Days before its FDA decision, Telix Pharmaceuticals’ kidney-cancer imaging agent TLX250-CDx was trading at 88% to win approval. The science looked solid: 86% sensitivity, 87% specificity, and a successful Phase 3 clinical trial. But on August 28th, 2025, the FDA rejected it for a second time.
The drug was first denied in 2024 over a manufacturing issue. Telix fixed what they could, resubmitted, and got another rejection. The reason cited was “deficiencies” at two of the drug’s outside factories.
Was this a black swan? A random manufacturing issue that would have been impossible for anyone on the outside to spot?
In fact, there was a clue hiding in plain sight - something within the drug’s timeline that would have stuck out to anyone who had examined a larger FDA drug approval dataset:
TLX250 had already been rejected once, and its second review came barely a year later.
In our first piece for The Oracle on FDA forecasting, we hand-checked 36 resolved polymarkets and found that the most common rejection reason was the factory, not the drug. We also admitted the sample was too small to build real base rates, and promised a follow-up on the full historical record.
Here it is. We pulled the most comprehensive dataset available of 446 FDA Complete Response Letters to find the base rates across several different dimensions.
What we found are several clear trends that will be helpful for anyone looking to forecast future approvals.