Bihormonal artificial pancreas reimbursement denied in the Netherlands

Zorginstituut Nederland has ruled that bihormonal artificial pancreas reimbursement stays outside the Dutch basic health insurance package. In its final position dated 22 September 2026, the institute found that the Inreda AP, a fully closed-loop system delivering both insulin and glucagon, lacks sufficient proof of effectiveness in adults with type 1 diabetes who do not reach agreed glucose targets on standard treatment.

The patient group behind the ruling is not small. In 2024 about 100,000 people with type 1 diabetes were known to Dutch general practitioners, 52,800 men and 47,200 women, and roughly 2,700 people receive the diagnosis each year. At age 45, life expectancy is about 13 years shorter than for people without diabetes, mostly because of cardiovascular complications. Most patients never reach their targets, and managing glucose day to day is a heavy burden.

The ruling rests on one randomised trial. The DARE trial received funding through the Veelbelovende zorg subsidy scheme, which began on 1 February 2019 to support research into treatments likely to enter the basic package. The Ministry of Health, Welfare and Sport ended new subsidies from 1 January 2026 but allowed ongoing projects to conclude. DARE started on 3 October 2023 and finished on 26 March 2026, when results reached the institute. Dutch rules required a position within six months of submission.

Ad

How the institute judged the evidence

Zorginstituut Nederland manages the content of the basic package under the Health Insurance Act (Zorgverzekeringswet). New care qualifies only if it meets the statutory test of “the state of science and practice”, which in practice means it has to be at least as effective as standard treatment. The institute frames that question as a PICO: adults aged 18 and over with type 1 diabetes who miss their targets, the bihormonal artificial pancreas as intervention, and, as comparator, an intensive insulin regimen with a glucose sensor, whether multiple daily injections (at least four a day) with a sensor, an insulin pump with a sensor, or a hybrid closed loop system.

Glucose control, quality of life and serious adverse events counted as critical outcomes, with all adverse events and psychological wellbeing as important ones. The institute required at least six months of follow-up for short-term results and a year for long-term results, and set thresholds for a clinically relevant difference: an HbA1c fall of at least 0.5 percentage points or 5 mmol/mol, at least 5% more time in range, a standardised mean difference of at least 0.5 for time below range, a rate ratio below 0.75 for hypoglycaemia events, and a risk ratio no higher than 1.25 for serious adverse events. Evidence quality was graded with GRADE, and risk of bias was checked with the Cochrane tool.

A search on 23 April 2026 across Embase, Medline and Cochrane Central returned 204 references. No published study met the inclusion criteria. Only the DARE trial answered the PICO question, and it has not been published, so the institute worked from the end report filed under the subsidy scheme. No guideline, no position from another authority and no ongoing randomised study matched the question either.

Inside the DARE trial

DARE ran in fourteen centres in the Netherlands and randomised 243 adults. One hundred and nineteen received the Inreda AP and 124 continued their existing insulin treatment. Participants were 18 to 75 years old, had lived with type 1 diabetes for at least a year, and had used multiple daily injections with rt-CGM or a hybrid closed loop system for three months. Their treatment goals had to be unmet in the previous eight weeks, with time in range (glucose 3.9-10.0 mmol/l) at or below 80% or time below range at or above 4%, and HbA1c no higher than 10.5% (91 mmol/mol).

The trial split into two cohorts by comparator: 70 patients compared the Inreda AP with multiple daily injections plus rt-CGM, and 173 compared it with a hybrid closed loop system. Time in range at 12 months was the primary endpoint, measured at 3, 6, 9 and 12 months alongside HbA1c, time below and above range, hypoglycaemia events, adverse events, and questionnaire measures of quality of life (EQ-5D-5L and EQ-VAS), psychological wellbeing (WHO-5, PAID-5 and HFS-II), treatment satisfaction and sleep.

At baseline the arms looked alike. Just over half of participants were men (58% on the Inreda AP, 55% on standard care), mean age was 49 in both, mean HbA1c was 7.4% (58 mmol/mol) against 7.3% (56 mmol/mol), and mean time in range was 63% in both groups.

The glucagon switch that shaped the trial

The trial hit a supply problem before it could collect clean data. In October 2023 the glucagon it was designed around, GlucaGen from Novo Nordisk, became unavailable, so the study switched to Glugon from United Biotech. The two were expected to behave alike. They did not. Glugon acted more slowly and less potently, and held glucose within target less effectively, possibly because it formed more fibrils in solution. The trial’s algorithm software was updated twice, in March 2024 and December 2024, to rebalance glucagon and insulin delivery, and a different glucagon infusion set arrived in April 2025 to reduce blockages.

Discontinuation tells part of the story. Half of the patients on the Inreda AP (59 of 119) stopped treatment, most of them within the first three months. They cited the burden of the device: frequent alarms, repeated finger-pricks to calibrate the sensors and the size of the unit. Dissatisfaction with glucose control, and recurring infusion set and sensor failures, also featured. In cohort A the discontinuation rate was 43% (16 of 37); in cohort B it was 52% (43 of 82). No patient in the control arm stopped standard treatment.

What the trial showed

The institute’s verdict was direct. DARE did not show that the device improves glucose control over six months or a year. HbA1c differences stayed below the 0.5 percentage point threshold and the gains in time in range fell short of the 5% marker. Time below range barely moved at either cutoff. Fear of hypoglycaemia and diabetes-related distress did improve, in the direction that matters to patients, but those count as important outcomes, and the institute does not lean on important outcomes when the critical ones point elsewhere. Serious adverse events went the wrong way, and the institute called the safety signal very uncertain.

