Of all the problems regarding large scale clinical trials cataloged by James Penston, the most compelling is the inverse relationship between practical value and trial size. This could almost be formulated as a law:
The clinical value of a randomized controlled trial is inversely related to its size
Of course, clinical is used in the original sense, meaning at the bedside.
It is a testimony to the effect of propaganda promoting “powerful” clinical trials that this law may sound counter-intuitive when in fact it is so obvious: if it takes 18,000 patients to demonstrate an effect, how relevant or useful is the information likely to be for a clinician dealing with an individual patient? And there is now some empirical evidence that the phenomenon of conducting, reporting, and inflating trials with “tiny effects” is getting more and more common.
Trial designers and drug companies are well aware that when it comes to mega-trials, the NNT will most certainly be very large from the outset. A lot of pre-existing data is utilized in making power analyses to predetermine the size of N, and large trials are for the most part (if not invariably) post-approval studies that seek to expand indications for a drug already on the market. And I suspect that if it was up to the market alone, many such trials would yield little gain for the pharmaceutical industry. That’s when guidelines qua mandates become so useful. The database doctors who sit on guideline committees have long ago decreed on the “hierarchy of evidence” and decided that a p-value is all that is needed to reward a given treatment with the prized “class I” indication. Again, Penston’s books have good references on this phenomenon.
We should also keep in mind that the value of randomized trials is not always to establish therapeutic efficacy. The 1965 VA Cooperative Study, for example, was able to settle that hypertension is not so “essential” after all. But it did not necessarily provide an easy answer to the clinically pertinent questions regarding which patient with hypertension should be treated, or at what level of blood pressure should treatment be initiated. Forty five years and hundreds of RCTs later, these questions continue to generate debate…
Meanwhile, the N-of-1 trial model is still largely ignored, even though it most suitably addresses the problem of “individual medicine.” Originally proposed by Guyatt et al. in the the mid 80’s, the methodology is barely registering a blip on the radar screen (less than 300 citations in Pubmed) despite its tangible benefits. New economic realities may change this, hopefully. And if it’s any indication, a number of papers have been authored in the last couple of years aiming to revive it (with one authored by no other than Eric Topol himself!).
What about N-of-9 being enough? Just a bad French pun, I’m afraid…