207_Combined course Presentations

Adaptive Trials

• Good  :

Adaptive algorithm

– Pick-a-winner – Can adapt on drug or biomarker – Smaller, conserve resources • Bad  : – Interim estimates=  error risk – Complicated! Continuously collecting response data – If biomarker-based • Must be validated.

Early/iterative analysis (drug or biomarker working?)

Stopping rule met?

Continue data collection

Yes

No

• Need real-time results • Cannot do discovery

Stop trial or begin next phase

Revise allocation per algorithm e.g. randomize more to Drug A arm

Example: ISPY2 - novel biologics in combination with chemotherapy