast cancer. NICE has since recommended abemaciclib plus an aromatase inhibitor (NICE technology appraisal guidance TA810, published in 2022) and ribociclib plus an aromatase inhibitor (NICE technology appraisal guidance TA1086; published in 2025) as adjuvant treatment options for early breast cancer at a high risk of recurrence. Because of this, the proportion of people with locally advanced or metastatic cancer who have adjuvant treatment with a CDK 4 and 6 inhibitor plus an aromatase inhibitor is expected to increase over time. So, the EAG noted it was unclear if the trial results were generalisable to the NHS. It also noted that it was unclear whether prior CDK 4 and 6 inhibitor plus aromatase inhibitor treatment affects the relative effectiveness of inavolisib plus palbociclib and fulvestrant, or a CDK 4 and 6 inhibitor plus fulvestrant. A clinical expert noted that, in the NHS, people can have retreatment with a CDK 4 and 6 inhibitor if there is at least a 12-month interval between stopping CDK 4 and 6 inhibitor treatment and detecting recurrence. They noted that the same principle would be expected for inavolisib plus palbociclib and fulvestrant. The NHS England Cancer Drugs Fund clinical lead (from now, CDF clinical lead) added that around 7,500 people per year in England are starting adjuvant treatment with a CDK 4 and 6 inhibitor. A clinical expert noted that adjuvant CDK 4 and 6 inhibitor use is likely to reduce the number of recurrences by around a third. They noted that PIK3CA-mutated cancer is associated with a poorer prognosis and rapid progression. So, it is more likely to relapse within a year of treatment and would not be eligible for retreatment with a CDK 4 and 6 inhibitor. The EAG also highlighted that people in the trial were younger (with a mean age of 54 years), fitter (with a lower Eastern Cooperative Oncology Group [ECOG] performance status) and weighed less than the expected NHS population. Also, the trial included more people with Asian ethnicity and fewer people with Black or African American ethnicity than the expected NHS population. The committee noted that hyperglycaemia was apparent in 60% of people in the inavolisib plus palbociclib and fulvestrant arm of INAVO120, with 74% having treatment with an antidiabetic medicine. The clinical experts explained that people with insulin-dependent diabetes are unlikely to have inavolisib because of concerns with glycaemic control. Also, because hyperglycaemia is a reversible side effect, often happening early in treatment, preventive measures to control blood glucose or adjust dose can be implemented. They also noted that similar issues occur with alpelisib and capivasertib, which they already have experience with. So, the clinical experts did not think this would impact uptake of inavolisib plus palbociclib and fulvestrant. The CDF clinical lead added that real-world evidence for starting a CDK 4 and 6 inhibitor plus fulvestrant in England shows that:

• around 45% of people are prescribed palbociclib plus fulvestrant, with an average age of 73

• around 40% are prescribed ribociclib plus fulvestrant, with an average age of 63.
  
  The clinical experts explained that palbociclib is favoured for people who are frailer and may be on multiple drugs. This is because it has fewer side effects than ribociclib and because ribociclib may not be suitable for some people, such as people with cardiac comorbidities. They also explained that a CDK 4 and 6 inhibitor can be used later in the treatment pathway as a rescue option, often in older people. As a result, people switching to a CDK 4 and 6 inhibitor plus fulvestrant may be older on average. Because inavolisib plus palbociclib and fulvestrant is a triplet treatment regimen for established endocrine resistance, the clinical experts expected it to be used by people who are sufficiently fit for triplet treatment. The committee noted the increasing number of people having adjuvant treatment with a CDK 4 and 6 inhibitor in the NHS compared with the trial. It also noted the large difference in age of people in the trial compared with people in the NHS having treatment with the comparators. It concluded that the generalisability of the trial population to the NHS population was uncertain.

