Amgen — 11% replication success of 53 landmark cancer biology studies

Amgen's hematology and oncology team attempted to replicate 53 landmark preclinical cancer studies and succeeded on 6 — an 11% success rate (S-0048).

The studies were not random. They were the foundational cancer biology that was guiding drug development at major pharmaceutical companies. The 11% figure is therefore a measurement of the reliability of the most consequential preclinical literature, not of mediocre work selected for failure analysis.

The case is one of the strongest empirical inputs for C-0022. The architectural reading: replication attempts that failed when raw data was unavailable terminated as inconclusive because the diagnostic step (examine the original measurements) was unavailable. Even where replication clearly failed for methodological reasons, the inability to inspect the original data prevented diagnosis of which methodological reason. The 89% failure rate compounds across decades of pharmaceutical R&D investment.