26 February 2025 3 4K Report

Hello beautiful and brilliant minds of the internet,

I don’t come from a biostatistics or informatics background, so your insights on this topic would be greatly appreciated!

??? The Question: Is there a systematic way to determine the optimal number of in-measurement replicates (within one subject) for a given parameter, based on its coefficient of variation (CV)?

We know that different echocardiographic parameters exhibit varying levels of measurement variability:

  • lets's say for the sake of argument- Left ventricular (LV) wall thickness has relatively low intra-measurement variability.
  • Mitral valve flow has higher intra-measurement variability due to hemodynamic fluctuations.

If I want to detect a true 1-unit change over time within the same subject, I need to ensure that measurement error (technical variance) does not obscure the change.

??? Can we establish a power-based approach (similar to standard sample size calculations-power analysis) to determine the number of in-measurement replicates needed for a given CV?

For example:

  • If a parameter has a CV of 15%, how many replicates do I need to confidently detect a 1-unit change?
  • If I choose to measure only 5 times instead of 10, how can I quantify the trade-off in terms of the minimal detectable change?

Would love to hear your thoughts, references, or any existing methods that tackle this problem

More Mila Bukova's questions See All
Similar questions and discussions