The John Radcliffe Hospital was opened in the 1970s and is Oxfordshire’s main accident and emergency site. With over 700 in-patient beds and nearly 100 day case beds, it also provides acute medical and surgical services, trauma, intensive care, cardiac, diabetes and women’s services. The largest of the
trust’s hospitals, it houses many departments of Oxford University Medical School and is the base for most of the medical students training throughout the trust.
The Quo-Test assay is primarily intended for the measurement of HbA1c at the point of care by clinicians or other healthcare staff using fingerprick or venous blood samples. The purpose of this evaluation is to establish the performance characteristics of the Quo-Test system when used by an independent evaluator. To that end, 40 patient samples were analysed on two Quo-Test analysers and the results are compared with those reported by the Menarini HA-8160 analyser, which is the system used by the hospital laboratory. The results will then be analysed for accuracy, precision and bias.
When comparing two (or more) systems for the measurement of HbA1c, it is important to consider the methodological variability derived from the type of assay used to measure the patient samples.1,2 There are four main methods commonly used in the measurement of HbA1c: electrophoresis, high-performance liquid chromatography (HPLC), antibody affinity and boronate affinity. While all these methods are accepted and standardised for the measurement of HbA1c,3,4 it is also accepted that small discrepancies will be found between these methods when measuring HbA1c values in a patient population. The Menarini HA-8160 uses HPLC with reverse-phase cation exchange chromatography to separate HbA1c from other types of haemoglobin. The Quo-Test A1c assay is a boronate affinity method.
The acceptance criteria for the study are that the coefficient of variation (CV) for duplicate testing of the Quo-Test assay on two separate analysers should be less than 5%. The mean bias between the two analysers should be ±0.3% HbA1c. The correlation coefficient for all samples compared with the standard laboratory device (Menarini HA-8160) should be r>0.95. The 95% limits of agreement for the overall bias of the Quo-Test duplicate tests compared with the laboratory assay should be within ±0.85% HbA1c, according to guidelines issued by the National Glycohemoglobin Standardisation Program,5 when comparing a device with a secondary reference method.
Comparison with laboratory results
Two Quo-Test analysers and a single lot (No 13) of Quo-Test A1c test cartridges were used in the study. Forty venous blood samples were obtained from the biochemistry department and blinded results were obtained on the hospital’s Menarini HA-8160 analyser.
The 40 blood samples were run singly on each of the two Quo-Test analysers. The results from each analyser were then compared with the results obtained using the Menarini HA-8160 for accuracy and bias. The Quo-Test results for both analysers were then pooled and all 80 results were compared with the Menarini results. Finally, the results from the two Quo-Test analysers were compared to determine inter-analyser variation.
Within acceptable limits
In the first instance, the results obtained with analyser 1 (Figs 1a and 1b) and analyser 2 (Figs 2a and 2b) were compared individually with the results from the Menarini HA-8160 for accuracy and bias using a scatter diagram and a Bland Altman plot.
For analyser 1, Pearson’s correlation coefficient was r=0.987, the bias was –0.09% HbA1c and the 95% limits of agreement for the bias were –0.72 % to +0.55% HbA1c. Similarly for analyser 2, Pearson’s correlation coefficient was r=0.987, the bias was –0.06% HbA1c and the 95% limits of agreement for the bias were –0.7% to +0.58% HbA1c. All the results were within acceptable limits.
Comparable performance
The results for analysers 1 and 2 were combined and compared with the results obtained from the Menarini HA-8160 for accuracy and bias (Figs 3a and 3b). The results were combined to give an indication of the level of performance observed if a laboratory were running two Quo-Test analysers in the same or different laboratories. With Pearson’s correlation coefficient of r=0.987, bias of –0.07% HbA1c and 95% limits of agreement for the bias of –0.71% to +0.56% HbA1c, all the results were within acceptable limits. This indicated that running two analysers still produced excellent results.
When the two analysers were compared, no significant inter-instrument variation was found. Figures 4a and 4b showed that the mean bias (Quo-Test analyser 1 result – analyser 2 result) was +0.03 % HbA1c. In addition, the 95% confidence intervals for the bias include zero. Therefore, the difference between the two analysers was not statistically significant, indicating that the results produced were indistinguishable.
The coefficient of variation (CV) for the duplicate results was calculated for each pair, and overall CV was found to be 3.21%, which was acceptable.
Meeting performance criteria
Although the two systems use different methodologies for the measurement of HbA1c, the agreement between the two analysers was excellent. A summary of the results can be seen in Table 1. The results of the study indicated that the Quo-Test analyser gave results which were substantially equivalent to the Menarini HA-8160 analyser in accuracy and with no significant bias. In addition, even when the results of the two Quo-Test analysers were compared with the hospital’s method, the total error was within the exacting performance criteria of the National Glycohemoglobin Standardisation Program.