From Hardware to Intelligence: Three Phases of Smart Insulin Patch Innovation

Smart insulin patch technology has evolved across three distinct phases since approximately 2008, moving from basic hardware viability to AI-driven dosing and biochemical glucose-responsive delivery. The innovation arc — spanning from the OmniPod’s commercial debut to Insulet Corporation’s December 2025 machine-learning reservoir prediction patent — reflects a field accelerating toward full automation.

The Foundational Phase (2008–2015) established hardware viability. The OmniPod system — described at its launch as “the first commercially available patch pump… controlled wirelessly through a handheld device containing a built-in blood glucose meter” — set the tubeless paradigm. Design patents from this era, including a waterproof insulin pump case (US, 2013) and Medtrum Technologies’ disposable tubeless pump (US, 2017), reflect rapid hardware industrialisation. Korean assignee Enwiser Inc. filed platform-level blood glucose management patents as early as 2015–2016, signalling early IoT integration in East Asian medtech.

The Development and Integration Phase (2016–2022) brought clinical validation at scale. The EOPatch tubeless pump study (Samsung Medical Center, 2022) reported median usage times of 84 hours per patch with significant time-in-range improvement. Smart insulin pens proliferated: Novo Nordisk’s NovoPen 6 demonstrated a 43% reduction in missed bolus injections in real-world use, while Omnipod 5 outpatient trials demonstrated hybrid closed-loop feasibility in children and adults. According to WHO, diabetes affects over 500 million adults globally — a scale that makes automated delivery platforms commercially and clinically urgent.

The Emerging Intelligence Phase (2023–2025) is characterised by AI-driven dosing and glucose-responsive biochemical delivery. Insulet filed an EP patent in December 2025 applying machine learning to total daily insulin estimation and reservoir refill timing. The University of California filed in EP jurisdiction (2025) on a therapeutic hybrid microneedle patch with dual glucose-responsive release. Two Italian academic filings (University of Magna Graecia, 2024–2025) apply AI explicitly to dosing decisions at the device level, with one achieving active status by May 2025.

Four Technology Clusters Shaping Automated Insulin Delivery

Smart insulin patch innovation organises around four principal sub-domains: tubeless wearable patch pumps, closed-loop automated insulin delivery (AID) systems, AI and IoT-connected dosing platforms, and glucose-responsive biochemical delivery. Each cluster occupies a different position on the maturity curve — from commercial deployment to early-stage research.

Cluster 1: Tubeless Wearable Patch Pumps

Tubeless patch pumps integrate the insulin reservoir, actuation mechanism, and transcutaneous cannula into a single on-body adhesive unit, eliminating tubing. The Omnipod DASH system — with IP28 waterproof housing, automated cannula insertion, customisable basal and bolus delivery, and Bluetooth-enabled PDM control — is the canonical commercial example. A stepping-motor-based patch pump described by Institut für Diabetes-Technologie (Ulm, Germany) provides comparative accuracy data showing median basal and bolus delivery deviations comparable to durable pumps.

Cluster 2: Closed-Loop AID Systems

Closed-loop AID combines patch or conventional pump hardware with continuous glucose monitoring (CGM) and a control algorithm — PID, model-predictive control, or machine learning — to automate basal and correction dosing. The Omnipod 5 AID system demonstrated superior time-in-range across adjustable glucose targets (110–150 mg/dL) in a multicenter outpatient trial. Real-world data from 9,451 Control-IQ (Tandem Diabetes Care) users showed median automation time of 94.2% over 12 months. Open-source implementations including OpenAPS, Loop, and AndroidAPS demonstrate that the algorithmic layer is reproducible by the community.

Cluster 3: AI and IoT-Connected Dosing Platforms

This cluster covers the digital intelligence layer — smart pen attachments, mobile apps, and AI-driven dosing advisors — that fall short of full closed-loop automation but significantly augment therapy management. The NovoPen 6 real-world study (Novo Nordisk, 2020) demonstrated that logging connectivity alone reduced missed bolus injections by 43% and increased time-in-range. Korean platform patents from Enwiser Inc. (2015–2016) describe Bluetooth-connected insulin pumps linked to mobile apps for graphical glycaemic management. A 2024 KR-active patent from G2E Co., Ltd. describes a platform-based smart insulin pen control method with bolus recommendations delivered from a cloud platform.

