Clinical Documents - Software Consistency Crosswalk¶
| Field | Value |
|---|---|
| Scope | Software behavior and software evidence wording only |
| Software baseline | ide-software-freeze-2026-08-21 / 91c0e98a |
| Designated submission build | BionicLoop 1.0 (843) / 1455cda7 |
| Status | Final protocol, consent, and synchronized IFU 1.17 device label verified; not an FDA submission attachment |
| Date | 2026-09-08 |
Reviewed Source Identity¶
The source DOCX files are review inputs and are not published on this site.
| Source | Identity used for review | SHA-256 |
|---|---|---|
| Protocol | Clinical Protocol BLS Clean-final.docx; Optimizing Automated Insulin Delivery to Meet Pregnancy Glycemic Targets, Version 1.0, version date 1 September 2026, 41 pages |
4bfb09542230c94c00dc5b5331a969657b5961b1c4bbd5a2759519a00f71d264 |
| Informed consent | Consent Form BLS Clean -final.docx; consent version date 1 September 2026, 19 pages |
6a0e1759997944522fa5dd9ba8dd9e9928cccb64ed7e6e4455b65f40c56a22cf |
| Device labeling | 008_Device_Labeling.pdf; BionicLoop 1.0 Build 843 and IFU-BL-001 revision 1.17 dated 2026-09-04, 3 pages |
561c4cd26171888454c27d2350162680a867190754a448fd4096d1dd8b10fe0c |
Final Document Review¶
Software review verified the final protocol's material descriptions of the distinct fingerstick mode, CGM-interruption meal gate, Standard/Pregnancy and Temporary Targets, Pregnancy meal setting, supported weight range, Same-Participant Reset, G7 startup, official study-data sources, and software-evidence boundary.
The final consent provides the concise participant-facing fingerstick
disclosure; the protocol and IFU carry the detailed backup-basal and meal-gate
instructions. The synchronized device labeling identifies BionicLoop 1.0 Build
843 and IFU-BL-001 revision 1.17 dated 2026-09-04. No further
software-consistency action is required for the protocol, consent, or device
label. The detailed review is retained in
Internal/IDE_Clinical_Document_Final_Candidate_Review_2026-09-03.md.
Historical September 2 Returned Drafts¶
The study team returned revised Word drafts and a cover-letter PDF in the controlled Dropbox folder. The Software Developer acquired them read-only, verified their hashes, inspected revision/comment state, and rendered all pages. The protocol, consent, and labeling remain active redlines and are not clean controlled copies.
| Returned draft | Review identity | SHA-256 |
|---|---|---|
| Protocol | 43 pages; 261 insertions, 176 deletions, 42 unresolved comments | 55b6b4650099e38f82b6ca264f62b4b24a827608ac0a590b12a01b6694f982a3 |
| Informed consent | 19 pages; 38 insertions, 24 deletions, 8 unresolved comments | 07fc8797d40eaf23661e28b1703ccf765d3f27b5442a1370f4b04d0cc4c42f1b |
| Device labeling | 4 pages; 80 insertions, 46 deletions, 9 unresolved comments | f05c62eb0dca9d00fc73c5a43b5ae160144656389a2c57cfe831e6636e072a7c |
| Cover letter | 2-page PDF dated 2026-08-31 | 0c84870e930f6f6cc00a5667b45a2c59cb099ccaa31744aedb0af86e28cb00ec |
The detailed returned-draft findings are retained in the internal controlled
record
IDE_Clinical_Document_Returned_Draft_Review_2026-09-02.md. No Dropbox source
was modified.
