Sample size and power determination when limited preliminary information is available

Area of Science:

• Biomedical Imaging
• Medical Technology Assessment
• Quantitative Imaging

Background:

• Investigational technologies often lack preliminary data for power and sample size calculations.
• Diffuse optical spectroscopic imaging (DOSI) is an experimental noninvasive technique for assessing mammographic density.
• Tamoxifen treatment efficacy in reducing breast cancer risk is linked to decreased breast density.

Purpose of the Study:

• To develop a novel strategy for power and sample size determination for studies using investigational imaging biomarkers.
• To assess changes in DOSI imaging biomarkers reflecting breast density fluctuations in premenopausal women undergoing tamoxifen treatment.
• To address the lack of preliminary data for DOSI in tamoxifen treatment studies.

Main Methods:

• A statistical simulation approach was developed, leveraging existing magnetic resonance imaging (MRI) data.
• MRI data from 16 women before and after tamoxifen treatment were used to simulate DOSI biomarkers (water, ctHHB, lipid).
• Simulations generated 5,000,000 pairs of DOSI values based on varying correlation coefficients (0.5, 0.8, 0.9).

Main Results:

• Simulations successfully generated DOSI values corresponding to MRI data.
• Minimum sample sizes required per group were determined for clinically relevant effect sizes.
• The study provides a framework for estimating sample size in the absence of direct preliminary data.

Conclusions:

• The described simulation techniques are applicable to other experimental technologies.
• This approach provides crucial preliminary data for power and sample size calculations.
• Enables robust study design for novel imaging biomarker investigations.