P120: An Exploratory Meta-Analysis on the Effect Size of FDA-Approved and Not-Approved Anti-Cancer Drugs Over 25 Years

Poster Presenter

Koji Ishizuka

Senior Director, Oncology Alliance Management
Daiichi Sankyo, Inc
United States

Objectives

The objective of this study is to clarify the average performance of FDA approved anti-cancer drugs over 25 years and identify a potential boundary on effect size between FDA approved and not-approved drugs.

Method

We collected the registrational trials with survival measurement initiated from Jan 1, 1990 to Jan 1, 2021 on the Clinical Trial.gov and confirmed the FDA approval status. We estimated the effect size of FDA approved and not-approved drugs using random effect model by Stata ver. 17.

Results

3,463 clinical trials were screened. Out of them, we identified 209 interventional, randomized Phase III studies conducted in the United States with the purpose to test superiority of new anti-cancer drugs using overall survival(OS). The analyzed dataset contained 102 drugs indicated for 27 tumor types (20 solid tumors, 7 hematologic malignancies). Of them, 79 studies (63,396 patients) were classified in FDA-approved group whereas 130 studies (97,524 patients) were in FDA not-approved group. In this data set, the probability of success (POS) from Phase 3 start to the FDA approval appeared 37.8% (79/209 trials), which was in agreement with the previously reported POS [33.0% - 41.3%, Thomas, D.W., et. al. (2016), Zhou S. and Johnson R (2016), respectively]. The estimated effect size in OS in FDA approved group was HR 0.71 [95 % CI: 0.69 – 0.72], which generally aligned with the suggested target HR by ASCO(American Society of Clinical Oncology) 2014 guideline. FDA not-approved group showed HR 0.95 [95% CI: 0.93 – 0.97]. The histogram of the two groups showed distinct two peaks in observed HR in OS. However, they were overlapping around HR 0.75 to 0.90 which suggested a potential boundary between the two groups. Notably, the probability of FDA approval appeared drastically dropping down from 82.4% to 8.3% at this boundary.

Conclusion

The average performance of the FDA-approved anti-cancer drugs was generally in line with the target HR suggested by ASCO (< HR 0.80). The probability of FDA approval substantially dropped in case HR was around 0.75 to 0.90. This suggests FDA approval is sensitive to the effect size in OS. Our analysis was based on the registrational trials to test superiority of the new anti-cancer drug over the standard of care. In such a situation, it is recommended to select the compound which is reasonably expected to achieve less than HR 0.73 in OS to attain more than 80% likelihood of approval.