Sources of variation in the determination of distal blood pressure measured using the strain gauge technique.

Abstract

Using the determination of distal blood pressure (DBP) measured using the strain gauge technique as an example of a routine clinical physiological investigation involving many different observers (laboratory technicians), the present study was carried out to assess (1) the influence of the number of observers and the number of analyses made by each observer on the precision of a definitive value; and (2) the minimal difference between two determinations to detect a real change. A total of 45 patients participated in the study. They were all referred for DBP determination on suspicion of arterial peripheral vascular disease. In 30 of the patients, the DBP curves were read twice, with a 5-week interval, by 10 laboratory technicians. The results were analysed using the variance component model. The remaining 15 patients had their DBP determined twice on two different days with an interval of 1-3 days and the total day-to-day variation (SDdiff) of DBP was determined. The inter- and intraobserver variations were, respectively, 5.7 and 4.9 mmHg at ankle level and 3.5 and 2.7 mmHg at toe level. The index values as related to systolic pressure were somewhat lower. The mean day-to-day variation was 11 mmHg at ankle level and 10 mmHg at toe level, thereby giving a minimal significant difference between two DBP determinations of 22 mmHg at ankle and 20 mmHg at toe level. To decrease the value of SD (standard deviation) on a definitive determination of DBP and index values, it was slightly more effective if the value was based on two observers performing one independent DBP curve reading than if one observer made one or two DBP curve readings. The reduction in SDdiff was greatest at ankle level. The extent of the Sddiff decrease was greatest when two different observers made a single DBP reading each at both determinations compared with one different observer making two readings at each determination. Surprisingly, about half of the maximum reduction in the SDdiff was achieved just by increasing the number of observers from one to two. We have found variance component analyses to be a suitable method for determining intra- and interobserver variation when several different observers take part in a routine laboratory investigation. It may be applied to other laboratory methods such as renography, isotope cardiography and myocardial perfusion single-photon emission computerized tomography (SPECT) scintigraphy, in which the final result may be affected by individual judgement during processing.