Review Summary¶
| ID | Topic | Documents affected | Current disposition |
|---|---|---|---|
SW-01 |
Fingerstick BG is a distinct safeguarded input mode | Protocol, consent | Final document set verified |
SW-02 |
Meal availability during CGM interruption | Protocol, consent, IFU | Final document set verified |
SW-03 |
Standard, Pregnancy, and Temporary Targets | Protocol, consent, labeling | Final document set verified |
SW-04 |
Standard-to-Pregnancy meal setting | Protocol, labeling | Final document set verified |
SW-05 |
Supported participant-weight range | Protocol | Final protocol verified |
SW-06 |
Same-Participant Reset for Phase 3 | Protocol, consent | Final document set verified |
SW-07 |
G7 startup/replacement terminology | Protocol, consent, IFU | Final document set verified |
SW-08 |
Official study-data sources | Protocol and data plan | Final protocol hierarchy verified |
SW-09 |
Software testing claim | Protocol, labeling | Final documents match the controlled evidence boundary |
SW-10 |
Participant-facing CGM-outage disclosure | Consent, protocol, IFU, and training | Final document set verified |
SW-11 |
Software build and IFU labeling | Device labeling | Build 843 / IFU 1.17 synchronized label verified under D14 |
SW-01 - Fingerstick BG During CGM Interruption¶
Historical document issue: The June protocol said entered BG values were treated "just as if they were Dexcom CGM values."
Frozen software behavior: A confirmed fingerstick value enters a distinct BG-based closed-loop path with BG-specific timing, validity, alert, transition, and exactly-once consumption controls. It is not stored or processed as an ordinary Dexcom sample.
Accepted protocol wording:
If Dexcom CGM data are unavailable, a participant may enter and confirm a fingerstick BG. BionicLoop processes the value through a distinct BG-based closed-loop mode with BG-specific timing, validity, alert, and transition safeguards. The value is available to the next eligible algorithm step, is consumed once, and may result in automated insulin delivery. The Dexcom G7 should be restored or replaced as soon as possible.
SW-02 - Meal Availability During CGM Interruption¶
Historical document issue: The June protocol said meal announcements could be managed "as per usual" while the CGM is offline.
Frozen software behavior: A meal announcement is accepted only when runtime preflight can execute the associated algorithm step. A pending qualifying fingerstick may be consumed by that meal-triggered step. If current glucose or another safety precondition does not permit the step, the app states that the meal was not announced and does not command meal insulin.
Accepted protocol wording:
During a CGM interruption, a meal announcement may be made only when BionicLoop has a qualifying current glucose input and can execute the associated algorithm step. A confirmed fingerstick BG may provide that input. If the requirements are not met, the app tells the participant that the meal was not announced and no meal insulin was delivered.
SW-03 - Glucose Targets¶
Historical document issue: The earlier protocol described permanent and recurring targets but does not describe the implemented participant-facing Temporary Target.
Frozen software behavior:
- Standard configuration exposes permanent targets of 110, 120, or 130 mg/dL.
- Pregnancy configuration exposes permanent targets of 90, 100, or 110 mg/dL.
- Pregnancy configuration allows a participant to choose a Temporary Target of 120 or 130 mg/dL for 30, 60, 90, or 120 minutes.
- Temporary Target returns automatically to the permanent target, including after relaunch.
- Temporary Target changes the primary-controller target only. The safety target and programmed backup-basal schedule remain based on the permanent Pregnancy configuration.
Accepted protocol wording:
The Standard configuration provides permanent glucose targets of 110, 120, or 130 mg/dL. The Pregnancy configuration provides permanent targets of 90, 100, or 110 mg/dL. In Pregnancy configuration, a participant may select a Temporary Target of 120 or 130 mg/dL for 30, 60, 90, or 120 minutes. The app automatically returns to the permanent target when the selected time ends. Temporary Target does not change the permanent Pregnancy safety target or backup-basal schedule.
SW-04 - Pregnancy Meal Setting¶
Historical document issue: The earlier documents described 75% and 90% meal settings but do not state when the investigational app changes the setting.
Frozen software behavior: Changing a draft clinical configuration from Standard to Pregnancy proposes 90%. Clinical staff may select 75% before saving. The stored setting does not change until the reviewed configuration is saved.
Accepted protocol/labeling wording:
When clinical staff change a configuration from Standard to Pregnancy, the reviewed draft defaults to a 90% meal setting. Clinical staff may select 75% before saving. The active configuration changes only after the reviewed settings are saved.
SW-05 - Supported Participant-Weight Range¶
Historical document issue: The earlier protocol said the algorithm could be initiated up to 255 kg (561 lb).
Frozen software behavior: BionicLoop accepts participant weights from 50 through 500 lb inclusive (22.7 through 226.8 kg). Entries outside that range prevent the algorithm from starting. The algorithm applies an additional final safety range of 20 through 230 kg.
