DRAFT FOR REVIEW BY APPOINTED CLINICAL SAFETY OFFICER — MUST BE REVIEWED, AMENDED AND SIGNED BY A GMC/NMC/HCPC-REGISTERED CLINICIAN WITH FORMAL DCB0129 CLINICAL SAFETY OFFICER TRAINING BEFORE USE.
CSO06 — Clinical Safety Case Report (CSCR): Trial Concierge
Document reference: BRITI-CSCR-TRIAL-001 Version: 0.9 (Draft for CSO review) Manufacturer: BritiAI Limited Applicable standard: DCB0129:2018 Status: Draft pending CSO sign-off
1. Executive Summary
Trial Concierge is an agentic clinical trial recruitment and protocol Q&A assistant. It supports research teams and clinicians involved in trial recruitment by surfacing potentially eligible patients against active trial protocols and by answering protocol questions with citations. Trial Concierge does not enrol patients. Every eligibility determination requires sign-off by a registered clinician with appropriate authority within the trial’s delegation log. Patient-facing communications are out of scope.
This CSCR documents intended use, environment, claims, hazards (CSO07, TRI-01 to TRI-10), and the basis on which release is proposed subject to CSO sign-off and trust DCB0160 assurance, with explicit reference to Good Clinical Practice (GCP) obligations that remain with the trial sponsor and Principal Investigator (PI).
2. Solution Description
- Protocol ingestion. Ingests trial protocols, eligibility criteria, schedules of assessment and amendments with version tracking.
- Patient-screening agent. Operates over the trust EHR (read-only) to produce a candidate list of patients who may meet eligibility criteria, with explicit per-criterion evidence and confidence flags.
- Protocol Q&A. Answers clinician questions about protocol content with citations to the protocol section and version.
- Clinician sign-off gate. Every candidate patient must be reviewed by a delegated clinician before any onward action (e.g. invitation to discuss the trial).
- Audit trail. Records every screening query, every candidate determination, every clinician decision and every protocol Q&A response.
3. Intended Use
Trial Concierge is intended to support clinical research teams in trial recruitment by reducing manual screening burden and improving protocol comprehension. It is not intended to:
- Enrol patients in trials.
- Communicate with patients directly.
- Replace the PI’s responsibility for eligibility determination.
- Operate over trial protocols not formally ingested and version-tracked.
- Provide eligibility determinations for trials with regulatory requirements that prohibit AI-assisted screening absent specific approval.
4. Intended Users
- Registered clinicians delegated by the PI for screening and recruitment under the trial’s delegation log.
- Research nurses and trial coordinators authorised by the PI.
End patients do not interact with Trial Concierge.
5. Environment of Use
Trust-managed devices within the trust network. Use is constrained to trials for which the trust has confirmed governance approval (HRA, REC, trust R&D) and for which protocol ingestion has been completed under formal version control.
6. Clinical Claims
BritiAI claims that Trial Concierge:
- Produces a candidate list of patients who may meet ingested protocol eligibility criteria, with per-criterion evidence and confidence flagging.
- Answers protocol Q&A queries with citations to the ingested protocol version.
- Maintains a complete audit trail of screening and Q&A activity.
Explicit non-claims
Trial Concierge does not claim to:
- Be a medical device.
- Determine eligibility absent clinician sign-off.
- Communicate with patients.
- Replace the PI’s responsibilities under GCP or trust R&D governance.
- Reflect protocol amendments not yet ingested.
7. Risk Envelope
Principal hazard categories (CSO07, TRI-01 to TRI-10):
- Incorrect eligibility flagging — false positives leading to inappropriate approach to a patient, or false negatives leading to missed recruitment opportunity (with possible loss of access to investigational therapy).
- Protocol misinterpretation — Q&A response misrepresenting protocol content.
- Stale protocol — screening or Q&A against a superseded protocol version.
- EHR data drift — screening against incomplete or out-of-date EHR data.
- Privacy escalation — surfacing patient data to staff outside the trial delegation log.
- Automation bias in clinician sign-off of candidate lists.
- Mis-attribution of EHR evidence to the wrong patient.
- Cross-trial contamination — evidence or criteria from one trial influencing another.
- Loss of audit trail affecting later inspection or investigation.
- Drift in screening or Q&A behaviour following model update.
8. Risk Control Strategy
- Elimination by design. No patient-facing communication. No enrolment action. Read-only EHR access scoped by trial.
- Reduction by design. Per-criterion evidence panel for every candidate. Hard version pinning of ingested protocols with explicit warnings on supersession. Role-based access enforced against the trial delegation log. Hard separation of trial-specific contexts.
- Protective measures. Mandatory clinician sign-off gate; confidence flagging with conservative thresholds; on-screen banner stating “Candidate only — requires delegated clinician review against current protocol”.
- Information for safety. Mandatory training for trial teams; clear release notes; documented known limitations including GCP-specific considerations.
9. Residual Risks Summary
Subject to CSO judgement:
- Incorrect eligibility flagging carries the most consequential residual risk. False positives are mitigated by the clinician gate. False negatives are mitigated by transparency on the evidence used and the option for clinicians to broaden the search; however, no automated screen replaces a clinician’s manual review. Residual risk amber.
- Protocol misinterpretation is mitigated by citation-anchored responses and version pinning; residual risk amber.
- Privacy escalation is mitigated by delegation-log enforcement; residual risk amber.
10. Assumptions and Dependencies
- The trust R&D office maintains current trial delegation logs and ensures Trial Concierge access is consistent with them.
- Protocol amendments are ingested promptly via the documented process.
- EHR access is governed by appropriate data sharing and information governance agreements specific to each trial.
- The trust CSO performs DCB0160 assessment reflecting GCP and trial governance context.
11. Clinical Safety Verification
- Functional testing against safety-critical requirements.
- Screening evaluation against curated synthetic cohorts with known eligibility ground truth.
- Q&A evaluation including citation-grounding and version-awareness probes.
- Penetration and access-control testing including delegation-log enforcement.
12. Post-Deployment Monitoring
- Per-trial false positive and false negative rates against clinician decisions.
- Q&A citation-grounding rate.
- Frequency of supersession warnings and user response.
- Any incident, near-miss or complaint relating to trial recruitment supported by Trial Concierge.
A safety review per deploying trust per trial is conducted at three months post go-live and annually thereafter.
13. Change Control
Material changes triggering re-assessment include: change to screening model or Q&A model; change to protocol ingestion schema; change to delegation-log enforcement; introduction of new trial classes (e.g. paediatric, mental capacity-relevant, advanced therapy medicinal products).
14. Statement of Conformance
Subject to CSO review and sign-off, BritiAI confirms that the clinical risk management activities undertaken for Trial Concierge have been performed in accordance with DCB0129:2018, and that the solution operates within the boundary of supporting (never replacing) GCP-required PI responsibilities.
Linked artefacts: CSO01, CSO07 (TRI-01 to TRI-10), CSO08, BRITI-CRMP-TRIAL-001, BRITI-IFU-TRIAL-001.
CSO name: _________________________ Registration body and number: _________________________ Signature: _________________________ Date: _________________________