CGM Regulation: Achieving Standardisation for Outcomes

A simple question.

I wish the answer were yes. But, as Ben Goldacre (Bad Science)—and my favourite colleague in diabetes, the ever-curious and truth-seeking Prof Rob Andrews—both remind us, “I think you’ll find it’s a little more complicated than that.”

In reality, it’s a lot more complicated than that.

Right now, I’m deep in the weeds developing a model that aims to make sense of it all. It’s not quite ready yet, but I expect it to be in solid shape by early 2026.

Quick primer of what will be ready soon-ish:

A-T-B model of CGM user glucose exposure: Above-True-Below

Above True (A): The CGM device, on average, reads higher than your actual blood-glucose exposure. This can delay hypoglycaemia warnings and lead to unnessary glucose exposure, making hyperglycaemia look more frequent.
True (T): The CGM device, on average, reads closely with real biological exposure—this is the “honest” view of what your body experiences.
Below True (B): The CGM device, on average, reads lower than your actual blood-glucose exposure. It might look bad at first, but it’s often the safer pattern — it protects against hypos by being cautious. The trade-off is that you’ll need a higher time-in-range to reach the same HbA1c as CGM systems that reflect true glucose levels more accurately.

These behaviours shift depending on the time-in-range zone:

In time below range (TBR), a Below-True bias can exaggerate lows.
In time in range (TIR), small shifts up or down may have little consequence if the sensor is stable.
In time above range (TAR), an Above-True readings can make highs look worse than they are, and cause potentially dangeroues insulin corrections.

Consistency matters most. A CGM that’s steady—staying predictably Above, within, or Below True—is often more useful than one that drifts between the two.

So don’t take what follows as a verdict on which sensor is “good” or “bad.” A sensor that under-reads is not necessarily wrong—it usually means it carries a higher safety margin against hypoglycaemia, but you may need to set a slightly higher target range. That’s very different from assuming it’s underestimating your glucose exposure.

I used to think the same—that under-reading meant inaccuracy—until a deeper look at the evidence a couple of weeks ago changed that view. The picture is subtler, and I’ll share clearer guidance as the model matures.

HARD STOP

What follows is a simplified version of a more complex reality.
Helpful? Yes. Gospel truth? No.
By 2026, this will mature into the full A–T–B model you can discuss with your healthcare team.

How to read this now (November 2025 onwards):

Approach with caution, curiosity, and understanding. It is useful, but incomplete.
My current work is focused on enabling safe switching between CGM and AID systems while achieving equivalent long-term risk.
The mechanism is simple in principle: adjust time-in-range targets to match the calibration behaviour of the system you use.
We are not there yet. Until then, build your literacy, and treat this as a primer, not a verdict.

Stay involved

I’m iterating the model in real time and will share updates as it stabilises.
Thoughts and ideas are welcome: john@theglucoseneverlies.com.

Now, please read on for the context and why “70% will never mean 70%” across devices. But, the A-P-B model will change to the A-T-B model before long.

Understanding CGM zones, where your CGM device really measures

CGM systems are calibrated in the factory or by the user to measure glucose in one of three zones:

Zone A: CGM systems calibrated to read in Zone A produce glucose readings, on average, 5–10% higher than the body’s true physiological levels found in Zone P.
Zone P: CGM systems reading in Zone P produce glucose values that represent physiological blood glucose exposure — this is the level of glucose the body is truly experiencing.
Zone B: CGM systems calibrated to read in Zone B produce glucose readings that are, on average, 5–10% lower than the true physiological levels of Zone P.

These 5-10% differences may appear small, but they fundamentally change how glucose data should be interpreted, how targets are set, and they impact automated insulin delivery (AID) systems adjust insulin.

At present, all commercially available CGM systems with peer reviewed clinical data read in Zone P or Zone B.

Hint, the aim should not be 70% time in range for all CGM systems.

After reviewing CE-marking data, iCGM approvals, and the latest published accuracy datasets, I have been able to group most CGM devices into Zone P or B.

