When multiple comparators are needed for reimbursement

The pivotal trial showed that inavolisib improved survival against one existing treatment: palbociclib plus fulvestrant.

But standard care in Scotland involved three relevant treatments. The Scottish Medicines Consortium did not recommend inavolisib for NHSScotland.[1]

This is a compelling case study of why one trial comparator may not be enough for reimbursement.

The pivotal trial answered one comparison

Inavolisib, sold as Itovebi, is a targeted treatment for a specific type of advanced or metastatic breast cancer. The cancer must be oestrogen receptor positive, HER2 negative and have a PIK3CA mutation.[2]

In the INAVO120 trial, every patient received palbociclib and fulvestrant. Half also received inavolisib. The other half received a placebo.[3,4]

The updated results were strong.

Median progression-free survival was 17.2 months with inavolisib and 7.3 months without it. Median overall survival was 34.0 and 27.0 months.[1,4]

The trial therefore showed that adding inavolisib was better than palbociclib plus fulvestrant alone.

EMA concluded that the benefits outweighed the risks. SMC also accepted palbociclib plus fulvestrant as a relevant comparator.[1,2]

But it was not the only relevant comparator.

Scottish reimbursement required three comparisons

Patients in Scotland also received ribociclib plus fulvestrant and abemaciclib plus fulvestrant.[1]

Public Health Scotland identified 164 patients who received one of the three combinations during 2025. Ribociclib was used in 79 patients, abemaciclib in 45 and palbociclib in 40.[5]

That means the treatment used in the pivotal control group accounted for 24.4% of use across the three relevant combinations.

The control was not wrong. It answered a clinically relevant question and supported regulatory approval.

It just did not answer the whole Scottish reimbursement question.

SMC also needed to know how the inavolisib combination compared with the treatments used in the other three quarters of patients.

A network meta-analysis built the missing comparisons

There were no head-to-head trials comparing the inavolisib combination with ribociclib or abemaciclib plus fulvestrant.

The company therefore submitted a network meta-analysis.[1]

A network meta-analysis links randomised trials through treatments they have in common. In this case, other trials compared palbociclib, ribociclib or abemaciclib plus fulvestrant with fulvestrant-based controls. Those links were used to estimate the comparisons that had never been tested directly.

We explain these methods more fully in our field guide to indirect treatment comparisons.

The Scottish network contained seven studies. It produced estimates for progression-free and overall survival against the two additional comparators.[1]

The treatments connected.

The harder question was whether the studies were comparable enough for those connections to be trusted.

A connected network did not mean comparable patients

The studies differed in several ways that could affect patient outcomes or the relative treatment effect.[1,6]

INAVO120 included patients with PIK3CA-mutated disease. Other trials included mixed populations or did not report mutation status. PIK3CA mutation is associated with poorer progression-free and overall survival, although that does not automatically mean it changes the relative effect of every treatment.[7]

The studies also differed in endocrine sensitivity, previous treatment, menopausal status and follow-up.[1,6]

FLIPPER, for example, enrolled patients with endocrine-sensitive disease. They either had newly metastatic disease or had relapsed more than 12 months after completing endocrine treatment.[8]

MONARCH plus cohort B enrolled patients whose disease had progressed on endocrine treatment. Its own authors noted that its patients had poorer prognostic characteristics than those in MONARCH 2, including worse performance status and more primary endocrine resistance.[9]

The EU methodological guidance calls the required assumption exchangeability. In plain English, patients from one study should be sufficiently comparable with those in another that they could, in principle, have been included in either trial.[10,11]

A line connecting two treatments on a network diagram does not establish that assumption.

The broader network answered a less specific question

The company’s main analysis combined endocrine-sensitive and endocrine-resistant patients. This produced a larger and more connected network.[1]

In that mixed population, the progression-free-survival estimates favoured the inavolisib combination over both additional comparators:

• versus abemaciclib plus fulvestrant: hazard ratio 0.42, with a 95% credible interval from 0.21 to 0.87
• versus ribociclib plus fulvestrant: hazard ratio 0.38, with a 95% credible interval from 0.18 to 0.81

The overall-survival estimates were much less certain. Their credible intervals included the possibility of no difference.[1]

The company also analysed endocrine-resistant patients separately. This population was more closely aligned with the licensed indication.

In that analysis, the progression-free-survival comparison with abemaciclib had a hazard ratio of 0.46, but its credible interval ranged from 0.18 to 1.26. The overall-survival interval was also wide. A comparison with ribociclib was unavailable.[1]

This does not show that the treatments work equally well. It means the analysis could not establish a difference with sufficient certainty.

The broader network answered more comparisons. The narrower network answered a more relevant question.

Neither advantage made the other problem disappear.

