EMA finalizes clinical anticancer therapeutic guidance update

Regulatory Background

• Cancer is the second leading cause of death in the E.U. after ischemic heart disease and stroke. In 2019 more than one in four (28%) of deaths in the E.U. were from cancer. Even under the shadow of the Covid-19 pandemic, 1.2 million deaths, or 23%, were due to cancer in 2020.
• “ Europe’s Beating Cancer Plan” was released in 2021 as a political commitment to address rising cancer rates in the E.U., putting forth proposals for better options to detect cancer earlier, better understanding of cancer and to ensure equal access to cancer medication across the E.U. Additionally, the plan aims to improve the quality of life for the over 12 million cancer survivors who benefited from earlier detection, effective therapies and supportive care. The European Commission also implemented its Mission on Cancer, which aims to further cancer research and innovation through funding in the same areas as the cancer plan.
• The EMA started providing guidance on the clinical development of anticancer therapeutics in 1996. The document has been updated over time to implement advances in understanding cancer and drug development. For example, the first three editions focused on cytotoxic compounds. In 2005, the guidance was updated to include non-cytotoxic drugs were gaining in importance. The fifth update accounted for the impact that changes in the therapeutic landscape had on safety data reporting requirements, which also inform the benefit-risk evaluation.
• In general, the EMA guideline is meant to provide guidance for oncology developers across the drug development span, according to the agency’s landing page for the guideline. Areas covered include “identifying target population with optimised benefit risk in exploratory trials, design of confirmatory trials, choice of endpoints, the impact of adverse drug reactions on the benefit-risk.” In its latest iteration, the soup-to-nuts guidance runs to 43 pages, including a section on definitions and abbreviations but excluding appendices.
• The current update expands the discussion on biomarker-guided therapeutic development and recent oncology study designs, in addition to a variety of less consequential updates throughout. EMA floated these updates through a Concept Paper in 2019 and subsequently issued a draft updated guideline for public consultation in late 2020. The finalized guideline released this week expands concepts even further than suggested in the draft. This analysis will look at the new considerations offered in version 6 compared to version 5 of this guideline. The guideline is supported by four appendices: 1) methodological considerations for use of progression-free or disease-free survival, 2) use of patient-reported outcomes, 3) summary of product characteristics and 4) condition specific guidance. A specific pediatric oncology guidance is also available.

The rewritten and significantly expanded biomarker section reflects a dramatic increase in use of biomarkers since the last version of the guidance

