Analysis of the sample size used in clinical MRI studies

Abstract

Background Choosing the sample size in clinical MRI studies is a common, important, and challenging task, complicated by the substantial variation in potential study parameters. However, considering previously used sample sizes may provide a reference point for future studies. The purpose of the study was to systematically investigate and to provide orientation for sample size selection based on information from current practices in clinical MRI studies. Methods We assessed 1,046 research articles published in the Journal of Magnetic Resonance Imaging (JMRI) between 2020 and 2023. Only studies that involved patients were included. Review articles and studies using phantoms, animals, ex vivo samples, publicly available datasets, and non-imaging techniques (e.g., spectroscopy) were excluded. The included studies were categorized according to various criteria including anatomical region, field strength, contrast category (e.g., T1 mapping or diffusion-weighted imaging), retrospective vs. prospective and single vs. multicenter studies, automatic or manual segmentation, and quantitative or qualitative evaluations. Results The median sample size (=number of patients) of the 734 studies included in the analysis was 74.5 (retrospective studies =  129, prospective studies =  41) and varied between the investigated categories. Sample size clusters were found in multiples of ten (e.g., 20, 30, 40), and 90.3% of the studies had less than 350 patients with 50.5% having less than 75, while 1.6% had more than 1,000 patients. Conclusion There is wide variation in the sample sizes of studies published by JMRI between 2020 and 2023, depending on study type, content category, or evaluation method. In clinical MRI studies, balancing statistical power and minimizing patient involvement is crucial, necessitating carefully choosing the sample size.