AI
Thursday, April 07, 2016 9:58:47 PM
This is how to get outperforming control arms OR study drug arms from oncology trials:
1. Small randomized trials, no stratification during randomization based on baseline characteristics like geographic regions, prior therapies, histologies etc resulting imbalances between the arms.
And/Or
2. Enroll LARGE portion of patients in Eastern Europe or Asia especially India where SOC differ from the west.
Example, striking difference in mOS from SNTA 2nd line NSCLC ph2 trial between East and West. It was very nice of SNTA providing detailed explanation from ph2 - look how different the patient population of two Eastern European countries from rest of the trial and typical/historical 2nd line NSCLC trials - at least SNTA tried to correct in ph3 but unfortunately not helpful because the great ph2 result/signal (4 months better than control arm in mOS) was a mirage. The end result, no surprise, futility at 1st interim:
http://phx.corporate-ir.net/phoenix.zhtml?c=147988&p=irol-newsArticleIR&ID=2098806
This is why talking about mOS without looking at actual patient population is useless. Historical control means nothing if your trial population differ significantly.
1. Small randomized trials, no stratification during randomization based on baseline characteristics like geographic regions, prior therapies, histologies etc resulting imbalances between the arms.
And/Or
2. Enroll LARGE portion of patients in Eastern Europe or Asia especially India where SOC differ from the west.
Example, striking difference in mOS from SNTA 2nd line NSCLC ph2 trial between East and West. It was very nice of SNTA providing detailed explanation from ph2 - look how different the patient population of two Eastern European countries from rest of the trial and typical/historical 2nd line NSCLC trials - at least SNTA tried to correct in ph3 but unfortunately not helpful because the great ph2 result/signal (4 months better than control arm in mOS) was a mirage. The end result, no surprise, futility at 1st interim:
http://phx.corporate-ir.net/phoenix.zhtml?c=147988&p=irol-newsArticleIR&ID=2098806
This is why talking about mOS without looking at actual patient population is useless. Historical control means nothing if your trial population differ significantly.
