Variability and reliability of diurnal cortisol in younger and older adults: implications for design decisions

The extant research is inconclusive regarding the best sampling methods to construct reliable measures of between-person differences in derived parameters of diurnal cortisol, and no study provides such recommendations for detecting within-person changes. These studies determined how many days of sampling are necessary to assess between-person differences and within-person changes over multiple occasions in diurnal mean, diurnal slope, and area under the curve (AUC). Generalizability and decision analyses were conducted on diurnal salivary cortisol data from two separate longitudinal studies, one with younger adults (N=124) and one with older adults (N=148). In both studies, results indicated that 3 days of data collection provided the minimal level of reliability in mean cortisol to detect between-person differences; 4-8 days were necessary to reliably assess AUC, and 10 days for cortisol slope. Similarly, in order to reliably characterize within-person changes across occasions, at least 3 days of data collection were needed for mean cortisol and AUC and 5-8 days for slope. Results also indicated that only two samples per day, taken morning and evening, could faithfully reproduce the diurnal slope calculated from 3 or 4 samples (r=.97-.99). Instead of having participants provide many samples per day over the course of a few days, we recommend collecting fewer samples per day over more days.