Some CGMs don’t have data available to make this assessment. They may sit in Zone A, P or B. Not being able to make this assessment is simply unacceptable! I hope you feel the same way too!

You can find the full references used to determine the groupings at the end of this page.

If you’d like to see the detailed alignment tables or learn how these values were derived, please email me directly at john@theglucoseneverlies.com. .

Measuring in Zone P and B is neither good nor bad. What matters is knowing where your CGM system reads, because that determines how you interpret the numbers and what “time in range” you should be aiming for.

Hint, the target is not 70% time-in-range for all CGM systems.

You now understand not all CGMs measure glucose in the same way, some read within the true physiological range, while others read below it.

Therefore, it’s time to see how this affects your clinical time-in-range targets.

CGM Systems reading in Zone P

CGMs calibrated to read in Zone P display glucose values that, on average, mirror what the body is truly exposed to. That is a glucose level between capillary and venous blood glucose exposure.

IMPORTANT NOTE: The evidence base for all current CGM targets — including the 70% time-in-range (TIR; 3.9-10.0 mmol/L or 70-180 mg/dL) benchmark, the link between CGM TIR and HbA1c, the finding that a 5 % change in TIR predicts future complication risk, and the improved pregnancy outcomes seen with tighter control — was built almost entirely on systems reading in Zone P.

In other words, the datasets connecting CGM metrics to HbA1c, long-term outcomes, and pregnancy health all come from CGM Devices aligned with true physiological glucose exposure.

If your CGM reads in Zone P, these standard clinical benchmarks apply directly — because they reflect the body’s real glucose exposure and the evidence that defines safe, effective diabetes management.

Time in Range (TIR): 3.9–10.0 mmol/L (70–180 mg/dL), goal ≥70%
Time in Tight Range (TITR): 3.9–7.8 mmol/L (70–140 mg/dL), goal ≥50%
Pregnancy TIR (TIRp): 3.5–7.8 mmol/L (63–140 mg/dL), goal ≥70%

Again, because it is so important to understand. The well known clinical targets used across the world for all CGM devices, are the correct targets, only if your CGM measures in Zone P.

CGM systems calibrated to read in Zone P

Device	Manufacturer	Bias to venous blood glucose	Comparator
FreeStyle Libre 2 / Libre 2 Plus / Libre 3 / Libre 3 Plus	Abbott Diabetes Care	+5-6%	Venous
Dexcom G7 (adult & paediatric)	Dexcom Inc.	+3-5%	Venous
Dexcom G6 / ONE / ONE+	Dexcom Inc.	+3%	Venous
Roche Accu-Chek SmartGuide	Roche Diagnostics	+3% (inferred)	Capillary
Medtrum TouchCare Nano A8	Medtrum Technologies	+0.5%	Venous
i-SENS CareSens Air	i-SENS Inc.	+0-0.5% (inferred)	Capillary

CGM Systems reading in Zone B

On average, these CGM systems report glucose about 5–10 % lower than venous blood glucose, meaning:

Displayed CGM glucose values climb more gently after eating and fall more gradually after a correction.
Low alerts trigger earlier, which can help protect against hypoglycaemia.
The body is typically exposed to more glucose than the CGM values suggests, again by about 5–10 %.

CGM systems calibrated to read in Zone B

Device	Manufacturer	Bias (%) to venous blood glucose	Comparator
Sinocare iCan	Sinocare Inc.	−0.3%	Venous
Infinovo GlucoNovo	Infinovo Medical Ltd	−1%	Venous
Senseonics Eversense E3 / 365	Senseonics Inc.	−5%	Venous
Medtronic Simplera / Guardian 4	Medtronic MiniMed Inc.	−10%	Venous

For CGM systems that read in Zone B, reaching 70% TIR can seem easier because these CGM devices report lower and smoother glucose values. They underestimate true physiological exposure, meaning glucose levels don’t appear to rise or fall as sharply.

To reflect true physiological exposure, the TIR target should be set about 5–10% higher — giving the same real-world glucose exposure as 70% TIR on a Zone P device.

