IDE Software Device Description for the Frozen Baseline¶
Updated: 2026-09-08
| Field | Value |
|---|---|
| Device software | BionicLoop investigational automated insulin-delivery controller |
| Baseline | ide-software-freeze-2026-08-21 / 91c0e98a9bc9429a0486bebdebffc7d8dbbe300e |
| Last shipping-product commit | ae00e754 |
| Status | Frozen engineering baseline; exact-freeze execution and final software-package approval complete |
FDA Documentation Level¶
BionicLoop is documented at the Enhanced Documentation Level described in FDA's Content of Premarket Submissions for Device Software Functions (June 2023). A software failure or latent defect could contribute to incorrect or delayed insulin delivery and serious injury. The determination is based on the potential consequence of software failure, not on the investigational status or the current study size. The controlled content crosswalk is Technical Appendix A00.
Intended Function¶
BionicLoop is an investigational iPhone application that receives glucose data from a Dexcom G7, hosts the sponsor-controlled Algo2015 dosing logic, and commands insulin delivery through an Omnipod DASH. It provides participant and study-staff workflows for meals, fingerstick BG, device status, alerts, Clinical Settings, temporary targets, insulin suspension/resumption, and study record recovery.
The app performs algorithm-driven dosing only. It does not provide a manual test-dose path.
Major Software Components¶
| Component | Function |
|---|---|
| BionicLoop app | User interface, application lifecycle, study configuration, alert presentation, and integration composition |
| BionicLoopCore | Runtime policy, state, persistence, safety controls, telemetry models, and algorithm host |
| Algo2015 / Algo2015Safety | Two symbol-isolated builds of the same unmodified dosing source for primary and safety-controller tracks |
| G7SensorKit / G7SensorKitUI | Dexcom G7 acquisition and study setup/settings surfaces |
| OmniBLE | Omnipod DASH communication, status, and delivery commands |
| LoopKit | Shared device/data support used by the app and integrations |
| Local evidence stores | Protected telemetry outbox, complete active-session step archive, alert history, and clinical-gated recovery export |
Principal Data and Control Flow¶
- The first algorithm step requires a Dexcom G7 sample. After that anchor, the app accepts a fresh Dexcom G7 sample or an eligible validated fingerstick BG for the due algorithm step.
- Current pump status and any accepted delivery feedback are assembled with glucose, meal, and persisted algorithm state.
- The primary and safety-controller instances execute independently.
- Runtime policy evaluates output, pump availability, suspension, fallback, forced-open, reconciliation, and session state.
- A permitted command is sent to the active Pod and its delivery status is retained for next-step reconciliation.
- Inputs, outputs, commands, status, alerts, and recovery evidence are stored locally and may be uploaded or exported according to the approved data scope.
Safety-Relevant Behavior¶
- Pump commands are blocked or deferred when current state does not support a safe command.
- CGM loss invokes algorithm-native degradation, participant guidance, BG-specific operation, and programmed backup-basal behavior as applicable.
- The app programs and masks backup basal during connected operation and reconciles bounded exposure when the mask expires or is released.
- Interrupted automatic boluses retain request-step identity and are consumed exactly once using confirmed, corrected, canceled, or explicitly labeled assumed evidence.
- Pod and sensor identity checks prevent cross-device reconciliation or staleness-based adoption.
- Clinical configuration is open only before the first usable save; later changes require a subject-scoped single-use offline unlock.
- Pregnancy Temporary Target affects the primary controller for a bounded period while the permanent safety/fallback basis remains unchanged.
- Insulin suspension remains active until manual resume and prevents meal execution while suspended.
External Interfaces¶
- Bluetooth Low Energy interfaces to Dexcom G7 and Omnipod DASH
- iOS notifications and app lifecycle services
- supportive/optional authenticated telemetry and study-support interfaces to BionicScout; local dosing does not depend on Scout availability
- clinical-gated local recovery export
Development and Verification References¶
| Topic | Controlled reference |
|---|---|
| Documentation level | Enhanced Documentation Level and FDA Content Crosswalk |
| Requirements | Software Requirements Specification, version 1.82 |
| Design | Software Design Description, version 1.84 |
| Risk | Risk Analysis, version 1.53 |
| Verification plan | Software Verification and Validation Plan, version 1.91 |
| Traceability | Requirements Traceability Matrix, version 1.86 |
| Cybersecurity | Cybersecurity Plan, version 1.03 |
| Execution | IDE Freeze Execution Report - 2026-08-21 |
| Labeling | IFU-BL-001, revision 1.17; approved for IDE submission 2026-09-04 |
Baseline and Approval Status¶
Exact-freeze algorithm, automated app, simulation, and scoped local cybersecurity execution is complete. The 24 IDE-scope risk-analysis rows have execution evidence mapped. Build 843 and IFU-BL-001 revision 1.17 are approved for IDE submission, and the final protocol, consent, and device labeling are software-consistent. Evidence, residual-risk, cybersecurity, and final software-package approvals are recorded in the Decision Register. The supportive Scout role, official study sources, physical-use claim boundary, and human-factors evidence description were clinically approved on 2026-08-27. The controlled multi-G7 exercise remains a predeployment condition; final IDE assembly remains under study-team control.