I gave a talk this morning to the American College of Allergy, Asthma & Immunology (ACAAI). I documented my preparations for this talk on my webpages and wanted to share some thoughts I had during and after the talk.
This was by far the largest audience I had ever given a talk for. I don't have an official count, but the talk was in one of those large ballrooms (the Texas Ballroom, both A and B!) and more than half of the seats looked† to be filled. It was easily more than five hundred people. No doubt many of the attendees were there for the other two speakers, but it was still flattering to have such a large group to talk to.
In reviewing my notes prior to the talk, I found out to my dismay that I don't have a good web page outlining the five phases of testing to establish the validity of a diagnostic test. These five phases were developed by Margaret Pepe and are outlined in her book. I allude briefly to these five phases in a June 15, 2007 weblog entry, but I need to expand on this and place it on its own page.
I also need to elaborate on the sources of bias that can affect the calculation of sensitivity and specificity in a diagnostic test, spectrum bias, verification/workup bias, and bias caused by an imperfect gold standard. These are also covered briefly in the same weblog entry. I did mention prominently in my a talk a fourth source of bias, failure to blind knowledge of the diagnostic test from the person conducting the gold standard and vice versa. This point came up in the second speaker's talk.
There were several good references given by the other speakers, which I need to review.
I was asked about how prevalence can influence a diagnostic test, and although I covered this well on my webpages, I did not include an appropriate discussion of this on my PowerPoint slides. If I give this talk again, I'll be sure to add this to the slides.
(Update: November 14, 2007) At the beginning of my talk, I briefly discussed my concerns with the conflict of interest requirements that the ACAAI and most other organizations have. I believe that a statement that no conflict exists needs to be an active statement that explicitly states whether any commercial product is mentioned, whether regulatory changes are advocated, and what financial ties are associated with companies that sell the commercial product or which benefit from the proposed change in regulations. The statement should also disclose any financial ties with competing companies. I have tried to put my thoughts down in a November 12 weblog entry.