Endpoint Adjudication: Preparing a Charter and a Complete Case Packet

When a clinical trial designates an event-driven primary outcome—such as an adjudicated composite of cardiovascular death, myocardial infarction, and hospitalization for heart failure—the operational integrity of that endpoint does not depend on the clinical prominence of the adjudicators. It depends entirely on the operational rules locked into the adjudication charter and case-packet specification before the first participant is randomized. If an investigative site submits an ambiguous event report—for instance, a three-line hospital discharge summary lacking electrocardiograms (ECGs) and biomarker assays—the coordinating office cannot improvise how that packet is investigated, queried, or adjudicated. Without explicit pre-specified rules, the team faces an immediate methodological breakdown: informally querying the site creates an unmonitored communication channel, guessing whether the event occurred compromises data integrity, and silently dropping the case distorts the protocol's primary estimand.

An endpoint assessment or adjudication committee (commonly termed a clinical events committee, or CEC) is formally optional under United States Food and Drug Administration (FDA) guidance and International Council for Harmonisation (ICH) standards. Under Section 3.3 of FDA's in-force March 2006 guidance, Guidance for Clinical Trial Sponsors: Establishment and Operation of Clinical Trial Data Monitoring Committees, sponsors may choose to establish an endpoint assessment committee to review important endpoints reported by investigators against protocol-specified criteria. Similarly, ICH E6(R3) Section 3.9.8 states that sponsors may establish an endpoint assessment or adjudication committee where appropriate. Neither authority mandates a committee for every trial.

Once a sponsor chooses to use an adjudication committee, ICH E6(R3) Section 3.9.9 applies to any trial-related committee that could affect participant safety or the reliability of results. That section says the committee should include members with relevant expertise and managed conflicts of interest, should have written operating procedures (for example a charter), and should document its decisions. Appendix B.8.2 adds that where a committee such as an independent data monitoring committee or an adjudication committee assesses efficacy data, its procedures, timing, and activities should be described in the protocol or a separate document. The essential-records table lists two further items to retain when they are produced: a signed agreement between the sponsor and the independent data monitoring committee and/or adjudication committee members, and the procedures, meeting minutes, and submissions to that committee. Section 3.6.1 says those agreements should be documented before the activities begin. None of this language makes a committee mandatory for every trial.

FDA and ICH describe the adjudication committee and the data monitoring committee as separate bodies. Under the in-force March 2006 FDA guidance (Section 3.3), an adjudication committee is typically masked to the assigned study arm whether or not the trial itself is blinded. Information reviewed on each presumptive endpoint may include laboratory, pathology, and imaging data, autopsy reports, physical descriptions, and other relevant records. When those assessments are made at frequent intervals and entered into the database promptly, the data monitoring committee is more likely to review endpoint data that are accurate and free of ascertainment bias. FDA's February 2024 draft guidance, Use of Data Monitoring Committees in Clinical Trials (a draft, not for implementation and not in force), states the separation directly: a data monitoring committee with access to unblinded comparative data should not adjudicate trial endpoints, and adjudication committees are expected to be blinded to the assigned intervention regardless of whether the trial is blinded. Adjudication results should be completed and transferred to the data monitoring committee in time for its review.

Adoption is regional. In the European Union, the overarching principles and Annex 1 came into effect on 23 July 2025 (EMA/CHMP/ICH/135/1995). Annex 2, which covers decentralized and pragmatic elements, comes into effect on 15 January 2027. The consolidated Step 5 guideline that combines the principles with both annexes (EMA/CHMP/ICH/162879/2026) carries that same 15 January 2027 legal effective date. In the United States, E6(R3) is FDA guidance, not a statute. The committee duties in Sections 3.6.1, 3.9.8, 3.9.9, and Appendix B.8 sit in the Annex 1 text already in effect for EU trials; Annex 2's later date does not postpone them. Aligning protocol endpoint language under ICH M11 protocol endpoint definitions and verifying trial-wide governance against the ICH E6(R3) endpoint implementation checklist are essential prerequisites before authoring the charter.

An adjudication charter is the foundational operating contract governing all trial-specific event classification activities. Drawing upon consensus practitioner models—such as the 14-section charter framework published in Applied Clinical Trials by Widler et al. (2016) and the operational review by Held (2018)—a complete, audit-ready charter must specify nine distinct procedural pillars before the first event is reported.

