The Voice of Epidemiology

    
    


    Web EpiMonitor

► Home ► About ► News ► Job Bank Events ► Resources ► Contact
 
NIH Gordon Lecturer Highlights Enormous Waste In Epidemiologic and Biomedical Research And Suggests  7 Preventive Measures

The topic is not a new one and has been tracked for some time by Michael Bracken, professor of epidemiology at the Yale School of Public Health. Speaking as the Robert S Gordon Jr Lecturer on “Inefficiency and Waste in Biomedical Research: How Prevalent Is It, What Are Its Causes, and How Is It Prevented? Bracken described ways in which research work is wasted. For example, 50% of research is never published, another 50% is flawed, and a final 50% may be redundant or unnecessary, according to one estimate. Thus, only 12.5% is not wasted.

The Gordon Lecture at NIH has been given for over 20 years as a tribute to Robert S Gordon for his outstanding contributions to epidemiology and his distinguished service to NIH. A list of previous lecturers is provided in an accompanying article in this issue.

False Positives

One important finding about research waste relates to false positives. Early research in an area is often later found to be wrong or exaggerated. “Repeat studies tend to no effect,” Bracken told the NIH audience, “and the risk of being misled is very high.”

Bracken discussed the topic of false positive results in many observational studies whose results were overturned in clinical trials, and said that implementing larger studies did not automatically remove bias. He reflected on the angst these false positives have created in the epidemiology community and even offered the opinion that the much discussed and criticized article several years ago by Gary Taubes in Science entitled “Epidemiology Faces Its Limits” actually rendered a service to the profession.

Wasteful Replication

An important contributor to the waste in research is the work done to replicate previous findings. According to Bracken, such replication reaches a point at which the replication is no longer necessary because the evidence is in. Yet investigators continue to repeat studies unnecessarily. One potential solution to this problem is to track research prospectively so that the accumulating body of evidence in any area becomes more apparent.

SIDS Example

A striking example of the potential public health impact of excessive research and the delay it causes is the one related to sudden infant death syndrome. By some estimates, many follow up studies about the baby sleeping position were redundant and delayed public health recommendations. Had the evidence been properly tallied, an estimated 10,000 babies in the United Kingdom, and another 50,000 in Europe, Australia, and the US could have been saved.

Preventive Measures

Among the measures advocated by Bracken to help prevent wasted research are the following:

1. Reduce the play of chance by designing studies with more stringent alpha and beta values in much larger studies. Bracken recommends using alpha errors of 0.01 or 0.005 rather than 0.05 and a beta error of 0.10 not 0.20

2. Pay more attention to threats from multiple comparisons. Bracken reminded the audience that when Bradford Hill was describing standards for relying on research findings, there was often only one hypothesis per study being tested. Today, it is not uncommon for 200 associations to be examined. “A different world” said Bracken.

3. Strive harder to avoid bias, which subverts all research.

4. Comply with appropriate standards and guidelines.

5. Systematically review extant evidence in a published protocol before starting new research.

6. Follow the protocol outcomes when reporting findings or declare the differences.

7. Systematically update new findings in the context of all available research

To listen to the videocast of the Gordon Lecture, visit:

https://tinyurl.com/gnfssar  ■

 


Reader Comments:
Have a thought or comment on this story ?  Fill out the information below and we'll post it on this page once it's been reviewed by our editors.
 

       
  Name:        Phone:   
  Email:         
  Comment: 
                 
 
       

           


 

 
 
 
      ©  2011 The Epidemiology Monitor

Privacy  Terms of Use  Sitemap

Digital Smart Tools, LLC