In the next issue

Abstract

Discussion The meter iscompact, easy to use, and has a large, clear display. Its construction appears robust and it is stable in use on a flat surface. The provisional instruction booklet supplied was clear and comprehensive, but the final version was not available at the time of this study. The nature of the analytic process with the Glucometer 4 is such that contamination of the optical surface by blood should not be possible. Correctly applied samples should not contact the strip carrier either, but this is easily removable for cleaning ifnecessary. The reagent strips are individually foilwrapped for stability, and the foil packets open easily for use. Sample application has been improved by use of a raised plastic sample cup plus spreading area. Once a small drop of blood contacts this layer it is automatically absorbed onto the reagent pad with no need for the patient to ensure its even application. On two occasions during the initial familiarisation period, the analyst managed to apply the sample in such a way that it formed a film across the top of the cup without contacting the spreading layer. However, this gave an error message rather than a false reading and, with experience, it becomes obvious if this has occurred. Correct sample application can easily be checked by turning the strip over to examine the reagent pad. It is understood that the sample application instructions will now be modified to minimise any chance of this occurring in routine use. Timing of the analytic process is no longer critical and introduction of the reagent strip into the meter is straightforward, such that the measurement process is relaxed compared with usage of previous models. Whilst accuracy and precision goals for laboratory glucose measurement have been derived from biological variation datav (CV<2.2%, zero bias), these standards are not routinely met by over 90% of chemical pathology departments, and it would be unrealistic to expect such performance from small bedside analysers used by untrained staff. The precision data achieved in this study on real patient samples in a clinic setting (overall CVof 6.7% on 145 paired patient samples) would seem to be a significant improvement on previous performance when compared with published data, that attained locally during prior studies, and also that which was measured during this study from the routine meters used in the diabetes clinic. The accuracy goal calculated by Tonk's method? is 10% bias, which is also that stated as desirable by the American Diabetes Association'', This was achieved in 86% measurements made on the new meter during this study, which compares well with 67% on other meters used in the local clinic by the same staff. In fact, 96% of all results obtained on the Glucometer 4 were within 15% of the YSI values and no difference exceeded 20%. Comparison ofthe Glucometer4 results with those obtained by analysis of the plasma fraction of the same sample showed a negative bias of 7%. This expected difference between plasma and whole blood should be borne in mind when comparing results from ward-based meters with those obtained on plasma samples sent at the same time for checking purposes. In summary, the new meter performed well on real patient samples in a clinical situation. While it is not possible from this study to describe the system as operatorindependent, it would seem much less prone to user error than other models, and this should lead to greater confidence in near-patient glucose testing results. Following this evaluation the new system will be introduced for ward use in this hospital, and it will be interesting to See how it performs in local quality assurance schemes.