1. Endpoint Definitions and Protocol Alignment: The charter should copy the endpoint definitions from the protocol and the statistical analysis plan, including the same event criteria and the same handling of events that cannot be classified. Where the protocol follows the ICH M11 template, those definitions sit in template Section 3. Adjudicators do not apply a personal diagnostic standard that drifts from the protocol.

2. Scope Exclusions and Event Triggers: The charter must delineate exactly which reported adverse events, laboratory abnormalities, or diagnostic findings trigger adjudication, and which events fall outside committee scope. Ambiguity regarding trigger criteria leads to selective reporting or massive dossier backlogs.

3. Committee Membership, Expertise, and Conflict Nullification: The charter must identify voting members and clinical alternates possessing relevant medical specialty credentials. It must enforce regular conflict-of-interest disclosures. Crucially, the charter must contain a strict conflict nullification rule: if an adjudicator was involved in treating the participant, enrolling the patient, or reporting the event at an investigative site, their review of that event is automatically nullified and re-assigned to an alternate reviewer.

4. Blinding Architecture: The charter must specify how masking is preserved. In double-blind studies, committee members remain blinded to treatment assignment. In open-label or pragmatic trials, the charter establishes an independent blinded review firewall to eliminate differential ascertainment bias between arms.

5. Anonymized Event Coding: To prevent geographical or institutional bias, each reported case is assigned a unique anonymized event identification code by the coordinating office. Site numbers, investigator names, and participant identifiers are stripped before dossier generation.

6. Package Completeness and Query Loop: The coordinating center or endpoint office must verify that each packet satisfies a pre-specified minimum source-document checklist. If mandatory records are missing, the charter defines a structured Request for Information (RFI) workflow with binding turnaround timelines.

7. Multi-Reviewer Workflow and Disagreement Resolution: The charter specifies the review architecture. Standard practice routes each case packet to two independent primary adjudicators. If both agree, the decision is locked. If discordant, the charter dictates whether resolution occurs through mutual discussion, referral to a third senior adjudicator, or evaluation by a plenary consensus committee.

8. Re-Adjudication Precedence: The charter must define objective criteria under which a finalized adjudication can be reopened—such as the discovery of previously unavailable autopsy findings or corrected angiographic recordings. It must establish that a formal re-adjudication explicitly supersedes the prior decision in the audit trail.

9. Data Flow, Timelines, and DMC Feeding: The charter must specify operational turnaround windows: days from site reporting to packet compilation, days for reviewer assessment, and schedules for transmitting validated determinations into the clinical database to ensure timely feeding of DMC meetings.

flowchart TD
  A["Investigative Site Event Report"] --> B["Endpoint Office Completeness Check"]
  B --> C{"Packet Complete?"}
  C -- "No" --> D["Issue Structured RFI to Site"]
  D --> E{"Records Obtained?"}
  E -- "Yes" --> B
  E -- "No" --> F["Mark Source Data Exhausted"]
  F --> G["Redaction and Anonymization"]
  C -- "Yes" --> G
  G --> H["Blinded Case Packet Issued"]
  H --> I["Reviewer 1 Assessment"]
  H --> J["Reviewer 2 Assessment"]
  I --> K{"Concordant Determination?"}
  J --> K
  K -- "Yes" --> L["Documented Final Decision"]
  K -- "No" --> M["Consensus or Chair Review"]
  M --> L
  L --> N["Transfer Locked Decision to Database and DMC"]

To prevent operational ambiguity during study conduct, each charter section must be paired with concrete operational rules that eliminate regulatory inspection vulnerabilities. The following comparison table maps core charter sections to their operational execution and the specific regulatory findings they avert.

The operational unit of endpoint adjudication is the case packet. FDA's March 2006 guidance, Section 3.3, says the information reviewed on each presumptive endpoint may include laboratory, pathology, and imaging data, autopsy reports, physical descriptions, and any other data the charter deems relevant. It does not publish a mandatory document list. Held (2018) describes what an endpoint office actually assembles: hospital notes, relevant electrocardiograms, laboratory reports, imaging such as computed tomography or angiography, and a pre-selected electronic case-report-form patient profile, checked for completeness before two independent reviewers see the packet. The charter's job is to turn that menu into a minimum source-document set for each event type the committee will classify.