Accepted protocol wording:
BionicLoop accepts participant weights from 50 through 500 pounds inclusive (22.7 through 226.8 kg). Entries outside this range prevent the algorithm from starting. The algorithm also applies an internal final safety range of 20 through 230 kg.
SW-06 - Phase 3 Reset¶
Historical document issue: The earlier protocol and consent called the Visit 10 action a "factory reset" and say the system memory is erased.
Frozen software behavior: The appropriate same-person workflow is Same-Participant Reset. It clears learned algorithm state and prior algorithm-session history while preserving the participant identity, weight, target configuration, meal setting, and insulin-action setting. New Participant Reset is a separate phone-reassignment workflow that erases participant-scoped data and must not be used for the Phase 3 transition.
Accepted protocol/consent wording:
Before Period 3, study staff will perform a Same-Participant Reset. This clears the algorithm's learned state and prior algorithm session while preserving the participant identity and clinical configuration. New Participant Reset is reserved for preparing a study phone for a different participant and will not be used for this visit.
SW-07 - G7 Startup and Replacement¶
Historical document issue: One earlier protocol phrase referred to a replacement G7 as having been "calibrated."
Frozen software behavior: BionicLoop waits for the intended replacement sensor to finish starting and provide usable readings. Routine calibration is not the software gate.
Accepted wording: Replace "before the new sensor has been calibrated" with "before the intended replacement sensor has finished starting and is providing readings."
SW-08 - Official Study-Data Sources¶
Historical document issue: The earlier protocol said CGM data were automatically stored in the study app and downloaded from it, without identifying the approved software source hierarchy.
Approved software/data boundary:
- Dexcom Clarity is the official source for CGM outcomes.
- The BionicLoop exported CSV is the official source for insulin-delivery outcomes.
- The electronic case-report form is the official source for adverse events and other safety information.
- The BionicLoop recovery ZIP and Scout telemetry support reconstruction, monitoring, troubleshooting, and corroboration; they are not the official outcome or safety-report sources.
The data-management plan should distinguish the glucometer measurement record from the fingerstick value entered into BionicLoop and used by the software.
SW-09 - Software Testing Claim¶
Historical document issue: The earlier protocol and labeling said the app underwent "risk-based software verification and validation" under simulated and representative-use conditions.
Controlled evidence boundary: The package contains risk-based software verification, simulation, UI testing, engineering representative-use testing, and working real-device evidence. It does not claim a separate formal hardware validation study or summative participant human-factors validation study.
Accepted protocol/labeling wording:
BionicLoop has undergone risk-based software verification and engineering representative-use testing under simulated and real-device conditions to support its intended investigational use.
SW-10 - Participant-Facing CGM-Outage Disclosure¶
Final document treatment: The consent states that a confirmed fingerstick may be used once at the next eligible dosing step with additional safeguards. The protocol and IFU provide the detailed backup-basal and meal-availability instructions during CGM interruption.
Detailed IFU/training wording:
If sensor readings are unavailable, the app may ask you to enter and confirm a fingerstick blood-glucose value. The app may use that confirmed value to calculate and deliver automated insulin with additional safeguards. Backup basal insulin may continue while current sensor information is unavailable. A meal announcement may be temporarily unavailable until the app has an acceptable fingerstick or sensor reading. Follow the app instructions and restore the sensor as soon as possible.
SW-11 - Software Labeling Identity¶
Historical document issue: Earlier supplied labeling did not identify the designated study build or approved IFU revision.
Verified labeling identity: Build 843 was approved on 2026-09-03 and IFU
revision 1.17 was approved on 2026-09-04. The synchronized
008_Device_Labeling.pdf identifies BionicLoop 1.0 Build 843 and IFU revision
1.17 and retains:
- the exact BionicLoop version/build and environment approved for study use
- the approved IFU document identifier and revision
- the study-phone/device label, as applicable
- the in-app investigational-use banner specimen
- software-relevant warnings and precautions or a controlled cross-reference to the IFU
Disposition¶
Camille Powe, MD approved implementation of all eight recommended change groups on 2026-08-31. The final protocol, consent, and synchronized Build 843 / IFU 1.17 device labeling were reviewed as documented above. Software consistency is complete for the material protocol behaviors, participant-facing disclosures, and software-label identity.