Protocol-aware adverse event attribution, documents that inform grading, and a patient app

What shipped in the week of April 6, 2026

Burna AI added protocol-aware adverse event attribution this week: a coordinator can enter the adverse events a protocol expects for each drug in a regimen, and the multi-drug attribution workflow weighs every event against those expectations. Uploaded protocols, investigator brochures, drug labels and institutional guidelines began informing CTCAE grading, EHR settings moved to the organization, and the mobile app trial patients use to report symptoms between visits was completed.

Coordinators on combination-therapy trials gain the most: the platform now starts from the same reference documents they do.

Protocol-aware adverse event attribution

In a combination regimen, whether an event is expected for a given drug is central to deciding which drug caused it. Attribution is also hard to hold steady. In the placebo arms of two phase III trials, when the same adverse event was reported in the same patient at more than one visit, its attribution category changed at least once 31% and 36% of the time (Hillman et al., Journal of Clinical Oncology, 2010).

Coordinators can now record a protocol's expected adverse event profile by hand. For each drug in the regimen, the profile holds:

• The expected adverse event terms, entered with CTCAE term auto-complete
• How often each event is expected, from very common to rare
• The maximum grade expected for each event

When a profile is active, the multi-drug attribution workflow receives it alongside the clinical note, so the WHO-UMC and Kramer attribution for each suspect drug starts from what this protocol expects rather than from general drug safety profiles alone. Profiles are optional; attribution runs without one.

Every change to a profile is written to an audit log at field level: the trial identifier, sponsor and status, each drug, and each expected event, with who changed it and when. Profiles are scoped to the organization that created them.

Since then (September 2026): protocol reading for research sites has its own product, Protocol Studio. For a protocol whose drug profile is confidential sponsor information, see how protocol-aware attribution handles sponsor IP.

Investigator brochure upload: documents that inform CTCAE grading

An organization can now upload the documents its clinicians grade from: protocols, investigator brochures, drug labels and institutional guidelines, across nine supported document types. Each upload runs through a background pipeline:

1. Parse the document and classify its type.
2. Extract drug names and CTCAE terms.
3. Split the text into passages and index them for semantic search.
4. Store the index for retrieval during grading.

During a CTCAE grading run, a dedicated step searches three categories in parallel: protocol passages, drug safety information and institutional guidelines. It keeps the passages that pass a relevance threshold and match an allowed document type, and passes them to the grading judgment, so the model reasons from the same reference material the clinician uses.

• Grading does not depend on it. With no documents uploaded, or if retrieval fails, grading runs as before.
• Organization-scoped. Documents and their search index belong to one organization, and deleting a document removes its index entries.
• Visible status. Each document shows whether processing succeeded, is in progress or failed.

Since then (August 2026): the document library lives in Files as a Documents category, with a Retrieval column showing whether each document is indexed, indexing, failed or not indexed.

Organization EHR settings: one endpoint for the whole team

EHR connection settings moved from each user's account to the organization.

• One endpoint per EHR. An admin chooses the organization's Epic or Oracle Health FHIR endpoint from a searchable list of published endpoints, and every member connects through it. For Oracle Health, that is the Oracle Health connection added in Week 14.
• Clean reconnection. When an admin changes an endpoint, every member's connection to that EHR is disconnected, so each person reconnects against the new one. Only admins and owners can make that change.
• Discovery at request time. Token endpoints now come from each EHR's published SMART configuration when a request is made, with a configured endpoint as the fallback, so a change on the vendor's side is picked up without a code change.

The same change narrowed the EHR list to Epic and Oracle Health.

A mobile app for trial patients

The mobile app for patients enrolled in a trial was completed this week, so they can report symptoms between clinic visits:

• Sign in with an email address or phone number and a one-time passcode
• Enroll in a trial by entering the token the site provides
• See trial details and recent diary entries on the home screen
• Keep a symptom diary, with history to look back on
• Complete check-ins covering pre-visit questions, symptoms and safety questions

When a safety answer is flagged, the app lists the flagged responses and puts a button to call the study site's urgent line on screen, along with an emergency call option. Symptoms recorded as they happen, in the patient's own words, give the site more to work from than recall at the next visit.

Since then (August 2026): the trial participant app is Healthnotes, which a sponsor deploys through its sites. The PRO token a site issues on app.burna.ai is the participant's login.

Smaller changes

• The web app's search, patient queries and OTP sign-in redirect were made more reliable on a fresh sign-in.
• Drug attribution shows confidence in a clearer format, and a run's full output reads as labeled fields instead of raw JSON.
• PDF downloads now start reliably in the browser.
• The provider mobile app's settings, policy screens, home screen and patient edit form were updated.

To see protocol-aware attribution on a combination regimen, book a 15-minute conversation.

The AI platform for accelerated clinical research, from first-in-human trials through postmarket surveillance. Burna AI performs automated adverse event grading and attribution in real time, providing clear rationale and audit-ready evidence, including generation of IRB, sponsor and FDA reports. Burna AI delivers CTCAE v5 and v6 grades and MedDRA LLT codes on every encounter; enabling standardization, cross-site consistency and accuracy of clinical research data. AI suggests, clinicians decide. Human-in-the-loop, always.