For acute cardiovascular outcomes, a complete dossier typically comprises five distinct evidentiary layers:

• Clinical Hospital Records: Emergency department admission notes, inpatient progress notes detailing symptoms (nature, onset, and duration of chest discomfort or dyspnea), discharge summaries, and transfer documentation.

• Objective Diagnostic Tracings: Serial 12-lead ECG recordings (baseline, admission, peak symptom, and post-event) displaying voltage calibrations and paper speeds, evaluated for ST-segment elevation, ST depression, T-wave inversions, or new pathological Q-waves.

• Quantitative Laboratory Assays: Serial cardiac biomarker reports—specifically cardiac troponin I or T (cTnI, cTnT, or high-sensitivity hs-cTn) and CK-MB—accompanied by local laboratory assay manufacturer documentation, reference ranges, and the upper reference limit (URL, 99th percentile).

• Imaging and Procedural Documentation: Echocardiogram reports assessing regional wall motion abnormalities; coronary angiography reports or catheterization cine recordings documenting vessel occlusion, thrombosis, or revascularization; CT or MRI scans.

• De-Identified eCRF Patient Profile: A standardized electronic Case Report Form export summarizing relevant baseline medical history, cardiovascular risk factors, concomitant medications, and adverse event timelines, stripped of site and subject identifiers.

The endpoint office also has to keep treatment assignment out of the packet. The 2006 guidance says these committees are typically masked to the assigned arm, and the February 2024 draft says they are expected to be blinded to the assigned intervention even when the trial is open-label. A practical charter therefore names who checks redaction before release and which clues are removed: investigational-product lot numbers, dosing schedules that differ by arm, packaging descriptions, and monitoring instructions that exist only on the active arm. That scrub list implements the blinding rule. It is not a document list in the FDA guidance or in Held's review.

To illustrate the critical necessity of pre-specified charter rules, consider a clearly hypothetical operational scenario from a fictional clinical trial.

Context: A fictional 4,000-participant, multinational, double-blind phase 3 cardiovascular outcomes trial evaluates a novel oral small-molecule inhibitor against matched placebo in people with established atherosclerotic disease. The primary endpoint is the composite named in the opening: cardiovascular death, myocardial infarction, and hospitalization for heart failure, classified by an independent clinical events committee. No count or laboratory value in this example comes from a real trial.

Incident at Week 7: An investigative site reports a presumptive primary endpoint: 'Suspected acute myocardial infarction' in a 64-year-old participant. However, the site uploads only a brief three-line hospital discharge summary stating: "Patient presented with retrosternal chest discomfort; treated with antiplatelet therapy and statins; discharged in stable condition after 48 hours." No serial 12-lead ECG tracings, no cardiac biomarker lab sheets (troponin or CK-MB), and no coronary angiographic reports are attached.

Operational Pressure: The twice-yearly data monitoring committee meeting has a data-cut date in exactly 10 business days. Clinical operations leadership asks the endpoint office whether the packet can be forwarded immediately to the two assigned adjudicators so that a decision can be recorded before the database snapshot.

How the Pre-Specified Charter Governs the Case:

1. Completeness Gating: The endpoint office coordinator evaluates the uploaded dossier against the charter's mandatory checklist for suspected myocardial infarction. Serial troponin assays and ECGs are designated as non-negotiable core diagnostic evidence. The packet is formally flagged as 'Incomplete / Missing Mandatory Source Data.' Under charter rules, the endpoint office is prohibited from issuing the packet to clinical adjudicators.

2. Structured RFI Workflow: The endpoint office issues an automated, tracked Request for Information (RFI) to the investigative site's principal investigator with a mandated 7-calendar-day response window. The RFI specifies: (a) certified laboratory reports of all cardiac troponin assays performed during the index admission, including assay-specific 99th percentile upper reference limits; (b) photocopies of all 12-lead ECG tracings obtained at admission, peak pain, and discharge; and (c) cardiology consultation notes.

3. Branching Scenario A (Source Records Delivered): On day 5, the site uploads certified laboratory reports demonstrating high-sensitivity cardiac troponin T (hs-cTnT) values rising from 18 ng/L at admission to 142 ng/L at 6 hours (URL: 14 ng/L), alongside ECGs exhibiting 2 mm ST-segment elevation in leads V2–V4. The endpoint office scrubs all site identifiers and hospital headers, packages the dossier, and delivers it to Reviewers 1 and 2. Both adjudicators independently confirm acute ST-elevation myocardial infarction under Fourth Universal Definition criteria. The concordant determination is locked in the database on day 8, well in advance of the DMC snapshot.

