Image via Wikipedia
A link on the Drudgereport proclaims “Study shows aspirin causes blindness” and that headline alone might kill hundreds of people who quit using aspirin because they didn’t read the entire article:
Researchers who tested more than 4,000 elderly people across Europe found that those who took the drug every day were twice as likely to be diagnosed with late stage age related macular degeneration as those who did not.
While the study provided no evidence of a causal link between aspirin and the condition, experts are now examining whether a regular dose somehow exacerbates the disease.
…
During a recent major study, researchers based in Holland found that of 839 people who took aspirin each day around 4 per cent had an advanced form of the disease called wet macular degeneration, which leads to the most profound blindess.
In comparison, just two per cent who took aspirin less frequently had the same type of macular degeneration.
Does anyone see the big problem with this study?
In fact, there are two big problems.
The first is the sample size. Normally, you’d think 839 people would be enough to be confident about the results. But look at the size of the effect. Four percent? That means we’re not talking about 839 people, we’re talking about 33.56 people.
Yes, you’d expect to find just 16.78 people with this type of macular degeneration in the same sample size. But, are you really willing to give up potentially life-saving aspirin on the basis of 16 people in one study? You shouldn’t.
There’s another huge problem, and that is the types of people who use aspirin every day of their lives are likely very different from the types of people who live a long time without using aspirin daily. Those on the aspirin, along with cardiovascular disease, are also more likely to have diabetes, exercise less, eat less healthy or weigh more than the non-aspirin group. And that’s before genetics gets thrown into the works.
This is an apples-oranges sampling problem. It can’t be taken seriously and should not be posted around the internet like it’s a meaningful study. I doubt very much you could accurately control for all the variables that separate habitual aspirin users from those who don’t even with a much larger sample size.
And how, in this study, with just 17 people above and beyond the expected level of macular degeneration, could you control for even one factor, that of heart disease?
It’s possible there is some underlying mechanism and aspirin takers should be aware of the problem. But they should also be aware that the real risk of suffering this consequence is at most 1 in 50, and it’s not lethal. Not taking aspirin can be very lethal. Be smart kids. Do the math.
Related articles
- Regular aspirin users at higher risk of sight problems, research suggests (telegraph.co.uk)
- Daily Aspirin Tied to Risk of Vision Loss (nlm.nih.gov)
- Is macular degeneration terminal (wiki.answers.com)
Filed under: Health, Numbers and Studies, Statistics | Tagged: Aspirin, DrudgeReport, Health, Macular degeneration, Old age, Vision loss |
Marty Andrade 
What a wonderful take on this news story! You are right on the mark with this; I hope people taking aspirin for heart health will talk to their doctors before halting therapy. Aspirin thins the blood; that’s why people with heart problems take it. If you have wet AMD and the blood vessels in the eye break and leak, guess what? They will leak more because of the aspirin. That doesn’t mean the aspirin caused the leak. I wish this was not getting so much play.
That’s basically my argument, Aspirin use might exasperate an underlying weakness that might not manifest itself without help; you have to be open to all things and I’m no expert. You have to properly weigh the risks against the evidence, which is weak, imho.
Umm, 4000 people is actually a very large sample. The difference between 2 and 4% isn’t enough to base your behavior on, but the two populations are very probably different.
As for the concern as to the two groups being fundamentally different for other reasons, we don’t know if that is the case. If the treatment groups were not randomized, I would reject the study outright; selection bias would be an obvious problem, as you pointed out. However, the news articles don’t say whether or not aspirin use was randomized among the people studied. You’re probably right, since the groups aren’t of a similar size, but I wouldn’t accept or reject the study outright without knowing for sure.
To clarify, while you say that the size of the effect is the important thing, the size of the sample is just as important. Its sort of like if you encounter 2 albinos in a crowd of 20 people. Freak occurence, maybe. 20 albinos in a group of 200 is absolutely bizarre, way more bizarre than 10 albinos in a group of 200.
4000 people is definitely enough to be confident with the effect size; the p-value is somewhere around .001.
You’re always vulnerable to data-mining errors as there are always some random patterns that manifest themselves.
This was not a randomized trial, it was a survey of people over the age of 65.
And 65 year old people who require a daily aspirin are different from those who don’t. Which was my point.
There was an article on this survey that I didn’t link to that included more details about the study.
The big problem with assigning a statistical measure of confidence in this survey’s results is the data is meaningless if the two populations are fundamentally different. Assuming aspirin takers are the same as non-aspirin takers is a leap of faith I simply can’t take.
The “4000” number is not as important the 800 number. The fact it was hard to find a lot of daily aspirin takers (They represent about 20% of those surveyed) tells me this population probably has a death problem. If you had 2000 aspirin takers randomly selected from a total population of 4000 tracked over 20 years, sure, p-value is going to be small and I’d accept the results of the study. If you go around to a bunch of retirement homes, survey aspirin habits and throw the values into a t-test machine without controlling for other factors, not a good use of statistics.
I’m now repeating myself, so I’ll just say it wouldn’t surprise me if there was some real effect. The effect is small and there’s a good possibility the only people affected are already prone to macular degeneration and the aspirin just exasperated the problem or expedited the process. Nevertheless, there’s no reason this survey should result in people changing their behaviour.