Interference Testing Protocols

Interference Testing Protocols for [Diabetic Test Strips](/sell-diabetic-supplies/)

An interference study asks a narrow question: if a given substance is in the blood, does the strip report a different glucose number? Every major protocol answers it the same way. You split one blood sample in two, add the substance to one half and the same volume of plain solvent to the other, then compare what the meter reads on each [1]. The FDA's 2020 guidance documents, ISO 15197:2013 and CLSI EP07 differ on the details: which substances to test, how much of each, and how big a gap counts as interference.

Three rulebooks, one basic design

In the US, the rules come from two final FDA guidance documents issued in September 2020, one for over-the-counter (home) meters and one for prescription meters used at the point of care [1][2]. Both describe the same interference study and the same minimum list of substances [1][2].

ISO 15197:2013 is the international standard for self-testing meters, and it sets its own interference requirements; the European Union accepts it for showing that a self-testing meter conforms to its rules [3]. CLSI EP07, the Clinical and Laboratory Standards Institute's general guideline for interference testing in clinical chemistry (3rd edition, 2018), supplies the underlying lab method. A companion document, EP37, lists recommended test concentrations for many substances [3]. ISO 15197 points to EP07's screening procedure but adds the specific test steps and pass limits that EP07 leaves to the user [3].

The FDA doesn't simply adopt ISO. Its home-meter guidance notes that many manufacturers design their studies around ISO 15197, then says the standard's criteria aren't enough to protect lay users, citing performance at low glucose and differences between strip lots [1].

How the FDA wants the study run

The FDA asks for testing at three glucose levels, about 50 to 70, 110 to 130 and 225 to 270 mg/dL, which it calls clinically relevant decision points [1]. Each blood sample is first measured at least four times on a laboratory comparator method to pin down its true glucose. It is then split into a test sample with the substance added and a control sample that gets only the solvent [1].

Both halves are run on the meter 30 times: 10 replicates on each of at least three strip lots [1]. The manufacturer reports the bias in mg/dL and as a percentage, with 95% confidence intervals [1]. Each substance starts at the highest concentration that could plausibly appear in whole blood. If it interferes there, the maker steps the dose down to find where the effect begins. The FDA's own example: if 20 mg/dL of acetaminophen interferes, test 15, 10 and 5 mg/dL, and if 5 is clean but 10 is not, report 5 mg/dL as the highest concentration with no interference [1].

Hematocrit (the share of blood made of red cells) and altitude get separate studies. The FDA recommends a 20 to 60% hematocrit range tested in 5% steps, and testing at 10,000 feet or higher unless the maker can justify why altitude doesn't matter [1]. We cover how meters handle hematocrit in Interference and Hematocrit Correction.

The FDA's minimum substance list

Both guidances list 26 entries that every new strip should be tested against, each with a recommended test concentration [1][2]. A sample:

The last entry, "sugar alcohols," means testing each common one separately, sorbitol and xylitol among them [1]. EDTA and heparin appear as drugs given to patients, not as the anticoagulants used to prepare samples [1]. The list is a floor. Makers are also expected to run a risk analysis and bench-test other drugs common among their users, and to keep evaluating new drugs after the meter is on the market [1].

What counts as "significant"

Here's what surprised us: the FDA guidance sets no single pass/fail number. It says acceptable interference may vary by substance and by the people the device is meant for, so the maker submits the data along with the meter's expected imprecision, and any labeling for an interference is worked out during the 510(k) review [1].

FDA decision summaries show what manufacturers use in practice. The review of Sinocare's TRUEness meters, cleared in December 2023, follows the guidance design at its three glucose ranges and counts a result as "no significant interference" when the average bias between test and control stays within ±10% [4]. Its list went past the FDA minimum to include icodextrin, metformin, glipizide, caffeine, ethanol, tetracycline and warfarin [4]. One limitation reached the labeling: don't test during or soon after a xylose absorption test [4].

The lab study described below, built on ISO 15197, used a two-part limit instead: ±10 mg/dL when glucose is 100 mg/dL or lower and ±10% above that. Its authors took those numbers from ISO's criteria for interferences that must be described in the instructions for use [3].

One lab screen, start to finish

A 2024 study from an accredited test lab in Ulm, Germany, shows the full routine. Researchers screened 70 substances common among people with diabetes on two Terumo glucose oxidase systems, one for point-of-care use and one for self-testing, using three reagent lots of each [3]. Following ISO 15197, they tested at two glucose ranges, 50 to 100 and 250 to 350 mg/dL, in venous blood from volunteers who hadn't taken vitamin C or most medicines in the previous 48 hours [3].

Test concentrations came from EP37 where available. Otherwise they followed EP07's rule of at least three times the blood level expected after a therapeutic dose [3]. The team also corrected each result against a lab analyzer, which it notes departs from ISO and EP07, to cancel the glucose that blood cells consume while samples wait [3].

Sixty-nine of the 70 substances passed. The exception was N-acetyl-L-cysteine, a mucus-thinning drug that is also the antidote for acetaminophen overdose, and it failed only at low glucose and only in one lot per system, by −10.6 and −14.3 mg/dL [3].

What a bench test can't show

The authors list the limits themselves. A purified compound spiked into a tube may not behave like the same drug in someone's blood after they take it, and some effects only appear at particular combinations of glucose and substance levels [3]. Lab screens can't fully replace testing real patient samples [3].

That is the gap post-market studies try to fill. A study by LifeScan researchers at three UK hospitals checked the OneTouch Select Plus strip against a lab method in 1,023 people with type 1 or type 2 diabetes who were taking 538 different medications between them, and found no evidence of interference from more than 500 prescription drugs [5].

For any one strip, the results end up in the limitations section of the package insert. Federal labeling rules require it to name known interfering substances, and the FDA asks makers to state the most extreme conditions the device was validated under [1]. If a medicine you take shows up there, whether a different meter would suit you better is a question for your care team.