Dear Ms. Method Matters, In one of your recent articles, you mentioned using the BlandAltman test for comparison between two methods, rather than using Spearman’s (or Pearson’s correlation) [1]. The… Click to show full abstract
Dear Ms. Method Matters, In one of your recent articles, you mentioned using the BlandAltman test for comparison between two methods, rather than using Spearman’s (or Pearson’s correlation) [1]. The statistic of interest, when using correlation, is the rvalue. If a high r-value (for example, 0.998) is obtained, we may be tempted into thinking that there is good agreement between the two methods. But r measures the strength of a relationship, not agreement, and even if two sets of values were obtained using different scales (or expressed in different units), the strength of the relationship would be high. We know that the BlandAltman analysis is the correct method if we want to look at agreement, however, you didn’t go into details with regard to how the results from a Bland-Altman test should be interpreted. For example, in an Anaesthesia paper comparing the results from a non-invasive monitor of cardiac output (Nexfin , BMEYE, Amsterdam, The Netherlands) with the gold-standard thermodilution technique, it states that the mean (SD) bias was 0.60 (1.62) l.min 1 with limits of agreement 2.67 to 3.86 l.min 1 [2]. The precision error was 46%. How are these results to be interpreted? If I were to conduct a similar study, how many subjects would I be required to recruit?
               
Click one of the above tabs to view related content.