• “Biomarker is commonly defined as a characteristic that is measured as an indicator of normal biological processes, pathogenic processes, or biological responses to an exposure or intervention, including therapeutic interventions,” according to the EMA. Biomarker assays used in oncology measure “alterations specific to tumours.” Radiographic biomarkers can be used in combination with histological and biological characteristics.
• A section on general considerations provides a list of items developers should include in regulatory filings when referencing their chosen biomarkers. These include the pharmacological rationale for the biomarker and its function; purpose and context of use, and relationship between the biomarker and outcomes; sample collection and analysis considerations; analytic methods of measurement; and data to support clinical validity. The population for biomarker use should be clearly defined and pre-specified. Biomarkers for use with special populations may require clinical validation in that target population.
• When biomarkers are used for patient selection, they may be predictive, prognostic, or both, note the guideline authors, offering the example of HER2 expression in patients with breast cancer. Predictive biomarkers enable the identification and selection of patients who are more likely to benefit from a specific treatment and would achieve a better clinical outcome in response to a specific treatment. Prognostic biomarkers enable patient selection that would have a better clinical outcome irrespective of treatment. This enables patient selection, stratification and other adjustment in clinical trial design planning.
• Patient selection criteria should be widened as far as possible within the target population, though sufficient homogeneity should be maintained to allow for treatment effect estimation, according to the guideline. The patient population enrichment or selection strategy for pivotal clinical studies should be adequately justified. The EMA also expects a documented rationale for defining the target population if more than one predictive biomarker is used. Additionally, a decision not to investigate the treatment effect in the non-selected population (e.g., biomarker negative) must also be supported by a sound rationale and clinical data.
• When biomarkers are used as clinical trial endpoints, developers should establish a strategy for biomarker development and validation as early as possible. Regulatory requirements will depend on the category and intended use (e.g., measure of drug activity versus patient selection).
• EMA notes that for any novel endpoints used as surrogate endpoints supporting the benefit-risk evaluation, clinical validity must be comprehensively established with respect to “the relationship with a treatment effect.” Although this requirement is not new, EMA highlights the pathway for qualification of novel methodologies that could result in a letter of support to aid further development or full qualification, which then supports the acceptability of the method for its intended purpose. [See AgencyIQs analysis of a workshop on novel methodology qualification.]
• Biomarker assays must be adequate for the intended context of use which should be validated by supportive data. The appropriateness of a particular assay is limited by the respective platforms, technology and analytes measured. The clinical validity of the assay should be assured. EMA recommends that the analytical validity of biomarkers, including those used in early clinical studies, should be “sufficiently assured.” Developers should, where possible, switching assays at different development stages Changing assays at different development stages or through the course of identifying the patient population should be minimized, but developers should demonstrate assay concordance if a change is needed or more than one assay was used.
• Generally, centralized testing is recommended for pivotal clinical trials. Although local testing may be used for secondary analysis or where an assay is well-established, the sponsor will still have to show concordance between the local and centralized tests. The protocol and statistical analysis plan should clearly specify the plans for specimen collection, sample analysis and related clinical outcomes depending on the biomarker assay development and validation strategy.
• Briefly addressing companion diagnostics, the guideline advises that sponsors consider co-development of a companion diagnostic and the medicinal product.
• Sample collection, processing, transport, storage and disposition should generally follow the principles discussed in the ICH E18 guideline. “The source and quality of the tissue samples should be appropriately justified” with respect to the intended purpose of the biomarker analysis. Biological variability and other sources of variability leading to heterogeneity of biomarker measurements should be considered. Although the guideline notes that collecting tumor tissue in progressive disease is important, the authors also address collection of circulating tumor DNA (ctDNA) and circulating tumor cells (CTC) as approaches “that allow easy and repeated sampling.” Developers should establish the connection between these sample types and tumor tissue DNA, so that ctDNA or CTC can be used as surrogates to detect molecular alterations in tumors.
• For use of biomarkers in confirmatory studies, the revised guideline first refers sponsors to other guidelines that discuss general methodological considerations. Citing “frequently non-established predictive roles of biomarkers used in confirmatory/pivotal studies,” the guideline authors advise “careful pre-specification and statistical error control” that reflects how the biomarker is being used in the investigation. EMA recommends that sponsors consider seeking specific biomarker scientific advice based on the role of the biomarker and associated uncertainties. Where the biomarker’s predictive or prognostic role is known before a confirmatory Phase 3 study, sponsors should pre-specify subgroup analyses with adjustment for multiplicity, as necessary.
• When a continuous biomarker used for patient selection is then assigned a cut-off value so patients can be judged biomarker-positive or -negative, that cut-off value must be justified. The cut-off should be based on the functional relationship between the marker and the clinical outcome, which should be described for the selected target population, notes the updated section. If the cut-point needs to be changed based on data from Phase III trials, “availability of independent clinical data to validate the change is usually necessary,” writes the EMA.

