Capivasertib with fulvestrant for treating hormone receptor-positive HER2-negative advanced breast cancer after endocrine treatment

The clinical experts explained that initial treatment for HR‑positive HER2‑negative advanced breast cancer is usually a cyclin-dependent kinase (CDK) 4 and 6 inhibitor plus an aromatase inhibitor (a type of endocrine treatment). After this initial treatment, the main options are alpelisib plus fulvestrant (if the cancer has a PIK3CA alteration) or everolimus plus exemestane (unless chemotherapy is needed because of symptomatic visceral disease; see NICE's technology appraisal guidance on alpelisib with fulvestrant for treating HR-positive HER2-negative PIK3CA-mutated advanced breast cancer [TA816] and on everolimus with exemestane for treating advanced breast cancer after endocrine therapy [TA421]). The clinical experts explained that there is an unmet need because current endocrine-based treatment is relatively ineffective after initial endocrine treatment plus a CDK 4 and 6 inhibitor. They said alpelisib plus fulvestrant is associated with toxicity that has substantially limited its use in the NHS. And that everolimus plus exemestane has low response rates and exemestane is less effective for ESR1‑positive cancer, which is up to 50% of cancers in this population. They said that treatments that make endocrine-based treatment more effective after progression on endocrine treatment plus a CDK 4 and 6 inhibitor are important. The clinical experts explained that if endocrine-based treatment fails, chemotherapy is the main option, but delaying this is highly important for patients because of the toxicity it involves. They added that they believed capivasertib plus fulvestrant is a highly effective and well-tolerated targeted treatment that represents a step change in managing breast cancer with PIK3CA, AKT1 or PTEN alterations. The clinical experts said that the toxicity is much lower with capivasertib plus fulvestrant than alpelisib plus fulvestrant, so quality of life is likely to be better on treatment. The committee noted that side effects including hyperglycaemia, rash and stomatitis (an inflamed and sore mouth) were listed in the summary of product characteristics for capivasertib. The clinical experts explained that these side effects are more significant with alpelisib plus fulvestrant or everolimus plus exemestane. They said that there are substantial challenges with managing hyperglycaemia associated with alpelisib plus fulvestrant. For example, treatment requires learning to use a glucose monitor because of the risk of diabetes. The clinical experts added that rash was a more serious problem with alpelisib and that stomatitis could be a serious problem with everolimus. The patient experts also highlighted the value of having a well-tolerated treatment that delays progression and targets AKT1 or PTEN gene alterations, noting that there are currently none available. The committee recognised the limitations with existing treatments, the advantages of capivasertib plus fulvestrant, and the importance of having a targeted treatment for HR‑positive HER2‑negative advanced breast cancer with PIK3CA, AKT1 or PTEN alterations.

The marketing authorisation for capivasertib plus fulvestrant is for people whose cancer has recurred or progressed on or after an endocrine-based regimen. The company submitted evidence for a narrower population: people whose cancer has recurred or progressed after a CDK 4 and 6 inhibitor plus an aromatase inhibitor. The company said that this positioning reflected the anticipated use of capivasertib plus fulvestrant in the current UK treatment pathway. It did not anticipate capivasertib being used for people who had not already had a CDK 4 and 6 inhibitor. The committee noted the clinical experts' comments that a CDK 4 and 6 inhibitor plus an aromatase inhibitor is standard initial treatment for HR‑positive HER2‑negative advanced breast cancer (see section 3.2). It also noted that the company's positioning of capivasertib plus fulvestrant reflected that of alpelisib plus fulvestrant in TA816. The committee concluded that the company's positioning of capivasertib plus fulvestrant after progression on a CDK 4 and 6 inhibitor plus an aromatase inhibitor was appropriate and in line with expected clinical use.

The company submitted evidence against 2 comparators: alpelisib plus fulvestrant for PIK3CA-altered breast cancer and everolimus plus exemestane. It said that these were the standard treatments after initial endocrine treatment with a CDK 4 and 6 inhibitor plus an aromatase inhibitor, where capivasertib would be used. It said it expected capivasertib plus fulvestrant to be used mainly in place of alpelisib plus fulvestrant because most people with PI3K and AKT pathway-altered tumours have PIK3CA alterations. The clinical experts agreed that these are the standard treatments after initial endocrine treatment (see section 3.2). The committee concluded that alpelisib plus fulvestrant for PIK3CA-altered breast cancer and everolimus plus exemestane were the relevant comparators for the evaluation.

