Skip to content
type1.science

Diabeloop DBLG2

Diabeloop SA

Next-generation, smartphone-based Diabeloop control algorithm (FDA-cleared Dec 2025 as an interoperable automated glycemic controller; CE-marked Aug 2025). It builds on the DBLG1 self-learning algorithm — which raised time-in-range ~9 points and roughly halved time-low in trials — and is being integrated into Sequel's twiist pump as a second US algorithm option. Not yet shipping in any commercial product as of mid-2026; US launch targeted for 2027.

On the horizonStrong evidencehybrid-closed-loopphone-controltubedinteroperableself-learningsingle-hormoneOfficial site ↗

The scorecard

Time in range76

Achieved ~72% median TIR (70-180) in the largest real-world DBLG1 cohort (n=3,706 German adults), with supporting cohorts at ~70-72%; this is the same algorithm DBLG2 runs but on a phone, with no DBLG2-specific data, so it does not beat the 780G's ~76% real-world anchor.[1]

Hypo protection80

Achieved ~0.9% median time-below-70 (time-below-54 ~0.1%) in the 3,706-user German T1D cohort, and ~1.3% in the pooled 6,859-patient analysis — genuinely class-leading low-glucose protection.[1]

Automation level78

Auto-correction boluses for highs without user action; DBLG2 keeps controlling glucose even when meals go undeclared (real-world ~9% of days), though labelling still advises announcing meals for optimal control (3-6 point TIR cost if skipped).

Average glucose76

Achieved median GMI ~7.0% (HbA1c ~6.95-7.1% in smaller cohorts) in real-world DBLG1 T1D users — consistent with ~72% TIR and tracking a few points below the 780G's lower mean glucose.[1]

Low variability72

Achieved CV ~29.6% (in target) in the Spanish real-world cohort (n=62), with GMI 7.0% in the large German cohort indicating well-controlled variability for a good hybrid loop, though no DBLG2-specific CV is published.[3]

Exercise handling80

Announced-activity mode cuts insulin and recommends preventive carbs; in the WP7 physical-activity post-hoc, time-below-70 on activity days (2.0%) matched non-activity days (2.2%), preventing exercise-induced lows. Automatic activity detection is still in development.

Customizability82

Personalized glucose targets plus a self-learning module that refines parameters from each user's glucose and insulin history; setup needs only body weight and total daily insulin.

Access & cost40

Commercial in Europe (with Kaleido/Dana-i pumps and Dexcom CGM); FDA-cleared but not yet in any shipping US product — twiist integration targeted end-2026, US launch 2027.

Freedom (form factor)54

Pairs with compact pumps including the semi-patch Kaleido; less tubing burden than a classic pump, but not fully tubeless.

Glycemic criteria are scored on the levels actually achieved in large real-world Type 1 diabetes cohorts — not the headline improvement over a trial's baseline (an improvement that looks bigger when the starting population was doing poorly). Type 2 diabetes trial data is never used to score a Type 1 system; where only improvement data exists, it informs the rationale, not the score. Freedom captures form factor and wearability, so a tubeless system is rewarded for the mobility a tubed one can't match.

The full picture

Diabeloop DBLG2 is not a pump or sensor — it is the control algorithm, the "brain" of an automated insulin delivery (AID) system, delivered as a smartphone app. It reads continuous glucose monitor (CGM) values every few minutes and automatically tells a connected insulin pump how much to deliver, aiming to keep glucose in a healthy target range while reducing the daily mental load of type 1 diabetes.1

Components. DBLG2 is "interoperable": it is cleared to pair with compatible CGMs and pumps rather than locked to one device. In Europe it works with Dexcom sensors (G6, with G7 planned) and the ViCentra Kaleido and Sooil Dana-i pumps.2 In the US, Diabeloop has partnered with Sequel Med Tech to add DBLG2 as a second algorithm option on the twiist pump (alongside the Tidepool Loop algorithm twiist ships with today).3 twiist is a small tubed pump controlled from an iPhone and Apple Watch that uses sound waves to measure each insulin micro-dose directly.4

Trial outcomes (DBLG1 lineage). DBLG2 inherits the self-learning DBLG1 algorithm, whose evidence base is substantial. In the pivotal 12-week randomized crossover WP7 trial (68 adults), DBLG1 raised time-in-range (70–180 mg/dL) to 68.5% versus 59.4% on a sensor-augmented pump — a 9.2-point gain (about 2.2 extra hours/day in range).5 Pooled real-world data from ~6,800 adults showed TIR improvements of 12–17 points and time-below-70 cut by roughly half, reaching ~1.3% in the largest cohort.6 In people with dangerous lows at baseline, time-below-70 fell from 7.9% to 3.2% and time-below-54 from 1.9% to 0.8%.7 For very unstable ("brittle") diabetes, the DBLHU version reached 73.3% TIR versus 43.5% on a predictive-low-glucose-suspend pump.8 DBLG2 itself adds a further >3 points of TIR over DBLG1 from continued algorithm refinement.9

Automation level. DBLG2 delivers automatic correction boluses when glucose rises above a threshold, with no user action.9 Announcing meals still gives the best control (skipping them costs roughly 3–6 TIR points), but uniquely it keeps managing glucose even on undeclared-meal days — which happened ~9% of the time across 800,000 real-world patient-days.9 Setup needs only body weight and total daily insulin, and a self-learning module then personalizes parameters from each user's history.1

Exercise. When a user announces physical activity, the algorithm lowers insulin and can recommend preventive carbs. In a WP7 sub-analysis, time-below-70 on activity days (2.0%) matched non-activity days (2.2%), effectively preventing exercise-induced lows.10 Automatic activity detection is still in development.9

Ages & indications. The FDA clearance covers type 1 diabetes; for the twiist integration the labeled population is people aged 12 and older (twiist itself is cleared from age 6).3 Diabeloop is pursuing extensions to children 4+, pregnancy, and type 2 diabetes.2

Access. DBLG2 received CE marking in August 2025 and FDA 510(k) clearance in December 2025 as a Class II "interoperable automated glycemic controller," including a Predetermined Change Control Plan that lets the algorithm add compatible devices without a new submission.2 It is launching in Europe; in the US it is cleared but not yet in any shipping product.

