EMA Conditional Marketing Authorisations Based on Surrogate and Intermediate Endpoints: Precedents and Specific Obligations

EMA conditional marketing authorisation on surrogate and intermediate endpoints: precedents and specific obligations

EMA has granted conditional marketing authorisations (CMAs) on surrogate or intermediate endpoints in two main settings. The first is oncology and haematology, where single-arm or uncontrolled trials supported approval on response rates. The second is hepatology, haemostasis and HIV, where approval rested on biochemical, histological, biomarker or virological endpoints. In each case the CHMP required confirmation through specific obligations set out in Annex II.E. The track record runs both ways. Several products converted to standard authorisations once confirmatory studies reported (Intelence, Blincyto, Pixuvri, Adcetris). Others lost their authorisation or an indication when confirmatory studies failed (Blenrep monotherapy, Ocaliva, Rubraca treatment indication). Blenrep was later authorised again for combination regimens under a new procedure.

This article sets out the legal basis, how the CHMP has described these endpoints, the specific obligations attached to each precedent, and what happened next.

Legal basis and how specific obligations work

Under Article 14-a of Regulation (EC) No 726/2004, a CMA may be granted before comprehensive clinical data are available. The medicine must treat, prevent or diagnose a seriously debilitating or life-threatening disease and meet an unmet medical need: either no satisfactory authorised method exists in the Union, or the medicine offers a major therapeutic advantage. The benefit of immediate availability must outweigh the risk that additional data are still needed ⁶⁹. Under Regulation (EC) No 507/2006, the scope also covers orphan medicines and medicines for use in recognised public-health emergencies. The benefit-risk balance must also be positive, and the applicant must be likely to provide comprehensive data ⁷⁰.

The mechanics that matter for sponsors:

• Specific obligations are binding conditions. They are set out in the conditions of the authorisation, with deadlines where appropriate. They include completing ongoing studies or running new ones to confirm the favourable benefit-risk balance ⁷¹. EMA's post-authorisation procedural advice places them in Annex II.E of the Commission decision ¹⁶⁶.
• Annual review and renewal. The Agency reviews the specific obligations every year ⁷⁴⁴. Renewal depends on a reassessment of the benefit-risk balance, including compliance with the specific obligations, through the annual renewal procedure ¹⁶⁷. The CMA is valid for one year and is renewable ⁷².
• Labelling disclosure. The SmPC and package leaflet must state clearly that the authorisation is subject to specific obligations ⁷⁴⁵.
• Conversion. Once the obligations are fulfilled, the Commission may grant a standard authorisation valid for five years, on the holder's application and after a favourable Agency opinion ⁷². The route only runs one way: an existing standard authorisation cannot later be converted into a CMA ¹⁶⁸.

How the CHMP characterises surrogate and intermediate endpoints

EPAR language varies a great deal by therapeutic area. That matters when a sponsor frames its own endpoint justification.

Explicit "surrogate" or "intermediate" framing (non-oncology).

• Ocaliva (obeticholic acid), primary biliary cholangitis (PBC): the CHMP described the biochemical changes as a surrogate for clinical efficacy rather than shown improvement in clinical outcomes. It said the link to liver outcomes remained uncertain ⁷⁶⁵⁷⁶⁶.
• Iqirvo (elafibranor), PBC: the biochemical biomarkers were called surrogate parameters reasonably likely to translate into clinically relevant outcomes, with long-term benefit still to be confirmed ⁴⁸⁸⁴⁸⁹. The CHMP also noted that the effect was mainly on alkaline phosphatase (ALP), with minimal effect on bilirubin ⁴⁴⁵.
• Lyvdelzi (seladelpar), PBC: the biochemical response was called a surrogate reasonably likely to translate into clinical benefit. Its relevance to liver-related outcomes was unknown ^55455525. The CHMP noted that the evidence linking ALP and bilirubin reductions to outcomes came from PBC natural history and UDCA. It had not been shown generally for new treatments ⁴⁴⁶.
• Rezdiffra (resmetirom), MASH: the liver-histology endpoints were called intermediate endpoints, not fully validated surrogates in the strict sense, though reasonably likely to predict liver outcomes ⁶⁶²⁶⁶³.
• Kayshild (semaglutide), MASH: the histology endpoints were called intermediate and reasonably likely to predict long-term clinical outcomes ²⁶⁰.
• Ondexxya (andexanet alfa): the CHMP did not regard reversal of anti-factor Xa activity as a validated indicator of clinical efficacy, because its correlation with haemostatic efficacy had not been established ⁷²⁸.

Implicit framing (oncology response endpoints). In single-arm oncology CMAs, the CHMP has usually not labelled objective response rate (ORR) or duration of response (DoR) as "surrogate". Instead it has treated clinical benefit as an assumption that needs comparative or time-to-event confirmation.