The updates include expanded study design considerations for pivotal trials

• The general structure of this section remains the same in the update, with considerable expansion in some areas. Some of the wording addressing use of biomarkers has been removed from this section, since the earlier biomarker section was rewritten and greatly expanded. Additionally, new wording clarifies when an investigator has discretion in choosing the reference regimen, that choice “should be declared and reported prior to randomization,” with the protocol as detailed as possible about any patient-level choices.
• An expanded section on crossover designs explains that crossover trials allow participants in the control arm to switch to the experimental treatment after their disease is progressing on control treatment. However, EMA warns that a crossover design prevents direct observation of the effect the intervention has on overall survival (OS), a key endpoint, compared to control. The guideline observes that statistical efforts to correct for the treatment switch “may have important limitations, especially when the crossover happens after patients have already progressed. One situation where trials “should normally incorporate crossover” is where the treatment is already approved for a later line of treatment and is being tested for an earlier line. Here, the question of interest is whether earlier use of the product improves OS compared to use in a later line, and a crossover design should be employed to avoid harm to participants in the control arm, and to avoid exaggeration of the OS benefits observed in early versus late therapy.
• The updated section on randomization and blinding addresses the question of functional unblinding, as can happen where two treatment regimens show markedly different toxicity or when the physician has a choice among several treatments in the control arm. Blinding may also be infeasible in some instances where use of a double-dummy approach might “cause an unreasonable burden.” The guideline also gives considerations for circumstances when a trial must be conducted open-label. Still, the guideline acknowledges that blinding will remain still be the best approach in most clinical trials. EMA notes that “blinding can reduce early drop-outs in the control arm and investigator’s bias in progression endpoints.”
• Typical primary efficacy endpoints for pivotal trials include overall survival, progression free survival (PFS) and disease-free survival (DFS), notes the updated guideline, which has dropped “cure rate” from this list. Of these, OS “is often considered the most important efficacy endpoint to establish a positive balance of benefits and harms,” notes the EMA. PFS is generally a “less convincing” endpoint, writes the agency, although in some indolent cancers, the delay in progression could be reasonably linked to improved quality of life or increased OS. When PFS or DFS are the primary endpoint(s), OS should be the secondary one, and vice versa. In any case, however, the EMA makes clear that patients should receive standard-of-care treatment that could improve OS.
• Patient-reported outcomes (PROs) such as symptom control could be clinically relevant and even serve as primary endpoints if the data is of high quality, the guideline notes in newly added language. Additionally, other endpoints could be appropriate if they point to treatments that are likely beneficial. Additional new language encourages sponsors to include “relevant PRO outcomes,” referring sponsors to Appendix 2, devoted to the topic.
• The guideline now also briefly touches on circumstances where scientific advice should be sought, if sponsors plan to deviate from the guideline. The EMA also refers sponsors to Appendix 4 for condition-specific guidance.
• Additionally, the updated guideline includes advice that the trial design include pre-specified rules for early stopping for efficacy. These planned rules should take into consideration the power required to demonstrate effect in “important subgroups” and that required to show an effect on longer-term secondary endpoints.

In a new section 8 entitled “Study designs for special situations,” EMA discusses studies in very small populations for rare cancers, and addresses master protocols, which are gaining traction in oncology trials

• Clinical trials for rare cancers: Studies in small populations for very rare cancers or narrow indications may present situations where “recruitment of a large enough study population for a robust” benefit-risk decision based on conventional standards is not feasible, notes the EMA in the new section. Still, a randomized study “is methodologically more convincing” and generally recommended where feasible; the EMA recommends sponsors consider unequal randomization favoring the interventional arm in a blinded study, for example. Randomized trials also allow for blinding, and provide the advantages of unbiased estimate of relative treatment effect, as well as clearer interpretation of efficacy endpoints and comparative safety data.
• “Within the limitations associated with the rarity of the disease,” the EMA advises that the trial’s objectives and clinical questions should be formulated to address the question of benefit/risk, and the trial should be designed so these objectives can be met. One consideration when considering a randomized trial The lower number of participants in each arm may reduce the estimate precision. Choice of endpoints can help balance the precision with clinical interpretability of results.
• Study endpoints should be selected based on their clinical meaningfulness, sensitivity to detect an effect, and feasibility. PFS and OS should always be reported, even if interpretability is hampered by uncertainty of the estimates. Overall response rate (ORR), or other stringent depth of response measures could be used if the standard endpoints are not feasible. Additionally, if the population size is too small for standard endpoints, pharmacodynamic endpoints as a continuous variable (e.g., percentage of tumor change from baseline) could add sensitivity in detecting an effect. The selection and order of assessment for endpoints should be pre-specified at the planning stage if it is apparent that conventional pivotal hypothesis testing may be challenging. The clinical relevance of each endpoint is a key discussion to include in the protocol.
• If a single-arm or other non-randomized trial is used, sponsors must justify the choice, potentially based on the predictability of the disease course combined with a large treatment effect on evaluated endpoints (EMA guideline). However, notes the agency, “Long-term efficacy and safety should always be collected unless otherwise justified.” Any uncertainties on treatment effect would be evaluated through indirect comparison and then further investigated in post-authorization studies.
• External control arms may be used only if the “treatment effect is dramatic and the usual course of the disease is highly predictable,” according to the ICH E10 guideline on choice of control in clinical trials. The revised EMA guideline and the ICH guideline also cite the inability to control bias when using external controls.
• Contextualizing results from single-arm studies, where this approach is justified, is a key issue, according to the EMA. External controls should be sought unless the effect is dramatic and follows treatment rapidly so it couldn’t have occurred spontaneously.
• Within-patient analysis is an option as external control. This would involve comparing time to progression under the prior treatment with PFS on the trial treatment. Although this comparison reduces some of the potential bias and heterogeneity of external controls, it still has several weaknesses. The conditions and measurement of progression during prior and experimental therapy application must be comparable to avoid bias, notes the guideline, and sponsors will have to address any uncertainty related to less stringent assessment or adjudication of progression during earlier lines of treatment.
• Data from previous clinical trials, meta-analysis and registries may be used provided the data is of sufficiently high quality. Patient-level data is generally expected, and patients should have received standard of care. The rationale and methodology in collecting and analyzing external data should be discussed; the guideline gives a list of considerations to account for, including statistical considerations, criteria for matching patients from the external to the control arm, and more. “It is imperative that these steps are pre-specified prospectively in the protocol to avoid any convenient selection of external controls once the endpoint has been observed in the experimental arm,” advises the updated guidance.

