What is it about?
If 700 out of 1000 patients 'responded' to a treatment for migraine, it is often assumed that it must work all the time for 70% of sufferers. An alternative explanation, which is hardly ever considered, is that it works for 100% of sufferers 70% of the time. Intermediate explanations are possible. This piece describes a number of statistical pitfalls, such as this one, that have lead to the scope for personalised medicine being exaggerated.
Featured Image
Photo by Louis Reed on Unsplash
Why is it important?
If we make better attempts at establishing where there is scope for personalising treatment, we shall allocate resources better and make more progress in doing so than if we always naively accept that lack of personalisation is the explanation for every treatment failure.
Perspectives
Read the Original
This page is a summary of: Statistical pitfalls of personalized medicine, Nature, November 2018, Springer Science + Business Media,
DOI: 10.1038/d41586-018-07535-2.
You can read the full text:
Resources
Responder Despondency
A discussion of the inadequacy of statistical basis for the FDA's assumptions about the scope for personalised medicine in their report Paving the Way for Personalized Medicine https://www.fdanews.com/ext/resources/files/10/10-28-13-Personalized-Medicine.pdf
Numbers Needed to Mislead
A discussion as to how Numbers Needed to Treat are misunderstood
Is Precision Medicine Terminally Ill
Commented presentation showing how and why the scope for personalised medicine is being overestimated
Mastering Variation
A more technical article covering many aspects of the Nature Commentary in more depth and with more justification but freely available online
30 Years of Numbers Needed To Treat
Explains why Numbers Needed to Treat do not tell one what proportion of patients respond to treatment and links to various other blogs that expand on this.
Contributors
The following have contributed to this page