Source: Zorginstituut Nederland, Standpunt Zvw, definitief 22 september 2026, GRADE evidence table and appendix 4. Clinical relevance thresholds: HbA1c 0.5 percentage points, time in range 5%, standardised mean difference 0.5, rate ratio 0.75, risk ratio 1.25.

Certainty was downgraded twice for the glucagon switch in the analyses of HbA1c, time below range and serious adverse events, because the switch could have biased the estimates in both directions: it may have hidden a benefit for glucose control and hypoglycaemia, and it may have inflated the safety signal. Five per cent of patients on the Inreda AP (4 of 119) had a level 3 hypoglycaemia, against 0% of the standard care group (0 of 124). The rest of the serious events did not appear to be linked to the device.

Subgroup results: a signal against injections, not against hybrid closed loop

The cohort analyses are where the trial gets interesting. Against multiple daily injections plus rt-CGM, the Inreda AP held an HbA1c advantage of 0.4 percentage points at both six months (-0.7 to -0.2) and one year (-0.7 to -0.1), and time in range improved by 6.8% (-0.8 to 14.5) at six months and 8.3% (1.0 to 15.6) at one year. Against a hybrid closed loop system the picture changed: HbA1c differed by -0.2% (-0.3 to 0.0) at six months and 0.0% (-0.2 to 0.2) at a year, and time in range by -1.6% (-2.0 to 5.1) and -0.4% (-3.7 to 2.9).

The institute used that split to explain why the comparison matters more than the headline result. Patients on multiple daily injections who miss their targets do not move straight to a bihormonal artificial pancreas; they can still improve on an insulin pump with a sensor or a hybrid closed loop system. The new device is meant for people who have already tried those and still miss the mark, which makes the hybrid closed loop the relevant comparator. Serious adverse events followed the same pattern: 8% (3 of 37) against 3% (1 of 33) in cohort A, risk ratio 2.68 (0.29 to 24.49), and 17% (14 of 82) against 5% (5 of 91) in cohort B, risk ratio 3.11 (1.17 to 8.25).

The baseline HbA1c also drew comment. At 7.4% and 7.3% the trial population sat above target but was not the poorly controlled group that reimbursement criteria for rt-CGM define, where HbA1c starts at 8%. People with an HbA1c of 8% or higher may show a larger benefit, and the institute said future research should focus on them.

Bihormonal artificial pancreas reimbursement and the budget question

On cost, the institute worked from an estimate that around 1,000 patients a year in the Netherlands would qualify, a figure supplied by the Dutch Internists Association (NIV) in a written statement of 29 May 2026. Direct costs for those 1,000 patients would come to 29,189,000 euros on current treatment and 40,878,000 euros on the new device, a budget impact of 11,689,000 euros a year. Complications and indirect costs were left out because precise data are missing. Keeping the device out of the package avoids up to 11.7 million euros in direct spending a year.

Insurers do not need to amend their policies. Once the institute establishes whether care meets the state of science and practice, that finding flows straight into every individual health insurance contract. The position took effect on 22 September 2026, its date of adoption, and the institute did not give it retroactive force, which is its usual practice for a negative position.

For patients and clinicians, the practical consequence is narrower than the headline suggests. Patients with type 1 diabetes who miss their targets on standard treatment are not entitled to reimbursement for the device from the basic package. Health insurers may still fund it from their own resources, whether as a gesture of goodwill or from innovation budgets, and that choice is theirs.

Objections from the manufacturer

The institute consulted the NIV, Diabetes Vereniging Nederland (DVN) and Zorgverzekeraars Nederland (ZN). Inreda Diabetic responded on its own initiative with letters on 8 and 21 September 2026. The company argued that impaired glucagon regulation is part of the pathology of type 1 diabetes, not a side issue, and that adding glucagon to treatment addresses a counter-regulatory response that is diminished or absent during hypoglycaemia. It also pointed out that the trial tested a different system configuration from the one used in practice, because Glugon stood in for GlucaGen, and said a registration dossier for a generic version of GlucaGen went to the Dutch Medicines Evaluation Board in August 2026, with performance data on the combination expected by mid 2027. Inreda asked the institute to hold off on a final conclusion, warning that a negative decision would push 127 existing patients to pay for the device themselves or switch to treatment that had worked less well for them.

The institute declined. It replied that very low certainty evidence caused by unforeseen circumstances is not a reason to withhold a position, and that it had explained this in a meeting on 17 September 2026. The studies Inreda submitted did not fit the PICO or the required study design, so they did not change the conclusion. The institute also noted that it had already downgraded certainty for the glucagon switch under risk of bias, which is why it says the device is not yet sufficiently proven effective rather than proven ineffective. Its Scientific Advisory Board backed the conclusion on 22 June 2026, and DVN, while accepting the finding, asked for careful communication with current users.

The institute will watch future publications and considers a reassessment if new data justify one. Whether the money stays in the package or not, the dossier now turns on a single question the trial could not answer: whether a bihormonal system configured for the intended glucagon can beat a hybrid closed loop system in the patients who need it most.

Source: Zorginstituut Nederland, Standpunt Zvw: bihormonale kunstalvleesklier bij volwassen patiënten met diabetes type 1, definitief 22 september 2026