Biostatistics Dr Larry Rubinstein National Cancer Institute Phase

Biostatistics Dr. Larry Rubinstein, National Cancer Institute

Phase 0 Trial Statistics for PD Assay v Endpoints are PD assay, baseline (ideally, just before Tx) vs. post-Tx, in tumor tissue and in blood v Baseline measures: 1 pre-treatment tumor biopsy measurement and at least 3 pre-treatment blood measurements (ideally, taken over the baseline to post -Tx time span) per patient v 1 post-Tx measurement used as primary endpoint for tumor and blood (at same time post. Tx) v Declare significant Tx-related change, for an individual patient, for observed 2 -fold change combined with 90% statistical confidence of a true difference (10% false positive rate) ¬ For blood this is 1. 8 x intra-pt pre. Tx SD ¬ For tumor tissue this is 1. 8 x inter-pt pre. Tx SD

Minimal Phase 0 Trial Design v Accrue 3 patients to each of the dose levels v For either endpoint, tumor or blood, significant Tx-related response for a dose is defined as significant change for at least 2 patients out of the 3 v For a dose, if there is 80% likelihood of observing change at the patient level, there is 90% power to detect significant response for the dose level v There is an overall 6% false positive rate, for both endpoints combined, for a dose level with no biologic effect (because of the 10% false positive observed change at the patient level)

Enhanced Phase 0 Trial Design v There may be a need to detect a lower rate of response per dose level—use a 2 -stage design v Accrue 3 -5 patients per dose level, allowing for the expansion of a cohort to 5 patients if the results are not definitive for the initial 3 v Declare significant change, for an individual patient, for an observed 2 -fold change combined with 95% statistical confidence of a true difference ¬ For tumor tissue this is 2. 3 x inter-patient pre-Tx SD ¬ For blood this is 2. 3 x intra-pt pre. Tx SD

Enhanced Phase 0 Trial Design Details v Accrue 3 patients to each dose level ¬ If a significant change is observed for exactly 1 patient, for either endpoint, expand the dose to 5 patients ¬ For either endpoint, significant Tx-related response for a dose, for cohort size 3 or 5, is defined as significant change for at least 2 pts v For a dose, if there is 60% likelihood of observing change at the patient level, there is 89% power to detect significant response for the dose level v There is an overall 4% false-positive rate, for both endpoints combined, for a dose level with no biologic effect (because of the 5% falsepositive observed change at the patient level)

The Next Speaker is: Dr. Holly Taylor