Astrology and the interpretation of data from clinical trials


Over the weekend, an  interesting article on clinical trials appeared (somewhat unexpectedly) in the New York Times color supplement.

The article, by Siddartha Mukherjee, MD, is entitled “A failure to heal“, and for those who don’t know (or have forgotten), Dr. Mukherjee is a medical oncologist and the author of a highly regarded book entitled The Emperor of All Maladies: A Biography of Cancer.

In this new article Dr. Mukherjee writes about two things in particular:

  • How physician-researchers like him are affected by clinical trials that fail to meet their hoped for outcomes, and
  • Why so-called “post hoc” analysis of data from subsets of patients in clinical trials that failed can lead to both
    • The possible identification of subsets of patients in whom the tested drug might actually work quite well
    • Mistaken beliefs that the tested drug really does work well in certain subsets of patients when there is no actual justification for such a belief

The development, execution, and analysis of the results of clinical trials are all somewhat more complex than many of us would think. Such trials are designed to test very specific hypotheses under very highly defined circumstances.

The over-interpretation of data from such trials (sometime referred to as “data-dredging”) can therefore be very dangerous, and Dr. Mukherjee reminds us of a famous study carried out some 30 years ago that is still being used to make this point with a high degree of clarity for all concerned. (Astrologists please note.)

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: