BG Meter Accuracy: 10 Popular Meters Put to the Test!

Are blood glucose meters accurate?

A few months ago, I read a post on blood glucose meter (in)accuracy that was quite alarming. Chris Hannemann, T1D and OpenAPS’er, tested five popular blood glucose meters for accuracy and found major discrepancies. The overall variability between the meters was roughly ±11 %, and two of the meters — both from the same manufacturer — showed major bias. Chris had been using one of those meters to calibrate his CGM and consistently found that his lab-measured A1C would come in a full percentage point higher than his CGM average would predict (e.g., the meter-calibrated CGM data would correspond to an A1C of 6%, but laboratory-measured A1C would actually be 7%).

As someone who wears a CGM and always strives to achieve an A1C of ~6.0% or lower, this post freaked me out. If I’m calibrating my CGM with a meter whose results are always off by that much… how can I be sure about reaching my targets?!?

But I was also skeptical when I saw Chris’ post, because my A1C has always matched my predictions (based on my CGM average). So, I decided to run my own test on 10 popular meters from various manufacturers. Much to my relief, all 10 meters produced very similar results for each blood sample, with an overall between-meter variability of just under 6%. Below I describe the test and discuss the results. And, in another post, I explain some reasons why your average BG on your meter might not always correspond with your A1C results. (Key point: It doesn’t necessarily mean your meter is bad!)

The Test

The Meters

• Accu-Chek Aviva Connect
• Contour Next EZ
• OmniPod PDM FreeStyle Meter
• FreeStyle Lite
• Livongo InTouch
• OneTouch Ultra Mini
• One Touch Ultra 2
• Walgreens True Metrix Air
• Walgreens True 2 Go
• Wal-Mart ReliOn Confirm

These 10 meters varied in age and wear. Some were old, some were new… one was my own personal meter that I used to calibrate my CGM and make mission-critical decisions each day. All of them passed their respective control solution tests, so it’s safe to assume that they were in good working order. I tried to match the testing method employed by Chris (author of the original post) as closely as possible.

The Procedure

Eight rounds of testing were performed over the course of 24 hours according to the following procedure:

1. Wash and dry hands
2. Arrange meters on table in random order
3. Insert new lancet into lancing device
4. Remove test strip from each meter’s strip vial and insert strips into meters
5. Wipe fingertip with alcohol pad & wait for it to dry
6. Prick fingertip & squeeze out large drop of blood
7. Apply blood to each test strip in order
8. Record results

Notes:

• Order of meters was randomized for each round.
• Tests were performed only when CGM readings were stable (i.e. no insulin on board and CGM showing a slope of ~0 mg/dL/min).
• I didn’t do anything special to stabilize my blood glucose — just tested as I went about a normal day.
• The test strips used for each meter all came from their own unique vials.
• Before and after completing the eight testing rounds, the meters were checked using their respective control solutions. They all passed the control solution tests.

The Results

BG Results For Each Round, Compared

Unlike Chris, I didn’t have an alarming spread in my results for any round. The overall between-meter variability (“% Error”, or “%CV” for you stats folks) was only 6%.

In plain English: My treatment decisions wouldn’t have varied much at all, regardless of the meter I was using.

One unit of rapid-acting insulin brings my BG down by ~80 mg/dL, and I correct whenever I’m over 100 mg/dL. I’ll usually correct down to 70–110 mg/dL, depending on my plans for the next couple hours (big meal = correct to 70; workout = correct to 110).

I was relieved to see that even if I took a correction bolus for the maximum BG of each round, I still would have been brought down to a desirable blood glucose level.

For example, take Round 1. The highest reading I saw was 182 md/dL, and I’d take 1 unit for that. Even if we assume the true glucose was the lowest value from that round, 149 mg/dL, I’m still in good shape taking 1 unit because I’d only go down to ~70 mg/dL.

What does this mean for A1C?

To figure out what this means for A1C estimates, let’s take a look at how each meter’s average value over the 24 hours of testing compares with the overall average across all meters. (NOTE: I realize that the average across all meters may not accurately represent true A1C, but this comparison is useful to show that, no matter what meter you use, you’re getting roughly the same results for averages and, thus, roughly the same estimated A1C.)

Looking below at the deviation from the mean — i.e, how much each meter’s average value (solid gray line) differed from the overall average value (dotted gray line down the center) — I like what I see!

The average BG for each meter (solid gray line) was pretty close to the overall average (dotted line down the center), with the greatest deviation at around 6% (FreeStyle Lite & Walgreens True 2 Go). (Compare Chris’ results, where max deviation was 14%.)

