Why this episode exists

The DSN Forum’s CGM comparison chart has become one of the most widely used practical resources for diabetes nurses navigating a crowded device market. But the chart — and the conversation around CGM more broadly — started from features and MARD. It took time, frustration, and some uncomfortable conversations with device companies to recognise that feature comparison only means something if the evidence behind the device is sufficient in the first place. This episode is the story of how that shift happened, what it revealed, and what it means for anyone — nurse, clinician, or person with diabetes — who is trying to make a well-informed CGM decision.

It is also a conversation about what came out of ATTD 2025: three DSNs with front-line clinical perspective on fully closed loop systems, GLP-1 in type 1 diabetes, continuous ketone monitoring, and what optimising AID system settings actually looks like in practice.

In this episode

John Pemberton is joined by Amanda Williams, Beth Kelly and Tamsin Fletcher-Salt — three leading members of the Diabetes Specialist Nurse Forum UK — for a wide-ranging conversation about CGM accuracy, device choice, and the evidence that should sit at the foundation of both. Starting from how the DSN Forum’s comparison chart was created, the conversation works through why CE marking and MARD proved insufficient as quality markers, how the five-point accuracy scoring system was built, and what it takes for a device to earn a place on the chart. The episode is frank, practical, and grounded in years of front-line clinical experience.

The second half of the episode turns to ATTD 2025, where John and the three DSNs share what genuinely excited them: the MiniMed Flex fully closed loop system, the MyLife Liberty algorithm, GLP-1 use in type 1 diabetes, Abbott’s upcoming continuous ketone sensor, and what the GNL AID System Explorer reveals about optimising different AID algorithm settings in practice. This episode connects directly to the GNL CGM Guide series and to Episode 35 with Professor Othmar Moser, which laid out the research basis for the accuracy standards discussed here.

What this episode explores

• Why the Diabetes Specialist Nurse Forum created its CGM comparison chart and how it has evolved
• Why CE marking is not a quality standard for CGM accuracy and what that means for insulin dosing decisions
• Why MARD alone is insufficient when evaluating a CGM for insulin-dosing indication
• How the five-point accuracy scoring system was built and what it requires from a device
• Why there are currently only four devices that meet the data sufficiency standard, and why two more are imminent
• What 20/20 and 40/40 agreement rates mean in real-world terms and why the one-percent tail matters
• Why people using CGM still need an in-date glucose meter and why the lost art of finger pricking still has a role
• The calibration debate: whether factory calibration was a step forward or a step away from safety
• What ATTD 2025 revealed about fully closed loop systems — who benefits and who may not
• The case for GLP-1 in type 1 diabetes and why John believes it deserves priority consideration
• Abbott’s continuous ketone monitor: the opportunity, the unknowns, and the risks for specific patient groups
• How different AID systems register insulin on board and what that means for optimisation — including the GNL AID System Explorer

Watch or listen

Episode chapters

• 00:00 — Introduction: meeting Amanda, Beth and Tamsin
• 02:14 — Why the DSN Forum CGM comparison chart was created
• 06:47 — Turning complex CGM evidence into a practical scoring system
• 16:23 — How CGM choice works in real clinical settings
• 20:36 — Why finger-prick testing still matters — and the calibration debate
• 29:26 — What stood out at ATTD 2025
• 44:03 — AID systems, insulin on board, and the GNL AID optimiser
• 49:49 — Where to find the DSN Forum, the chart, and closing thoughts

Key themes

1. The DSN Forum comparison chart — from features to data sufficiency

The chart was created to give diabetes nurses a practical way to compare the growing number of CGM devices on market — days wear time, alarms, interoperability, licensing for specific groups. It was useful and widely adopted. But it began from an assumption that a device with CE marking and a published MARD had already passed a meaningful quality threshold. What John and the DSN Forum discovered was that this assumption does not hold: CE marking and MARD alone do not tell you whether a device has sufficient, appropriately designed data to support insulin dosing decisions in the population you are prescribing it to. The chart has since been restructured with a front-end gate: data sufficiency comes first, features come second.

2. CE marking is not a CGM quality standard

CE marking confirms that a device meets the minimum regulatory requirements for sale in the EU and UK. It does not require that the accuracy data supporting an insulin-dosing indication was collected in the population that indication covers. John describes discovering that devices were being approved for insulin dosing — including in children from the age of two — based entirely on accuracy data from adults with type 2 diabetes. This is not a regulatory loophole in an obscure part of the system; it is a structural feature of how CE marking works for medical devices. The contrast with the glucose meter ISO standard is stark: meters have a defined, independent accuracy benchmark. CGMs do not have an equivalent. This episode explains why that gap matters and what the DSN Forum is doing to provide the practical shorthand that regulation does not.

3. The five-point accuracy scoring system

Over two weeks of early-morning messages, John and the DSN Forum team distilled the complex evidence requirements for CGM accuracy into five questions that can be answered quickly by anyone on the front line.

4. What 20/20 and 40/40 mean in clinical practice

MARD gives an average. It does not tell you how often a reading is dangerously wrong.

5. The lost art of finger pricking — and the calibration debate

The episode explores a tension that runs through clinical practice but rarely surfaces explicitly: CGM has advanced so fast that the habits and equipment that provided a safety net — working glucose meters, in-date strips, regular calibration — are being quietly dropped. Amanda describes patients whose ketone meters have expired, whose glucose strips are out of date, and who have no backup insulin pens. Beth describes primary care removing glucose test strips from prescriptions because a CGM is on record. The group is split on calibration: John and Tamsin believe the option to calibrate adds safety; Amanda argues pragmatically that calibration done badly — using the sensor reading to calibrate the sensor — is worse than no calibration at all. The one point of full consensus: regardless of position on calibration, every person using CGM for insulin decisions should have a working, in-date finger-prick meter.

6. Fully closed loop systems — who benefits and who does not

ATTD 2025 featured the MiniMed Flex (FDA approved, awaiting UK launch) and the MyLife Liberty (coming very soon). Both are truly closed loop — no carbohydrate announcement required. The clinical data presented shows meaningful improvement in time in range for people struggling on hybrid closed loop: those not pre-bolusing, not counting carbs, with HbA1c around 8.5–9%. Tamsin asked Roman Havorka directly whether someone already achieving 80%+ time in range would see a drop if they switched. His answer was yes — almost certainly. John’s position is more pointed: a fully closed loop that works by being maximally aggressive with insulin creates a system that performs well for sedentary users and poses an underappreciated risk for anyone who exercises. The algorithm cannot see the difference between high glucose from food and high glucose from any other cause.

7. GLP-1 in type 1 diabetes — the case for priority access

John describes GLP-1 as the development he is most excited about from ATTD 2025 — more so than fully closed loop. The evidence base from the US, where hundreds of patients have been treated at single centres, shows 20–30% reductions in total daily insulin dose, better post-prandial control, and weight loss that is genuinely difficult to achieve in type 1 because of the anabolic effects of exogenous insulin. The current consensus in the UK suggests starting at 20% insulin reduction and not exceeding 7.5–10 mg. John’s argument for prioritisation is personal and direct: people with type 1 diabetes who have been taking exogenous insulin since childhood face a structural weight management disadvantage that people with type 2 do not. If GLP-1 access is rationed, he argues that type 1 patients should be near the front of the queue.

8. Continuous ketone monitoring — opportunity, unknowns, and risks

Abbott’s upcoming combined glucose-ketone CGM prompted one of the episode’s most nuanced conversations. The potential applications are real: DKA prevention in frequent flyers, safety monitoring for frail elderly patients, and for the first time a continuous view of what ketone levels actually look like across the day in people with type 1 diabetes who are not in crisis. But the unknowns are significant. Nobody yet knows what a normal continuous ketone profile looks like, or what threshold and duration on a continuous sensor should prompt clinical action — as opposed to the episodic ketone readings teams currently act on. And Amanda raises a concern rarely discussed in technology conversations: continuous ketone monitoring in patients with insulin restriction behaviours and eating disorders, where visibility of ketone levels could reinforce dangerous patterns rather than prevent them.

9. AID system optimisation and insulin on board

John describes what he learned from an advisory board presentation at ATTD: the key to getting the best out of any hybrid closed loop algorithm is ensuring that the algorithm-delivered insulin constitutes approximately 60% of total daily dose, with bolus ratios relaxed accordingly. This means the algorithm has enough latitude to respond to glucose changes without the user accruing large manual bolus IOB that blocks automated correction. The same principle applies across systems — 780G, Omnipod 5, Control-IQ, MyLife — but expressed differently in each. Tamsin provides a real-world illustration: switching from a 780G to MyLife using the same settings produced hypoglycaemia in 30% of readings within 48 hours because the algorithms behave fundamentally differently. Setting up fresh, as if starting from scratch, is the right approach. The GNL AID System Explorer allows DSNs and patients to model how each system behaves at different responsiveness levels before making decisions.

10. Critical appraisal training — the gap nobody fixes

A thread that runs through the whole conversation: nurses are not trained in how to read clinical evidence. They are trained in practice. When device companies present MARD data and CE marking as sufficient justification for an insulin-dosing indication, nurses have no framework to push back — not because they are incurious, but because the tools for questioning are not taught in undergraduate training. John describes the same dynamic from the other side, having worked in industry: sales teams are trained on the good numbers and the approved claims, not on how to handle a clinician who has read the study design questions. The chart, the scoring system, and this series of conversations are a partial answer to that gap. What the episode makes clear is that the gap itself needs to close at a structural level.

What the team actually uses — and what’s coming next

A recurring question after any CGM accuracy conversation: which devices do the guests actually use or prescribe? Here’s what came out of this episode.

The DSN Forum team — mainly Freestyle Libre 2 and Libre 3

John — mainly Dexcom G7, watching FSL3 and Roche SmartGuide

Medtronic — now adopting the Instinct sensor (Libre 3 platform)

Practical exploration checklist

For people with type 1 or type 2 diabetes:

• Before comparing CGM features, check whether the device has a score of five out of five on the DSN Forum chart — features only matter once the data standard is met
• Keep an in-date glucose meter and test strips, regardless of how reliable your CGM feels — the 40/40 outlier rate of around 1% is rare but real
• If you notice a reading that does not match how you feel, trust the symptom and use your finger-prick meter — the sensor is most likely to be wrong when it matters most (rapid glucose change, hypoglycaemia)
• If you are using a CGM that permits calibration and you choose to calibrate, always calibrate from a finger-prick blood glucose — never from the sensor glucose reading
• If you are switching AID systems, expect to reconfigure your settings from scratch rather than transferring directly — the algorithms behave differently and direct transfer can cause dangerous hypoglycaemia
• Use the GNL AID System Explorer to understand how your current settings compare to optimised settings for your system, and what changes would mean for your algorithm’s behaviour

For diabetes specialist nurses, dietitians, and clinicians:

• Apply the DSN Forum five-point scoring system as a first filter whenever evaluating a new CGM device — treat it as a data sufficiency gate before any feature discussion
• At annual reviews and device initiations, confirm that patients have a working glucose meter and in-date strips — this has become a gap as CGM adoption has grown
• When patients switch AID systems, initiate a fresh settings configuration rather than a direct settings transfer — Tamsin’s case demonstrates why direct transfer is a patient safety issue
• Understand the difference between device-reported IOB and physiological IOB when advising AID users about exercise — the GNL Exercise IOB Calculator models the gap between the two
• When considering GLP-1 in type 1 diabetes, note that US experience supports a starting insulin reduction of 20% with a ceiling of 7.5–10 mg, and that weight-related outcomes may be particularly meaningful in this population
• Approach continuous ketone monitoring in patients with insulin restriction behaviours with care — continuous visibility of ketone levels may have unintended effects in this group

About the guests

Amanda Williams is Lead Diabetes Nurse in East Kent and a founding member of the Diabetes Specialist Nurse Forum UK. Her clinical focus includes CGM initiation, device comparison, and building practical education frameworks for DSNs working across primary and secondary care. Amanda has been involved in developing the DSN Forum CGM comparison chart since its first iteration.

Beth Kelly is Clinical Lead Diabetes Specialist Nurse in Wiltshire, with a particular focus on community diabetes care and primary care education. She has been working to improve CGM literacy across community nursing teams and is one of the DSN Forum’s most active contributors to practical clinical guidance.

Tamsin Fletcher-Salt is Lead Diabetes Specialist Nurse at University Hospital North Midlands, where her work spans type 1 diabetes technology, AID system initiation, and clinical education. Tamsin currently uses an AID system herself, which gives her direct experience of the device-switching issues and algorithm optimisation challenges she describes in this episode.

All three are core members of the Diabetes Specialist Nurse Forum UK, whose CGM comparison chart is one of the most widely used practical resources in UK diabetes nursing.

Related reading on GNL

• Episodes 10–12 — CGM Series Part 1: How to Choose a CGM — where the GNL CGM Series started: the foundations of CGM choice, accuracy, and what the evidence means
• CGM Guide Part 2: Assessing CGM Accuracy and Performance — the written companion to the accuracy framework discussed in this episode, with detail on G7, Libre 3, Roche SmartGuide, and Medtronic
• CGM Guide Part 1: How to Choose a CGM
• CGM Guide Part 3: CGM Bells and Whistles
• Episode 35 — CGM Accuracy and Study Design with Professor Othmar Moser — the research basis for the five-point scoring system
• Episode 19 — iCGM vs eCGM Regulatory Standards (Dr Guido Freckmann)
• Episode 20 — Standardisation of Testing CGM Performance
• The CGM Black Swan: Pursuit of Simplicity and Asymmetric Risk
• GNL AID System Explorer — model how different AID algorithms behave at different responsiveness settings
• DSN Forum CGM Comparison Chart — the five-point accuracy scoring system and full device feature comparison including G7, Libre 3, Roche SmartGuide, and Medtronic