There was no clinical trial directly comparing inavolisib plus palbociclib and fulvestrant with abemaciclib plus fulvestrant or ribociclib plus fulvestrant. So, the company used a network meta-analysis (NMA) to indirectly compare PFS and OS for abemaciclib plus fulvestrant and ribociclib plus fulvestrant with that for inavolisib plus palbociclib and fulvestrant. The company did an NMA with a population of mixed endocrine resistance using 7 randomised controlled trials: CAPItello-291, FLIPPER, INAVO120, MONALEESA-3, MONARCH-2, MONARCH plus and PALOMA-3. The results suggested an improvement in PFS (HR 0.42, 95% CI 0.21 to 0.87) and OS (HR 0.58, 95% CI 0.24 to 1.42) for inavolisib plus palbociclib and fulvestrant compared with abemaciclib plus fulvestrant. The results also suggested an improvement in PFS (HR 0.38, 95% CI 0.18 to 0.81) and OS (HR 0.73, 95% CI 0.3 to 1.8) for inavolisib plus palbociclib and fulvestrant compared with ribociclib plus fulvestrant. The results imply that inavolisib plus palbociclib and fulvestrant is more effective compared with abemaciclib plus fulvestrant than with ribociclib plus fulvestrant. An NMA for only the endocrine-resistant population was also done, which excluded the FLIPPER and MONALEESA-3 trials. However, because ribociclib plus fulvestrant was only connected into the network through the MONALEESA-3 trial, the endocrine-resistant NMA does not have comparative efficacy results for ribociclib plus fulvestrant. The EAG noted that the endocrine-resistant NMA was more relevant to the scope population but having no results for ribociclib plus fulvestrant was an important limitation. The EAG also noted the heterogeneity between the baseline characteristics of the trials within the NMA which align with previously identified potential treatment effect modifiers of CDK 4 and 6 inhibitor regimens in this population. It highlighted that the PIK3CA mutation may also be a treatment effect modifier. The EAG considered clinical advice that the CDK 4 and 6 inhibitor plus fulvestrant regimens have similar efficacy and that TA836 concluded that palbociclib, abemaciclib and ribociclib, each with fulvestrant, were likely to provide similar health benefits for PFS and OS. The clinical experts agreed that, in clinical practice, the choice of CDK 4 and 6 inhibitor depends on comorbidities and side effect profiles, but it is assumed the treatment effects of all 3 are all similar. The company noted that because some trials in the NMA had unconfirmed PIK3CA mutation status, the clinical benefit of abemaciclib plus fulvestrant and ribociclib plus fulvestrant may be overestimated because of the poorer outcomes associated with having a PIK3CA mutation. A clinical expert noted that, although the overall outcomes with a PIK3CA mutation are worse, this does not impact the relative benefit of a CDK 4 and 6 inhibitor. They noted that the INAVO120 control arm is an appropriate reflection of what would be expected in clinical practice with a PIK3CA mutation. The committee noted the issues with the NMA including the mixed endocrine-resistant population, as well as potential treatment effect modifiers highlighted by the EAG including PIK3CA mutation status. The committee noted the expert testimony and previous TA836 conclusion. It concluded that it was not appropriate to use the NMA results when comparing inavolisib plus palbociclib and fulvestrant with abemaciclib, palbociclib or ribociclib, each with fulvestrant. Instead, it was appropriate to assume similar treatment efficacy for the CDK 4 and 6 inhibitors plus fulvestrant combinations.

The company used a partitioned survival model with 4 mutually exclusive health states: PFS (the model entry), first progressive disease (PD1), second progressive disease (PD2), and death. At each model cycle, people can remain in the PFS health state or transition to the PD1 or death health state. Transition to the PD1 health state risks further progression to the PD2 or death health states. Those in the PD2 health state remain here until transitioning to the death health state. The model had 7-day cycles with a half-cycle correction to health outcomes and costs. It used a time horizon of 40 years, which was assumed to be a lifetime time horizon. The committee concluded that the model structure was appropriate for decision-making.

To extrapolate PFS, the company did a piecewise approach using INAVO120 PFS Kaplan–Meier data, then applied a parametric extrapolation beyond the trial follow-up. Clinical advice to the company was that log-logistic or log-normal extrapolations are most clinically plausible because they estimate a 5% PFS at 10 years. The company preferred to fit a log-normal distribution to the entire dataset but only applied the log-normal distribution once 20% of people remained at risk, referencing recommendations from Pocock, Clayton and Altman (2002). The company stated that this approach avoids under- or over-estimation that may occur when using parametric distributions from the start of the model. The EAG questioned the need for a piecewise approach since the parametric distributions had good visual fit to the Kaplan–Meier data in both arms. It found it unclear whether there were statistically and clinically significant turning points in the hazards. The EAG noted that the 20% patient-at-risk threshold is an arbitrary value and debatable, and cost-effectiveness results were sensitive to this assumption. It preferred to apply a log-logistic distribution from the start of the model because of its statistical and visual fit to the Kaplan–Meier data and clinical plausibility, and because it was appropriate for both treatment arms. The committee considered the different parametric distributions and concluded that the log-logistic distribution offered a good fit to the data. It agreed that using the 20% patient-at-risk threshold was arbitrary and added uncertainty to the modelling, so it preferred to apply the log-logistic distribution from the start of the model.

