Uncertainties in the measurement of blood glucose in paediatric intensive care: implications for clinical trials of tight glycaemic control

In preparation for a tight glycaemic control (TGC) clinical trial we assessed the agreement between methods used to measure blood glucose in critically ill children. Service evaluation comparing blood gas and main laboratory analysers with point-of-care (POC) devices PCX, ACCU-Chek and Hemocue. Two hundred forty-five samples from 157 children measured on 2-4 devices provided 790 values. Marked variation was evident in glucose values between devices, time between tests, sample (whole blood/plasma) and source; 39% of paired values had > 20% difference. The decision to start insulin at 7 mmol/L differed depending on the device used for 33% of samples. At low glucose values (< 4 mmol/L), differences up to 1.8 mmol/L were evident. The blood gas analyser read lower than all POC models and the laboratory analyser (less risk of undetected hypoglycaemia). An inverse relationship was evident between haematocrit (Hct) and glucose error using POC devices. PCX values for samples with Hct < 30% were higher (85%), whereas those for Hct values > 38% were lower (66%). Glycolysis occurred during transfer of samples to the laboratory. Using the PCX at the bedside resulted in 0.5 mmol/L mean difference higher than laboratory values; locating the PCX in the laboratory reduced this to 0.2 mmol/L. Discrepancies between measurements may mask hypoglycaemia, and the potential benefits of controlling hyperglycaemia may not be achieved. Variation introduced by different devices, sample or source may have led to misclassification of treatment decisions contributing to the conflicting results of TGC studies.