That means:

Time in Range (TIR): 3.9–10.0 mmol/L (70–180 mg/dL), goal ≥75-80%
Time in Tight Range (TITR): 3.9–7.8 mmol/L (70–140 mg/dL), goal ≥55-60%
Pregnancy TIR (TIRp): 3.5–7.8 mmol/L (63–140 mg/dL), goal ≥75-80%

CGM Devices that we don not know which glucose zone they display on their CGM screens and downloads

Device	Manufacturer	Reason / Status
GS1 (SiBionics)	SiBionics Ltd	User guide only – no clinical performance or comparator data in uploads
Linx CGM	MicroTech Medical	CE IFU only – no paired venous data
Yuwell CT3	Yuwell / POCTech	Brochure only – no accuracy metrics or comparator type
Urathon CGM	Urathon / POCTech	User manual – no bias data or comparative study
Syai Tag	Syai Health Technology	Product PDF only – no validated in-clinic dataset
GlucoRx AiDEX	MicroTech	Not reported
GlucoMen Day & Medtrum A6	Obsolete devices – excluded from current market set

Practical implications

Targets for TIR must be achievable in real life. For CGM systems reading below physiological glucose (Zone B), reaching 75–80 % Time in Range (TIR) is not necessarily harder. In fact, the traces tend to move more smoothly, which can make it easier for both users and automated insulin delivery (AID) systems to stay on top of fluctuations.

What matters is tuning your targets and hypo thresholds to the zone your system measures in. AID algorithms act directly on the CGM values. If the CGM consistently reads lower than true physiological glucose, the algorithm may reduce or suspend insulin earlier than intended, increasing actual glucose exposure.

The same applies to manual decisions. A user may feel comfortable at 9.4 mmol/L (170 mg/dL) when their true physiological glucose is closer to 11.1 (200 mg/dL). That gap can change what action is taken, and when.

These are practical points for people using CGM systems that read in Zone B, and for healthcare professionals supporting them. Discuss these adjustments with your diabetes team, this is not medical advice:

Target glucose: consider setting the AID target towards the lower end of the range.
Low alert limit: consider reducing your hypo alert threshold by about 0.5 mmol/L (10 mg/dL) to prevent unnecessary treatments for readings that appear low but are still within the ideal physiological range.
High alert limit: consider lowering your high alert by 1–2 mmol/L (10-20 mg/dL) so that action is taken sooner and post-meal peaks are better controlled.

These small adjustments can help achieve the extra 5–10 % TIR needed for equivalent glucose exposure compared with a CGM system reading in Zone P.

Again, this is not “good” or “bad” — it’s simply about awareness, understanding, and appropriate action.

Always confirm changes with your diabetes team before adjusting settings.

Interpretation Caution for Researchers

When interpreting study results that employ CGM systems spanning both Zone P (physiological alignment) and Zone B (below-venous alignment), it is essential to recognise that these devices are not directly comparable.

Meta-analyses or pooled data sets that combine CGMs with differing physiological calibration anchors introduce systematic bias, particularly when outcomes such as Time in Range (TIR) are treated as equivalent across devices. Systems that under-read relative to venous glucose (Zone B) will artificially inflate apparent TIR, whereas those aligned with capillary or venous reference (Zone P) reflect true physiological exposure.

This discrepancy is amplified when TIR is used as the primary endpoint without concurrent measures of HbA1c to contextualise CGM system. Consequently, researchers should avoid assuming equivalence across CGM systems in pooled analyses and must report comparator type and bias zone explicitly to preserve interpretive validity.

Still here?

A special shoutout to Kirsten de Klerk from South Africa Diabetes Advocacy 🇿🇦 who is leading the charge in getting CGM for all in South Africa. Check out her Podcast episode from Burnout to CGM Access for All in South Africa.

Kirsten reached out to give three powerful upgrades that will make this message hit home.

