An Epidemic of Bad Research and Reporting

An Epidemic of Bad Research and Reporting

By: Sarah Kunkle

Original Publication found here:

Earlier this summer, the popular science and technology blog io9 ran a story that caught the eyes of many: “I Fooled Millions into Thinking Chocolate Helps Weight Loss.” Over the course of the article, John Bohannan, a science journalist, describes his elaborate hoax and laments the state of both nutrition research and science reporting. Unfortunately, this is all too common.

Although the study was real, it was intentionally plagued by methodological and analytical flaws, including an extremely small sample size and large number of measurements that gave the study a greater than 60 percent chance of finding at least one statistically significant result. To address these issues, some journals are considering getting rid of p-values (a measure indicating how likely it is that study results are due to chance) and many do not accept studies with fewer than 30 subjects. Nevertheless, many low quality studies still end up published in peer-reviewed journals.

Nutrition research is particularly vulnerable to biased results because of its dependence on self-reporting. A recent Mayo Clinic Proceedings article argued that memory-based dietary assessment methods were “fundamentally and fatally flawed” and should not be used to inform dietary guidelines. Organizations like the Nutrition Science Initiative are trying to combat these issues by funding more rigorous (and expensive) studies. While the evidence is inconclusive for some nutrition research questions, the 2015 Dietary Guidelines Advisory Committee seems to be a step in the right direction with its emphasis on minimally processed wholesome foods rather than specific nutrients.

In addition to poor quality research, bad reporting further complicates the issue. As Bohannon notes, reporters covering topics such as nutrition or broader scientific research should not merely echo what they read in press releases: “you have to know how to read a scientific paper – and actually bother to do it.” Readers should be especially weary of articles that do not mention sample size and effect size.

Bohannan is not alone in his views. Lancet editor Richard Horton recently published a commentary on bad scientific practices, claiming, “much of the scientific literature, perhaps half, may simply be untrue.” Increased public awareness and transparency are likely to ameliorate the problem. In the meantime, both reporters and readers should be cautious as they digest health headlines – if it sounds too good to be true, it likely is.

How to avoid getting duped by overblown health claims (via Quartz)

1. Are humans involved? If the claims are based on a study done in mice, the results are not necessarily applicable to humans.
2. What is the sample size? Be skeptical of results that involve less than 100 people and fairly skeptical of those that involve less than 1,000 people.
3. What type of study? Study design matters. Systematic reviews, meta-analyses, and randomized controlled trials are generally the most reliable for testing hypotheses.
4. Is it correlation or causation? Relevant to the point above on study design, most health studies draw mere correlations rather than direct causes.
5. What are the study’s limitations? A good health story will not only explain the results, but also discuss the study’s limitations and why you shouldn’t trust the claim fully.

Leave a Reply

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

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

Google+ photo

You are commenting using your Google+ 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 )


Connecting to %s