Beyond MARD: why adaptive sensing could reshape CGM accuracy

Image AI generated

For years, MARD has been one of the headline figures used to compare CGM accuracy.

Healthcare professionals at all levels discuss it and it's a hot topic within Facebook forums for those living with diabetes, often using the figures provided by medtech manufacturers to compare sensors when trying to understand which system might perform best.

At first glance, it seems a straightforward way to evaluate CGM systems. A lower percentage suggests a sensor is closer to laboratory reference glucose values.

However, the reality is more complicated.

As diabetes technology evolves, the conversation around accuracy is beginning to move beyond a single number. Increasingly, manufacturers are exploring AI for adaptive sensing  by applying more intelligent algorithms to respond to real-world conditions and individual users.

The shift is significant because accuracy is not simply about how a sensor performs in a controlled study. It is also about how much confidence people have in the readings they see every day.

Editor's comment: A recent survey by Love My Libre across online forums showed accuracy was the second most important factor to CGM users.

Why MARD does not tell the full story

MARD stands for Mean Absolute Relative Difference. It compares CGM readings against reference glucose measurements and provides an average indication of how closely the two align.

In simple terms, lower numbers are generally seen as better.

That sounds straightforward, but MARD is not yet a fully standardised international comparison measure. Diabetes technology experts have repeatedly called for more consistent reporting standards because published figures can vary depending on how studies are designed.

Participant numbers, glucose ranges tested, rates of glucose change, sensor placement and the reference method used can all influence the final result.

As a result, comparisons between brands are often less straightforward than they appear.

Published MARD comparison

Published manufacturer figures are not strict like-for-like comparisons because study methods differ.

Looking at the table, it is easy to start ranking devices. FreeStyle Libre 3 Plus appears to post the lowest figure, while Dexcom G7 and FreeStyle Libre 2 Plus sit broadly level.

Yet this is where things become more complicated.

Different manufacturers cite different studies, different participant groups and different testing conditions. A lower published percentage does not automatically mean a sensor will feel more accurate in day-to-day use.

For many people living with diabetes, the more relevant question is often much simpler: can I trust the number when it matters?

Accuracy is only part of the story

Dexcom recently highlighted that 97% of users rate accuracy as highly important.

That is hardly surprising. Accuracy remains one of the most important aspects of any CGM and is one of the easiest performance measures to compare.

But anyone who spends time reading diabetes forums will recognise that user conversations rarely focus on accuracy alone.

People certainly talk about false lows, overnight readings that do not seem believable, sensors that feel unreliable during the first day and numbers that look very different from finger-prick results.

At the same time, many of the most common frustrations have little to do with published accuracy figures.

Users talk about sensors being knocked off early, Bluetooth dropouts, signal loss, app problems, missed alerts, failed starts and sleep disruption. These are often the issues that shape confidence in a system over weeks, months and years of use.

The distinction is important because people living with diabetes tend to judge a CGM differently from the way manufacturers present them.

Companies focus on measurable improvements. Users focus on reliability.

They want to know whether a sensor behaves predictably during exercise, whether it remains dependable overnight, whether it settles quickly after insertion and whether it broadly matches how they actually feel.

Those experiences are much harder to capture in a single percentage.

What affects CGM accuracy in real life?

CGM performance is influenced by more than the sensor itself.

Because CGMs measure glucose in interstitial fluid rather than directly in blood, timing differences are normal, particularly when glucose levels are rising or falling quickly.

As explored in Love My Libre's guide to improving CGM accuracy, several everyday factors can affect readings.

Hydration can influence interstitial fluid balance. Exercise can create rapid glucose movement that sensors may struggle to track perfectly in real time. Pressure on a sensor during sleep can contribute to compression lows, while the first day after insertion can sometimes feel less predictable than the rest of the wear period.

Hot weather can also play a role, affecting both sensor wear and, in some situations, confidence in readings.

These factors help explain why two people wearing the same CGM can report very different experiences.

The move towards adaptive sensing

This is where adaptive sensing becomes interesting.

Rather than relying solely on smaller sensors or incremental hardware improvements, manufacturers are increasingly exploring ways to make sensors respond more intelligently to changing conditions.

The idea is that not every wearer, every glucose pattern or every situation behaves in exactly the same way.

Better adaptive algorithms could help a sensor tune performance more intelligently to the individual wearer over its life, helping distinguish genuine glucose movement from temporary signal noise, pressure effects during sleep, rapid rises or falls, exercise-related variability and the instability sometimes seen after insertion.

In other words, accuracy becomes less about a fixed specification and more about how effectively a system interprets what is happening in real time.

That represents a subtle but important change in thinking.

The question MARD cannot answer

For all the attention MARD receives, it cannot fully capture what it feels like to live with a CGM every day.

Most people with diabetes do not spend their time comparing percentage points. They remember the overnight false low that woke them unnecessarily. The sensor that felt unreliable for the first day. The alert that arrived too late, or the reading that simply did not match how they felt.

That helps explain why conversations about CGM performance often sound very different from manufacturer presentations.

Companies naturally focus on measurable improvements. Users tend to focus on confidence. They want to know whether a sensor behaves consistently during exercise, whether it stays reliable overnight, whether it recovers quickly after pressure-related lows and whether they can trust it when making everyday decisions.

Adaptive sensing is interesting because it attempts to address some of those real-world challenges rather than simply chasing another small improvement in a published accuracy figure.

If future systems become better at recognising changing conditions, adapting to individual wearers and filtering out some of the factors that create uncertainty, the conversation around CGM performance could begin to shift.

The question may become less about which sensor posts the lowest MARD and more about which sensor earns the greatest confidence over the course of everyday wear.

About the editor

Samantha is editor of Libre Life and founder of Love My Libre. She writes about diabetes medtech, CGM wear, hybrid-closed loop systems and practical real-world device use.