Here's a good article for today

Type 1 & LADA
1

Trusted – Until It Stops: Real‑World Failures in Diabetes Technology

Derek Brandt

Derek Brandt

Per Aspera ad Astra - Auf rauen Wegen zu den Sternen

December 5, 2025

An example what can happen, if there is a pump malfunction:

A vivid example is a teenager with type 1 diabetes who developed recurrent diabetic ketoacidosis (DKA) because an insulin pump silently stopped delivering insulin, and nobody initially recognized that the device itself had failed. The story illustrates how quickly pump malfunctions can escalate when people trust the device but miss subtle warning signs.

The clinical course

A 14‑year‑old girl on pump therapy was admitted with severe DKA and treated successfully with intravenous insulin and fluids; the initial assumption was “human error” (diet, insulin handling) rather than hardware failure. When she improved, subcutaneous pump therapy was restarted, but the pump later malfunctioned, insulin delivery again dropped off, and biochemical markers showed a relapse of DKA that required another prolonged course of IV insulin before a replacement pump was arranged.

What went wrong technically

Retrospective review concluded that the pump had a mechanical failure that intermittently stopped insulin despite appearing to function, so neither she nor the team immediately suspected the device. This kind of “silent interruption” is particularly dangerous in pump users because there is no long‑acting insulin in the background, so relative insulin deficiency can develop within hours and evolve into absolute deficiency and full DKA within a very short time window.

Lessons for people with diabetes

The case underlines the need to treat any unexplained, rapid rise in glucose and ketones on pump therapy as potential pump or infusion‑set failure until proven otherwise, not just as a behavioral issue. It also shows why education on troubleshooting (checking the line, changing set and site, considering a temporary switch to injected insulin) and on advocating for device evaluation in hospital is as critical as the pump technology itself.

Here are some further information about the case: https://pmc.ncbi.nlm.nih.gov/articles/PMC2769395/pdf/1757-1626-0002-0000008012.pdf

Diabetes Technology malfunctions

Common malfunctions in diabetes technology devices cluster around failures in insulin delivery (hardware and consumables), sensor accuracy and connectivity, and skin/adhesive problems, often amplified by environmental and human‑factor issues. These malfunctions can be “silent” (e.g., insulin under‑delivery without alarm) or very visible (e.g., device shutdown, sensor fall‑off), but in both cases they drive glycemic excursions and loss of trust.

Insulin pumps: device failures

Insulin pump malfunctions include failure to power on, generic “pump error” or malfunction alarms, unresponsive touchscreens or buttons, and speaker‑related defects that break communication with CGM or alarms. Physical damage (cracks, water ingress), mechanical or software faults, and reservoir/cartridge defects are frequent root causes in adverse‑event narratives. These issues can result in inappropriate insulin delivery and are associated with both hypo‑ and hyperglycemia in observational analyses of MAUDE data.

It is really interesting to search the MAUDE database: https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfmaude/search.cfm

Infusion sets and catheters

A large proportion of insulin pump problems trace back to the infusion set: kinked cannulas, occlusions, catheter knots or bending, accidental traction, and leakage at the insertion site. These failures may present clinically as “unexplained” hyperglycemia and failed correction boluses, sometimes without an occlusion alarm being triggered. Adhesive detachment of the cannula patch, bleeding, local inflammation, and skin problems (itching, pain) are also common and often lead to premature set changes and variable insulin absorption.

CGM sensors: accuracy, signal, and wear

For CGM systems, the dominant malfunctions are inaccurate readings, premature sensor failure, communication errors between sensor and transmitter/receiver, and generic error codes. Environmental factors such as heat, humidity, and physical activity contribute to adhesive failure, sensor lift‑off, and intermittent signal loss, which in turn produce missing data and unreliable trend information. Insertion problems (bleeding, partial insertion, incorrect placement) and repeated use of the same site can destabilize sensor performance and shorten effective wear time.

