ICH M11 Protocol Endpoints: Writing Definitions That Can Be Executed and Reported

A clinical trial protocol exists for two operational audiences: the investigator site teams who must administer assessments without improvising procedures, and the regulatory reviewers and biostatisticians who must reconstruct exactly what occurred when evaluating study findings. When protocol authors draft an ICH M11 Clinical Electronic Structured Harmonised Protocol (CeSHarP), they encounter a common structural hazard: Section 3 provides a standardized estimand table, but reduces the endpoint definition to a single blank cell labelled Endpoint | Definition of the endpoint. If authors enter only a loose concept or an isolated instrument acronym—such as "HAM-D17" or "ACR20"—they fail the primary operational standard of trial design. Site coordinators are left to guess testing windows, data managers must invent scoring derivations, and the statistical team is forced to define cutoff thresholds after unblinded data begin to accumulate.

This guide addresses exactly one practical challenge: how protocol authors write endpoint definitions in an ICH M11 protocol so sites can execute them reliably and later trial reports can fully reconstruct them. It is not an operational re-evaluation of computerized systems validation under ICH E6(R3), nor does it re-examine how biostatisticians select among intercurrent-event strategies under ICH E9(R1). It does not duplicate pre-freeze quality risk evaluations under ICH E8(R1), the timing of SAP locks analyzed in our guide on statistical analysis plans, sensitivity modeling for missing data versus intercurrent events, or family-wise error allocations across multiple trial endpoints. Most importantly, authors must recognize two non-negotiable boundaries: completing the M11 template does not constitute regulatory clearance or approval of an endpoint, and satisfying the completeness items of SPIRIT-Outcomes 2022 or SPIRIT 2025 does not establish a legal or GCP defense for an unexecutable trial protocol.

To make an endpoint fully operational, authors must lock five core outcome elements inside Section 3, then systematically wire that single definition into every downstream execution chapter: allowable visit windows in Section 1.3, continued follow-up protocols in Section 7, measurement citations and instrument controls in Section 8, and estimand-aligned estimators in Section 10.

The ICH M11 guideline, officially titled Clinical Electronic Structured Harmonised Protocol (CeSHarP), reached Step 4 final adoption by the Regulatory Members of the ICH Assembly on 19 November 2025. CeSHarP is not a single document or narrative outline; it is a tripartite harmonized package comprising three interrelated technical components:

1. The M11 Guideline: Sets forth the overarching design principles, objectives, and scope governing harmonized protocol authoring and maintenance across international drug development.

2. The M11 Template: Supplies a standardized table of contents, mandatory common headers, structured instructional text, and formalized estimand-characteristics tables designed to support interventional clinical trials.

3. The M11 Technical Specification: Defines the data elements and technical attributes (definition, conformance, cardinality) that enable interoperable electronic exchange of protocol content among sponsors, ethics bodies, investigators, and regulatory agencies.

The M11 Template applies broadly to interventional clinical trials of medicinal products across phases and therapeutic areas, including human pharmacology, exploratory, confirmatory, and post-approval trials. When authoring an M11 document, strict structural rules apply: Level 1 headings are mandatory and must be retained. If a specific Level 1 section is entirely unnecessary for a given trial design, the author must explicitly designate that section as Not applicable rather than deleting the chapter. Level 2 headings and black (universal) Level 3 and Level 4 headings follow the same retain-or-mark rule; they are not free to delete. Optional lower-level headings that the Template styles as optional, and red instructional text, may be adapted, reworded, or removed. Where the Template cites other ICH guidelines, the current version applies.

In the European Union, the Committee for Medicinal Products for Human Use (CHMP) adopted the Step 5 package on 11 December 2025, with the European Medicines Agency (EMA) publishing the Step 5 Guideline, Template, and Technical Specification (EMA/CHMP/ICH/778799/2022) on 15 December 2025. The official EMA Step 5 Template designated 11 June 2026 as the formal date for coming into effect across the European medicines regulatory network. As of 18 September 2026, implementation across ICH member jurisdictions remains uneven. The same ICH table lists FDA as implemented on 22 May 2026 and EC/Europe as implemented on 11 June 2026 (EMA/CHMP/ICH/778799/2022). Swissmedic treats ICH guidelines as applying automatically at Step 4 (19 November 2025), and SFDA (Saudi Arabia) is listed as implemented on 15 April 2026. Health Canada, HSA (Singapore), SAHPRA, and TFDA are listed as in the process of implementation. MHLW/PMDA (Japan), MHRA (UK), and several other members remain not yet implemented. Authors must verify the receiving authority's current expectation rather than assuming that electronic CeSHarP exchange is universally mandatory.