What's coming. The Sequel partnership targets product readiness by end of 2026 and a US launch in 2027 — the first Diabeloop algorithm in a commercial US pump.3 On the roadmap: Dexcom G7 support, automatic exercise detection, and indication extensions to young children, pregnancy, and type 2 diabetes.2

References

References

  1. Diabeloop. DBLG2: an automated insulin delivery algorithm in an app — FDA 510(k) clearance. Diabeloop News (2026). https://www.diabeloop.com/news/company/diabeloop-fda-clearance-en ↩ ↩2

  2. Diabeloop. Diabeloop receives FDA 510(k) clearance for DBLG2: a strategic turning point. Diabeloop Press Release (2026). https://www.diabeloop.com/media-press/press-releases/diabeloop-fda-510k-clearance ↩ ↩2 ↩3 ↩4

  3. Diabeloop. Diabeloop and Sequel Med Tech enter partnership to integrate FDA-cleared DBLG2 into the twiist Automated Insulin Delivery System. Diabeloop Press Release (2026). https://www.diabeloop.com/media-press/press-releases/dbl-sequel-pr-en ↩ ↩2 ↩3

  4. Sequel Med Tech. Sequel and Diabeloop enter partnership to integrate DBLG2 into twiist. Sequel Med Tech News (2026). https://www.sequelmedtech.com/news/sequel-and-diabeloop-enter-partnership-to-integrate-into-twiist-tm ↩

  5. Benhamou PY, Franc S, Reznik Y, et al. Closed-loop insulin delivery in adults with type 1 diabetes in real-life conditions: a 12-week multicentre, open-label randomised controlled crossover trial. Lancet Digit Health (2019). https://doi.org/10.1016/S2589-7500%2819%2930003-2 ↩

  6. Benhamou PY, Adenis A, Lablanche S, et al. First Generation of a Modular Interoperable Closed-Loop System for Automated Insulin Delivery in Patients With Type 1 Diabetes: Lessons From Trials and Real-Life Data. J Diabetes Sci Technol (2023). https://doi.org/10.1177/19322968231186976 ↩

  7. Benhamou PY, Adenis A, Tourki Y, et al. Efficacy of a Hybrid Closed-Loop Solution in Patients With Excessive Time in Hypoglycaemia: A Post Hoc Analysis of Trials With DBLG1 System. J Diabetes Sci Technol (2022). https://doi.org/10.1177/19322968221128565 ↩

  8. Benhamou PY, Lablanche S, Vambergue A, et al. Patients with highly unstable type 1 diabetes eligible for islet transplantation can be managed with a closed-loop insulin delivery system: A series of N-of-1 randomized controlled trials. Diabetes Obes Metab (2021). https://doi.org/10.1111/dom.14214 ↩

  9. Diabeloop. DBLG2: FDA-cleared and CE-marked advanced AID algorithm designed to manage unexpected glycemic excursions. Diabeloop News (2026). https://www.diabeloop.com/news/health-sector/dblg2-manage-unexpected-glycemic-excursions ↩ ↩2 ↩3 ↩4

  10. Franc S, Benhamou PY, Borot S, et al. No more hypoglycaemia on days with physical activity and unrestricted diet when using a closed-loop system for 12 weeks: A post hoc secondary analysis of the multicentre, randomized controlled Diabeloop WP7 trial. Diabetes Obes Metab (2021). https://doi.org/10.1111/dom.14442 ↩

Coming soon

ETA · FDA-cleared (Dec 2025) but not yet in any shipping US product; US launch targeted for 2027

  • →Sequel partnership to integrate DBLG2 as a second algorithm option on the twiist pump; product readiness targeted · end of 2026
  • →First Diabeloop algorithm in a commercial US pump at US launch · 2027
  • →Dexcom G7 support
  • →Automatic exercise/activity detection (still in development)
  • →Indication extensions to children 4+, pregnancy, and type 2 diabetes

Sources

  1. [1]One-year real-world performance of the DBLG1 closed-loop system: Data from 3706 adult users with type 1 diabetes in Germany · peer-reviewed · 2023-01-01
  2. [2]First Generation of a Modular Interoperable Closed-Loop System for Automated Insulin Delivery in Patients With Type 1 Diabetes: Lessons From Trials and Real-Life Data · peer-reviewed · 2023-01-01
  3. [3]Efficacy, Safety, and Satisfaction with the Accu-Chek Insight with Diabeloop Closed-Loop System in Subjects with Type 1 Diabetes: A Multicenter Real-World Study · peer-reviewed · 2023-01-01
  4. [4]Diabeloop DBLG1 Closed-Loop System Enables Patients With Type 1 Diabetes to Significantly Improve Their Glycemic Control in Real-Life Situations Without Serious Adverse Events: 6-Month Follow-up · peer-reviewed · 2021-01-01