• Blenrep: the original CMA rested on ORR and DoR in DREAMM-2, judged long enough to potentially translate into clinical benefit ⁵³³⁵³⁴. The trial had no reference arm, so effects on PFS and OS could not be isolated ⁵³⁵⁵³⁶.
• Retsevmo: the single-arm approach was accepted because the ORR effect was outstanding and durable in a rare population. The CHMP noted that time-to-event outcomes were hard to interpret ⁶⁸³¹³⁴.
• Aucatzyl: overall remission rate was accepted as the primary endpoint when read with durability of response ²⁸¹²⁸².
• Enhertu, HER2-mutant NSCLC: the CHMP was explicit here. It stated that response rate and DoR "are not surrogate endpoints for OS and PFS" in this setting, and therefore that the uncontrolled data were not comprehensive enough for full approval ³⁵³.

Minimal residual disease (MRD). An expert consultation cited in the Blincyto assessment found a strong patient-level association between MRD negativity and event-free and overall survival. It had not been validated as a treatment-effect surrogate, but it was considered useful for an accelerated benefit-risk decision where unmet need was high, subject to confirmation on clinical outcomes ⁴⁴⁴. The later Blincyto indication in MRD-positive ALL was granted as a standard Type II extension, not a CMA. The original CMA was for relapsed or refractory ALL ⁸¹⁴⁸¹⁷.

Precedents: endpoint basis and specific obligations

Hepatology (biochemical and histological endpoints)

The MASH CMAs share one design. Accelerated approval rests on a prespecified interim histology readout, and the specific obligation is the final, outcomes-driven analysis of the same trial. Whether the histological effects translate into fewer events such as decompensation, transplantation or death had not been shown at authorisation ^664665261.

Haemostasis and HIV

The Ondexxya obligations stand out because they target the surrogate itself. They require data to examine the biomarker-haemostasis relationship and to clarify thrombotic risk against a concurrent control ^733735729.

Oncology and haematology (response-based endpoints)

Other response-based CMAs follow the same pattern. Anktiva relied on a six-month complete response rate, with obligations to complete QUILT-3.032 (including five-year follow-up) and run the randomised QUILT-2.005 study ^137138139. Caprelsa's obligation was an observational study (D4200C00104) to evaluate benefit-risk by RET mutation status ¹³⁰¹³².

Typology of specific obligations

Across these precedents, Annex II.E obligations fall into a small number of types:

1. Randomised comparative confirmation against standard of care or an active comparator: TOWER ³⁷⁸, ARIEL4 ³²⁵, DREAMM-3 ⁵³⁹, PIX306 ⁶⁴⁹, LIBRETTO-431/531 ⁶⁹¹⁶⁹², DESTINY-Lung04 ³⁵⁷, ANNEXA-I ⁷³⁵.
2. Clinical-outcomes trials to validate a biomarker or histology effect: 747-302 ⁷⁶⁹, ELFIDENCE ⁴⁹¹, AFFIRM ⁵⁵⁶, MAESTRO-NASH OUTCOMES ⁶⁶⁷.
3. Final or long-term analysis of the pivotal study itself: MAESTRO-NASH ⁶⁶⁶, ESSENCE ²⁶², FELIX ²⁸⁶, LIBRETTO-001 ⁶⁹⁰, DP-02 ³⁵⁹, and Adcetris OS follow-up ⁵⁰⁴.
4. Studies aimed at the surrogate-outcome relationship and dose: Ondexxya's biomarker-correlation, comparative PK and PK/PD model obligations ^733734736.
5. Population and real-world gaps: special populations (Ocaliva 747-401, hepatic impairment ⁷⁶⁹), off-label regimen coverage (Intelence ⁵¹⁷), registries (Aucatzyl AUTO1-LT2 ²⁸⁸), and further single-arm studies in populations not covered at approval (Adcetris ⁵⁰⁶⁵⁰⁷).

Due dates often move. Pixuvri's PIX306 deadline moved from 2015 to 2018 ⁶⁵⁰⁶⁴⁹. Blenrep's DREAMM-2 obligation moved from April 2021 to February 2023 ⁵³⁷⁵³⁸. Ondexxya's ANNEXA-I moved from June 2023 to July 2024 ⁷³⁵⁷³⁷. Retsevmo's LIBRETTO-431 moved from October 2023 to December 2024 ⁶⁹¹⁶⁹³.

What happened next

Converted to standard authorisation after obligations were fulfilled

• Intelence: renewed as conditional each year from 2009 to 2013 ⁴⁵⁶⁴⁶⁰. The final retrospective cohort (1,115 people) found no significant difference between etravirine with darunavir/ritonavir and with other PIs. The CHMP recommended an authorisation no longer subject to specific obligations on 19 September 2013, about five years after the CMA ⁹⁶.
• Blincyto: TOWER showed longer overall survival than chemotherapy (7.7 vs 4.0 months) and higher MRD negativity (23.6% vs 9.0%), fulfilling the obligation ¹⁹⁴. The authorisation switched to standard on 18 June 2018 ^195196189.
• Pixuvri: PIX306 was submitted in November 2018 and was judged to corroborate benefit, including after prior rituximab ^6162305. On 28 March 2019 the CHMP recommended removing the specific obligations ³⁰⁶. The product's listed status is now "Expired", meaning the authorisation later lapsed after conversion and is no longer in force ²⁹⁰.
• Adcetris: renewed as conditional for nearly a decade ²²⁸²³⁵. The final obligation (C25006) showed ORR above 60% with durable responses ²³⁶²³⁷. On 24 March 2022 the CHMP concluded that all specific obligations were fulfilled ²³⁸²³⁹.