The key clinical trial, CAPItello‑291, was a randomised, double-blind, placebo-controlled phase 3 trial comparing capivasertib plus fulvestrant with placebo plus fulvestrant. It included people whose HR‑positive HER2‑negative breast cancer had recurred or progressed on or after treatment with an aromatase inhibitor with or without a CDK 4 and 6 inhibitor. People who had previous fulvestrant were excluded. A subgroup of people in the trial had PI3K and AKT pathway-altered (PIK3CA, AKT1 or PTEN) tumours. Results showed that capivasertib plus fulvestrant significantly improved progression-free survival (median 7.3 months; 95% confidence interval [CI] 5.5 to 9.0) compared with placebo plus fulvestrant (median 3.1 months; 95% CI 2.0 to 3.7) in people with PI3K and AKT pathway-altered tumours. Results were similar for people who had previously had a CDK 4 and 6 inhibitor (the company considers this data confidential, so it is not reported here). For overall survival, there was not enough data for a formal analysis. For people with PI3K and AKT pathway-altered tumours, overall survival was better on capivasertib plus fulvestrant, although this was not statistically significant (hazard ratio 0.69; 95% CI 0.45 to 1.05). The committee concluded that capivasertib plus fulvestrant improved progression-free survival compared with placebo plus fulvestrant.

Because there was no direct evidence comparing capivasertib plus fulvestrant with alpelisib plus fulvestrant or everolimus plus exemestane, the company did indirect treatment comparisons. The network meta-analysis (NMA) included 10 randomised controlled trials. Data for capivasertib plus fulvestrant came from the CAPItello‑291 and FAKTION trials; for alpelisib plus fulvestrant from the SOLAR‑1 trial, and for everolimus plus exemestane from the BOLERO‑2 and BOLERO‑5 trials.

To estimate long-term progression-free survival, the company fitted parametric survival models (exponential, log-normal, Weibull, log-logistic, gamma, generalised gamma and Gompertz) to the patient-level data of the placebo plus fulvestrant arm from CAPItello‑291 (the common comparator in the NMA). It chose the log-normal distribution in its base case (preferred by 1 out of 5 of the company's clinical experts), based on goodness of fit, visual inspection and clinical opinion. The company explored the log-logistic distribution in a scenario, noting that generalised gamma was also suitable (these 2 distributions were preferred by 3 out of 5 of the company's clinical experts). The EAG preferred the log-logistic distribution for its base case because it had the best fit according to the Akaike information criterion (AIC) and Bayesian information criterion (BIC) and it was preferred by most of the company experts. The committee concluded that the log-logistic model was the most appropriate to estimate long-term progression-free survival because it had the best fit to the trial data and was preferred by most of the company's clinical experts.

The company estimated progression-free survival and overall survival for all 3 treatments using the hazard ratios from the NMA. It applied these to the modelled placebo plus fulvestrant progression-free survival curve for people with PI3K and AKT pathway-altered tumours who had previous CDK 4 and 6 inhibitor plus endocrine treatment. This was done on the assumption that proportional hazards applied for the studies in the network (that is, the assumption that the relative hazards remain constant over time for each treatment comparison). The company said it had found no consistent evidence that this assumption did not hold. It said the fixed-effects model had the best statistical fit to the trial data, so it used the hazard ratios from this in its base-case model. The company also provided a time-varying approach, using a piecewise NMA with 2 different hazard ratios and time periods of:

• 0 to 3 months and over 3 months for progression-free survival (and 0 to 2 months and over 2 months in another scenario)

• 0 to 6 months and over 6 months for overall survival.
  
  The EAG said that a time-varying analysis was more appropriate because of evidence that the proportional hazards assumption did not hold for progression-free survival in some trials (and possibly for overall survival). It thought that the company's piecewise NMA was an improvement over the constant hazard ratio NMA and used this in its base case. But it noted that the reasons for the time periods chosen were not well justified. It also preferred an NMA using a time-varying parametric model. The committee agreed with the EAG and thought that the piecewise time-varying NMA did not properly explore a time-varying hazard ratio. It thought that neither the constant nor piecewise NMAs were appropriate for modelling relative treatment effect. It concluded that it would prefer to see fully time-varying analyses for modelling survival, such as using a fractional polynomial model. After consultation, the company modelled long-term relative treatment effect using a fractional polynomial approach in which the hazard ratio varied over time. The EAG said the company's fractional polynomial analysis and selection of best fit models were appropriate and used them in its base case. It noted that a different fractional polynomial model from the one the company chose to estimate progression-free survival could be more clinically plausible. But the company noted that they were both a good fit and equally plausible, and the hazard ratios over time were very similar. The committee was reassured that appropriate models had been used to estimate overall and progression-free survival. It concluded that the fractional polynomial approach was appropriate for decision making and reduced the uncertainty around the estimates of long-term treatment effect.

