Testing treatments
How do we know a treatment works?
Guess first. Then check. Seventeen short modules.
Start here
Module 0
Are they different?
Drug or placebo: is the difference real, or luck? Why flip a coin to decide who gets which?
✓Module 1
Before and after the trial records
How trial records change what you believe about a drug.
✓Module 2
What a p-value is (and isn’t)
If the drug does nothing, how surprising are these trial records?
Designing a trial
Module 3
Errors, power and sample size
How many patients before we can trust the answer?
✓Module 4
Many tests, false alarms
Could one of 20 useless drugs look helpful?
✓Module 5
Case: the ulcer surgery trial
Four operations, 100 patients each. Which one would you pick?
✓Module 6
Fair comparisons: placebo, blinding and who counts
Dummy pills, hopes and patients who quit: what can still go wrong.
Watching a trial
Module 7
Peeking at the data: interim looks
Check again and again and you can fool yourself.
✓Module 8
Check after every patient
A way to look at any time and stay fair.
✓Module 9
Adaptive and Bayesian designs
Change a trial while it runs, fairly.
✓Module 10
Platform trials
Ten drugs to test: ten trials, or one that keeps running?
Deciding
Module 11
What counts as proof
One trial wins. Would it win again?
✓Module 12
What blocking false alarms costs
Fewer false alarms means more missed good drugs.
✓Module 13
The regulator’s map
Who checks a new medicine, and when?
✓Module 14
Trial simulator: test the rule
Simulate 2,000 trials to test a stopping rule.
✓Module 15
Reading a Bayesian trial
“Probability of benefit 99%”: what does it mean?
✓Module 16
Capstone: your Bayesian protocol
Design a trial, check it by simulation, defend it.
Studio
Trial simulator studio
Free, open simulator: many arms, dose finding, borrowing, a full report.