Kirsten’s Summary: Top Three for CGM Integrity

Not all CGMs measure glucose the same way, and that matters!
Some CGM dispay vlaues that match what the body is exposed to (between capillary and venous blood glucose), while others read lower than what the body is exposed to (below venous blood glucose), and this is best shown in a graph showing a full day.
- The dashed red line represents CGM systems reding in Zone P, and the blue dashed line represents CGM systems reading in Zone B

3. We urgently need a single, global accuracy standard for CGMs, so in the future, 70% TIR means the same thing across all CGM systems.

Let’s put all of this into a pracical context.

Continuous glucose monitoring (CGM) has no rival in day-to-day glucose management.

The metrics generated by CGM have been proven to be among the strongest predictors of future diabetes complications and overall health.

But with a new wave of CGM systems gaining approval without publicly available performance data, and lots of existing devices measuring below venous glucose (Zone B), 5-10% lower than the levels used in the studies that built our current evidence base — we’re entering uncertain territory.

For some of these newer CGM systems, because their performance evaluations are not made publicly available, we simply don’t know whether their readings align more closely with capillary or venous glucose.

We also don’t know whether a lot of the new CGMs have been tested rigorously — or on a sufficient number of people with type 1 diabetes, including children. That information just isn’t publicly available.

And that raises a fair question: why not?

EU law — which covers almost all CGM systems currently on the market — doesn’t require manufacturers to share performance data. They’re simply not legally obliged to.

This is beginning to change under MDR (2017/745), which introduced plans for a central European database of medical device performance data. But that system still isn’t fully operational.

Then again… it was only agreed back in 2017 — so there’s still time, right?

Yet, the established CGM companies choose to publish their data openly in peer-reviewed journals and on regulatory websites.

That transparency is the foundation of trust — it’s what enables clinicians and people with diabetes to use CGM data with confidence.

At the same time, marketing claims are emerging from devices that under-read venous glucose, resulting in apparently higher time-in-range values — numbers that may look better on screen but don’t necessarily translate into the same long-term health benefits as those from CGMs proven to run close to venous glucose.

Here’s a glimpse into a potential future for many people using CGM — and a reality for some today. You might want to find out if that includes you.

Imagine this…..

You spend four years at university. Your tutor tells you you’re consistently averaging 70% on your assignments — so you’re confident you’ll graduate with a first-class degree.

But when the final marks come in, you discover you’ve actually been scoring 60% all along.

The tutor had been marking too generously. Instead of a first, you graduate with a second-class degree. That one number will follow you for years, closing doors to jobs and opportunities you thought were yours.

Now apply that to diabetes.

Imagine this….

You and your identical twin both have type 1 diabetes. For 20 years, you’ve each worn different CGM systems.

Both of you hit the gold standard—70% TIR—year after year.

You didn’t pay much attention to your HbA1c results because you were told 70% TIR was all that mattered.

You start to get complications first, and they are progressing much faster.

You cannot understand why. Was it just bad luck?

You review your diabetes history and find that your HbA1c has consistently been 7.5% (58 mmol/mol), while your sibling’s has been 6.5% (48 mmol/mol), despite both of you recording the same TIR on your CGMs for 20 years.

Why is my HbA1c 1.0% (10 mmol/mol) higher?

We both had 70% TIR?

There are biological reasons. HbA1c is not determined solely by average glucose; red blood cell lifespan and glycation efficiency vary between people.

For the definitive take, see James Hempe’s work on biological variation and the “hemoglobin glycation index.”

Believe it or not, the body’s ability to use vitamin C to recycle key antioxidants in the red blood cells may play a key role — but that’s a story for another day, and maybe a podcast with Dr Hempe.

In this hypothetical case of identical twins, we can safely set aside biological variation.

So, we must explore other explanations.

Could it be that your CGM system is reporting glucose differently from your identical twin’s CGM?

Important physiology interlude.

Capillary blood contains glucose freshly delivered from the heart to the tissues, before some is taken up by the body’s cells for energy. Venous blood, on the other hand, represents blood returning from the tissues after glucose has been partly used for energy. For this reason, capillary glucose levels are physiologically 5-10% higher than venous levels overall and up to 30% higher after eating.