4. Branching Scenario B (Source Records Unobtainable): The site reports that the participant was admitted to an unaffiliated rural hospital while traveling, and the non-study institution refuses to provide records despite signed participant authorizations. Two tracked escalation queries fail to yield diagnostic sheets. The charter's pre-specified missing-source rule takes effect. The endpoint office compiles the records it has, attaches a note that the query trail is exhausted, and still routes the redacted dossier to the reviewers. Both adjudicators classify the event as indeterminate because the clinical data are insufficient to confirm infarction. The event is not counted as a confirmed component of the primary composite. The classification, the query trail, and the exhaustion note stay in the file.

5. Statistical Alignment: Because the charter and the statistical analysis plan were aligned before the first participant, the handling of this indeterminate event is already written in Statistical Analysis Plans for endpoints. In this example the primary estimand counts only events the committee confirms. An indeterminate classification is not the same as a missing measurement, and dropping the case is not a regulatory default. The example analysis plan states the strategy in advance: confirmed events enter the composite, indeterminate events stay out of the primary count, and a tipping-point sensitivity analysis imputes those indeterminate events as primary outcomes to test whether that choice would change the conclusion.

The Failure Mode If Rules Were Missing: Consider the same week if the charter had none of these rules. To meet the data cut, the coordinator forwards the three-line summary. One reviewer treats the antiplatelet therapy as evidence of infarction; the other finds the record insufficient. An undocumented telephone call produces a consensus to confirm the infarction. Months later the audit trail shows a primary-endpoint myocardial infarction supported only by that summary, with no electrocardiogram, no troponin report, and no record of the call. The endpoint cannot be reconstructed. That is the failure the charter is written to prevent.

A rigorous endpoint strategy has to face a narrower finding than many protocols assume: centralized adjudication does not automatically change a trial conclusion or improve the ability to determine a treatment effect. The size of the change depends on the event definition, the design, and how much judgment the diagnosis requires.

In a seminal methodological critique published in Clinical Trials, Granger and colleagues argued that central adjudication has not been shown to improve the ability to determine treatment effects and that it may be overly complex and overused. They made a specific exception: in an open-label trial, adjudication that stays blinded to treatment assignment can protect against differential misclassification by investigators who know the assignment.

Where investigator reports and committee decisions diverge is also specific, and it should not be flattened into one percentage. Khan and colleagues (2023), reviewing cardiovascular trials in the European Heart Journal, use REDUCE-IT as the worked comparison: agreement between investigator-reported and committee-adjudicated events was nearly 100 percent for cardiovascular death and coronary revascularization, about 80 to 90 percent for myocardial infarction and stroke, and about 50 percent for unstable angina. Across the trials they review, reclassification often left the effect size for composite endpoints, all-cause mortality, all-cause hospitalization, and stroke essentially unchanged. Those figures are cardiovascular-trial results. They are not a general rate for heart-failure hospitalization, transient ischemic attack, or non-cardiovascular endpoints.

Where discordance occurs, does it alter trial conclusions? Historical and modern cardiovascular trials offer contrasting lessons:

• The PURSUIT Trial (Mahaffey et al., 2001): In the PURSUIT trial of eptifibatide in acute coronary syndrome (10,948 participants), the CEC systematically screened 5,005 patients with suspected myocardial infarction (46% of the trial population) through CRF logs, enzyme tracking, and routine ECGs, confirming 1,415 (28%) as protocol endpoint MIs. Site investigators and the CEC disagreed in 983 of 5,005 reviewed cases—a 20% discordance rate (816 MIs were identified only by the CEC, and 167 were identified only by sites). Crucially, the observed absolute and relative treatment-effect estimates for eptifibatide were smaller when evaluated using CEC-adjudicated event rates compared to site-reported event rates.

• The EMPEROR-Reduced Trial (Carson et al., 2023): In EMPEROR-Reduced (3,730 participants), the clinical events committee confirmed 76.3 percent of investigator-reported primary-outcome events. Confirmation was 89.1 percent for cardiovascular death and 73.7 percent for hospitalization for heart failure. The committee did not confirm heart failure as the primary reason for 324 investigator-reported heart-failure hospitalizations. The hazard-ratio point estimate for empagliflozin versus placebo on the primary composite was 0.75 either way: 0.75 (95 percent confidence interval 0.66 to 0.85) on investigator reports and 0.75 (0.65 to 0.86) on committee adjudication. These percentages come from one heart-failure trial and should not be reused as rates for other endpoint families.

