When science communication breaks down, people panic
Pills; not as dangerous as you might think
If you?ve been reading the news recently, you could be forgiven for thinking that ibuprofen is basically poison that is going to kill you with a heart attack. Scary headlines like ?Study links ibuprofen to heart problems? and ?Ibuprofen Linked To ?Increased Risk Of Cardiac Arrest?? have popped up everywhere, and people keep asking me the question;
?Should I stop taking ibuprofen??
Now that?s a complex question. Short answer is always the same; go and talk to your doctor. I?m an internet blogger, not your personal font of medical advice, unlike the person that you pay to take care of you.
I don?t even own a white coat!
But the more interesting question that I can answer is ?What should I do based on this new study??.
And the reason why it?s so interesting is because every article you?ve read has probably got the entire mess so completely wrong.
Every paper panicking about scary medications is referencing the same paper, recently published in the European Heart Journal. Danish researchers conducted a study to see if you were more likely to have a heart attack if you took a non-steroidal anti-inflammatory (NSAID ? ibuprofen is one of these).
There are some finicky epidemiological bits in here, so I?m italicizing anything that requires a lot of thinking because really, who wants that?
every article you?ve read has probably got the entire mess so completely wrong
They researchers conducted a modified case-control study, which is public health jargon for a study where you look back at a bunch of people who have had a medical problem (heart attacks, in this case) and compare them to healthy people to see if there are any differences. The Danish team identified every person who?d had a heart attack in Denmark between 2001 and 2010, and matched these with a group of control patients who were the same sex and age as the people who?d had a heart attack. They also checked to see whether people had filled in a script for a group of drugs 30 days before admission including ibuprofen (Nurofen), diclofenac (Voltaren), naproxen (Naprogesic), rofecoxib (Vioxx), and celecoxib (Celebrex).
They then did some cool statistical tests (look, time controlled regression is cool for me) which allowed them to compare the odds of someone having a heart attack if they?d had an NSAID in the last 30 days vs if they hadn?t.
The team found that the odds of having a heart attack were 30% higher for people who took ibuprofen and 50% higher for people who took diclofenac, as well as a 30% increase in risk from taking any NSAID at all.
Be afraid. Be very, very afraid.
Time To Freak Out?
So you?ve finally realized; the pills are out to get you. Tiny, scary bastards that they are, they?re going to creep inside you and make your heart stop beating.
There are a few very important caveats that makes this study pretty damn unlikely to mean anything at all to you, the individual. The first is the biggest irritation that I have with all science reporting ever; relative vs absolute risk.
Now, this study was retrospective (they looked back in time), so they actually used odds ratios instead of relative risk. This WHO guide gives a decent explanation as to why they had to do this.
Odds ratios are a different measurement than risk ratios. When you calculate a risk ratio, you are comparing the chance of an event happening in one group with the chance that it will happen in another. With odds ratios, you are comparing the ratio of one event happening in a population with the ratio of another event happening.
A simple example; imagine you have two groups with 100 hat-wearing people in each. The first group has your average baseball cap on, but the second group has taken to heart their grandparents? warnings about the wind stealing away their head-coverings and has literally glued their hats to their heads. On a windy day, 30 of group 1 hats are lost, but only 5 of group 2 hats fly away to find their own adventure.
The relative risk of losing your hat if it is not securely glued down is; 30/100 5/100 = 6. So you are 6 times more likely to lose your hat if it is not glued down.
The odds ratio of losing your hat, however, is; 30/70 5/95 = 8.14. So the odds of you losing your hat are actually quite a bit higher than your relative risk of losing it. Generally speaking, the lower the risk of either event happening the better the two measurements match up.
Unlike relative risk, odds ratios don?t allow you to calculate the absolute risk increase of taking an action. However, we do know that the risk of having a heart attack in any given year is about 0.3%. With that figure, the absolute increase in risk from taking NSAIDs would likely work out to be around 0.1%.
So when you look at a news article reporting this story, where they say ?30% increased risk? you can replace with ?I don?t understand odds, but it?s probably about 0.1% increased absolute risk?
With that figure, the increase in risk from taking NSAIDs would work out to be around 0.1%
Correlation vs Causation
I love stories like this, because it lets me trot out all of the tired old public health lines. In this case, correlation does not equal causation. Isn?t it wonderful?
This study looked at what happened after people filled scripts for a group of drugs. However, the researchers didn?t give people drugs and then see what happened; they just observed differences between groups of people who had taken them. This means that there were hundreds of factors that the study didn?t control for, despite their really cool time-sequenced design.
Which any journalist who had read the bloody thing would know, because the authors spell it out pretty clearly;
The main limitation of the study is inherent in the observational nature of the analyses. The treatment allocation is not randomized and the study reports only associations and therefore any conclusion on causality should be made with caution.
Step one for reporting on a study; READ THE DAMNED STUDY
If you?ve been reading carefully, you might?ve noticed something else; this study didn?t actually even look at whether ibuprofen was associated with heart attacks!
The researchers used a script for ibuprofen as a proxy measure, essentially saying that if you fill a prescription for the drug you are probably going to take it. However, they acknowledge that not having this information could ? reduce the association between exposure and outcome? ? public health speak for ?change our results completely?.
Journalists do a tough job; they have to produce a huge volume of work in a tiny amount of time, and do it whilst staying relevant and finding interesting things to write about.
Furthermore, this is not actually basic stuff. Without a course in epidemiology it?s pretty tough to understand the difference between odds ratios and relative risk, and proxy measures are a headache all of their own.
That being said, virtually every article used the same grab from a press release about the study. They all replicated the same speech by the lead author, which was part of a nuanced argument he?s been developing for years, but came across as ?don?t take ibuprofen or you?ll have a heart attack?. Every article said things that were just wrong.
This wouldn?t be a problem, except that scaremongering about drugs causes people to stop taking them. People who are worried about their hearts are going to look for another solution for their pain, and chances are they?ll find something more dangerous than ibuprofen. The nuanced discussion about the exact locations that NSAIDs should be sold has been lost in a blizzard of terrified screaming.
It?s a tough problem. For journalists, there?s a simple solution; read the study.
No one would write about a book without reading it.
You?d be ridiculed for critiquing a restaurant you?d never eaten at.
Can you even imagine a sports writer who never watched any games?
Every article said things that were just wrong.
It took me 20 seconds on Google Scholar to find the study. A quick glance at the abstract told me that all of the articles were wrong. None of this takes time, it just takes a bit of effort and the desire to care.
I care. Do you?
If you enjoyed, take a second to click on the heart below to let me know 🙂 You can also follow me here on medium, on twitter, or on facebook. If you didn?t enjoy, I?d love to know why down below, or @ me on twitter!