Tye Cummins lay in the dark, gasping for air and unable to move. He was slipping into a life-threatening, unresponsive state after too much insulin was pumped into his body as he slept.
The culprit was an Omnipod insulin pump – a device he wore as part of a clinical trial for patients with type 1 diabetes.
“It nearly killed me,” Cummins says of the device, which delivered insulin under the skin and is an alternative to injections. “I was the worst I had ever been. I couldn’t walk or talk for more than two hours.”
In the past decade, diabetes treatment – which long required finger-prick testing and insulin injections – has been revolutionised by automated insulin pumps such as the Omnipod and continuous glucose monitors (CGMs), which provide real-time glucose data on a patient’s smartphone.
“They have been life-changing,” said Dr Ashley Ng, head of education and professional standards at the Australian Diabetes Educators Association. High-quality evidence shows CGMs and insulin pumps cut complications and improve blood-sugar control.
The rollout has also been accompanied by a dramatic increase in the number of adverse event reports linked to the devices.
The Therapeutic Goods Administration tracks these problems in a public database but does not release that data in aggregate, making it impossible for consumers to know which devices are linked to the most reports.
An investigation by this masthead and 60 Minutes used automated website scrapers to crack open that database, extracting about 81,000 publicly available reports to reveal that diabetes management devices make up seven of the 10 devices with the most adverse-event reports submitted to the TGA since 2024.
They also hold five of the top 10 spots for reports of injury or death since 2024. The results highlight the tension between advancing technology and the adequacy of warnings and regulation.
Many of the complaints focus on the devices’ accuracy, alarm failures or difficulty connecting to readers or smartphones.
The TGA notes that an adverse event report, which can be submitted by a patient, health professional or any other member of the public, does not mean the device caused the event. Nor can such a report be used in isolation to assess the safety of a medical device.
Still, diabetes experts are concerned the TGA has eased the regulatory requirements for monitors, allowing a new generation of less-accurate devices to be sold here.
US regulators have also warned American patients that Bluetooth connections from smartphones or car stereos can mute crucial blood sugar alarms – a warning that Australian regulators did not pass on.
A Victorian coronial inquest is investigating five deaths resulting from a high-profile car crash possibly linked to a CGM that failed to raise an alarm because of a Bluetooth issue.
Widely used in Australia
The Omnipod 5 insulin pump, which is partially subsidised by the federal government, is associated with 429 adverse-event reports, including 330 injuries, 283 cases of hyperglycaemia and six cases of diabetic ketoacidosis – a life-threatening complication typically requiring treatment in intensive care.
Hundreds of deaths related to insulin pumps have been reported in the US. Just before this story was published, some batches of the Omnipod 5 were recalled in Australia over an insulin leaking issue associated with 24 serious events.
The five government-subsidised glucose monitors from Dexcom, Abbott and MiniMed are linked to 687 adverse-event reports, including 215 injuries and one death.
The device manufacturers pointed to the large number of Australians now using their technology to explain the number of incidents. More than 107,000 Australians are approved for monitor subsidy, the TGA said, and more than 146,000 people for insulin pumps.
The devices include warnings about smartphones or Bluetooth connections muting alarms, and instructions for patients to check their glucose levels with a finger-prick test if they suspect the reading is inaccurate.
“Omnipod is supported by extensive clinical evidence and real-world experience across geographies demonstrating its safety, effectiveness and quality-of-life benefits for people with diabetes,” a spokeswoman for the device’s manufacturer, Insulet, said.
Associate Professor Neale Cohen, director of clinical services at the Baker Heart and Diabetes Institute, is one of several leading experts who told this masthead all mainstream CGMs were reliable and accurate.
Professor David O’Neal, director and founder of the Diabetes Technology Research Group at the University of Melbourne, said while the monitors were “not necessarily as accurate as some finger-prick meter, as a rule they are accurate enough, so that one can use them on their own without using a glucose meter to confirm the reading.”
Insulet asked O’Neal to speak to this masthead but has no commercial relationship with the academic.
In interviews with this masthead, many people with diabetes recounted highly inaccurate CGM measurements compared to finger-prick tests. Those concerns have also shown up repeatedly in academic surveys and systematic reviews.
“They are deadly,” says Yvette Petrie, who uses a CGM connected to an automatic insulin pump to manage her type 1 diabetes. “It was an absolute nightmare. When it was reading high – but I was only 13 mils, minimal – it was still dosing insulin. It could have overdosed me. I have told Dexcom a few times they are going to kill someone.”
Bec Ritchie, a diabetes educator, registered nurse and founder of SEE Diabetes, which offers support and education for people with type 1 diabetes, said reports of inaccuracy “do not surprise me”.
“I have had many a client who has had significant issues and injuries from these devices,” she said. “The key thing that I see most from clients is inaccuracy. And inaccuracy, when they are connected up to insulin pumps that are automatically delivering insulin, it can very easily over or under-insulin people and cause hypo or hyperglycemia, which is quite a risk.”
