NACB - Scientific Shorts
NACB - Scientific Shorts (formerly NACB Blog)
By Brad Karon, MD, PhD
PermalinkRecommendFacebookEmailComments

Glucose meter accuracy criteria as defined in ISO 15197 (1) and CLSI guidelines require that for glucose values above 75 mg/dL, 95% of glucose meter results be within 20% of the reference glucose value. There has been growing momentum towards revision of these guidelines, primarily because of the changing paradigm of glycemic control in the hospital. These glycemic control protocols may require greater accuracy of glucose monitors in order to avoid severe hypoglycemia (SH) in patients receiving intravenouss insulin. Even one episode of SH (glucose concentration < 40 mg/dL) is associated with an increased risk for death in the ICU and hospital. The observation that the most dramatic positive outcomes from well-performed trials of glycemic control were obtained in studies that used more accurate blood gas analyzers to measure glucose levels (2); while follow-up studies that relied upon less accurate handheld glucose meters resulted in neutral or negative findings (3,4,); has fueled interest in the issue of what level of glucose monitor accuracy is required for safe and effective glycemic control in the hospital. The FDA has taken an interest in the question, holding a public forum in March 2010 to solicit input from the general public and healthcare experts on what level of accuracy should be required for hospital-use glucose meters.

It now seems likely that the FDA will tighten accuracy requirements for hospital use glucose meters, though the level of accuracy that will be required is not yet known. An error simulation model study suggested that glucose meters than maintain a 15% (as opposed to 20%) total error tolerance will prevent large insulin dose errors even with tight glycemic control protocols (5). However other experts suggest that a 5% or 10% error tolerance limit might be required to safely manage hospital patients on glycemic control protocols. In the midst of these debates many facilities have moved towards more moderate glycemic control protocols, which presumably will not demand as much glucose monitor accuracy as the original tight glycemic control protocols.

Within three ICUs in our institution we have recently measured rates of severe (< 40 mg/dL) and moderate (40-60 mg/dL) hypoglycemia for one month, both before and after modifying our glycemic control protocol from a more traditional tight glycemic control (target 80-130 md/dL) to a more moderate (target 110-150 mg/dL) glycemic control protocol. Rates of severe hypoglycemia went from 1.4 to 0%; while rates of moderate hypoglycemia went from 7.7 to 0.5% by changing the glycemic target. Causes of SH episodes were investigated during the tight glycemic control era, and 90% were found to be due to protocol violations (glucose reading not taken, glucose reading not acted upon appropriately). Glucose meters that at best met current ISO 15197 criteria were used for all glucose monitoring. This is not to say that glucose monitor accuracy is not important, only that other factors (glucose target range, protocols and systems in place to insure protocol compliance) may be relatively more important than glucose monitor accuracy in avoiding hypoglycemia in the ICU. Based upon available evidence and experience, I believe that bedside glucose meters that maintain a 15% total error criteria will be safe and effective for hospital use. Do you agree or disagree? What do you feel should be required of glucose monitors for use in the ICU or for glycemic control protocols?

 

Please log in to leave a comment!

Posted by
On 5/26/2011

Thank you to everyone for your comments. I think that David is right on in that specimen choice is critical. Most studies have found that in the ICU arterial catheter blood is most reliable and consistent, but now we are facing (for good reasons) pressure to get the arterial catheters out earlier in ICU patients. Venous catheter blood is more problematic because of contamination but also because some property of venous catheter blood interferes with the function of some glucose meters (we published this an a Diabetes Tech Ther paper a few years back if you are interested). The nature of the outliers (beyond the 95% within whatever percent) is also important as is the "sensitivity" to user errors which can also be looked at as degree of built-in error proofing in the device. Fortunately both have tended to improve with newer devices but more work needs to be done. I get a lot of questions about normal range for meters in hospital but one needs to keep in mind that these are essentially all random glucoses. So in my opinion action limits are more appropriate than true reference ranges, and will vary and be defined by instituional protocol. The "normal range" in the ICU probably changed after NICE-SUGAR, for example (not meant to start another debate on NICE-SUGAR study design, just using as an example). The question of precision shoud probably not be addressed by looking at classifying normal as normal, but by looking at relationship of precision to error dosing errors within any given protocol. Precision is main driver or larger dosing errors, we and others have done this thru error simulation but can be done empirically as well. Brad

Posted by
On 5/26/2011

3 opinions: 1. Based on control characteristics of the Glc-insulin loop: infrequent sampling & loose adherence to protocols (timing, correct sampling) are a more relevant source of error and maltreatment then the analytical accuracy of the measuring instrument. 2. Based on auditing experience: use of sensored (flagged as unreliable results by the measuring instrument) can be a relevant source of errors. 3. Based on auditing experience: Incorrect blooddraws from catheters are a frequent source of significant error. 4. The question how good analytical precision has to be in order to classify normals as normals, may have an easily explainable solution, but why is it a relevant question ?

Posted by
On 5/15/2011

I THINK IN ADDITION TO ACCURACY ISSUES, IT WILL BE PROPER TO DETERMINE GLUCOSE REFERENCE RANGES FOR GLUCOSE METER ANALYSED SAMPLES IN EACH CENTER BEFORE SUCH METERS ARE USED TO MONITOR CRITICAL-CARE PATIENTS. SEGUN GHAZALI. MB, CH.B; FMCPath.

Posted by
On 5/15/2011

The notion that there is a safe and effective total error tolerance is a myth. If 15.1% is unsafe, how is 14.9% safe? And the unspecified 5% is a much bigger problem. I agree with Dr. Alter “the more accurate the better” and would add to that “within financial constraints.” Jan S. Krouwer, Ph.D. Sherborn, MA

Posted by
On 5/12/2011

Of course, the more accurate the better. However, I am more concerned whether or not the proper specimen is being used. Arterial or venous specimens tested on glucose meters should be the standard of care in the critical care setting. Fingerstick specimens will have a a natural layer of inaccuracy secondary to decreased distal perfusion in the "sick" patient. David Alter MD Grand Rapids MI

About the Author
Brad Karon, MD, PhD
Brad Karon, MD, PhD 
 

References

1. International Organization for Standardization. In vitro diagnostic test systems-Requirements for blood glucose monitoring systems for self-testing in managing Diabetes Mellitus. Report Number ISO 15197 2003.

2. Van den Berghe G, Wouters P, Weekers F, et al. Intensive insulin therapy in critically ill patients. N Engl J Med 2001;345:1359-67.3. Van den Berghe G, Wilmer A, Hermans G, et al. Intensive insulin therapy in the medical ICU. N Engl J Med 2006;354:449-61.

4. NICE-SUGAR Study Investigators. Intensive versus conventional glucose control in critically ill patients. N Engl J Med 2009;360:1283-97.

5. Karon BS, Boyd JC, Klee GG. Glucose meter performance criteria for tight glycemic control estimated by simulation modeling. Clin Chem 2010;56:1091-7.