The analytical model changed how certain the results appeared

Canada’s Drug Agency reviewed a closely related six-trial network for the same medicine.[6]

The Canadian review is useful because it shows how much the findings depended on an analytical choice.

The company’s main analysis used a fixed-effect model. This assumes that every study is estimating the same underlying treatment effect.

A random-effects model allows the underlying effect to differ between studies. This can be more realistic when populations and study methods vary, but it is difficult to estimate reliably when a network contains very few studies.

The fixed-effect model produced narrower credible intervals. When random-effects models were used as sensitivity analyses, several comparisons that had appeared to favour inavolisib could no longer rule out little or no difference.[6]

This does not mean that a random-effects model is always correct or that a fixed-effect model is always wrong.

It means the apparent certainty depended partly on a defensible analytical choice. Decision-makers needed to see that sensitivity.

One hazard ratio could not summarise every survival pattern

A hazard ratio is commonly used to compare how quickly an event occurs between treatment groups over follow-up.

It is easiest to interpret when the relative event rate remains broadly stable over time. This is called the proportional-hazards assumption.

That assumption was not met for some outcomes in the network. Follow-up also ranged from 11.1 months in MONARCH plus cohort B to 80 months in MONARCH 2.[1,6]

When relative effects change over time, the average hazard ratio can depend on how long patients were followed. Combining hazard ratios from studies with very different follow-up can then become difficult to defend.[10,11]

The method cannot be chosen separately from the outcome data and the question being asked.

The clinical uncertainty entered the economic model

The indirect comparison was not produced only for the clinical section of the submission.

Its hazard ratios were applied to the inavolisib survival curves to estimate progression-free and overall survival for ribociclib and abemaciclib in the economic model.[1]

The progression-free-survival hazard ratios were also applied to second progression and time off treatment. The network meta-analysis had not estimated either outcome.[1]

Each step added another assumption.

SMC also considered a scenario in which palbociclib, ribociclib and abemaciclib plus fulvestrant were assumed to have equal efficacy. That scenario increased the estimated cost per health benefit gained from inavolisib.[1]

The Canadian review reached the same broader conclusion. Missing comparative evidence against nontrial comparators limited its ability to validate the economic model. It also found that 52% of the modelled survival benefit occurred after the observed trial period.[6]

An indirect estimate does not stop being uncertain when it enters an economic model.

The indirect comparison was not the only reason SMC said no

SMC summarised the comparative problem clearly:

“Indirect evidence against the two other relevant comparators has limitations which raise uncertainty.”[1]

But the network meta-analysis was not the only issue.

SMC also questioned survival-curve selection, quality-of-life values used after second progression and assumptions about the doses patients would receive. It concluded that the economic analysis was not sufficiently robust and that the treatment’s cost was not justified by the expected health benefits.[1]

The decision does not mean that inavolisib failed to show efficacy. It demonstrated progression-free and overall-survival benefits against a relevant comparator.

The problem was that the complete clinical and economic case against all relevant Scottish comparators remained too uncertain.

Different markets asked for different comparators

The Canadian review adds another important lesson.

Scotland considered abemaciclib, ribociclib and palbociclib plus fulvestrant relevant. Capivasertib plus fulvestrant was included in the Scottish network but was not considered a relevant comparator.[1]

Canada asked a different question.

Abemaciclib plus fulvestrant was not publicly funded and was not considered relevant to the Canadian public-payer decision. Capivasertib had received a positive reimbursement recommendation and completed price negotiations, so the Canadian reviewers considered it relevant. The sponsor had not included it in the economic analysis, and comparative evidence was unavailable.[6]

The Canadian clinical experts also reported that ribociclib was the most frequently used CDK4/6 inhibitor in Canada in 2025.[6]

The same medicine and global evidence programme therefore faced different comparator requirements in Scotland and Canada.

Mapping the lead market is not enough.

Every market needs its own comparator map

Start by identifying what patients receive in each intended market. Include treatments that are funded, recommended and used in practice.

Then define the exact population and outcomes for every comparison. A network that answers a broad first-line question may not answer the narrower reimbursement question in the licensed population.

Search the published literature and relevant real-world sources systematically. Establish which trials and datasets exist, how the treatments connect and whether the variables needed to assess comparability have been reported.

Run a feasibility assessment before relying on an indirect comparison. Test population alignment, common controls, treatment history, outcome definitions, follow-up and the assumptions required by the proposed method.

Finally, plan how the remaining uncertainty will enter the economic model. Test alternative defensible assumptions rather than treating one comparative estimate as fixed.

Finding the studies, choosing the method and interpreting the result are not separate jobs. They are one evidence plan.

Plan, map and generate evidence for every market

This is what Evidax is built to do.