‘It can be deadly’
Tye Cummins, who struggled with the Omnipod pump, also ran into issues with Dexcom’s G6 monitor. In 2024, a year after having his Omnipod pump removed, his Dexcom monitoring device showed that his sugars were elevated, even though he didn’t have the telltale signs of thirst and lethargy. The 50-year-old injected himself with insulin to lower his glucose levels.
Something didn’t feel right, so he decided to check his glucose using an old-school finger-prick test. He placed the drop of blood on a test strip and found his blood glucose was now dangerously low.
This meant the reading on the Dexcom G6 monitoring device had been wildly inaccurate. He had now lowered his glucose to such a level that he was at risk of seizures, coma or death.
“If this had happened when I was going to bed, I may have never woken up,” the Geelong man said.
Cummins complained to the manufacturer, and said it told him the inaccuracy was probably caused by something he had done, such as taking painkillers.
“You put a fair bit of faith in these companies and devices,” he said. “The advancements in diabetes are so good, but it is only easier until something goes wrong. Then it can be deadly, and that’s my worry.”
Amy Wright used Dexcom’s G6 continuous glucose monitor for years without incident, before switching to the newer G7. “It’s inaccurate, it disconnects. It is dangerous,” she said.
“The numbers were way out. I had the G6 on at the same time. There was a massive difference. And the G7 kept cutting out all the time. It kept saying it wasn’t in range, even though my phone was next to me.” Wright has started a petition calling for Dexcom to stop phasing out the G6.
“An inaccurate CGM can lead to death. It can make your pump overdose you. If it’s reading 12, but you’re actually six, it might overdose you, and you end up with a hypo.”
The accuracy of Dexcom’s new G7, which won government subsidy in March 2025, is a major flashpoint among people with diabetes.
In that month, US regulators sent Dexcom a letter warning it was relying on data from the G6 to determine the G7’s accuracy – and the G7 was using a new, unapproved component in sensors that internal Dexcom testing showed to be less accurate. The Food and Drug Administration’s warning letter noted “significant variation in the clinical performance of sensors”.
“The larger inaccuracies in (the new) sensors cause higher risks for users who rely on the sensors to dose insulin or make other diabetes treatment decisions,” it said.
That has caused consternation here.
“I am definitely worried about that,” said Dr Ng, from the Australian Diabetes Educators Association. “If I had the choice, I wouldn’t go with the G7.”
In a statement, Dexcom said it was working closely with US regulators to address the issues identified and had made “meaningful progress toward resolution”. It noted US regulators had not demanded a recall of the G7.
“There is a large body of clinical evidence that demonstrates the use of Dexcom CGM leads to improved glycaemic control compared to the use of finger sticks,” the company said. “Dexcom G7 has been recognised globally for its contributions to diabetes care.”
The TGA issued a product correction last year for users of Dexcom’s G6 receiver, warning it might fail to alert for high or low blood glucose or signal loss.
Some batches of another monitor, Abbott’s FreeStyle Libre 3, which won access to an Australian government subsidy in December, were recalled by US regulators last year due to sensor inaccuracies that caused 860 serious injuries and seven deaths.
The company said Australian customers were not affected by the issue. Its US parent said it had resolved the problem and continued to produce the sensors.
How technology and warnings fail
Many modern CGMs can link to a smartphone to display real-time blood glucose levels and alarms. This combination of a life-saving medical device with a frequently updated smartphone has introduced a new weakness to the system.
The TGA and Dexcom have issued six product corrections for its app since 2025, for delays or failures to update glucose levels, alarms not going off and silent loss of connection.
Abbott’s FreeStyle app had a similar issue in 2023, after a software update caused some Android phones to silently lose connection to their sensors, according to a TGA alert.
Antonia Watson, 40, from Melbourne, has had “heaps and heaps” of connection issues between her phone and FreeStyle Libre 2. “If I leave my phone on charge in the kitchen and walk 10 metres to the laundry, I have to come back and manually reconnect it. And if you put your phone on silent, you don’t get the alarms.”
And when the sensor worked, it was frequently inaccurate. “How many people are out there who are treating themselves for lows and highs that don’t exist, and are sickening themselves in the process?”
This link between CGMs and smartphones is also at the centre of a major coronial investigation.
In November 2023, William Swale suffered a severe low-blood-sugar attack and crashed his BMW into families dining in the beer garden outside the Royal Daylesford Hotel in Daylesford, Victoria, killing five people.
Swale was initially charged over the deaths, but a magistrate struck out those charges in 2024 after finding his actions had been involuntary.
Swale, who has type 1 diabetes, wore an Abbott FreeStyle Libre 2 CGM on the day of the crash. It sent 10 low-blood-sugar alarms in the hour leading up to the crash, but Swale told the Coroners Court he never heard them.
Following the fatal crash, Swale’s doctor pointed him to a 2025 warning from US medical regulators that connecting a phone linked to a CGM to a car Bluetooth can mute blood sugar alarms.
Australia’s TGA has never passed on the American warning about car Bluetooth to patients in this country using CGMs, the regulator confirmed.
After police turned off his car at the crash scene, disabling the Bluetooth connection, they were able to hear the alarm beeping from his phone, Swale told the coroner.