The company did not initially incorporate treatment waning into its model. That is, the model assumed the relative benefits of treatment continued at the same level over time. It said it believed that the piecewise, time-varying NMA accounted for treatment waning because of the 3‑month (progression-free survival) and 6‑month (overall survival) cutoff points, with the hazard ratio increasing at the second time point. But the EAG did not think it would do so in the longer term. It felt it was important to explore treatment waning because there was some evidence of it with the comparators. The EAG set the hazard ratios for progression-free survival and overall survival to 1 for all treatments (that is, the risk of death was similar for all treatments) at 2 years in its base case, and at 3 years in a scenario analysis. The company said the analyses were not reasonable, evidence based, or aligned with TA816, which applied treatment waning at 5 years. The committee noted the lack of data beyond 5 years. It thought that an indefinite duration of treatment effect was unlikely and that the higher hazard ratios at the second time point indicated treatment waning. It also noted that only a small number of people were still having treatment at 2 years, so it was not implausible to set a hazard ratio of 1 at 2 years. A clinical expert said that it was reasonable to assume the same treatment-waning characteristics would apply for capivasertib plus fulvestrant and its comparators. After consultation, the company incorporated treatment waning at 5 years into its model, in line with TA816. This was on the basis that alpelisib is a kinase inhibitor similar to capivasertib that also targets part of the AKT signalling pathway. It said that an earlier treatment-waning time point would be arbitrary, and did not reflect the treatment effect in the trials. It also said it was not consistent with the trends in the fractional polynomial NMA. The company added that, although only a small number of people were still having treatment at 2 years, starting waning at this point was overly conservative. It said capivasertib's treatment effect could continue for some time after. The EAG said that there was no evidence of a prolonged treatment effect after 2 years of treatment. But because the hazard ratio for capivasertib plus fulvestrant was below 1 at 2 years, it chose to start waning at 3 years. The committee found the evidence on continued treatment effect uncertain, but noted some indication that it continued beyond 2 years. Given the similar mode of action between capivasertib and alpelisib, it concluded that a 5‑year waning assumption was reasonable in the absence of evidence suggesting otherwise.

Treatment with capivasertib plus fulvestrant requires genomic testing for PIK3CA, AKT1 or PTEN alterations. The company did not include the cost of testing for these alterations in its base case. This is because testing for PIK3CA alterations (the most common PI3K and AKT pathway alteration) is done routinely in UK clinical practice since NICE recommended alpelisib plus fulvestrant in 2022 (see NICE's technology appraisal guidance on alpelisib with fulvestrant). The NHS England Cancer Drugs Fund clinical lead said that although PIK3CA alterations are tested for, AKT1 and PTEN are not, so the model should include additional costs to test for them. The committee agreed with this. After consultation, the company included testing costs in its base case, estimated at £114.71 per person. This was based on staffing costs to analyse and check the genetic alterations, using NHS pay scales. The company estimated that 2.45 people would need to be tested to identify 1 person with the relevant genetic alteration, based on the proportion of people with PIK3CA, AKT1 or PTEN alterations. This results in a total testing cost per eligible patient of £281.02. The committee concluded that these costs were reasonable.

In the company's model, people on all treatments were assumed to continue until disease progression, unacceptable toxicity or until they withdrew consent. To model time to treatment discontinuation for capivasertib plus fulvestrant, the company calculated the ratio between time to treatment discontinuation and progression-free survival in CAPItello‑291. It applied this ratio to the modelled progression-free survival curve for all treatments. The company considers the figure for the ratio confidential and so it is not reported here. This was a pragmatic approach for alpelisib plus fulvestrant and everolimus plus exemestane because there was no publicly available data on time to treatment discontinuation. The EAG noted that discontinuation rates because of disease progression and adverse events differed substantially in the relevant trials:

• capivasertib plus fulvestrant (CAPItello‑291); 58.9% disease progression, 13% adverse events

• alpelisib plus fulvestrant (SOLAR‑1); 37% disease progression, 25% adverse events

• everolimus plus exemestane (BOLERO‑2); 55% disease progression, 19% adverse events.
  