Master protocols – Basket and umbrella trials

• Master protocols for designs such as basket and umbrella trials allow for the assessment of multiple indications or investigational products in parallel. However, the EMA notes that these are complex study designs, adding that sponsors considering such designs should consider seeking scientific advice.
• Master protocol trials must still follow the general methodological and statistical considerations for clinical trials. These include, but are not limited to, control of type 1 error, biomarker validation and dose-finding. Any deviation from existing guidelines should be planned and discussed with EMA in a scientific advice procedure.
• Umbrella trials investigate multiple drugs or drug combinations in one disease. They are generally used for exploratory trials to aid the design of standard confirmatory studies. However, they may also be used as pivotal trials in marketing authorization applications as long as the standard requirements (e.g., randomized control arms) are met.
• Complex clinical trial designs can pose statistical interpretation challenges. EMA explains that trials with a common control arm or overlapping populations fall into this category. Umbrella trials allocating treatment arms based on biomarker status may result in different prognosis for treatment arms, “conferred by the biomarker,” notes the agency. This circumstance could mean that different treatment arms have different prognoses, potentially precluding the use of a common comparator arm.
• Basket trials investigate the efficacy and safety of one drug or combination in a diverse study population with a variety of diseases. Study sub-populations are often defined by the presence of a response-predictive biomarker in oncology basket trials. These patient groups are defined by “different conventional histology- and anatomy-based tumour types” and called baskets. Basket trials, explained the updated guideline, “can be used for early phase trials aimed to identify patient populations likely to respond to the treatment for further development,” and to see how a therapy performs across baskets.
• Sufficient homogeneity across subpopulations must be demonstrated if an analysis across those subpopulations is planned. This should be supported by a sound rationale and adequate data during the planning phase to ensure a homogenous treatment effect. Data to support an analysis that pools baskets could stem from mechanistic rationale, or from pre-clinical and pharmacodynamics data. However, this may not be feasible if the participant number in each subgroup is insufficient. EMA advises that “sponsors must justify and make it convincingly plausible by clinical and/or pre-clinical data that the interaction with tumour site or histology is negligible and this should also be supported by the final results.”
• The guideline lays out a situation where looking at one therapy across several different histologies could include “primarily at least one relatively common sub-population” as defined by histology that would have sufficient power to detect a treatment effect. In this case, data from the larger sub-population could then be used to extrapolate efficacy to the smaller sub-populations so long as specific elements of heterogeneity between the various trial baskets didn’t preclude this extrapolation.

