Abstract
Portable, hand-held glucose meters are widely used by both patients and health-care professionals alike. Patients with diabetes may undertake self-blood glucose monitoring for a variety of purposes such as self-titration of insulin dose, the avoidance of hypoglycaemia in potentially hazardous situations and as a component of a broader self-management programme. Glucose meters are also used by health-care teams in the assessment and treatment of patients with diabetes in a range of clinical settings including ambulatory care, management of diabetes during periods of hospitalization with other concomitant illness, treatment of diabetic ketoacidosis and the monitoring and treatment of hyperglycaemia associated with an acute stress response in non-diabetic patients. Each of these very disparate applications may pose differing analytical challenges in obtaining a glucose result of sufficient accuracy to support optimum clinical management.
One particular area of interest is the use of intensive insulin regimens to maintain tight glycaemic control (TGC) in critically ill patients in the intensive care unit (ICU). Such patients commonly develop hyperglycaemia and insulin resistance as part of the stress response. Many studies have demonstrated a strong association between hyperglycaemia and adverse outcomes in acutely ill patients. The reasons for this are unclear but there is evidence of increased sepsis and multiorgan failure in hyperglycaemic surgical ICU patients. These observations inevitably provided a stimulus for clinical trials to investigate whether the maintenance of normoglycaemia in critically ill patients improved outcomes. The use of TGC regimens in the surgical ICU patient was proposed over a decade ago following an early study which suggested that maintaining the blood glucose concentration at <6.1 mmol/L was associated with a 34% reduction in in-hospital mortality when compared with conventional treatment in which the blood glucose concentration was maintained at around 10–11 mmol/L. 1 The lower mortality was explained by a reduction in deaths from multiorgan failure in patients with proven sepsis. Despite the promise of early trials and the endorsement of TGC regimens by a number of professional associations, subsequent studies in both medical and surgical ICU patients have generated conflicting results with some evidence pointing to higher mortality rates in the TGC treated patients. A particular concern has been the much higher prevalence (up to 12-fold) of severe hypoglycaemia in TGC patients compared with conventional treatment. Hypoglycaemia is known to be associated with worse patient outcomes and the increased risk of hypoglycaemia associated with TGC regimens has been proposed as a possible cause of the higher mortality seen in some studies. 2
A typical TGC regimen requires frequent measurement (at one to four hourly intervals) of blood glucose concentration and titration of insulin infusion rates (or bolus dose administration) according to a dosing algorithm to maintain blood glucose concentrations within the desired range. Dosing algorithms may require titration of insulin doses in response to small changes in glucose concentration (<2 mmol/L). Point-of-care (POC) measurement of glucose is a necessary component of TGC regimens since the turnaround times associated with central laboratory testing would in most cases be too long to allow timely insulin dose adjustment. TGC clinical trials have used a range of POC devices, from portable hand-held glucose meters to desktop analysers which measure glucose in addition to blood gases and electrolytes.
The accurate measurement of whole blood glucose in critically ill patients poses major challenges for POC analysers. ICU patients may have extremes of haematocrit, reduced arterial partial pressure of oxygen (pO2), acid–base disturbance and poor capillary perfusion, all of which may affect glucose measurement.
Given the very tight degree of glycaemic control desired, and the possible adverse consequences of unwanted hypo- and hyperglycaemia, what degree of accuracy is required for POC glucose measurement in this setting? Although numerous performance standards have been proposed, few have related specifically to TGC regimens. For example, the ISO 15197 (2003) requires that 95% of results fall within ±20% at glucose concentrations ≥4.2 mmol/L and within ±0.83 mmol/L at glucose concentrations <4.2 mmol/L. 3 It is important to stress that the ISO standards specifically relate to blood glucose monitoring systems used by lay persons and may not therefore be appropriate for TGC. Karon et al., 4 using simulation modelling of TGC regimens, suggested that POC glucose measurement operating to a total allowable error of <15% was unlikely to produce large insulin dosing errors. In contrast, total allowable error of ≥20% (the ISO 15197 standard) was associated with a frequency of serious insulin dosing errors of >0.2%. The 2011 American Association for Clinical Chemistry (AACC)/National Academy of Clinical Biochemistry (NACB) guidelines have proposed an intermediate goal for manufacturers of limiting total allowable error for 95% of samples to ≤15% at glucose concentrations ≥5.6 mmol/L and to <0.8 mmol/L at glucose concentrations <5.6 mmol/L. 5 However, the guidelines have the important added caveat that lower (although unquantified) total allowable error may be desirable and indeed necessary for TGC applications.
How do currently available POC glucose analysers perform? A study by Freckmann et al. 6 of 27 blood glucose monitoring systems reported that more than 40% did not even meet the less stringent ISO 15197 standards. 6 Further evidence is provided in the current edition by Watkinson et al. 7 who studied the analytical performance of three POC glucose measuring systems (two portable hand-held glucose meters and a desktop blood gas/glucose analyser) in an ICU setting, compared with a plasma glucose reference method. Strengths of the study included the stratified design that recruited patients with a broad range of glucose concentrations and the fact that duplicate samples were taken to allow an assessment of the precision of the instruments under study. The authors found that only one of the three instruments evaluated met the AACC/NACB 2011 performance standards: two were affected by haematocrit and one of these was also affected by pH and pO2.
It is clear, therefore, that not all currently available POC instruments for measuring glucose meet the intermediate AACB/NACB goal for analytical performance (which in itself may be insufficiently rigorous) and are unsuitable for use in TGC regimens. It is essential that teams with responsibility for POC understand fully the way in which POC glucose results will be applied in different clinical areas. The analytical performance of an instrument which may be acceptable (if suboptimal) in one clinical setting, for example self-monitoring of blood glucose in non-insulin requiring patients with type 2 diabetes, may be unacceptable in another setting, for example, TGC regimens. This requires close liaison between the POC and clinical teams to ensure that the analytical performance of the instrument used is appropriate to the clinical application of the results.
DECLARATIONS