Now, imagine if your CGM consistently reports glucose values that are lower than venous glucose, whereas your sibling’s CGM reads very close to capillary glucose.

If that was the case, the natural question.

Wait… the data that linked glucose to risks for my eyes, feet, heart, and brain — was that based on CGMs reading close to capillary glucose, or on ones that run below venous glucose?

Every landmark study in type 1 diabetes — most famously DCCT and EDIC — relied on HbA1c as the key measure of glucose exposure. HbA1c represents how much glucose binds to red blood cells and reflects your average glucose levels over roughly three months.

From these data, robust models were built showing that as HbA1c increases, the risk of both short- and long-term diabetes complications—such as retinopathy, kidney disease, heart attack, and stroke—rises in parallel. The reverse is also true: as HbA1c decreases, complication risk falls.

While HbA1c is not a perfect causal marker, the association is very strong and has been validated over more than 30 years of follow-up. Each 10 mmol/mol (≈1.0% HbA1c) change in either direction modifies complication risk by roughly 30–40%, across both microvascular and macrovascular outcomes.

CGM researchers have shown that a 5% change in TIR corresponds, on average, to about a 5 mmol/mol (≈0.5%) change in HbA1c. As TIR rises, HbA1c tends to fall, and vice versa. The relationship is very strong.

Encouragingly, more recent studies are bypassing HbA1c as the middleman and mapping CGM data directly to complication risk. These data suggest that a sustained 5% improvement in TIR is clinically meaningful and 70% is a good target.

Better still, data looking over 7 years shows 5% changes in TIR modify the risk of eye and kidney disease by around 20%.

Please note;

All the research establishing 70% TIR as the target has been based on data from CGM systems shown to align closely with capillary glucose, not reading below venous glucose.

Remember, venous runs 5-10% below capillary.

What does that mean in practice?

It means that if you record 70% TIR on CGM consistently reading below venous glucose, your HbA1c will be higher than someone getting 70% TIR on a CGM reading close to capillary glucose.

How much higher?

Well, the recent big study of all commercially available Automated Insulin Delivery (AID) systems reports differences of up to 10% in TIR between systems, but almost identical HbA1cs, aka, future health risk!

See below, it should not be a surprise that AID 5 is the only system driven by CGM readings in Zone B.

AID System – See the 2025 analysis	End-of-trial TIR % (95% CI)	End-of-trial HbA1c % (95% CI)
AID 1	72.4 (57.8–87.0)	7.1 (6.4–7.7)
AID 2	71.8 (69.4–74.2)	7.0 (6.9–7.1)
AID 3	67.9 (66.2–69.6)	7.3 (7.1–7.5)*
AID 4	69.0 (67.5–70.5)	7.1 (7.0–7.2)*
AID 5	74.4 (69.7–79.1)	7.1 (6.8–7.4)
AID 6	63.1 (59.4–66.8)	7.2 (7.1–7.3)

To be clear, in our identical twin thought experiment, that would translate to a 20–40% higher risk of eye and kidney disease, and around a 30% higher risk of heart attack and stroke.

So, when the manufacturer of AID 5 publishes ‘real-world data’ showing their AID system achieves 5–10% more time-in-range than the others, does that mean it results in lower diabetes complications risk?

Shall we inspect the science?

Remember, the CGM data used to show 70% time in range = HbA1c of 52 mmol/mol or 7% was from in CGM systems reading glucose levels closely aligned to capillary glucose (Beck et al 2019 and Beck et al 2023), not CGM data reading below venous glucose.

Why is this important?

It’s been shown that a current market leading CGM device reads glucose levels below venous glucose. This resulted in a systematic over-reporting of TIR by 8% compared to CGM devices reading close to capillary glucose.

What does this mean?

CGM data from systems that report glucose levels below venous glucose are not directly comparable with data from systems that read close to capillary glucose.

So, these types of claims of superiority may hold true for CGM systems don’t hold for what matters the most, your future risk of diabetes complications.

Are these AID systems driven by CGM sensor glucose levels reading below venous glucose inferior?

Not for the average person with type 1 diabetes (if there is such a thing).

