ISO 15197:2013 Accuracy Standards

ISO 15197:2013 Accuracy Standards

ISO 15197:2013 is the international standard that says how close a home blood glucose meter and its test strips have to come to a laboratory result. Its core rule has two parts. At least 95% of a system's results must fall within ±15 mg/dL of the lab comparison when glucose is below 100 mg/dL, or within ±15% at 100 mg/dL and above, and at least 99% must land in zones A and B of the Consensus Error Grid [1]. In the US, the standard is a reference point rather than the bar a meter is cleared against. The FDA has its own accuracy recommendations, and they are stricter.

What changed from the 2003 edition

The first edition, ISO 15197:2003, allowed more error. It asked for 95% of results within ±15 mg/dL below 75 mg/dL and within ±20% at 75 mg/dL and above [2]. The 2013 edition narrowed the percentage band from 20% to 15%, and it raised the point where the rule switches from mg/dL to percent, from 75 to 100 mg/dL [2]. Meters tested around the changeover were often checked against both editions in a single study; a 2015 evaluation of the Contour Plus system reported results for each [2]. We trace the longer history in ISO 15197 Accuracy Standards Evolution.

Here's what surprised us: even the looser 2003 bar wasn't met by every meter on sale. Researchers in Ulm, Germany, tested 43 CE-marked systems against it. They could fully assess 34 of them, and 7 of those 34 failed [3]. Their conclusion was blunt. The CE mark, they wrote, apparently does not guarantee that a system is as accurate as the standard requires [3].

How the Consensus Error Grid works

A percentage tells you how far off a reading is. It doesn't tell you whether the miss matters, and the same 20% error can be harmless at one glucose level and risky at another. The Consensus Error Grid, often called the Parkes error grid after its lead author, was built to judge that. Its developers asked 100 diabetes experts to sort measurement errors into five risk categories, then drew separate grids for type 1 and type 2 diabetes, each split into concentric zones [4].

On the grid, the laboratory value runs along one axis and the meter's value along the other. Each zone stands for one of the panel's risk categories, so where a reading lands says how much its error could matter [4]. Published evaluations describe zones A and B together as clinically acceptable [5]. The grid was developed in 1994 and published in 2000, and ISO 15197:2013 wrote it into the accuracy test [6] [4]. In practice, a system can pass with up to 1 result in 100 outside zones A and B, as long as it also meets the ±15 rule [1].

What else a 2013 evaluation measures

System accuracy is only one part. The published studies we read describe three:

• System accuracy (clause 6.3). Capillary blood samples, spread across a required distribution of glucose levels, are measured on the meter and on a laboratory comparison method [5]. One evaluation ran this with three reagent lots [1].
• Precision. Repeat measurements of control solution and venous blood show how closely results agree with each other. The standard asks for these numbers but sets no pass or fail limit [5].
• User performance (clause 8). Accuracy is checked again in users' hands, with one strip lot in the study we read [1].

Where the FDA parts ways with ISO

ISO 15197 is a voluntary consensus standard. The FDA has a guidance on how manufacturers can use such standards in premarket submissions [7], and its guidance for over-the-counter meters notes that many manufacturers design their studies around ISO 15197 [8]. Then it draws a line. The agency wrote that the ISO criteria "are not sufficient to adequately protect lay-users," in part because the standard doesn't adequately address performance in the low-glucose range or across test strip lots [8].

So the FDA asks for more. For home-use meters, it recommends that 95% of results fall within ±15% and 99% within ±20% of the comparison method across the device's whole claimed measuring range [8]. There's no mg/dL allowance at the low end. The study should use at least 10 vials of strips covering at least 3 lots, shipped under typical conditions before testing [8]. If a meter can't stay within 15% at very low concentrations, the FDA says the maker should raise the low end of the claimed measuring range [8].

Prescription meters used by health care professionals at the point of care face a tighter bar: 95% within ±12% (±12 mg/dL below 75 mg/dL) and 98% within ±15% (±15 mg/dL below 75 mg/dL) [9]. For these meters, the FDA doesn't recommend the ISO criteria at all [9]. Both guidances are dated September 2020 and replaced versions issued in October 2016 [8] [9]. They are recommendations, not binding rules, and the FDA notes that a meter can still be found substantially equivalent when it misses them if other features make up for it [8].

Do meters on sale meet these bars?

Not always. For a study published in 2018, researchers bought 18 FDA-cleared meters from consumer outlets, systems that represented about 90% of those in use from 2013 to 2015, and tested them on 1,035 people [10]. A reading counted as compliant if it fell within 15 mg/dL below 100 mg/dL or within 15% at 100 and above, the same bands ISO 15197:2013 uses [10] [1]. Each meter went through three separate studies. Six of the 18 met the accuracy standard in all three, and four met it in none, even though the study's pass mark was more lenient than the regulatory one [10].

Reading an accuracy claim on a box

When a meter's materials say it meets ISO 15197:2013, that tells you the maker's study reached the 95% and 99% marks under study conditions [1]. It doesn't say whether the meter also meets the FDA's tighter home-use targets, how the strip lot in your hands performs, or whether anyone outside the company checked. Independent studies like the two above are where that last answer comes from, and lot-to-lot differences get their own article in Post-Market Surveillance and Lot-to-Lot Variability. If your meter's readings don't seem to match how you feel or a lab result, that's a conversation to have with your care team.