Showing posts with label Statistics. Show all posts
Showing posts with label Statistics. Show all posts

Wednesday, February 27, 2013

Dialog on Drug Baron's post "The Primacy of Statistics: In defense of the pivotal Phase 3 Clinical Trial"


The Primacy of Statistics: In defense of the pivotal Phase 3 Clinical Trial

By Drug Baron (David Grainger)

My comment on the above post;

A very compelling argument indeed - One factor though I think needs to be taken into account is the type of indication for which the drug is being pursued.
While for the more prevalent indications such as metabolic disorders, cardio-vascular diseases et al where the sample size is large statistical rigor is highly relevant & decisive while approving the drug, can it be simultaneously argued that for orphan & other highly specific indications, where the sample is small, a solely statistical model will eliminate a lot of potential treatment options to statistical bias? – particularly since its being increasingly noted that an individual’s genetic make-up (presence or absence of mutations on a specific gene et al in the healthy or diseased tissue) can determine how the patient responds to the drug under evaluation? (case-in-study, Vemurafenib working for BRAFV600-Mutation Positive Metastatic Melanoma patients)  

Drug Baron's reply t my above comment:


davidgrainger Mod  Murali Apparaju  an hour ago

Thanks for your comment. This goes right to the nub of the argument.
You are exactly right that the slavish adherence to statistics will deny people (particularly in small indications) access to medicines that do actually work. In the limit, unless you are an identical twin, you are the only person with your genotype and maybe the drug would work brilliantly for you and for no-one else. With statistics as the gatekeeper, you will never get access to that drug.
BUT the key point of the piece is to point out that without statistics there is no way to know if that drug really did work for you. There is no control. At present (and maybe always) there is no alternative to a statistical test to be sure that a drug works at all.
Unless you are happy to approve drugs that MIGHT work, then we have no choice but to accept that with a statistically significant phase 3 trial as the "gatekeeper" we will reject drugs that actually work, but which we cannot prove they work.
For me, I would rather have the current system - where drugs have to be proven to work - than the one that existed prior to regulators, when snake-oil salesmen could sell anything as long as they could assemble a compelling enough argument to persuade the purchaser. That was a bad model - but allowing drugs through that havent passed a statistical test simply because they may work in some people, and there arent enough people to do the proper test, is a big step backwards.
Yet I see that happening more and more, particularly in the orphan drugs space that you plead as a "special case" - which is precisely why I wrote this article!