Intraclass correlation coefficient rather than correlation coefficient to examine agreements among different methods measuring valvular area

B I Kuo
1994 Circulation  
the ankle-brachial index (ABI) so that it can be easily performed in a busy physician's practice. Indeed, this was a major concern of the workshop participants. We would like to address the important issues raised. First, using the Doppler device for measuring the arm pressure should not complicate or lengthen the procedure unduly as, of course, this device is already being used for the ankle pressures. The end point for the systolic pressure is much sharper by Doppler than it is by
more » ... In the clinical setting, the arm blood pressure is usually measured by either a nurse or paramedical person before the physician sees the patient. In the experience of the panelists, there is often a considerable discrepancy between the arm systolic pressure recorded by the stethoscope method and that obtained by Doppler. Since the accurate assessment of the ABI is critical, it was the view of the panel that both the arm and ankle systolic blood pressures be done at the same time and with the same device -the Doppler. Second, many of the panelists believed that the "drift" in arm pressure while performing an ABI was often quite significant. For example, in the Pittsburgh Epidemiology of Diabetes Complications Study,' the mean difference in arm pressure over approximately 5 minutes was 5 mm Hg (based on 650 subjects). Given the greater variability due to other causes, it would seem particularly desirable to minimize variation in those aspects where it can be controlled. Third, we do not formally recommend duplicate measurements and list only seven pressures to be made (not 18, as Dr Fowkes suggests). Although we do observe that variability can be reduced by using duplicate measures, for the very reasons Dr Fowkes raises, this was not made a formal recommendation. The Edinburgh Claudication Questionnaire-developed by Dr Fowkes unfortunately was not fully published at the time of the workshop. However, we concur that it is a clear improvement on the WHO/Rose Questionnaire.3 The Edinburgh revision is similar to the assessment suggested by the workshop, particularly in deleting the question on pain disappearance while walking. and is likely to be more specific in retaining the question on time to disappearance of pain. However, the determination of laterality by leg diagrams in a self-administered questionnaire allows for considerable error. The Edinburgh revision is said to have a 90.C sensitivity. However, this sensitivity estimate was derived from a symptomatic group. Many cases of lower extremity arterial disease are asymptomatic. The sensitivity of any claudication questionnaire will drop sharply when such cases are included in the study' group. Finally, we are grateful for the additional references to Dr Fowkes' Edinburgh Artery Study4 and to the Israeli Lipid Research Clinic Study,5 which provide further data on the sex differential in ABIs. We agree that while height partially explains these findings, further investigation of the sex difference is still needed. LaPorte RE, Kuller LH. The prevalence of complications in insulindependent diabetes mellitus by sex and duration: Pittsburgh Epidemiology of Diabetes Complications Study II. Diabetes. 1990;39: 1116-1124. 2. Leng GC, Fowkes FGR. The Edinburgh Claudication Questionnaire: an improved version of the WHO/Rose Questionnaire for use in epidemiological surveys. Gotsman MS. Peripheral vascular disease in a middle-aged population sample: the Jerusalem Lipid Research Clinic Prevalence Study. Isr J Med Sci. 1987;23:157-16'7 Intraclass Correlation Coefficient Rather Than Correlation Coefficient To Examine Agreements Among Different Methods Measuring Valvular Area The correlation coefficient is a common statistic applied to examine the agreements of valvular areas calculated by different methods. For example, the correlation coefficient was calculated by Rodriguez et all to examine the agreements of mitral valve areas derived from planimetry. Doppler pressure half-time, and Gorlin equation. The application of the correlation coefficient in this situation was not appropriate. The Pearson product-moment correlation coefficient (r) is a measure of association but not a measure of agreement or concordance. A relation or trend between two quantitative variables refers to the tendency for the variables to vary together. The correlation coefficient provides a quantitative measure of the dependence between two variables Agreement, on the other hand, refers to the extent of correspondence between two methods for measuring the same quantity. The correlation coefficient measures the strength of a relation between two variables, not the agreement between them. Lee et a12 showed an excellent example of data that were in poor agreement but produced quite high correlations. Regression is a useful statistical technique for pre dicting the value of y from the value of x, Therefore regressio,n cannot help in the agreement issue either. If the Gorlin formula or any other method is good enough to be taken as a -gold standard' for valvular area measurement, the authors should present the difference between the results of other methods and the 'gold standard.' Readers could evaluate the accuracy of these methods by examining the distribution of differences. If no method is good enough to be taken as a "gold standard" in measuring orifice area. intraclass correlation coefficient (rl) is an appropriate statistical method that was recomi mended by Lee. This statistic is most often used in interobserver reliability studies. Bland and Altman3 suggest plotting the differences against the average of the two methods. This graphic presentation makes it easier to assess the magnitude of disagree ment by visual examination. We may want to see whether a new, cheap, and quick method produces results that agree with those from an established method. Intraclass correlation coefficient, rather than correlation coefficient, is an appropriate statistical method for these agreement problems.
doi:10.1161/01.cir.89.4.1910 pmid:8149563 fatcat:x54f6gh2tjavxbyvjw4nlm2kru