Cluster 4: Glucose-Responsive Biochemical Delivery

The most research-stage cluster pursues elimination of external CGM by embedding glucose-sensing chemistry directly into the delivery matrix. The University of California’s EP-filed microneedle patch (2025) uses a co-polymerised matrix that is “dually-responsive” to both hyperglycaemic and hypoglycaemic glucose concentrations — releasing insulin in high-glucose conditions and glucagon in low-glucose conditions, demonstrated in a type 1 diabetic mouse model. Nanoscale PAM-PAspPBA-b-PEG glucose-sensitive carriers (Tsinghua University FIESTA Center, 2018) similarly achieve insulin release proportional to ambient glucose without real-time external sensing. Research published in Nature and related journals has highlighted glucose-responsive polymer systems as a foundational enabling technology for this cluster.

What the Clinical Evidence Shows: Time-in-Range, HbA1c, and Real-World Use

Clinical evidence across patch pump and AID system trials consistently demonstrates improvements in time-in-range (TIR) and HbA1c reduction, with real-world data from thousands of patients corroborating pivotal trial outcomes. The evidence base spans Type 1 diabetes (T1D) as the primary domain, Type 2 diabetes (T2D) as a growing secondary domain, and paediatric populations as a distinct cohort with dedicated data.

For T1D, the evidence base is deepest. A retrospective study of 4,738 patients initiating the Omnipod DASH system (Baylor College of Medicine, 2023) demonstrated consistent HbA1c reduction across paediatric and adult cohorts, confirming real-world efficacy at scale. The Omnipod 5 pivotal outpatient trial enrolled participants aged 6–70 years, demonstrating superior time-in-range across adjustable glucose targets (110–150 mg/dL). NovoPen 6 paediatric data from Swedish clinics — 39 children across three clinics — showed reduced nocturnal hypoglycaemia and improved time-in-range over 12 months.

“Open-source AID implementations — OpenAPS, Loop, AndroidAPS — demonstrate that the algorithmic layer is reproducible by the patient community, with Loop showing TIR improvement of 6.6 percentage points in 558 participants.”

For T2D, the V-Go wearable insulin delivery device (Valeritas) demonstrated HbA1c reductions of −1.5 ± 1.79% and reduction in total daily insulin dose in 139 patients. A randomised controlled trial (Calibra Medical / Johnson & Johnson, 2019) compared bolus insulin delivery via wearable patch versus pen across 278 T2D adults over 48 weeks, with the patch arm showing comparable HbA1c outcomes and superior patient preference metrics on convenience and lifestyle interference dimensions.

Cost-effectiveness evidence is also emerging. A health economics analysis (Ossian Health Economics and Communications, 2020) found that smart insulin pens are associated with improved clinical outcomes at lower cost versus standard-of-care treatment of Type 1 diabetes in Sweden. The University of Canterbury’s 2022 open-source pump design — built for USD 89.85 in materials — demonstrates that affordable hardware architectures are technically feasible, with implications for middle- and low-income market penetration where diabetes prevalence is growing fastest. Standards bodies including ISO and regulatory agencies such as the FDA have published guidance frameworks relevant to the safety and interoperability requirements these systems must meet.

Geographic and Assignee Patterns in the Patent Landscape

Commercial patch pump innovation in this dataset is highly concentrated in a small number of established players — primarily Insulet Corporation — while academic and startup-stage biochemical delivery innovation is broadly distributed across US, European, and Asian universities and research institutions.

United States: Dominant in commercial hardware filings and active IP. Insulet Corporation holds multiple active US and EP filings, including the December 2025 EP reservoir prediction patent. Verily Life Sciences holds an active US design patent for a CGM skin patch transmitter (2018). Medtrum Technologies Inc. holds an active US disposable tubeless pump design (2017). Academic innovation includes the University of California’s EP microneedle patch (2025).

South Korea: Two active KR filings from Enwiser Inc. (2015, 2016) on integrated blood glucose and insulin pump management platforms indicate early IoT-first innovation. A 2024 KR-active patent from G2E Co., Ltd. on a platform-based smart insulin pen control method reflects continued Korean engagement in connected device software.

Italy and Europe: The University of Magna Graecia di Catanzaro holds two filings (2024 pending and 2025 active) on AI-based insulin dosing devices — representing emerging European academic IP in algorithmic control. The University of California’s EP microneedle patch and Insulet’s EP reservoir prediction patent signal that leading US innovators are aggressively filing in European jurisdictions. According to EPO filing trends, medical device AI patents have been among the fastest-growing application categories in recent years.

