Roughly half of conducted clinical trials are never published

Goldacre's foundational statistic is stark: across the treatments currently prescribed, only about half of the trials ever conducted on them make it into the published medical literature that doctors consult. This is not a minor gap in an otherwise complete record — it means that when a doctor or a systematic review tries to assess whether a treatment works, they are working from a coin that has, in effect, only been allowed to land on one side roughly twice as often as it should.

Crucially, this missing half is not a random, unbiased sample of all trials conducted. Trials with positive, flattering results are substantially more likely to be published than trials with disappointing or negative results, which means the visible literature systematically overstates how well treatments work. Goldacre stresses this has been documented for roughly three decades without being structurally fixed.

Takeaway: An evidence base missing half its results, non-randomly, isn't incomplete — it's biased in a predictable and dangerous direction.