CMR Core Labs: Pros and Cons for Multi-center Clinical Trial-Topic Video

Video Transcription

Video Summary

The talk introduced basic biostatistical ideas relevant to medical imaging studies, especially MRI and cardiovascular applications. The speaker explained how p-values are interpreted as evidence against the null hypothesis, using one-sided and two-sided tests, and emphasized that statistical significance depends on effect size and sample size. A key message was that study planning must happen before data collection: researchers should define the outcome type (dichotomized, continuous, or survival), expected effect size, number of groups, available enrollment rate, and acceptable type I and type II error rates.<br /><br />The talk then covered power calculations, showing that power increases with larger effect sizes and sample sizes, and decreases when variability is high. The speaker also distinguished univariate from multivariate analysis, warning that including too many or highly correlated covariates can make models unstable or non-convergent. Multivariate models are useful for adjusting confounders and testing whether a marker remains significant after accounting for age, tumor size, and other factors.<br /><br />Finally, the speaker discussed survival analysis, highlighting that Kaplan-Meier curves depend on time and event status, with censoring as an important feature. The talk ended with net reclassification improvement, which evaluates whether adding a new marker improves risk classification, while noting its advantages and limitations.

Keywords

biostatistics

p-values

power calculation

multivariate analysis

survival analysis

Kaplan-Meier

censoring

net reclassification improvement