Monte Carlo vs. Bootstrap: How Uncertainty Is Quantified in a Virtual PK Trial

Answers: “Monte Carlo bootstrap confidence interval PK simulation”

A single Cmax number tells a reviewer nothing about confidence. Two techniques give a virtual PK trial its credibility: Monte Carlo simulation (to generate the subject population) and bootstrap resampling (to put confidence bounds on the summary statistics).

How they work together

• Monte Carlo draws 500 virtual subjects from the covariate distributions (CYP2A6, weight), producing a realistic spread of Cmax/Tmax/AUC.
• Bootstrap resamples that cohort to compute a 90% confidence interval around the means — the uncertainty FDA reviewers expect to see.
• Combined, you report both a 5–95% prediction interval (individual variability) and a 90% CI (estimate confidence) — two different, both necessary, uncertainty statements.

Pros vs. cons. Pro: this is standard, defensible statistical practice; VPC-style diagnostics let a reviewer check the model isn't fabricating structure. Con: simulation quality is bounded by input-distribution quality — garbage in, garbage out; state your priors.

Who should NOT buy: teams that only need a point estimate for internal screening and don't require reviewer-grade uncertainty reporting.