Azita Emami
1,2* 
, Jeehye Jun
3 , Gabriella Engström
4 , Lars Berglund
5,6,7 , Töres Theorell
8,9 1 Yale School of Nursing, Yale University, PO Box 27399, West Haven, CT, 06516 USA
2 Division of Occupational Therapy, Department of Neurobiology, Care Sciences, and Society, Karolinska Institutet, Fack 23 200, 141 83 Huddinge, Sweden
3 Red Cross College of Nursing, Chung-Ang University, 84 Heukseok-ro, Dongjak-gu, Seoul, Republic of Korea
4 Florida Atlantic University, Charles E. Schmidt College of Medicine, 777 Glades Road, Boca Raton, Florida, 33431 USA
5 Department of Public Health and Caring Sciences, Geriatrics, Uppsala University, Box 564, 751 22 Uppsala, Sweden
6 Epistat AB, Kungsängsvägen 27, 753 23 Uppsala, Sweden
7 School of Health and Welfare, Dalarna University, Högskolegatan 2, 791 88 Falun, Sweden
8 Stress Research Institute, Department of Psychology, Stockholm University, Frescativägen 8, 106 91 Stockholm, Sweden
9 Division of International Public Health, Karolinska Institutet, K9 GPH Ekström, 171 77 Stockholm, Sweden
Abstract
Introduction: Accurate estimates of intra-individual variability are necessary for proper design of clinical trials and epidemiological studies where the stress biomarkers cortisol and dehydroepiandrosterone sulfate (DHEA-S) are measured for dyads of persons with dementia (PWDs) and their family caregivers (FCGs). The aim is to determine the number of consecutive sampling days required to detect effect differences in clinical trials, and to accurately estimate regression coefficients in epidemiological studies where stress biomarkers are exposure variables in regression models with future disease as outcome.
Methods: Clinical trial data from dyads of PWDs and their FCGs were used. Salivary cortisol and DHEA-S samples were collected five days a week, for eight consecutive weeks. From this data, we created formulas and graphical tools for the number of required sampling days needed to detect effect differences, and we calculated number of days needed for regression coefficients to be estimated with<10% bias.
Results: A total of 5791 salivary samples from 34 dyads were used. For morning cortisol, five consecutive sampling days at baseline and an equal number of days at study termination is sufficient to detect a treatment difference>5% of baseline level with>20 dyads per group. When stress biomarkers are used in epidemiological studies at least six consecutive sampling days are required.
Conclusion: Based on a large number of consecutive measurements of stress biomarkers we calculated the sufficient numbers of sampling days for clinical trials and for epidemiological studies to produce credible results. Our findings will aid researchers in the study design phase.