Does manual analysis of CMR imaging improve repeatability and reduce sample size requirements compared to semiautomated analysis in PAH patients?
Manual analysis of CMR in PAH patients is more robust than semiautomated analysis, leading to lower sample size requirements for clinical trials.
PURPOSE: To calculate the sample size for a theoretical pulmonary arterial hypertension (PAH) randomized controlled trial (RCT) by using cardiovascular magnetic resonance (CMR) imaging to determine the repeatability of measures between two scans. MATERIALS AND METHODS: Two same-day examinations from 10 PAH patients were analyzed manually and semiautomatically. Study size was calculated from the standard deviation (SD) of repeatability. Different approaches to right-ventricle (RV) mass were investigated, agreement between methods tested and interobserver reproducibility measured by Bland-Altman analysis to explore how the PAH heart might be best measured. RESULTS: Repeatability was good for almost all manually-measured indices but poor for semiautomated measurement of RV mass and left-ventricle (LV) end-diastolic volume (EDV). Thus, for an RCT (power, 80%; significance level, 5%) analyzing "outcome" indices (RVEDV, LVEDV, RV ejection fraction, and RV mass; anticipated change: 10 mL, 10 mL, 3%, and 10 g, respectively) manually, 34 patients are required compared to 78 if analysis is semiautomated. RV mass was repeatable if the interventricular septum was divided between ventricles or if wholly apportioned to the LV. Limits of agreement between manual and semiautomated analyses were unsatisfactory for RV measures and interobserver reproducibility was worse for semiautomated than manual analysis. CONCLUSION: Manual is more robust than semiautomated analysis and at present should be favored in RCTs in PAH as it leads to lower sample size requirements.
Bradlow et al. (2009) studied this question.