1. PLAN: Evidax Six-Step IEP Method. A six-step method that identifies the evidence gaps gating market access and builds an Integrated Evidence Plan to close them.

2. MAP: Evidax TRACE. An AI-augmented workflow that synthesises the entire relevant published literature and available real-world data, identifying every comparison that external sources can credibly support. Experts remain in control and every decision is auditable.

3. GENERATE: Evidax Comparison Blueprints. HTA-grade workflows for indirect comparisons and external control arms, built by leading methodologists. Each Blueprint defines the decision question, data requirements, defensible method, assumptions, diagnostics and limits of interpretation before the analysis is run.

Earlier planning creates more options. But a completed pivotal trial is not the end of the opportunity. Existing literature and real-world evidence can still provide credible comparisons when the evidence bridge is identified and tested properly.

Take home

• A successful pivotal trial may answer only one of the comparisons needed for reimbursement.
• Palbociclib plus fulvestrant was relevant in Scotland, but it represented only 24.4% of use across the three relevant combinations in the available 2025 data.
• A connected evidence network is credible only when its patients, treatment settings, outcomes and follow-up are sufficiently comparable.
• The population and analytical model can materially change how certain an indirect comparison appears.
• Comparative uncertainty must be carried into the economic model, including outcomes and time periods not observed directly.
• Comparator requirements differ between markets. Map the treatments used and funded in every intended market.
• Plan the evidence gaps, map the available evidence and generate each comparison through one integrated evidence plan.

References

1. Scottish Medicines Consortium. Inavolisib 9 mg film-coated tablets (Itovebi®), SMC2902. Published 12 October 2026. https://scottishmedicines.org.uk/medicines-advice/inavolisib-itovebi-full-smc2902/

2. European Medicines Agency. Itovebi: EPAR public assessment report. EMA/CHMP/122122/2025. 22 May 2025. https://www.ema.europa.eu/en/documents/assessment-report/itovebi-epar-public-assessment-report_en.pdf

3. Turner NC, Oliveira M, Howell SJ, et al. Inavolisib-based therapy in PIK3CA-mutated advanced breast cancer. New England Journal of Medicine. 2024;391:1584-1596. https://doi.org/10.1056/NEJMoa2404625

4. Jhaveri KL, et al. Overall survival with inavolisib in PIK3CA-mutated advanced breast cancer. New England Journal of Medicine. 2025;393(2):151-161. https://doi.org/10.1056/NEJMoa2501796

5. Public Health Scotland. Treatment of adults with ER-positive, HER2-negative metastatic breast cancer following recurrence on or after adjuvant endocrine treatment: SMC2902. Published 30 June 2026. https://publichealthscotland.scot/publications/cancer-medicines-outcomes-programme-public-health-scotland-cmop-phs-report-for-the-scottish-medicines-consortium-smc/treatment-of-adults-with-er-positive-her2-negative-metastatic-breast-cancer-following-recurrence-on-or-after-adjuvant-endocrine-treatment-smc2902/

6. Canada’s Drug Agency. Inavolisib (Itovebi): Reimbursement Review. 2026. https://canjhealthtechnol.ca/index.php/cjht/article/download/PC0382r/3264?inline=1

7. Fillbrunn M, et al. PIK3CA mutation status, progression and survival in advanced HR-positive/HER2-negative breast cancer: a meta-analysis of published clinical trials. BMC Cancer. 2022;22:1002. https://doi.org/10.1186/s12885-022-10078-5

8. Albanell J, et al. Randomized phase II study of fulvestrant plus palbociclib or placebo in endocrine-sensitive, hormone receptor-positive/HER2-negative advanced breast cancer: GEICAM/2014-12 (FLIPPER). European Journal of Cancer. 2022;161:26-37. https://doi.org/10.1016/j.ejca.2021.11.010

9. Zhang Q, et al. MONARCH plus: abemaciclib plus endocrine therapy in women with HR-positive/HER2-negative advanced breast cancer: the multinational randomized phase III study. Therapeutic Advances in Medical Oncology. 2020;12:1758835920963925. https://doi.org/10.1177/1758835920963925

10. Member State Coordination Group on Health Technology Assessment. Methodological Guideline for Quantitative Evidence Synthesis: Direct and Indirect Comparisons. Adopted 8 March 2024. https://health.ec.europa.eu/publications/methodological-guideline-quantitative-evidence-synthesis-direct-and-indirect-comparisons_en

11. Member State Coordination Group on Health Technology Assessment. Practical Guideline for Quantitative Evidence Synthesis: Direct and Indirect Comparisons. Adopted 8 March 2024. https://health.ec.europa.eu/document/download/1f6b8a70-5ce0-404e-9066-120dc9a8df75_en?filename=hta_practical-guideline_direct-and-indirect-comparisons_en.pdf