Converted, but only after the indication was restricted

• Rubraca: the ARIEL4 overall-survival analysis raised concern about a potential detrimental effect compared with chemotherapy ⁴⁷⁴. The CHMP recommended a change on 21 July 2022 and the treatment indication was removed ^474475476. Rubraca switched to a standard authorisation on 9 November 2022 and remained authorised for maintenance treatment ⁴⁶⁸⁴⁷⁷. Submitting the confirmatory data fulfilled the procedural obligation. It did not confirm the original surrogate-based benefit.

Not renewed or revoked after confirmation failed

• Blenrep (monotherapy): the CHMP considered the DREAMM-2 obligation fulfilled. DREAMM-3 then failed to show PFS or OS benefit, so the DREAMM-3 obligation could not be considered fulfilled ^427428429. The Commission decided on 23 February 2024 not to renew the CMA ⁷⁷⁸. Blenrep was later authorised again on 23 July 2025, under a new procedure (EMEA/H/C/006511), for combination regimens ⁷⁷⁸⁷⁷⁹.
• Ocaliva: in 2020 the independent monitoring committee judged 747-302 unlikely to show efficacy. Both obligation studies were terminated early ⁷⁷¹. The primary clinical composite showed no difference (HR 1.01, 95% CI 0.68 to 1.51) ⁷⁷². The CHMP found conversion to a full authorisation not approvable ⁷⁷³, and the product is listed as revoked ³⁹¹. This is the clearest EU example of a biochemical surrogate that did not hold up against clinical outcomes ⁴⁴⁸⁴⁴⁹.
• Lartruvo: strictly, this was not a surrogate case. Its CMA rested on PFS and an OS difference (26.5 vs 14.7 months) in a small randomised phase 1b/2 study ⁸⁰⁵⁸⁰⁶. The confirmatory ANNOUNCE study failed its survival objective. An Article 20 review began on 25 January 2019 and the CHMP recommended revocation ⁸¹¹⁸¹². It shows that a randomised but exploratory basis carries a similar risk.
• Translarna: a related case. Its 2014 CMA relied on a post hoc six-minute walk distance signal. Phase 3 studies 020 and 041 did not replicate it, and the CHMP recommended non-renewal ^405406407409.

Still conditional at the most recent published renewal

Ondexxya has been renewed as conditional every year from 2020 to 2025. Its 2025 renewal still lists the specific obligations ⁶²⁶⁶³⁰. Retsevmo was renewed as conditional through a January 2025 Commission decision ⁷⁰⁹⁷¹⁰. An April 2025 assessment notes outstanding Enhertu obligations for the gastric and NSCLC indications ⁵⁶⁴⁵⁶⁵. Rezdiffra had a one-year renewal opinion on 23 April 2026 ⁵⁹⁸⁵⁹⁹. Iqirvo, Lyvdelzi, Kayshild and Aucatzyl have long-dated obligations (2029 to 2030) with no confirmatory readout reported in their published EPARs as of October 2026 ^{491556262286288}.

Practical implications for sponsors

• Justify the endpoint in the CHMP's own terms. In hepatology and haemostasis, the CHMP has explicitly said whether an endpoint is a surrogate, an intermediate endpoint or not validated ^554662728. In oncology, the debate is about whether an uncontrolled response rate can stand in for time-to-event benefit ³⁵³⁵³⁵. "Reasonably likely to predict clinical benefit" can support a CMA, but it is not validation ⁴⁴⁴.
• Have the confirmatory study under way at submission. Most successful conversions relied on a randomised study that was already running at the time of CMA (TOWER, PIX306) ³⁷⁸⁶⁴⁹. In MASH, the obligation is the outcomes phase of the same pivotal trial ⁶⁶⁶²⁶².
• Expect obligations to be realistic about feasibility. For Adcetris, the CHMP accepted further single-arm studies because a randomised HL trial would have been difficult and a randomised second-line sALCL trial was not feasible ⁵⁰²⁵⁰³.
• Plan for an unfavourable result. A failed confirmatory study has led to non-renewal (Blenrep) ⁴²⁹, revocation (Ocaliva, Lartruvo) ³⁹¹⁸¹², or loss of the surrogate-based indication before conversion (Rubraca) ⁴⁷⁴⁴⁶⁸. Under Article 14-a, annual renewal is the point at which these consequences take effect ⁷²¹⁶⁷.
• Budget for a long conditional period. In the precedents above, the time from CMA to removal of specific obligations ranged from about two and a half years (Blincyto, 2015 to 2018) to nearly a decade (Adcetris, 2012 to 2022) ^366189492238.