Carotid artery intima-media thickness (CIMT) assessed by B-mode ultrasound is an accepted marker for subclinical atherosclerosis commonly used in clinical trials. Their sample size and power calculations apply 2-sample independent t-tests and within group variance in progression rates from the literature. However, this approach obscures the impact of differences in study designs including length of follow-up and differences in the number of and interval between ultrasound scans. These effects can be assessed using common sample size formula for longitudinal models, but this approach requires decomposition of the total variance into between and within subject components that have not generally been reported in the literature. Here, we derive these variance components for the Measuring Effects on intima-media Thickness: an Evaluation of Rosuvastatin (METEOR) study, a randomized, double-blind trial that demonstrated treatment with 40 mg rosuvastatin significantly slowed CIMT progression in middle-aged patients with a low Framingham risk of coronary heart disease and subclinical atherosclerosis (baseline maximum CIMT ≥1.2-<3.5mm). We examined the impact of differing follow-up periods, use of intermediate scans, and use of duplicate scans using both sample size calculations and actual analyses based on subsets of the METEOR data. Reductions in study length or number of scans result in increased variances and larger sample sizes to detect a given treatment effect. Table
shows the impact of duplicate scans at baseline and end of the 2-year study, with and without intermediate scans performed every 6 months, on the sample size required to detect a treatment effect of 0.012 mm/year. These results underscore the importance of considering the number and spacing of ultrasound exams explicitly during study design, and suggest that reductions in scanning frequency may seriously erode study power and/or increase costs by requiring recruitment of additional subjects.