This means an estimated A1C calculated based on the average value from any of these meters would be roughly the same.

For example:

• If your average blood glucose on your meter were 154 mg/dL, that would translate to an estimated A1C of 7.0%.
• If you arrived at that average using a Walgreens True 2 Go, which appears to consistently report lower values, your actual A1C might be closer to 7.3%.*
• If you arrived at that result using a FreeStyle Lite, which appears to consistently report higher values, you may be pleasantly surprised by an actual A1C of 6.7%.*

I think most of us would agree that this is an acceptable degree of variation between estimated and actual A1Cs.

* Important note: This is just based on the data I collected for the specific meters I had in my possession. I can’t say whether these trends would be true for all meters of any particular brand that I tested.

My Own Experience With Self-Measured BG Averages & A1C

My CGM averages have always been spot-on when it comes to predicting A1C.

Last Summer & Fall, I was using the Omnipod PDM FreeStyle meter (same one tested here) to measure my blood glucose values & calibrate my Dexcom CGM. For September, my Dexcom showed a 30-day average of 132 mg/dl, which correlates to an A1C of 6.2%. And that’s exactly what my lab-measured A1C was back in September — 6.2%.

For the past 3 months, I’ve been on MDI and using a Freestyle Lite to calibrate my CGM. Right before my latest A1C (~1 week ago) Dexcom’s 30-day average showed 117 mg/dl, which correlates to an A1C of 5.7%. My lab-measured A1C was 5.8%.

And this is pretty much how it’s always been for me. Not the A1Cs — I wish! — but the match between estimates and actuals. 😉 I’ve never had any reason not to trust my BG meter readings, because the results always lined up. BUT I do want to take some time now to explain that even if the numbers don’t line up, it doesn’t necessarily mean your meter is to blame…

READ: Why doesn’t my average BG match my A1C?!

What about the post that started all this? Why were those results so off?

Well, the truth is, I don’t know. I actually own the meter that Chris uses, OneTouch UltraLink — I used it for ~4 years back when I was on a Minimed pump. Unfortunately, I didn’t have it with me at the time of this experiment, but I’ve tested with it a bit recently, just to see…

And the results are generally within 10 mg/dL of my FreeStyle Lite readings!

So, I’m not sure why Chris’ results were so different. He’s been using his OneTouch UltraLink for about 7 years, and mine only got about 4 years of use before being carefully stored at my parents place with all my other diabetes “antiques.” I wouldn’t be surprised if after a significant amount of wear, these things just don’t work as well. However, that doesn’t explain why some of the other meters also deviated significantly from the mean glucose value…

It may be the test strips

The FDA requires that all new meters meet particular standards (shown here in Table 7), but once a meter is approved, the FDA does nothing to monitor accuracy. This is significant because, although the meters may have functioned perfectly when approved by the FDA, that was with a particular set of test strips.

Test strips can vary from batch to batch.

Test strips contain an enzyme that converts glucose into an electrical current that runs through the test strip and is read and displayed on your meter as a glucose concentration. Enzymes are proteins and can breakdown due to humidity, temperature, and many other factors.

So, differences in the environment in which the test strips are manufactured, stored, and used can lead to differences in the blood glucose measurements they provide. Those differences may be OK — the blood glucose values provided by these strips may still fall within the FDA’s standards — but, when we are comparing one batch of strips to another, things can get hairy…

For example, suppose one batch of test strips yields BG readings that are 10% too high, and another batch (for a different meter, by a different manufacturer, or even for the same meter by the same manufacturer) yields readings are 10% too low. Even though each of these batches of strips might have an acceptable difference from the true BG value, there is a 20% discrepancy (bias) between them. The differences could be clinically significant, especially when combined with a 7% random error on top of the 10% systemic bias.

So, where does this leave us?

More testing should be done! I feel pretty good about meter accuracy based on my own experiment and personal experience, but I’d feel even better if more people conducted similar experiments and got similar results.

Results like Chris’ are scary — they make us feel like no matter how hard we try we may still be missing the mark. And the consequences go beyond the emotional toll — meter and test strip inaccuracy may lead to results that fool us into thinking everything is A-OK, when it’s really not.

Bottom line: we need to be able to trust the devices we use to make critical decisions about our health every day. If meter accuracy really is an issue, it needs to be addressed right away.

Written by Rachel Sanchez (@onedroprachel). Originally published at onedrop.today on May 31, 2016.