The company used the same approach as used for PFS to extrapolate OS, but instead used a gamma distribution and a 30% patient-at-risk threshold. The company explained that the higher threshold was because data is often more immature at this stage, and a higher threshold would mean the extrapolations were informed by a more robust dataset. To prevent illogical survival estimates, both the PFS1 and PFS2 curves were capped by OS. The EAG also used a gamma distribution to extrapolate OS but applied this from the start of the model. The EAG noted that applying a cap causes kinks in the curves which imply a sharp change in hazards, which may not be clinically plausible. It preferred to constrain PFS1 and PFS2 hazards so they cannot fall below the mortality hazard estimated from the OS curve, shifting the tail of the survival curves downwards. This ensured the extrapolations were logically consistent and clinically plausible. The committee noted that this issue was associated with the limitations of partitioned survival models that independently model survival endpoints, especially when there is more than one progressed disease health state. It noted that it may be important to preserve information on PFS for longer in the model, rather than prematurely shifting down the curve before it crosses the OS curve. The committee preferred to apply the unconstrained log-logistic distribution from the start of the model for PFS (see section 3.7). For OS, it preferred to apply the gamma distribution from the start of the model, but to cap PFS by the absolute value of OS so that PFS cannot exceed mortality.

The company used time-to-off-treatment data from INAVO120 to model treatment duration for both arms. It preferred to use a different parametric distribution than the log-normal distribution used for the PFS curves, stating that around 17% of people stopped treatment for reasons other than progression. To align with clinical expert opinion, it used a gamma distribution to extrapolate time to off-treatment at a 20% patient-at-risk threshold. The EAG noted that the separation between time to off-treatment and PFS curves was largely driven by discontinuation because of toxicity or adverse effects. The company's modelling implies that treatment discontinuation because of toxicity increases over time when in the progression-free health state because the time-to-off-treatment and PFS curves diverge. But clinical advice to the EAG was that most discontinuations because of toxicity occur early, and drug-associated adverse events can be well managed, so after 2 to 3 years the discontinuation rate is expected to be negligible. So, the EAG preferred to use the same distribution as for PFS (log-logistic), applied from the start of the model. The EAG modelling implies that the rate of treatment discontinuation because of toxicity decreases over time. The committee noted that a higher discontinuation rate would be expected earlier on, but people may also still discontinue treatment before their cancer progresses. The committee noted that when using the generalised gamma approach, a higher discontinuation rate was seen earlier on as in the EAG's preferred distribution, but it maintained a separation of the treatment duration and PFS curves towards the end of treatment. The committee concluded that the generalised gamma distribution was the most appropriate to model treatment duration.

In the company model, treatment efficacy of abemaciclib plus fulvestrant and ribociclib plus fulvestrant was modelled by applying the PFS and OS hazard ratios from the mixed endocrine-resistant NMA to the PFS and OS curves for inavolisib plus palbociclib and fulvestrant. Treatment duration of abemaciclib plus fulvestrant and ribociclib plus fulvestrant was estimated by applying the PFS hazard ratios from the mixed endocrine-resistant NMA to the palbociclib plus fulvestrant time to off-treatment curves for inavolisib plus palbociclib and fulvestrant. In the company base case, the mean survival for abemaciclib plus fulvestrant was lower than that for palbociclib or ribociclib plus fulvestrant, reflecting the NMA results. The EAG preferred to assume equivalent efficacy of the CDK 4 and 6 inhibitors plus fulvestrant, so it modelled efficacy by matching PFS, OS and time to off-treatment with palbociclib plus fulvestrant using results from INAVO120. The committee had already concluded that all of the CDK 4 and 6 inhibitors have very similar efficacy (see section 3.5). It concluded that it preferred the EAG's approach to model the treatment efficacy of abemaciclib plus fulvestrant and ribociclib plus fulvestrant by matching PFS, PFS2, OS and time to off-treatment with palbociclib plus fulvestrant using the results from INAVO120.

The company applied the following relative dose intensities in the model to account for the impact of dose reductions and interruptions on the costs of treatment:

• inavolisib plus palbociclib and fulvestrant: mean relative dose intensity from INAVO120

• palbociclib and fulvestrant: mean relative dose intensity from INAVO120

• abemaciclib plus fulvestrant: 100% relative dose intensity

• ribociclib plus fulvestrant: 100% relative dose intensity.

The subsequent treatment options for the PD1 and PD2 health states in the model were capecitabine, elacestrant (for cancer with an ESR1 mutation), everolimus plus exemestane, paclitaxel and best supportive care (for PD2 only). In the company's model, the proportion of people that had each subsequent treatment option in the PD1 and PD2 health states was informed by healthcare professionals from the company's clinical advisory board meeting. The company assumed that fewer people have elacestrant after inavolisib plus palbociclib and fulvestrant than after a CDK 4 and 6 inhibitor plus fulvestrant. The company also assumed that fewer people would have everolimus plus exemestane after inavolisib plus palbociclib and fulvestrant than after a CDK 4 and 6 inhibitor. This is because less benefit was assumed from having further endocrine-based treatment after inavolisib plus palbociclib and fulvestrant. The EAG instead assumed that the distributions for elacestrant and everolimus plus exemestane would only differ if progression was before or after 12 months (applies to only PD1 for everolimus plus exemestane). The clinical experts noted that more people may be eligible for elacestrant after inavolisib plus palbociclib and fulvestrant because survival on this treatment would be improved, and a minimum of 12 months treatment with a CDK 4 and 6 inhibitor is needed before elacestrant eligibility. However, healthcare professionals may not choose elacestrant when there is also a PIK3CA mutation, so uptake may be low. Healthcare professionals may also be cautious of using everolimus plus exemestane after a treatment that targets the same signalling pathway. The committee agreed that, based on the clinical experts' testimony, it was appropriate to model subsequent treatment distributions using the proportions informed by the company's clinical advisory board meeting.