Neither Swale nor his doctor knew about the issue before the crash.
“I don’t much know that anyone was aware that this failure of alarms was an issue,” Swale’s doctor, Matthew Cohen, told the coroner. “We’re relying on low glucose alarms, and we weren’t aware that these could be muted.”
Speaking generally to this masthead, Cohen described CGMs as “life-changing, amazing pieces of equipment”.
“The main issue people complain of is, like any technology, they occasionally fail. You can get a faulty batch, some that don’t work. We tell people they are not 100 per cent accurate,” he said. “It is better than a finger prick. But you can get false lows and false highs. None of these things are perfect, but they are so useful. The benefits are so overwhelming.”
In a statement, the TGA defended not passing on the US regulator’s warning about Bluetooth devices potentially muting CGM alarms.
The regulator said it “publishes device‑specific or general safety information about medical devices on its website where appropriate”, and that was “informed by safety signals and any trends, including numbers of adverse event reports received from both overseas and the Australian market”.
Yet the TGA has received some safety signals.
On December 23, 2024, a year after Swale’s crash, the Australian regulator received an adverse-event report about a person potentially injured by a CGM: “Hypoglycaemia episode with no alarm due to Bluetooth issue”. The device was the FreeStyle Libre 2, the same one Swale used.
And the regulator has received more than a dozen adverse-event reports about faulty connections between CGMs and smartphones, including a FreeStyle 2 patient who could not get glucose readings on their phone, leaving them unable to treat themselves. They “experienced a seizure and loss of consciousness”.
In a statement, the manufacturer, Abbott, said the “connectivity of Libre sensors with smartphones may sometimes be impacted due to automatic operating system updates on smartphones.
“We recommend Libre users disable automatic operating system updates on any smartphone and wait to update to a new operating system until Libre mobile app compatibility is confirmed.”
It remains for the coroner to decide whether the CGM-Bluetooth issue played a role in the crash.
Endocrinologist Professor John Carter told the coroner Swale was probably aware he was hypoglycaemic in the lead-up to the crash, as he tried to buy food, but was turned away.
And if a CGM regularly alerts about high or low blood sugar, many patients can grow frustrated and turn it off or tweak the alarm level lower.
Professor Richard O’Brien, dean of medicine at the University of Melbourne, told the coroner: “I have many patients who have switched off their low-glucose alarm entirely because of frequent annoying alerts.”
Continuous glucose monitors are most accurate when glucose levels are stable, and least accurate when they are changing quickly. But people with diabetes are prone to rapidly fluctuating glucose levels.
“It is crucial to subject CGM systems to those rapid glucose fluctuations during testing,” said Dr Manuel Eichenlaub, a researcher at the Institute for Diabetes Technology in Germany.
“Unfortunately, most manufacturers do not share this view and actively try to avoid rapid glucose changes during testing. This is because CGM accuracy tends to deteriorate during rapid glucose changes, which would worsen their claimed accuracy.”
Eichenlaub published a paper last year testing CGMs from Dexcom, Abbott and MiniMed in patients whose blood glucose was rapidly changing – the first evaluation not funded by the manufacturers.
He compared the devices to a gold-standard test for blood glucose, known as YSI, and two other commercial tests.
All three devices performed worse than in their key clinical trials.
“The [MiniMed] device shows systematically lower readings than the other two, and can absolutely have clinical consequences,” said Eichenlaub. “However, I should stress that CGM systems by established manufacturers – mainly Abbott/Libre and Dexcom – are safe and have been a blessing for people living with diabetes.”
MiniMed said Eichenlaub’s claim was “without merit”, arguing the YSI data was similar across all three devices tested. “MiniMed has complete confidence in the evidence built across 280+ independent, clinical and real-world studies across the globe, demonstrating consistent glycaemic outcomes in comparable patient populations.”
Amid this complicated backdrop, the TGA in 2024 moved to lower the regulatory classification of CGMs from high to medium-risk devices.
New sensors, new concerns
Some diabetes experts are watching with alarm as new sensors with varying scientific evidence reach the market.
“There are many, many more coming onto the market that are really not reliable,” said Associate Professor Neale Cohen, director of clinical services at the Baker Heart and Diabetes Institute. “You can get approval for these gadgets from the TGA, but it does not mean they are accurate.”
Worried about these declining standards, a coalition of diabetes expert groups last year released guidance on what they argued should be minimum standards for CGMs.
MiniMed’s line of Guardian CGMs fails to meet those standards, confirmed Sof Andrikopoulos, group chief executive of the Australian Diabetes Society.
“If you give too much insulin, you can kill yourself. That’s the reality. We really need to make sure the devices being approved do meet standards,” he said.
In a statement, MiniMed criticised the new minimum standards.
“Applying the proposed approval standards would not address the concerns raised and could unintentionally limit patient access to and choice of essential diabetes technologies without improving CGM safety or quality.”
The Examine newsletter explains and analyses science with a rigorous focus on the evidence. Sign up to get it each week.
Read the full article here