The central structural innovation of ICH M11 is that it operationalizes the core mandate of ICH E9(R1) Section A.6. Under E9(R1), sponsors are obligated to define a primary estimand corresponding to each primary trial objective explicitly within the study protocol itself—not deferred to a downstream statistical analysis plan. Secondary estimands intended to substantiate confirmatory regulatory labeling claims must likewise be prespecified in full. If the clinical question or treatment-effect definition changes during study conduct, that alteration represents a fundamental shift in scientific intent that must be documented through a formal protocol amendment.

In M11, this documentation obligation is anchored in Section 3: Trial Objectives and Associated Estimands. Section 3 establishes a direct, structured hierarchy across three tiers of clinical objectives:

• Section 3.1 Primary Objectives: Requires a structured estimand-characteristics table for every trial designed to estimate or test a treatment effect. The table binds five interrelated attributes: population, treatment condition, endpoint (variable) definition, population-level summary measure, and individual rows for each prespecified intercurrent event along with its associated handling strategy.

• Section 3.2 Secondary Objectives: Utilizes the identical estimand table for all key secondary objectives that support regulatory decision-making, ensuring that secondary endpoints are not left as vague exploratory concepts.

• Section 3.3 Exploratory Objectives: Requires, at minimum, an explicit statement of the endpoint associated with each exploratory objective. ICH E9(R1) does not require an estimand for every exploratory objective; M11 still allows the estimand table when operational precision is needed.

The M11 Template incorporates E9(R1)'s foundational insight: precise specifications of the target population, treatment regimen, and endpoint variable will naturally resolve many potential intercurrent events, leaving remaining clinical disruptions to be assigned explicit strategies (such as treatment-policy, hypothetical, composite, while-on-treatment, or principal-stratum). Crucially, M11 links Section 3 to Section 1.1.1 (Primary and Secondary Objectives and Estimands), which requires a concise, nontechnical synopsis summary of objectives and estimands for ethics committees and public registries, and to Section 4.2.1 (Rationale for Estimand(s)), which requires clinical evidence demonstrating that the chosen endpoints are clinically meaningful and provide valid, reliable reflections of the treatment effect.

flowchart TD
  subgraph M11_Section_3["M11 Section 3: Trial Objectives & Associated Estimands"]
    Obj["Primary & Secondary Objectives (3.1 / 3.2)"] --> EstTable["Estimand-Characteristics Table"]
    EstTable --> Pop["Target Population"]
    EstTable --> Trt["Treatment Regimen"]
    EstTable --> Var["Endpoint (Variable) Definition"]
    EstTable --> Summ["Population-Level Summary"]
    EstTable --> ICE["Intercurrent Event Strategies"]
  end

  subgraph Cross_Section_Alignment["Downstream Operational Execution"]
    Var --> SoA["Section 1.3: Schedule of Activities<br/>(Visit Windows & Assessment Timing)"]
    Var --> Proc["Section 8: Assessments & Procedures<br/>(Administration Rules & Validation Citations)"]
    ICE --> Disc["Section 7: Participant Discontinuation<br/>(Mandatory Follow-Up Protocols)"]
    EstTable --> Stats["Section 10: Statistical Considerations<br/>(10.4.1.1 Estimator & 10.4.1.2 ICE Handling)"]
  end

  subgraph External_Standards["Completeness & Regulatory Fences"]
    SPIRIT["SPIRIT-Outcomes 2022 & SPIRIT 2025<br/>(5 Core Elements & Reporting Completeness)"] -.-> Var
    E9R1["ICH E9(R1) Documentation Mandate<br/>(Protocol-Locked Estimands)"] -.-> EstTable
    CFR["21 CFR 10.115 / ICH E6(R3)<br/>(Nonbinding Guidance / GCP Operational Integrity)"] -.-> Stats
  end

  classDef section fill:#f4efe6,stroke:#7a1c1c,stroke-width:2px,color:#1e293b;
  classDef downstream fill:#eef2f6,stroke:#1e3a5f,stroke-width:1.5px,color:#0f172a;
  classDef standard fill:#f8fafc,stroke:#475569,stroke-width:1px,stroke-dasharray: 4 4,color:#334155;
  class EstTable,Obj,Pop,Trt,Var,Summ,ICE section;
  class SoA,Proc,Disc,Stats downstream;
  class SPIRIT,E9R1,CFR standard;

The greatest vulnerability in an M11 draft is treating the Endpoint | Definition of the endpoint cell as a simple label. Writing "Change in Montgomery-Åsberg Depression Rating Scale (MADRS) total score at Week 12" may sound like a definition, but it omits critical operational parameters: Is the score baseline-adjusted? What exact visit window defines Week 12? Who administers the interview? What minimal change is considered clinically meaningful? When continuous scores are converted to responder proportions, what exact threshold is applied?