The honest truth is this: the largest and most robust systematic review shows that all AID systems deliver the same outcome when it comes to the only marker we can trust when comparing different AID systems, that is, HbA1c.

Despite AID 5 reporting 5-10% more TIR, across systems, average HbA1c levels consistently sit at ~52 mmol/mol (7.0%).

So when it comes to future health, AID 5 is not better than the rest—it’s on par.

And maybe for these type of CGM devices, the target is actually 75-80%, not 70% you need when using a CGM system reading close to capillary glucose.

But nobody is teaching this, well apart from me, and now you!

What about special populations, where small glucose measurement errors carry a bigger impact.

Pregnant ladies with type 1 diabetes?

A new real-world study of 137 women with type 1 diabetes (Quirós et al., 2025) reported some interesting data.

Women using an AID system paired with a CGM that reads below venous glucose ended up with higher HbA1cs and a greater chance of having large-for-gestational-age babies.
In contrast, women using systems with CGMs aligned reading close to capillary glucose had lower HbA1cs and fewer large-for-gestational-age babies.

The reason?

Put simply: when a CGM reading approaches hypoglycaemia, the AID algorithm suspends insulin delivery to prevent levels from falling further — and ideally nudges glucose back into the target range, but may cause insulin deficiency, especially just before eating, leading to higher after-eating levels.

Also, if the user sees the level trending low, what will they do? Treat with carbohydrate to push the glucose higher.

So, if the CGM readings are reading lower than venous glucose, then the true capillary glucose—the level that actually drives HbA1c and fetal growth—will be getting increased unnecessarily!

I’m not stating this as certain, but it is a very plausible mechanism explaining the findings.

Now, this wasn’t a randomised controlled trial, and we don’t yet have multiple studies pooled in a meta-analysis—so it’s not causal evidence.

But given the thread of evidence already discussed, you can make your own mind up about which CGM you’d want driving an AID system in pregnancy.

Can CGM be used in early type 1 diabetes in people with islet autoantibodies?

Would I use CGM to track the progression of early-stage type 1 diabetes for my son, Jude?

Possibly — but only if the CGM is closely aligned to capillary glucose.

A CGM device reading below venous glucose won’t reliably catch those after-meal capillary spikes with the fidelity needed to track progression.

So yes, I’d probably use CGM — but with conditions:

I’d choose one aligned to capillary glucose.
I’d calibrate it every morning against an accurate finger-prick meter aligned to capillary glucose (e.g., Contour Next or Roche Smart Guide ).
I’d focus on 30-day averages, not single blips in a day that are likely measurement error that every CGM has!.

That way, you get the best of what CGM can offer, while avoiding being misled by under-reporting sensors, and being stressed every minute of every day!

This screening and monitoring piece really needs a full FAQ in itself — and that’s coming soon.

⚠️ And Now for the Scary Bit

All of the CGMs we’ve talked about so far are market leaders backed by robust, publicly available clinical data.

But today, CGMs are being sold without any public clinical evidence at all.

We don’t know if they read closer to venous or capillary glucose, we don’t know their accuracy, and some are being prescribed for insulin dosing and used to make assumptions about future health risks.

This is exactly why standardisation and stronger regulation are urgently needed.

Bottom line: until CGM performance is standardised and we can see the data, 70% on one system may not be the same as 70% on another.

This is why 2025’s hottest topic is standardisation of CGM performance — to get clarity for people living with diabetes, clinicians, and future research determining the evidence base.

I’ve deliberately avoided mentioning specific CGM or AID system names, because this isn’t about calling out companies. It’s about highlighting a major issue that needs tackling.

Getting lost in petty arguments over small details is pointless. What we need is to come together to make the future clearer — and to standardise so that 70% TIR on one CGM system means at least 68–72% on another.

We’ll never achieve absolute perfection, but we can get far closer than the current 60–80% spread. That level of inconsistency simply isn’t acceptable in 2025, especially in high-resource countries.

Nice work — you made it to the end!

You’ve now got everything you need to teach others or share this 101 as a starting point.