  The EAG said this suggests that the relative proportion of people stopping treatment because of reasons other than progression differs by treatment. The EAG used the company's value for time to treatment discontinuation in its base case. But it modelled the impact of a shorter time to discontinuation for the comparators (on the basis that adverse events were worse with the comparators) in a scenario to test its impact. The clinical experts said that if people have side effects, usually treatment will continue at the same dose, as long as they are tolerable. If side effects are severe, treatment might stop temporarily and be restarted at a lower dose and only a small number will stop treatment completely. In the absence of evidence and, based on what the clinical experts said about the toxicity profiles of the treatments, the committee accepted the company's base-case assumption.

The company modelled mean relative dose intensity (RDI) for capivasertib plus fulvestrant and everolimus plus exemestane. RDI is a measure of how much of the planned dose of drug someone actually has in a trial. Someone may not have the full planned dose, for example, because of toxicity, which may mean the dose needs to be reduced or delayed. For alpelisib plus fulvestrant, only the median RDI was available from SOLAR‑1 (82.7%), so the company assumed a 100% RDI in its base case, with a scenario analysis applying the median RDI. It said this was because means and medians were not the same; the data could be skewed, so means were preferred. The EAG pointed out that if the median RDI for alpelisib plus fulvestrant was 82.7%, this showed that there were delayed or reduced doses. It said that assuming 100% RDI would overestimate its cost. The committee agreed with the EAG's comments. The company provided a scenario that assumed the same RDI for alpelisib plus fulvestrant as for capivasertib plus fulvestrant, which the EAG used in its base case. The company considers that this value is confidential and so it is not reported. The clinical experts said that a 100% RDI was not plausible and reiterated that alpelisib was the least well tolerated of the treatments (see section 3.2). A clinical expert said that alpelisib was likely to have the lowest RDI of all the treatments. The committee concluded that using a 100% RDI would overestimate costs for alpelisib plus fulvestrant. It also decided that it would prefer the median RDI of 82.7% to be used in the model, in the absence of any other evidence. After consultation, the company applied the RDI for capivasertib to alpelisib in its base case. The EAG changed this to 82.7% in its base case in line with the committee's preference at the first meeting. The committee concluded that the EAG's approach was appropriate.

NICE's manual on health technology evaluations notes that, above a most plausible incremental cost-effectiveness ratio (ICER) of £20,000 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 appreciated the high unmet need of people with HR‑positive HER2‑negative advanced breast cancer. It also recognised that capivasertib is an innovative treatment that the clinical experts considered to be a step change in managing breast cancer with PIK3CA, AKT1 or PTEN alterations. Although some uncertainty remained, the committee was satisfied that the company's fractional polynomial analyses considerably reduced this. It concluded that an acceptable ICER threshold was around the middle of the range NICE considers a cost-effective use of NHS resources (£20,000 to £30,000 per QALY gained).

The most likely ICER cannot be reported here because of confidential commercial arrangements for capivasertib, alpelisib, fulvestrant, and one of the post-progression treatments. But it was below £20,000 per QALY gained when the committee's preferred assumptions on the following were incorporated:

• a log-logistic model to estimate long-term progression-free survival (see section 3.9)

• a fractional polynomial model approach to model overall survival (see section 3.10)

• treatment waning at 5 years (see section 3.11)

• health-state utility values derived from CAPItello‑291 (see section 3.12)

• incorporating the costs of testing for AKT1 and PTEN alterations (see section 3.13)

• incorporating an RDI of 82.7% for alpelisib plus fulvestrant and evidence-based scenarios for alternative RDIs (see section 3.15)

• a severity weighting of 1.2 applied to the QALYs (see section 3.16).

The committee did not identify any equality issues.

The committee recognised that capivasertib plus fulvestrant is a clinically effective treatment after a CDK 4 and 6 inhibitor plus an aromatase inhibitor. It also recognised the tolerability benefits compared with alpelisib plus fulvestrant and exemestane plus everolimus. The committee concluded that the ICER that included its preferred assumptions (see section 3.18) was below the range that NICE considers an acceptable use of NHS resources. So, capivasertib plus fulvestrant is recommended.