Pharmacogenomics Will Prove That 30% of Drug 'Failures' Were Actually Successes in the Wrong Population

Drug development has a 90% failure rate. But this statistic hides something: many "failed" drugs worked brilliantly in a subset of patients. The average efficacy was dragged down by non-responders who should never have been given the drug in the first place.

Trastuzumab would have failed if tested in all breast cancer patients — it only works in the ~20% that are HER2-positive. Ivacaftor only works in the ~4% of CF patients with G551D mutations. These drugs were saved by biomarker-driven patient selection. How many drugs died because nobody looked for the responsive subgroup?

Hypothesis: Retrospective pharmacogenomic analysis of Phase II/III failures from the past 20 years will identify responsive subpopulations (defined by genetic, transcriptomic, or proteomic biomarkers) for >30% of "failed" drugs. Resurrecting these drugs with companion diagnostics will be more productive than discovering new molecules.

Prediction: At least 10 previously failed drugs will be approved in new, biomarker-selected indications by 2032, with the first major success in CNS (likely a failed Alzheimer's drug repurposed for a genetically-defined subtype).