Want to level up and become a CGM accuracy pro?

Ready for Part 2: The Deep Dive: Regulation & Study Design — how accuracy is defined, tested, and reported, and what that means for real-world decisions.

References

Alva, S., Bailey, T., Brazg, R., Budiman, E., Castorino, K., Christiansen, M. P., et al. (2020). Accuracy of a 14-day factory-calibrated continuous glucose monitoring system in adults and children with diabetes. Journal of Diabetes Science and Technology, 16(1), 70–77.

Alva, S., et al. (2023). Accuracy of the third generation of a 14-day continuous glucose monitoring system. Diabetes Therapy, 14(4), 767–776.

Bailey, T., et al. (2015). The performance and usability of a factory-calibrated flash glucose monitoring system. Diabetes Technology & Therapeutics, 17(11), 787–794.

Eichenlaub, M., Waldenmaier, D., Wehrstedt, S., et al. (2025). Performance of three continuous glucose monitoring systems in adults with type 1 diabetes. Journal of Diabetes Science and Technology. Advance online publication. https://doi.org/10.1177/19322968251315459

Freckmann, G., et al. (2025). iCan (Sinocare) versus arterialised venous YSI: head-to-head evaluation. EASD 2025 Poster.

Garg, S., et al. (2022). Accuracy and safety of the Dexcom G7 continuous glucose monitoring system in adults with diabetes: a prospective multicenter study. Diabetes Technology & Therapeutics, 24(11), 1–10. (Adult pivotal; venous YSI).

Hochfellner, D. A., et al. (2022). Accuracy assessment of the GlucoMen Day® continuous glucose monitoring system. Biosensors, 12(1), 1–13. (Pilot; venous YSI).

Ji, L., Guo, L., Zhang, J., Li, Y., & Chen, Z. (2021). Multicenter evaluation comparing a new factory-calibrated real-time CGM system (AiDEX) to an existing flash system. Journal of Diabetes Science and Technology, 15, 1–6. https://doi.org/10.1177/19322968211037991 (MARD 9.08% vs venous YSI).

Laffel, L. M., et al. (2022). Accuracy of a seventh-generation continuous glucose monitoring system in children and adolescents with diabetes. Journal of Diabetes Science and Technology, 16(6), 1240–1248. (Paediatric pivotal; venous YSI).

Mader, J. K., et al. (2024). Pivotal evaluation of the Accu-Chek SmartGuide continuous glucose monitoring system using a capillary comparator. Journal of Diabetes Science and Technology, 18(5), 1201–1213. (Mean bias −2.2% vs capillary; MARD 9.2–9.6%).

Meng, R., et al. (2021). Clinical accuracy of the Glunovo continuous glucose monitoring system versus venous EKF reference. Journal of Diabetes Science and Technology, 15(6), 1365–1373.

NCT04436822. Rostock (Simplera) in-clinic dataset (ClinicalTrials.gov study record and exported dataset). (Adults with type 1 diabetes; in-clinic frequent sampling).

Senseonics (2023). Eversense E3 / Eversense 365-day implantable CGM: evaluation of accuracy and safety (venous-calibrated). Journal of Diabetes Science and Technology (supplement/white paper package). (Uploaded evaluation document confirms 365-day program).

Wadwa, R. P., et al.; Shah, V. N., et al. (2018). Pivotal evaluations of the Dexcom G6 CGM system in adults and youth. Diabetes Technology & Therapeutics, 20(6), 428–433; Journal of Diabetes Science and Technology, 12(3), 1–9. (Venous/arterialised YSI comparators).

White paper — i-SENS (2024). CareSens Air CGM System: Clinical accuracy (8.7% MARD) vs venous YSI in clinic. (Manufacturer white paper).

Zhou, J., et al. (2018). Performance of a new real-time continuous glucose monitoring system (Medtrum A6 TouchCare): multicenter study with venous YSI reference. Journal of Diabetes Investigation, 9(4), 860–866.

CGM FAQ Part 1: THE most important thing you will read this year!