The company and EAG did not include the cost of PIK3CA testing in their economic models because the National Genomics Test Directory includes this testing. The clinical experts explained that PIK3CA mutations often occur early in the condition, so it is appropriate to test for the mutation at diagnosis of the metastatic stage of the condition. However, because PIK3CA-mutation testing is usually done at a later stage of treatment, together with ESR1-mutation testing, an additional step for testing would need to be implemented. The EAG reported that, for inavolisib plus palbociclib and fulvestrant, PIK3CA testing would need to happen earlier in the treatment pathway. The additional cost is uncertain because it is unknown how many people would have testing for other gene mutations in the second-line setting. It included a scenario in the economic model that included PIK3CA testing costs. However, the committee concluded that PIK3CA testing costs had been accounted for in the economic modelling, because it is already funded by its inclusion in the National Genomics Test Directory.

NICE's technology appraisal and highly specialised technologies guidance manual notes that, above a most plausible incremental cost-effectiveness ratio (ICER) of £25,000 per QALY gained, judgements about the acceptability of a technology as an effective use of NHS resources will take into account the degree of certainty around the ICER. The committee will be more cautious about recommending a technology if it is less certain about the ICERs presented. But it will also take into account other aspects including uncaptured health benefits. The committee noted the uncertainty associated with the generalisability of the trial to the NHS, including the age of the population. It also noted that the outcomes from the trial were mature and includes a direct comparison of inavolisib plus palbociclib and fulvestrant with palbociclib plus fulvestrant. So, the committee concluded that an acceptable ICER would be towards the upper end of the range NICE considers an effective use of NHS resources (£25,000 to £35,000 per QALY gained).

The company's corrected base-case deterministic and probabilistic ICERs were above the range normally considered an acceptable use of NHS resources. In the EAG's base case, the ICERs were substantially above the range normally considered an acceptable use of NHS resources. The exact ICERs cannot be reported here because of confidential commercial discounts.

The committee's preferred assumptions were to:

• apply log-logistic distribution from the start of the model time horizon to model PFS and cap these by the absolute value of OS (see sections 3.7 and 3.8)

• apply a generalised gamma distribution to model treatment duration (see section 3.9)

• assume equal treatment efficacy of abemaciclib plus fulvestrant, palbociclib plus fulvestrant and ribociclib plus fulvestrant in the model, and match efficacy results to INAVO120 (see section 3.10)

• use adjusted INAVO120 relative dose intensity values that exclude the effect of dose reductions for inavolisib and palbociclib, as well as INAVO120 relative dose intensities applied to abemaciclib, ribociclib and fulvestrant (see section 3.11)

• model subsequent treatments assuming that a lower proportion of people would have elacestrant and everolimus plus exemestane as a subsequent treatment after inavolisib plus palbociclib and fulvestrant than after a CDK 4 and 6 inhibitor plus fulvestrant (see section 3.12)

• exclude additional PIK3CA-mutation testing costs (see section 3.13)

• apply a severity weight of 1.2 to the QALYs (see section 3.14)

• apply the EAG's additional model changes (see section 3.15).

No equality issues were raised by the company, EAG or stakeholders. The committee did not identify any equality issues.

The committee considered whether there were any uncaptured benefits of inavolisib plus palbociclib and fulvestrant. It did not identify any additional benefits of inavolisib plus palbociclib and fulvestrant not already captured in the economic modelling. So, the committee concluded that all additional benefits of inavolisib plus palbociclib and fulvestrant had been taken into account.

The committee considered that inavolisib plus palbociclib and fulvestrant is an effective treatment in terms of PFS and OS compared with standard care. But, when applying its preferred assumptions, the cost-effectiveness estimates for inavolisib plus palbociclib and fulvestrant were above the range that NICE considers an effective use of NHS resources. So, inavolisib plus palbociclib and fulvestrant should not be used to treat recurrent oestrogen receptor-positive, HER2-negative, PIK3CA-mutated advanced breast cancer after adjuvant endocrine treatment.