To eliminate this ambiguity, protocol authors must cross-reference the clinical trial reporting standard established by Butcher et al. in the SPIRIT-Outcomes 2022 Extension (JAMA. 2022;328(23):2345-2356). SPIRIT-Outcomes restates five indispensable core elements that must be present to define any clinical trial outcome:

1. Domain: The precise clinical concept, health dimension, or pathophysiological construct being measured (e.g., major depressive symptom severity, exercise endurance, joint inflammation).

2. Specific Measurement Variable or Instrument: The exact scale, diagnostic assay, biometric sensor, or observation tool used to capture the domain, including version and scoring range where those are part of the measure (e.g., MADRS 10-item clinician-rated interview, 0–60 scale).

3. Participant-Level Metric: How the outcome is quantified for an individual trial participant over time (e.g., absolute value at a single visit, absolute change from baseline, percentage change, or time from randomization to first confirmed event).

4. Method of Aggregation: How individual participant metrics are mathematically aggregated across the study arm (e.g., mean difference between arms, median survival, proportion achieving response), including any explicit numerical cutoffs used to dichotomize continuous metrics into categorical rates.

5. Time Point: The precise schedule of observation, specifying both the full timeline of measurement collection and the exact primary time point selected for primary analysis and hypothesis testing.

In addition to these five core elements, the author must explicitly specify the Assessor (e.g., an independent central rater blinded to treatment assignment, an investigator clinician, a parent observer, or the trial participant via an ePRO device). Conflating the E9(R1) population-level summary (such as hazard ratio or difference in least-squares means) with the participant-level metric is a severe error; the summary describes between-group comparison, while the endpoint variable describes individual measurement.

Furthermore, Butcher and colleagues introduced nine outcome-specific protocol items that rigorous protocols should prespecify before study initiation. In April 2025, Chan and colleagues published the SPIRIT 2025 statement (BMJ 2025;389:e081477), updating the foundational SPIRIT 2013 guideline by integrating key recommendations from SPIRIT-Outcomes 2022, CONSORT Harms 2022, and TIDieR directly into the core randomized trial protocol checklist. The EQUATOR Network continues to list SPIRIT-Outcomes 2022 as an active, specialized extension (record updated 30 May 2025). For trials initiated in 2026, authors submitting to medical journals must verify compliance with SPIRIT 2025 and SPIRIT-Outcomes 2022, while structuring the regulatory master protocol under M11.

A clinically robust definition written in Section 3 will collapse in practice if the operational sections of the protocol fail to support its execution. ICH M11 demands strict cross-section harmonization across three primary chapters: the Schedule of Activities, Participant Discontinuation, and Trial Assessments.

Section 1.3 Schedule of Activities (SoA): The SoA must detail every clinical procedure, laboratory draw, and outcome assessment across each protocol contact. To execute the Section 3 endpoint, the SoA must establish unambiguous, allowable visit windows (for example, Week 12 Day 84 ± 3 days). Without explicit windows, site coordinators will capture observations outside the analytical landmark, creating data integrity debates during data review. The SoA must also clearly designate tests required for stratification, baseline adjustment, or protocol-mandated stopping rules.

Section 8 Trial Assessments and Procedures: Section 8 operationalizes the measurement tools cited in Section 3. For every endpoint, Section 8 must describe administration conditions, scoring instructions, and who will perform the assessment. Authors must include literature citations for the validation of scales, instruments, questionnaires, and assays.

Authors must respect a critical structural boundary within Section 8: Section 8.8 (Medical Resource Utilisation and Health Economics) does not apply to COAs. Clinical Outcome Assessments (COAs)—encompassing Patient-Reported Outcomes (PROs), Clinician-Reported Outcomes (ClinROs), Observer-Reported Outcomes (ObsROs), and Performance Outcomes (PerfOs)—must not be placed in Section 8.8. The Template directs remaining value-evidence measures there only when they are not already in the efficacy or safety sections. When measuring efficacy, COAs belong in Section 8.3 (Efficacy Assessments and Procedures); when measuring safety or tolerability, they belong in Section 8.4 (Safety Assessments and Procedures). Finally, if endpoints require independent verification—such as major adverse cardiovascular events (MACE) or radiographic progression—the existence of the independent adjudication committee is noted in Synopsis Section 1.1.2, while the operational charter and governance framework are detailed in Section 11.4 (Committees).

