If AI found better drugs for every disease, what would it cost to prove it?
If AI found better drugs for every disease, what would it cost to prove it?
An analysis estimates the cost of pivotal trials to prove that, for every deadly disease in the Global Burden of Disease list, a new therapy halves the risk of the worst outcomes versus today's best care, using one trial per disease and Moore et al.'s median cost of $53,904 per participant (2026 dollars). The author was surprised the total was small relative to the roughly $100 billion the industry spent on phase 3 trials in 2024, attributing part of the gap to the large assumed effect size (today's best-care trials cut bad-outcome risk by a median of less than half) and part to redundancy, such as thousands of trials targeting PD-1/PD-L1. The piece suggests better-coordinated shared trial platforms, citing RECOVERY, I-SPY 2 and SEMATECH as models.
Source: Hacker News AI · cvigoe.github.io