Adhesive and skin complications

Across both CGM and infusion sets, adhesive‑related issues are central: sensors or cannula patches falling off, accidental pull‑offs, and transmitter–sensor uncoupling. Many users develop contact dermatitis or other skin reactions to adhesives, plastics, or metals, ranging from mild irritation to severe rashes and blistering that limit long‑term device use. Site overuse can lead to scarring and altered tissue characteristics, which further degrades both insulin absorption and sensor signal quality over time.

Human factors and system behavior

Human‑factor and algorithmic issues intertwine with hardware malfunctions: errors in set insertion, inadequate training, ignoring or misinterpreting alarms, and suboptimal site rotation are all implicated in adverse events. In AID systems, mismatches between CGM input and real glucose, or algorithm behaviors such as unexpected suspensions or aggressive corrections, can compound hardware issues and contribute to severe hypo‑ or hyperglycemia. Overall, analyses emphasize that many serious events occur without clear device alarms, underscoring the need for robust failure detection, better user education, and more resilient hardware–software co‑design.

Conclusion

Diabetes technology needs to become radically more user‑led: safer, simpler, and truly designed around everyday life, not just regulatory checklists and glossy trial data. The only way to get there is to treat real‑world user feedback as hard evidence, not soft “anecdotes.”

What needs to improve

  • Reliability first: Fewer silent pump and CGM failures, better detection of interrupted insulin delivery, and clearer, smarter alarms that actually point to the root cause instead of generic error codes.

  • Human‑factor design: Interfaces, workflows, and training that match how people really live, including sickness, travel, burnout, shift work, and low digital literacy—not just idealized “perfect user” scenarios.

  • Interoperability and data integration: Devices and apps that talk to each other seamlessly so PwD can see one coherent picture of glucose, insulin, food, sleep, and activity instead of juggling islands of data.

  • Equity and access: Designs, interfaces, and support models that work for older adults, people with low income, different languages, and varying tech skills, or technology will widen, not close, the diabetes gap.

Why real user feedback must lead

Today, most safety and usability signals come filtered through under‑reported adverse‑event systems and short research studies, which systematically miss frustration, workarounds, and “near misses” that never reach a journal or database. In contrast, continuous, structured feedback from PwD—user panels, in‑app surveys, co‑design workshops, real‑world N‑of‑1 data—captures what truly matters: when people stop using a device, which alarms they mute, which features they never touch, and which small fixes would change adherence and outcomes.

If diabetes technology companies and regulators start treating this lived experience as a primary design input—weighted at least as highly as HbA1c curves and time‑in‑range graphs—the next generation of systems can be not only more advanced, but meaningfully safer, fairer, and less exhausting for the people who rely on them every single day.

For me it is clear: Nothing about us without us: diabetes tech must be built with real users at the table.

Here are some links, if you want to read more about device malfunction:

https://pmc.ncbi.nlm.nih.gov/articles/PMC12436335/

https://pmc.ncbi.nlm.nih.gov/articles/PMC8875056/

https://pmc.ncbi.nlm.nih.gov/articles/PMC4764227/

https://pmc.ncbi.nlm.nih.gov/articles/PMC10716844/

https://diabetesjournals.org/care/article/38/4/716/37581/Insulin-Pump-Risks-and-Benefits-A-Clinical

https://journals.sagepub.com/doi/10.1089/dia.2015.0434?icid=int.sj-abstract.similar-articles.5

https://www.liebertpub.com/doi/10.1089/dia.2022.0498

https://cardiovascularbusiness.com/topics/clinical/heart-health/insulin-pump-error-linked-700-adverse-events

https://blog.profil.com/blog/insulin-infusion-system-failure

https://www.liebertpub.com/doi/full/10.1089/dia.2021.0540

https://pmc.ncbi.nlm.nih.gov/articles/PMC5478030/

https://journals.sagepub.com/doi/10.1177/19322968221093362

https://cdn.clinicaltrials.gov/large-docs/94/NCT04113694/Prot_000.pdf

https://www.chuwi.com/uploads/healthy/the-most-common-mistakes-people-make-when-they-first-get-a-cgm-dQrBcx.html