Analysis, and what’s next

• In general, this guideline is attempting to cover a large amount of regulatory territory in a single document. The basic structure of the document hasn’t changed, and ranges from pharmacokinetics to exploratory studies, through to guidance on pivotal trials, safety, and more. With the expansion of the section on biomarkers, more information in the general considerations section, and an entirely new section 8 addressing study designs for special situations, developers could be left feeling as though many items are given short shrift. The guideline does frequently refer to other ICH and EMA guidance in various areas, but some of the cross-referencing is confusing. Additionally, updates of disparate documents may not be well synchronized with each other, running the risk of further confusion if the referenced document has not been updated along with the anti-cancer guideline.
• Biomarkers and the use of master protocols are two rapidly evolving areas in oncology clinical development. The update to this guideline clearly highlights the importance of regulatory advice in the area through these updated guidelines and seeking EMA scientific advice, frequently recommended throughout the guideline. The draft guideline for this revision was under public consultation about three years ago and included a significant expansion in these topics. However, the finalized guideline has been expanded further yet in these areas. EMA may choose to place this guideline on a cycle of more frequent updates as knowledge and technology in this field evolves.
• The FDA has also been paying attention to the many of the areas addressed in the EMA’s refresh of this guideline, but It’s currently difficult to compare approaches across the two agencies for a variety of reasons. First, the FDA currently has much more granular information in its guidance than the EMA is providing, with a March 2022 guidance on master protocols for oncology products, and a December 2023 draft guidance on master protocols for drug and biological product development.
• Second, the EMA is working on a separate guideline for platform trials, floating a concept paper in 2022. Though public consultation on the concept paper ended exactly a year ago on January 31, 2023, the EMA has not released a summary of the comments it fielded during the year-plus public consultation period.
• The pharmaceutical trade association EFPIA has released its comments on the concept paper, however. In a lengthy general comment, EFPIA advises the EMA to “consider the output from the IMI EU-PEARL project that is due to sunset in April 2023.” This collaboration examining trial formats involved “numerous discussions with regulators, including the US FDA and the EMA Innovation Task Force,” as well as workshops. “Thus, rather than trying to ‘reinvent the wheel’ and duplicate discussions that have already been held, the deliverables from the EU-PEARL project should be taken into consideration when drafting a CHMP document on platform trials,” EFPIA advises.
• Notably, the EU-Pearl Consortium has just published a summary of its insights from the EU-PEARL initiative. Appearing online December 26, 2023 in The Lancet, The review article’s “10 things you should know” approach highlights the importance of early engagement with regulators and establishing strong clinical research networks, and emphasizes how an adaptive trial design can boost efficiency, among several other points. The document’s many charts and tables also provide frameworks and approaches for important components such as statistical approaches for platform trials – underlining EFPIA’s point that the EMA’s platform trial guideline drafting committee could take many pages from the EU-PEARL playbook.
• Third, the master protocol section of the guideline is particularly difficult to parse, perhaps in part because the foundational work of writing the EMA’s general guideline on the topic has not been completed. For example, the characterization of what constitutes an umbrella trial implies that a common control arm might not be a desirable characteristic when designing these trials, with word choice and phrasing that makes the regulator’s intent challenging to understand clearly. Most characterizations of umbrella trials, including the FDA’s, explicitly refer to a common control arm as an integral feature of these trials. Whether this lack of clarity is related to relative lack of regulatory maturity on the part of the EMA, confusion in translation from another language, or something else entirely is not known – but it could present problems for oncology developers.
• A related question: Which standard of care should be chosen for global master protocols? The FDA guidance clearly notes that standard of care should be based on U.S. medical practice. The EMA guideline, though, doesn’t make a specific statement, most likely due to the variability of standard of care in the different Member States of the E.U. Although many cancers may have sufficient overlap in treatment options to enable master protocol trials in both jurisdictions under the same protocol, the issue might also come into play when considering external control arms for rare cancers.
• Both regulators recommend that investigators initiate a conversation about approaches to biomarker and master protocol use, to ensure regulatory requirements are met. Guidance documents can only provide an overview of considerations and naturally wouldn’t be able to address every single situation. Additionally, these documents take often years to develop and recommendations may already have been updated based on newer experience and/or knowledge. Therefore, it is imperative for sponsors to seek opportunities to discuss their approaches with regulators for example through meetings, scientific advice or the E.U. novel methodology qualification.

To contact the authors of this item, please email Kirsten Messmer ( [email protected]) or Kari Oakes ( [email protected]).
To contact the editor of this item, please email Kari Oakes ( [email protected]).

Key Documents and Dates

• EMA webpage – Evaluation of anticancer medicinal products in man, last updated January 23, 2024
• EMA/CHMP/205/95 Rev.6 – Guideline on the clinical evaluation of anticancer medicinal products, dated November 18, 2023, published January 23, 2024