Clinical Trial Statistics: Sample Size, Randomisation & Survival Curves

What you'll build

What you will be able to do

Each lab ends with a concrete result that you build yourself and the checker confirms, cell by cell.

• Calculate how many patients the fictional TL-HF01 trial needs per arm for a binary and for a continuous endpoint, and inflate the total for dropout.
• Turn the sample-size formula around: what power does the trial have with only 500 patients per arm? Then you build and chart the power curve.
• Allocate 24 participants in blocks of 4 from frozen random numbers, then compare the balance with simple coin-flip randomisation and calculate how likely such an imbalance is.
• Build "Table 1" of a trial report from patient rows: mean and SD of age and % female per arm, and a standardised difference to judge balance.
• Analyse the trial's primary endpoint: the difference in 12-month hospitalisation rates with a 95% confidence interval and a two-sided p-value from the two-proportion z-test.
• Translate the trial result into the absolute risk reduction and the number needed to treat, give the NNT a confidence limit, and weigh it against the number needed to harm.
• Build a Kaplan-Meier table for 20 patients (at risk, events, censoring, conditional and cumulative survival), read off the median, and draw the step-shaped survival curve.
• Run the planned interim look at half the data against a given O'Brien-Fleming boundary, convert the boundary into a nominal α, and adjust four secondary endpoints with Bonferroni.