Article Text
Statistics from Altmetric.com
Editor—Deshpande et al 1 have produced some interesting data on blood glucose measurement in a neonatal intensive care unit setting and have recommended one measurement technique over others. The descriptions of the methodological differences and limitations of each technique are most helpful.
The data presented show comparisons between blood glucose measurements in the range 3–10 mmol/l. Although they have recommended one particular method, it must be remembered that blood glucose tests in the newborn are largely performed to detect hypoglycaemia. The authors have not addressed this issue well. The graphs presented show few comparisons in the hypoglycaemic range, and those that do, are wildly discordant. Rather than encouraging a rush to reinvest in new technology it would be prudent to demonstrate that these techniques can indeed reliably identify babies who are hypoglycaemic. The bottom line must be if a blood glucose value on a bedside test is less than 2.6 mmol/l or whatever is chosen, what is the likelihood of this being accurate and what is the prediction range of the test?
… is something of a vexed issue Blood glucose concentrations are monitored on our neonatal intensive care unit and postnatal wards using a BM test strip 1-44 with a Reflolux II S Reflectance meter (Boehringer Mannheim). If the glucose concentration recorded is low, consistent with hypoglycaemia, a capillary or venous specimen is sent for laboratory analysis for confirmation before management is altered. A blood glucose value of 2.2 mmol/l or below is accepted in our unit as indicating hypoglycaemia.
Nursing staff have received instruction in the use of the BM Reflolux meter and the HemoCue B glucose analyser (HemoCue AB, Angelhorn, Sweden). Over five months, whenever BM blood glucose concentrations were below 3mmol/l, 0.25 ml of blood was obtained by the same route and sent in a Sarstedt (Leicester) fluoride tube for immediate analysis on the Hitachi 717 analyser. A HemoCue cuvette, requiring 5 μl of blood was also filled and a result obtained on the ward using the HemoCue B glucose analyser. The comparative results were recorded, in addition to the gestation, birthweight, age, haematocrit (if obtained within a 24 hour period).
The Hitachi 717 analyser uses a hexokinase method for determining glucose concentration and is subject to national external quality control. At an overall mean glucose concentration of 2.77 mmol/l, the local SD was 0.03 mmol/l and the coefficient of variation 1.1%. Using this and other performance data, the laboratory glucose concentrations were accepted as the most reliable.
During routine monitoring, low blood glucose concentrations (below 3 mmol/l) were recorded by BM strips on 123 samples, for which complete comparative data were available for 115. These results were obtained from 57 babies aged 0–4 days, one baby aged 11 days, and another aged 60 days. Birthweights ranged between 0.78–5.34 kg and gestation between 25–42 weeks.
Of the 36 hypoglycaemic samples identified by laboratory analysis, concurrence was found in 31 by BM and 14 by HemoCue. The derived data are shown in table 1.
Detection of hypoglycaemia (glucose concentration < or = 2.2 mmol/l) (n=115)
To compensate for whole blood glucose/serum glucose differences, a 10% higher cutoff glucose concentration of 2.45 mmol/l was applied to the Hitachi hexokinase assay, while retaining the 2.2 mmol/l cutoff for the BM and HemoCue results. The sensitivity and specificity were not affected by this adjustment.
Estimating the 95% confidence limits, using the Bland-Altman method, for the agreement between the blood glucose obtained by BM, HemoCue, and the Hitachi methods,1-1 the difference between the results can be expected to be, at the most, 2.6 mmol/l and 2.4 mmol/l, respectively. Increasing haematocrit did not affect the detection of hypoglycaemia (range of 0.36–0.69, mean haematocrit 0.52).
Analysis of all 115 sets of results revealed a mean difference between the BM and Hitachi glucose concentrations of −0.70 mmol/l, with a standard deviation of 0.80 mmol/l. The mean difference between the HemoCue results and the laboratory was +0.52 mmol/l (SD 0.81) mmol/l. The degree of scatter of the results is similar in both methods of near patient testing for hypoglycaemia, as shown by almost identical standard deviations. However, the HemoCue method overestimates blood glucose concentration compared with the hexokinase method using the Hitachi 717 analyser, and this is reflected in a 38.9% sensitivity for detecting hypoglycaemia. This overestimate of the glucose concentration concurs with the positive bias of the HemoCue system reported by Deshpande et al.1-2
Use of the HemoCue B glucose system routinely to measure neonatal blood glucose concentration at the bedside means that a significant number of babies may have unrecognised hypoglycaemia. While the BM reflectance strips give a greater number of false positives, with a specificity for detecting hypoglycaemia of 49%, they show greater sensitivity than the Haemocue (86.1% vs 38.9%), reducing the likelihood of missed episodes of hypoglycaemia.
The BM test strip with the Reflolux reflectance meter was not used for comparison by Deshpande et al. The Reflolux performs better than many other reflectance meters1-3 with a coefficient of variation of 1.5–3.1% across a range of glucose concentrations 3.3–1.9 mmol/l.1-4 The BM test Glycemie 20-800 R test strips with Reflolux reflectance meter have a reported precision of 1.3% that is unaffected by haematocrit between 0.20–0.60.1-5 There is currently no good evidence that the HemoCue system of near patient glucose testing should replace the BM Reflolux reflectance meter used in many neonatal units. Low BM glucose concentrations should be confirmed by rapid laboratory assay. Further comparative studies are required to determine the most appropriate bedside neonatal glucose monitoring system.