Expert: Brad Karon, MD, PhD

How should I approach evaluating a glucose meter for use in the ICU?

The answer to this question has both patient care and regulatory considerations. For patient care purposes, the choice of sample type may be as important as the choice of device: arterial whole blood offers the most accurate measurement in the ICU, while capillary sampling may lead to erroneous results in patients with poor tissue perfusion. Venous catheter sampling can lead to overestimation of glucose at higher concentrations depending on meter technology.

The ideal approach to evaluating a meter is to collect nearly simultaneously whole blood glucose meter samples and plasma or serum laboratory samples. Using the actual end users to perform bedside measurement—and the intended sample type from ICU patients—allows observation of the variables most likely to impact glucose meter accuracy, such as hematocrit effect, medication interferences, and user errors.

To complement a patient-based assessment of meter accuracy, residual serum or plasma specimens can be spiked with medications used in the ICU to detect interferences. While accuracy (bias) and interferences are the most important factors, meter precision at normal, high, and low glucose concentrations is also important. Fortunately, precision of most hospital-use glucose meters is quite good, with a coefficient of variation <5%.

Laboratories should also consider using multiple laboratory methods to determine the reference glucose value, as commercially available laboratory methods may differ significantly. Paired whole blood and reference glucose values can be used to determine the impact of observed differences on insulin dosing decisions according to the institution’s glycemic protocol. The data can also be compared to accuracy guidelines produced by groups like the National Academy of Clinical Biochemistry and the Clinical and Laboratory Standards Institute (CLSI). CLSI recommends a minimum of 200 measurements for such a comparison. While this may be challenging, keep in mind that small studies will not allow the laboratory to measure the frequency and magnitude of outlier results or fliers. These are equally if not more important to measure than mean or median bias. There is no consensus on the level of accuracy required or the optimal number or extent of outliers, though recent CLSI guidelines are probably a reasonable benchmark.

Major hospital-use glucose meters marketed in the U.S. are CLIA-waived devices. When used according to manufacturers’ indications and instructions, no in-house validation of analytical performance is required by the Centers for Medicare and Medicaid Services (CMS). However, using the device other than according to the intended use or manufacturer instructions—or with patients or under conditions described in the limitations section of the package insert—may make such uses off-label, and as such high complexity under CLIA. For some devices, off-label use applies to glucose meter measurement in patients who are “critically ill.” One hospital-use glucose meter has received FDA approval for glucose measurement in “intensively treated” adult patients, though only with arterial and venous whole blood.

As a result, the extent of analytical performance validation required by CMS depends upon the meter being used (intended use and limitations), sample types (arterial, venous, and/or capillary whole blood), and intended patient population. Note that neither FDA nor CMS have defined “critically ill” or “intensively treated” patient populations. Therefore, each institution must define these in order to understand when use of glucose meters may be off-label and therefore high complexity.

For high complexity tests, laboratories must validate the accuracy, precision, reportable range, reference range, analytical sensitivity, and analytical specificity of the method. The approach outlined above would cover many of the requirements. However, personnel requirements for high complexity testing also differ markedly from those for waived testing. If use of high complexity testing personnel is not practical, this may preclude use of glucose meters in the ICU or limit the sample types that can be tested. 

Brad Karon, MD, PhD, is director of the hospital clinical laboratories and point of care testing and an assistant professor of laboratory medicine, at Mayo Clinic in Rochester, Minnesota. He is a past-chair of AACC’s Critical and Point-of-Care Testing division.