Solutions · Protocol and eligibility
Every criterion cites its source. Nothing is approved without a person.
For protocol reviewers, research nurses and investigators who turn eligibility sections into checklists — and redo the work at every amendment.
The protocol stays the source of truth.
TrialBridge keeps the uploaded document, extracts its text with page and section anchors, and links everything it structures back to those anchors.
Upload
Born-digital PDF or DOCX. Files are scanned and stored encrypted. Scanned-image protocols need OCR — planned, not in the first release.
Extract with anchors
Text is kept with page numbers, section headings and character positions so any criterion can point to its exact passage.
Locate eligibility sections
Inclusion, exclusion and synopsis sections are located. If they can’t be found, the reviewer selects the page range.
Review proposals beside the text they came from.
Select a criterion to highlight its passage. A proposal that can’t cite a passage that exists verbatim in the extracted text is rejected before anyone sees it.
§5.1(a) · p.18Participants must be between 40 and 75 years of age, inclusive, at the time of screening.
§5.1(c) · p.18Estimated glomerular filtration rate (eGFR) between 30 and 59 mL/min/1.73 m², based on the most recent value within 90 days before screening.
§5.1(d) · p.18HbA1c of at least 7.0% and no greater than 10.5%, based on the most recent value within 90 days before screening.
Synopsis · p.3Key inclusion: adults aged 40–75 with type 2 diabetes, eGFR 30–59, and HbA1c 7.0–10.0%.
§5.2(b) · p.19Participants with any history of chronic dialysis or kidney transplantation will be excluded.
§5.2(c) · p.19Heart failure classified as New York Heart Association (NYHA) class IV.
§5.1(f) · p.18Able and willing to provide written informed consent and to comply with study procedures, in the opinion of the investigator.
§5.1(d) · p.18HbA1c of at least 7.0% and no greater than 10.5%, based on the most recent value within 90 days before screening.
- Analyte
- HbA1c
- Operator
- Between (inclusive)
- Range
- 7.0–10.5 %
- Value used
- Most recent
- Window
- 90 days before run date
- Review status
- Pending professional approval
Passages disagree: the synopsis (p.3) says ≤ 10.0 %; §5.1(d) (p.18) says ≤ 10.5 %. A reviewer must choose which applies.
Typed conditions, with explicit units and windows.
Not every protocol requirement can be checked against data. Anything that needs judgement — or data the dataset doesn’t hold — is kept as a screener-assessed criterion rather than forced into logic.
| Type | Represents | Example (synthetic protocol) |
|---|---|---|
| AGE | Age range at a reference date | Age 40–75 years at screening |
| CONDITION | Diagnosis present or absent, from a value set | Type 2 diabetes mellitus, ever |
| LAB | Laboratory value, operator, unit, look-back window | eGFR 30–59 mL/min/1.73 m², ≤ 90 days |
| MEDICATION | Exposure to an ingredient or class over a period | Glucose-lowering therapy ≥ 56 days |
| PROCEDURE | Procedure history from a value set | Dialysis or kidney transplant, ever |
| MEASUREMENT | Clinical measurement or score | NYHA class IV |
| TEMPORAL | Time since or before an event | ≥ 30 days since last investigational product |
| GROUP | ALL / ANY of several conditions (one level) | eGFR 30–59 AND (UACR ≥ 30 OR ACE inhibitor) |
| SCREENER | Needs judgement or data not in the dataset | Able to provide written informed consent |
If no value falls in the window, the result is Unknown — it is never treated as “does not meet”.
Plain-language preview
Include if the most recent eGFR within 90 days before the run date is between 30 and 59 mL/min/1.73 m², inclusive.
- Type
- LAB
- Computability
- Data-checkable
- Source
- Required before approval
- Status
- Draft · not approved
Problems are flagged before approval, not after a cohort run.
A version can’t be approved while any flag is unresolved.
| Flag | Example | Resolution |
|---|---|---|
| Passages disagree | Synopsis says HbA1c ≤ 10.0 %; §5.1(d) says ≤ 10.5 % | Reviewer chooses which applies and records why |
| Ambiguous term | “Stable dose” without a duration | Define the parameters, or mark Screener-assessed |
| Impossible range | Minimum greater than maximum | Correct the values |
| Requires judgement | “In the opinion of the investigator” | Automatically set to Screener-assessed |
| Unmapped concept | A condition with no code set | Map a value set, or mark Screener-assessed |
Amendments become diffs, not rewrites.
An amendment starts a new draft copied from the approved version. The diff shows what was added, removed or changed; approved versions can’t be edited, and earlier cohort runs stay pinned to the version they used.
- I1Age 40–75 years at screeningUnchanged
- I3eGFR 30–59 mL/min/1.73 m², most recent ≤ 90 dUnchanged
- I4HbA1c 7.0–
10.0 %10.5 %Changed - E4ALT > 3 × ULN, most recent ≤ 90 dAdded
Cohort runs and screening records made under v1.0 stay pinned to v1.0. Approving v1.1 flags them for review — it never changes their status automatically.
What the AI does — and what it never does.
AI assistance is optional per workspace. Without it, the same studio works with manually entered criteria.
MVP scope Criteria proposals are planned for the first release, behind a per-workspace setting.
The AI
- Reads protocol text
- Proposes typed criteria
- Cites a verified source passage
- Flags conflicts it finds
The AI never
- Sees patient data or evidence
- Approves a criteria version
- Ranks or screens candidates
- Shows a confidence score as certainty
Bring one of your own protocols to the conversation.
We’ll walk through how its eligibility section would be structured, which criteria would stay with screeners, and what an amendment would look like.