Section 10 of the M11 protocol governs statistical methodology, ensuring full compliance with ICH E9 and ICH E9(R1). The primary imperative of Section 10 is that statistical analysis text must operationalize the Section 3 estimand without altering or rewriting its scientific definition. The protocol must explicitly pre-specify the aligned main estimator, the designated analysis set (e.g., full analysis set, per-protocol set), and the mathematical framework used for hypothesis testing.

ICH E9(R1) requires a suitable same-estimand sensitivity analysis for the primary estimand. M11 provides optional headings 10.4.1.4 {Sensitivity Analysis} and 10.4.1.5 {Supplementary Analysis} for that text. A sensitivity analysis must evaluate the robustness of inferences from the main estimator to departures from underlying modelling assumptions and data limitations under the exact same estimand; an analysis that alters the treatment-effect definition or changes the intercurrent-event strategy is a supplementary analysis investigating a different clinical question, not a sensitivity check.

Under Section 10.10 (Multiplicity Adjustments), the protocol must pre-specify the formal family-wise error rate (FWER) control architecture (such as hierarchical gatekeeping, Hochberg, or graphical testing) across primary endpoints, coprimaries, and key secondary endpoints. It must state the precise decision rule by which the trial will be formally declared a positive study. Under Section 10.11 (Sample Size Determination), the calculation must align directly with the primary estimand, detailing effect size, variance, anticipated intercurrent event frequencies, and dropout rates.

While sponsors typically prepare a standalone Statistical Analysis Plan (SAP)—and Section 10.1 General Considerations asks whether one exists (Gamble et al. 2017 remains a public SAP-content checklist, not a regulation)—the protocol must lock the estimand, the endpoint definition, ICE data handling, the protocol-level missing-data approach, multiplicity, and sample-size assumptions. The SAP may add programming algorithms and mock table shells; the SAP cannot silently alter or re-engineer the endpoint locked in the protocol.

A completed, beautifully formatted ICH M11 protocol is an indispensable operational achievement, but trial leadership must never confuse procedural completeness with scientific or regulatory validation. Five critical regulatory boundaries govern the interpretation of M11 documents:

1. Template Adherence Is Not Agency Approval: Submitting an endpoint within an M11-formatted protocol does not constitute regulatory acceptance, qualification, or clearance by FDA, EMA, or any international authority. Under 21 CFR 10.115, FDA guidance represents nonbinding current thinking. Reviewers evaluate the scientific adequacy of the clinical concept, the psychometric robustness of the tool, and the clinical meaningfulness of the treatment effect entirely on their merits.

2. Reporting Guidelines Are Not Regulatory Law: SPIRIT-Outcomes 2022 and SPIRIT 2025 are vital reporting guidelines established by academic and methodological consensus under the EQUATOR Network. They ensure completeness and transparency for journal peer review, institutional review boards (IRBs), and ethics committees. They are not Good Clinical Practice (GCP) regulations, and checking every item on a SPIRIT checklist does not fulfill the statutory obligations enforced by health authorities.

3. Regional Implementation Is Not Uniform: While ICH Step 4 was finalized in November 2025 and regional guidance took effect in the U.S. and EU in mid-2026, global adoption is ongoing. Sponsors running multinational trials must confirm local health authority requirements in each participating country rather than assuming that structured CeSHarP exchange is universally accepted or required.

4. Electronic Exchange Does Not Eliminate Narrative Scrutiny: The M11 Technical Specification establishes structured data elements for digital protocol transmission, but regulatory reviewers continue to scrutinize the narrative clinical justification in Section 4.2.1 and the operational feasibility of Section 8. Digital exchange accelerates intake; it does not loosen evidentiary standards.

5. Estimand Modifications Require Formal Protocol Amendments: If accumulated operational experience or emerging safety signals require modifying an estimand, altering an endpoint cutoff, or reclassifying an intercurrent event strategy during trial conduct, that change cannot be quietly inserted into an updated SAP. Under ICH E9(R1) and GCP principles, any change to a trial estimand requires a formal, submitted protocol amendment prior to unblinding.

By anchoring the five SPIRIT outcome elements firmly within M11 Section 3, aligning downstream operational execution across Sections 1.3, 7, and 8, and locking rigorous statistical models in Section 10, clinical development teams build protocols that stand up to regulatory review, eliminate site ambiguity, and deliver reliable, reproducible clinical evidence.