Posted by Gary Schwitzer in Journal practices, Limits of observational studies
(From Open Clip Art Library - MajinCline)
Around the globe today, there are misleading headlines about a study in the BMJ, “Long term alcohol intake and risk of rheumatoid arthritis in women: a population based cohort study.”
Alcohol cuts arthritis risk by half in women – Times of India
Alcohol ‘lowers arthritis risk’ for women – The Independent
Alcohol ‘lowers arthritis risk’ – The Press Association
Rheumatoid arthritis risk reduced by moderate alcohol consumption – News-Medical.net
Alcohol reduces risk of arthritis: study – The Local.se
Frequent tipple ‘halves arthritis risk’ – Irish Independent
Drinking wine can keep arthritis at bay, scientists tell women – Scotsman
Drink Thrice a Week to Halve Risk of Rheumatoid Arthritis – HeartZine
ALCOHOL ‘REDUCES RISK OF ARTHRITIS’ – Express.co.uk
Risk of arthritis half for ‘moderate’ female drinkers – The Conversation
Moderate Drinking Reduces Rheumatoid Arthritis Risk among Women – French Tribune
Each headline used a definitive statement of a causal connection: cuts/lowers/reduces/halves risk or some variation thereof. And each is inaccurate when applied to the study in question.
The researchers simply concluded this: “Moderate consumption of alcohol is associated with reduced risk of rheumatoid arthritis.”
A statistical association, not proof of cause-and-effect. And that’s all you can accurately say.
And what may have kicked it all off? A BMJ news release:
BMJ Press Release
Moderate drinking may reduce risk of rheumatoid arthritis
Three drinks per week can halve the risk of developing the condition
In that news release, there was not one mention of the limitations of such an observational study – no emphasis on association versus causation.
HealthDay ended its story appropriately and accurately: “Although the study found an association between alcohol and rheumatoid arthritis risk, it did not prove a cause-and-effect relationship.”
Maybe more media around the globe would have gotten it right if the BMJ news release had helped them a bit. We’ve written about flawed BMJ news releases on observational studies before:
We refer the BMJ news release writers to our primer, “Does the Language Fit the Evidence? Association Versus Causation.”
Journals could help lift all ships – or they can (and sometimes do) help us all drown in a daily tsunami of global miscommunication about health news.
A study finds lower rates of gestational diabetes, but also more small babies, stillbirths, and newborn deaths in women who had bariatric surgery prior to their pregnancies compared with obese women who didn’t have surgery.
A generally complete report on a study of peer phone counseling for postpartum depression.
Many consumers and health care providers continue to hope that a pill, including a seemingly harmless one like vitamin D, will be the solution to a host of chronic diseases. This story explains why that thinking has a lot of holes in it.
We will not display your name or e-mail address on the site, nor will we share youre-mail address with any third parties.
For site updates and news from Gary, find us on Twitter:
HealthNewsReview: @HealthNewsRevu
Gary Schwitzer: @garyschwitzer
Copyright © 2015 Health News Review
Patricia Hartman posted on July 12, 2012 at 7:10 am
This is only one example of the current trend: do a statistical analysis and publish it, without any real effort to explain or validate the “association”. Thus Vitamin D deficiency is now associated with (and reported in lay press as cause of) asthma, heart disease, etc etc. Is this another symptom of publish or perish?