https://www.snaq.ai/blog/7-factors-that-affect-your-cgm-accuracy-sensor-placement-medications-and-more

https://journals.sagepub.com/doi/10.1177/19322968241267774

https://gluroo.com/blog/diabetes-101/replace-cgm-guide/

https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfmaude/detail.cfm?mdrfoi__id=23437933&pc=QFG

https://www.accessdata.fda.gov/scripts/cdrh/cfdocs/cfmaude/detail.cfm?mdrfoi__id=10534902&pc=MDS&device_sequence_no=5

https://www.diabetesaustralia.com.au/news/warning-about-insulin-pump-failure/

4 Likes
2

From my years of testing electronic circuitry, there’s always the possibility that the testing device is faulty. For this reason after getting a CGM and later a pump paired to it I always have backup insulin delivery and glucose meter at hand.

Stuff breaks and often without warning. We all need to be aware of how we feel and if that doesn’t correlate to CGM readings TEST, and act accordingly.

It amazes me at how well CGMs work but there’s too many areas they can fail- mechanically, biologically, chemically or electronically. Even so I want my diabetes tech. My life is better with it than without.

Because I don’t have unwavering faith in CGMs I am wary of any pump more automated than Tandem C-IQ. That’s too much authority given to a brain without the ability to discern problems.

It would be nice to have tech make all decisions, but not when my health is in the pot.

3 Likes
3

I’ve left app reviews for Sugarmate in particular hoping they’ll integrate t:slim data with the app, since Tandem owns both companies. It would mean so much less labor to track information in the same place!

1 Like
4

That has been my experience using Omnipod. My theory is that the shallow angled cannula is prone to infusion disruptions from getting jiggled, knocked or skin stretched at the site, etc., which sometimes means a totally ineffective bolus and basal and other times means partially effective bolus and basal. Usually the infusion disruption is accompanied by pain at the site but not always. I can’t remember a single time the Omnipod software alerted me to an infusion disruption such as described above, other than an occasional occlusion.

These failures definitely present as “unexplained” hyperglycemia and failed correction boluses". They are hard to diagnose, often taking maybe a half day to decide to abandon a pod and start a new one. This insulin that is pumped out but not absorbed throws off the Auto mode algorithm on Omnipod 5. That is why I have gone back to manual bolus and basal.

I’ve never used a tubed pump…I am curious if other infusion sets used for Tandem, Medtronic or Sequel Twiist are overall more reliable day to day than Omnipod.

1 Like
5

I definitely found the Medtronic Quickset infusion set much more reliable than the Omnipod Dash. I never had leaking around the cannula with the Quickset.

2 Likes
6

Many years ago, before I started on my current Loop DIY system, I tried the Omnipod system for about 5 months. I found, over time, that I developed a trend of unexplained third day glucose rises along with unexplained glucose rises in the time following a pod change. These two trends led me to dropping the Omnipod.

I concluded that my case was in the minority and that my long history of pump use perhaps wore out my dermis. I knew that there were thousands of many happy users of Omnipods and I accepted that I was the exception. Yet I continued to read about accounts that reminded me of my poor experience of the Omnipod system.

I switched back to a tubed pump (Animas Ping combine with another angled infusion set) and the absorption problems totally resolved. Now I think that the fixed Omnipod cannula depth was just at a bad layer of tissue for me. I’ve had sustained (9 years now) success, however, with MM Silhouette infusion sets, surprisingly an angled infusion set.

The experience I had with Omnipod taught me to be quicker in my decision to change an infusion site even if it is newer. I’ve also experimented with the “untethered” regimen where I combined a pump with a daily long-acting basal insulin. I remember it as a dependable system but it did come at the cost of adding another daily injection. I used Tresiba and found that forgetting a shot for up to 12 hours made little difference in my glucose levels

2 Likes