The emperor has no clothes

by Julie Fine, BS, legal specialist, Legal Division, Worldwide Research and Development, Pfizer Inc. (Please note: The views presented here are my own and do not reflect the positions or policies of Pfizer, Inc.)

During lunch on Tuesday, I took advantage of the opportunity to join other conference attendees for an intimate and informal question and answer session with Keynote Speaker, John P. A. Ioannidis, MD, PhD, who was as candid and provocative in this setting as he was earlier that morning during his talk. In his remarks during the Keynote Address, Dr. Ioannidis talked about bias in research reporting and faulty results, exposing an uncomfortable truth. During the lunch, Dr. Ioannidis entertained several questions from the group, but the following four were particularly notable for their challenging nature and his unflinching responses.

How would “Cloud” publications work and how would journals get to be “first” in this model? 
In the Keynote Address, one proposal for enhancing the transparency of reported results was a collective repository of publications, essentially “Cloud” publications. It would make published results accessible, searchable by interests, and available to journals. Dr. Ioannidis quipped that the journals would be able to make offers to authors (e.g., $500K from one journal, $750K from another, $1M from yet another) and authors could decide which offer to accept (the $1M, of course).

On a serious note, though, he commented that if journals were more apt to publish negative results it would be a valuable service and it might eliminate costly research into non-effective therapies and perhaps save lives. I imagine another positive outcome would be that researchers could redirect their efforts into more promising areas and redesign protocols based on findings from previous trials.

What are your thoughts on the registration of observational data? 
Dr. Ioannidis suggested that, historically, researchers publish their work in the field, one outcome at a time, narrowly showing what’s been tested. In what he described as his somewhat utopian view, all observational data would be registered and categorized into different levels. One level might be registration of data sets – what it is, what was published, how it was analyzed – thus, researchers would “show their cards.” A second level might involve registration of protocols in a highly pre-specified way, showing the exact protocol and outcome, and indicating if it was a prospective study.

What more can institutional review boards (IRBs) do? 
Dr. Ioannidis believes that IRBs should insist on registration in support of retrospective review. He suggests a systematic process for evaluating whether or not an investigator’s study is “worth it.” This evaluation would include questions such as:

  1. What research has been conducted previously?
  2. How is this study unique?
  3. Has the researcher’s prior work (for example, a randomized trial completed two years ago) been published? And if not, should additional research be approved? 

What can be done about the whole medical enterprise? 
Dr. Ioannidis recommended that medical training ought to offer more training to enable physicians to better perform and conduct research. He suggested instruction on optimal research design and methods (which would help tackle the issue of bias), understanding risk, and appropriate interpretation of results. In her remarks at the end of the session, Joan Rachlin, JD, MPH, executive director of PRIM&R, complimented Dr. Ioannidis for having the courage to be like the boy in the children’s story who boldly called out the emperor who had no clothes – reminding us not to be fooled or go along blindly, but rather to look clearly, deeply, and honestly at the research enterprise