Japan Study: FreeStyle Libre's New Algorithm More Consistent

A Study from Japan Reveals Important Differences in FreeStyle Libre CGM Accuracy

A study from Japan's NHO Kyoto Medical Center has shed new light on how software updates to one of the world's most popular continuous glucose monitors (CGMs) — the FreeStyle Libre — affect the accuracy of blood sugar readings. For the millions of Americans who rely on this device to manage diabetes, these findings carry practical, day-to-day implications.

What the Japanese Researchers Did

Researchers at the Department of Clinical Nutrition in Kyoto recruited 11 people with type 1 diabetes and asked each participant to wear two FreeStyle Libre sensors simultaneously — one on each arm. One sensor ran on the older first-generation (Gen. 1) algorithm, while the other used the newer third-generation (Gen. 3) algorithm. Both were then compared against traditional fingerstick blood glucose readings, considered the gold standard for accuracy.

This side-by-side design, unusual in CGM research, gave scientists a rare direct comparison between the two software versions in real-world conditions.

What They Found: Consistency vs. Raw Accuracy

Neither algorithm was perfect, but they failed in different ways:

• Gen. 3 algorithm: Showed a bias (consistent offset) of 7.4 mg/dL and a Mean Absolute Relative Difference (MARD) of 11.9%. Importantly, it showed no proportional bias — meaning its errors stayed roughly the same whether blood sugar was low, normal, or high.
• Gen. 1 algorithm: Showed a smaller average bias of 4.4 mg/dL and a slightly better MARD of 9.7%. However, it did show proportional bias — meaning its errors grew larger as blood sugar levels moved higher or lower.

In plain terms: the older algorithm was slightly closer on average, but became less reliable at the extremes — exactly when accurate readings matter most, such as during hypoglycemia or post-meal spikes.

Comparing to US Guidelines

The American Diabetes Association (ADA) recommends CGM devices achieve a MARD below 10% for clinical reliability. Both algorithms in this study hovered near or above that threshold (9.7% and 11.9%), reinforcing that CGMs — while highly useful — should not entirely replace fingerstick confirmation in critical situations, such as before treating a low or making a major insulin adjustment.

Why This Matters for US Patients

The FreeStyle Libre is one of the most widely used CGMs in the United States. Abbott, its manufacturer, has been rolling out algorithm updates globally, and many US users may already be using the Gen. 3 algorithm without realizing it. This Japanese research is valuable because:

• It highlights that a lower average error doesn't always mean better real-world performance — consistency across blood sugar ranges matters just as much.
• It suggests that users on Gen. 3 should be especially cautious at blood sugar extremes and consider a fingerstick confirmation when readings seem off.
• It reminds patients to ask their care team which algorithm version their device currently uses, as software updates can change device behavior.

If you use a FreeStyle Libre, this study is a good reminder that your CGM is a powerful trend-tracking tool — but understanding its limitations helps you use it more safely.

Citation

Murata, Sakane, Hirota. "Difference in the accuracy of the third-generation algorithm and the first-generation algorithm of FreeStyle Libre continuous glucose monitoring device." The Journal of Medical Investigation (JMI), Vol. 71, pp. 225–231, August 2024. PMID: 39462556. DOI: 10.2152/jmi.71.225