India: A 2025 pending filing from Noida Institute of Engineering and Technology on a wearable insulin patch pump with dose feedback and BLE connectivity signals emerging innovation activity in South Asia.

Five Emerging Directions Redefining the Smart Patch Architecture

Based on filings and literature dated 2023–2025, five directions are ascendant in smart insulin patch technology — collectively pushing the field from hybrid closed-loop toward fully autonomous, biochemically intelligent delivery.

1. Dual-Hormone Glucose-Responsive Patch Systems. The University of California’s 2025 EP patent describes a microneedle patch with co-polymerised matrices responsive to both hyperglycaemia and hypoglycaemia — releasing insulin in high-glucose conditions and glucagon in low-glucose conditions. This “bihormonal patch” concept potentially eliminates severe hypoglycaemia risk without user input, and would render external CGM redundant if translated to clinical use.

2. Machine Learning for Insulin Reservoir Management. Insulet’s December 2025 EP filing moves beyond dosing algorithms into predictive logistics — using historical delivery data to estimate total daily insulin needs and optimise reservoir refill timing, reducing waste and preventing premature depletion.

3. AI-Embedded Device-Level Dosing. Two Italian academic patents (University of Magna Graecia, 2024–2025) describe AI at the device level rather than cloud or app level, suggesting a trend toward on-device intelligence for latency-free dosing decisions. One of these achieved active status by May 2025.

4. Multi-Sensor Feedback Integration. The 2025 Indian filing (Noida Institute of Engineering and Technology) describes a patch pump integrating flow sensors, pressure sensors, NFC authentication, and BLE connectivity into a single wearable unit with haptic and visual alerts — reflecting a trend toward safety-critical embedded feedback loops within the patch itself.

5. Platform-Level Bolus Recommendation via Cloud AI. The 2024 Korean smart insulin pen control patent (G2E Co., Ltd.) describes a cloud platform receiving device log data and returning bolus recommendations to the user’s electronic device — previewing a service-layer business model overlaid on hardware.

Strategic Implications for R&D and IP Teams

The smart insulin patch landscape presents distinct strategic imperatives depending on whether an organisation is competing in commercial AID hardware, developing algorithmic IP, or pursuing biochemical delivery research. Several cross-cutting signals from this dataset are relevant to R&D prioritisation and IP strategy.

Closed-loop AID is the competitive battleground. Insulet Corporation’s portfolio of tubeless hardware combined with algorithm optimisation patents — including the 2025 EP reservoir ML filing — positions it as the integration leader. R&D teams in this space must stake out differentiated algorithmic or sensing IP, as hardware commoditisation is accelerating.

Biochemical glucose-responsive delivery is the highest-risk, highest-reward frontier. The University of California’s bihormonal microneedle patch (EP, 2025) and similar nanoscale systems require 5–10 year development runways but, if translated, would render external CGM redundant — disrupting the entire sensing layer of the AID stack. According to WIPO, academic institutions are increasingly active filers in medical device AI and drug delivery, making freedom-to-operate analysis in this space essential before committing R&D resources.

Open-source AID systems are a market intelligence signal. Data from OpenAPS, Loop, and AndroidAPS demonstrate patient-driven demand for algorithmic customisation that commercial systems have not yet fully satisfied. Product developers should treat DIY AID feature requests as a validated market requirements document.

Accessibility and cost architecture will determine global market penetration. Evidence in this dataset — including an open-source pump designed for USD 89.85 in materials cost (University of Canterbury, 2022) and analysis showing smart insulin pens are cost-effective in Swedish societal models — indicates that next-generation smart patch platforms must address affordability from the design stage, especially for middle- and low-income markets where diabetes prevalence is growing fastest.

“Biochemical glucose-responsive delivery systems require 5–10 year development runways but, if translated, would render external CGM redundant — disrupting the entire sensing layer of the automated insulin delivery stack.”

For IP teams specifically: Korea and Italy show active filings in platform software and device-level AI that are underweighted relative to their innovation activity. Freedom-to-operate analysis in KR and IT jurisdictions is advisable before commercialising connected dosing platforms in those markets. PatSnap’s patent analytics tools and freedom-to-operate workflows are designed to surface exactly these non-obvious geographic risks at scale.