Letter to the Editor: Two formulas for computation of the area under the curve represent measures of total hormone concentration versus time-dependent change: A Letter to the Editor in response to J.C. Preussner et al. (2003) Psychoneuroendocrinology 28
In a recent issue of Psychoneuroendocrinology, an excellent and timely article by [Pruessner et al., 2003] discussed the use of area under the curve to represent hormone concentration versus time data. We wish to expand upon the interpretation of the AUC and point out some potential limitations. As the authors noted, the AUCI may be negative if the hormone levels continue falling after the first (basal) measurement, suggesting that such a negative area provided valuable information as an "index of decrease". We believe this approach may also miss valuable information. For example, cortisol levels normally fall in a circadian fashion in mid-morning, the afternoon or evening. If exogenous CRH is given, cortisol levels will continue to fall for some minutes, then rise, peaking approximately 15-30 min later, but this rise will be superimposed on the underlying circadian fall. The response pattern will then depend on the rate of circadian fall, and the magnitude of the stimulus. Rarely do we see a curve like Pruessner's Figs. 1 or 2, where hormone levels are above the basal level at the final measurement, allowing a straightforward calculation of a positive AUCI. The final hormone level may be below the basal level, or even below the nadir before the rise. The peak level may not even reach the basal level. There may clearly be a response to the stimulus, but application of Pruessner's formula for AUCI may yield a negative area or an artificially reduced positive area. While it is impossible to extract the underlying circadian curve from the hormone level versus time curve, in these circumstances our approach has been to approximate this by calculating the area above a straight line drawn between the nadir before the rise and the end point of the curve. Like Pruessner's AUCI, we believe that this area represents a measure of the system's response to the intervention. As stated in their discussion, the number of repeated measures is irrelevant when statistical analysis based on AUC is performed. However, it is not irrelevant in the AUC calculation. In clinical experiments where hormone responses to an intervention are measured, sampling times must be chosen to prevent response peaks being missed. For example, in studies using IV AVP to stimulate ACTH and cortisol release, we measured plasma AVP levels 5-minutely and cortisol and ACTH at 10, 15, 20, 30, 45 and 60 min after AVP injection [Nye et al., 1997], in order to define the peaks. In [Pruessner et al., 2003. J.C. Pruessner, K. Kirschbaum, G. Meinlschmid and D.K. Hellhammer, Two formulas for computation of the area under the curve represent measures of total hormone concentration versus time-dependent change. Psychoneuroendocrinology 28 (2003), pp. 916-931. Article | PDF (128 K) | View Record in Scopus | Cited By in Scopus (255)Pruessner et al., 2003] waking salivary cortisol studies, however, the AUC could be accurately defined by the 15- or 30-minutely sampling described. We endorse the approach of [Pruessner et al., 2003], with the proviso that the nature of the experiment will determine how the AUC should be computed.