The 2-tri­al hur­dle for drug ap­provals is steadi­ly los­ing ground. Are we de­valu­ing the in­dus­try's R&D gold stan­dard?

Over the week­end I had a lengthy dis­cus­sion with Acor­da CEO and for­mer BIO chair­man Ron Co­hen on Twit­ter about p val­ues and FDA stan­dards that I’d like to share with you.

It re­lat­ed to a sto­ry we ran a few days ago cit­ing Stan­ford’s John P A Ioan­ni­dis on the like­li­hood of false pos­i­tives when you set a stan­dard for suc­cess at p<0.05. Ioan­ni­dis and sev­er­al oth­er sta­tis­ti­cal ex­perts have voiced the need of rais­ing that sta­tis­ti­cal bar — low­er­ing the cut­off — to pre­vent stamp­ing some fail­ures as a suc­cess.

Co­hen, who felt that the FDA’s “typ­i­cal” stan­dard re­quire­ment of two “ad­e­quate, well con­trolled” stud­ies made it vir­tu­al­ly im­pos­si­ble to get an ap­proval based on a sta­tis­ti­cal­ly shaky read­out, wasn’t ac­cept­ing any anec­do­tal analy­sis from me. He want­ed hard num­bers. So I got some. The FDA is quite ex­plic­it about the da­ta used to ap­prove a drug. It’s easy to check how many drugs were ap­proved last year us­ing the two-tri­al stan­dard.

Endpoints Premium

This article is available only to Premium subscribers

Upgrade to Premium for unlimited access and Premium exclusives.