Sample Size Guidance and Justification for Studies of Biological Variability.

Abstract

Background: Good estimates of analytical, within-subject, and between-subject variation, from appropriately designed biological variability studies, allow the potential for a biomarker to diagnose and monitor disease to be assessed in addition to setting goals for test performance. Sample sizes for these studies require consideration of the numbers of participants (n1), observations per participant (n2), and replicates of each observation (n3). Little guidance exists to compute these values.

Methods: Data was simulated to examine the effect of changes in each component of sample size on analytical, within-subject, and between-subject variability, and on common measures of variability including the coefficient of variation (CV), reference change values (RCV), and index of individuality (II).

Results: Simulations showed the precision of estimated results (CVs, RCVs, and IIs) for varying sample sizes. Greater numbers of participants increased the precision of estimated analytical, within-subject, and between-subject variability; increasing the number of observations per participant increases the precision of estimates of analytical and within-subject variability; and increasing the number of replicates of observations per participants increases precision of only analytical variability.

Conclusions: The sample sizes for biological variability studies can be planned better, with consideration of the primary estimate evaluated in a study and resource used to increase the precision of this estimate. Studies with sample sizes too small to achieve reasonable precision can be redeveloped or abandoned, minimizing research waste. If the desired precision of variability components is known, values for the sample size (number of participants, observations and replicates of observations) can be determined.