The importance of using the correct bounds on the Bland-Altman limits of agreement when multiple measurements are recorded per patient.

The limits of agreement originally derived by Bland and Altman (Lancet i:307-310, 1986) are the most commonly used method for investigating statistical agreement between two medical devices. Bland and Altman describe a confidence interval for these limits that should always accompany the limits themselves. However, this interval presumes that the recorded differences between the two devices in question are independent. This is a reasonable assumption when only one measurement is recorded per device per patient. Bland and Altman (StatMethods Med Res 8:135-160, 1999) subsequently derived bounds for the more general case where multiple observations are recorded within each patient. Unfortunately, in practice, the bounds assuming independence are typically reported when in fact the repeated measures bounds are more appropriate. This communication illustrates the dangers of using the "original" (independence-based) bounds derived in Bland and Altman (Lancet i:307-310, 1986) in the presence of repeated measures per patient.