The cardiovascular evidence above supports a narrow decision rule. A sponsor should not create a clinical events committee for every endpoint by default. Adjudication has changed event counts, and in PURSUIT it also shrank the treatment-effect estimate, when the endpoint is a complex diagnosis such as myocardial infarction. It has not reliably changed the treatment-effect estimate when the event is less ambiguous: Khan et al. report that reclassification often leaves the effect size for all-cause mortality essentially unchanged, and EMPEROR-Reduced's primary hazard ratio was 0.75 under both methods. Open-label trials are the clearer case for blinded adjudication, because investigators know the assignment. The charter should name which endpoints are adjudicated and record that reason.

ICH E6(R3) Section 3.9.9 calls for documented decisions, and Appendix B.8.2 calls for the committee's procedures to be described in the protocol or a separate document. Appendix C treats the agreement with committee members, and the procedures, meeting minutes, and submissions, as essential records when they are produced. A final event tally alone does not show that those records exist. The reconstructable file is the charter version in force for that event, the packet the reviewers actually saw, and the decision record.

An inspector must be able to reconstruct five core documentary pillars from the trial archives:

1. Pre-Activity Written Agreements (ICH E6(R3) Section 3.6.1): ICH E6(R3) Section 3.6.1 says written agreements with an adjudication committee should be documented before the activities begin. The essential-records table includes a signed agreement between the sponsor and independent data monitoring committee and/or adjudication committee members. The guideline does not prescribe a compensation clause. The file needs to show who was appointed, that the agreement predates the first review, and that conflicts of interest are managed under Section 3.9.9.

2. Charter Version Control and Protocol Alignment: A documented history of the adjudication charter with formal approval signatures. If protocol amendments modify endpoint definitions or diagnostic criteria, the charter must undergo a synchronized, version-controlled amendment prior to evaluating events occurring under the amended protocol.

3. Member Qualification and Training Logs: Section 3.9.9 asks for relevant expertise. It does not require a named set of board certificates or medical licenses. A practical file includes a curriculum vitae and a record that the member was trained on the protocol definitions before reviewing events, plus a conflict-of-interest declaration refreshed on the schedule the charter sets. The TMF Reference Model files member curricula vitae and training evidence under artifact 01.03.04. On-study rater monitoring is a different control, described in ClinRO rater training and drift.

4. Individual Event Dossiers and Contemporaneous Decision Records: For every reported presumptive event: the original site reporting CRF, the endpoint office completeness verification checklist, tracked RFI query trails, the redacted case packet issued to reviewers, individual electronic scoring forms completed independently by primary reviewers, and consensus conference minutes or third-party adjudicator rulings for discordant evaluations.

5. Electronic Audit Trails and Database Reconciliation: System audit logs showing exact dates and timestamps for packet compilation, reviewer access, initial vote entry, consensus resolution, and the transmission of locked adjudication records into the clinical database. Reconciled data feeds delivered to the DMC must match the locked database state.

In the CDISC TMF Reference Model, trial management is Zone 01, and trial-committee records are section 01.03. The charter files with committee process (artifact 01.03.01, whose recommended sub-artifacts include the committee charter). Minutes, correspondence, and the data package file with committee output (artifact 01.03.03). Member contracts and curricula vitae have their own artifacts in the same section (01.03.06 and 01.03.04). Zone 04 is IRB or IEC and other approvals; section 04.02 is other approvals, not the committee file. How this site maps endpoint decisions across zones is covered in TMF Reference Model zones. An inspector who selects one adjudicated event from the clinical study report should be able to follow that event from the site report through redaction, the two independent reviews, any consensus step, and the locked record in the clinical database. A break in that chain, such as an undocumented phone call between reviewers, leaves the endpoint unreconstructable.

Before enrolling the first trial participant, study teams should audit their adjudication readiness against the following comprehensive operational checklist. Each domain maps directly to regulatory provisions and practitioner consensus, identifying the critical data failure prevented.

Clinical endpoint adjudication is a way of turning heterogeneous hospital records into events the protocol can count. Locking the charter, the case-packet rules, and the decision record before the first participant is enrolled keeps those events classifiable under the protocol, blinded to treatment assignment, and reconstructable at inspection.