Lies, damn lies and sta­tis­tics: A Stan­ford wiz says P<0.05 of­fers de­cep­tive ev­i­dence of bio­phar­mas' drug claims

The bio­phar­ma R&D world re­volves around one sim­ple for­mu­la: A P val­ue of less than 0.05 in a piv­otal study. But a top pro­fes­sor of med­i­cine and sta­tis­tics at Stan­ford says it’s a poor mea­sure of val­ue, and he wants to scrap it for some­thing far more de­mand­ing — and far more valu­able.

Writ­ing in the Jour­nal of the Amer­i­can Med­ical As­so­ci­a­tion, John P A Ioan­ni­dis notes that the P val­ue cut­off “is wrong­ly equat­ed with a find­ing or an out­come…be­ing true, valid, and worth act­ing on. These mis­con­cep­tions af­fect re­searchers, jour­nals, read­ers, and users of re­search ar­ti­cles, and even me­dia and the pub­lic who con­sume sci­en­tif­ic in­for­ma­tion.”

This article is for premium subscribers only

Upgrade to a premium subscription plan for unlimited access, and join our community of key biopharma players.

UPGRADE