Lies, damned lies, and statistics

by David Laing Dawson

Mark Twain said that long before we had computers and a few dozen algorithms we could apply to random numbers to find ‘meaningful’ patterns.

Data mining and scientific studies that find nothing or negative results seldom get published. So it behooves all academics to find something. To find at least an association that can be inflated by the manner the data is reported. Then it will get published, and the press might even pick it up if it is startling enough.

I am writing this because an article on the front page of our local paper tells us that people who take antidepressants are at risk of premature death. This is based on a local academic’s data mining and meta-analysis. The figure quoted is 33% higher risk of premature death and 14% more likelihood of death from cardiovascular disease. They also have to explain away the fact that if you have previous cardiovascular disease the use of antidepressants does not increase risk.

First, these are associations, not cause and effect. Secondly the variables are numerous. And the first variable that comes to mind is that the people who take antidepressants probably suffered from anxiety and depression, undoubtedly felt unwell, and did ask their physicians for help. The people who never took antidepressants did not. The only way these figures can be clarified would be to take 10,000 people who attend doctors complaining of anxiety, OCD, and depression and give antidepressants to 5000, and nothing to the other 5000, (randomly selected) and follow over 20 or 30 or 50 years.

Then we have the startling 33%. Well, if 3 people out of 1000 die in one group and 4 in the second group, that is a 33% increase, looking at it one way, but really a 0.1% difference looking at it in a real life way. These kinds of statistics are often misused in the press. When the actual risk (sorry, not actually RISK, just different finding) of contracting something increases from 1 in a million in one study to 2 in a million in another study that can be reported as a 100% increase.

I am sure antidepressants are both underused and overused. Underused in the rush of clinical practice when severe depression is not recognized or not reported, underused when the person is already self-medicating with marijuana, alcohol or opioids, underused when the dosages used are too small for severe depression – and overused as the go-to-drug for angst and unhappiness.

I am also sure any drug should be avoided if it can be. That goes for anti-hypertensives, statins, antibiotics, and aspirin.

So I did a little data mining of my own. It turns out that the people of Australia, Iceland, and Sweden rank in longevity 2, 3, and 4 in the world. Canada and New Zealand follow closely. Japan holds the number one spot but antidepressant data (for interesting cultural reasons) can’t be found so I have excluded Japan. On average the people in spots numbers 2,3, and 4 live between 82.4 and 82.8 years. Let’s average that to 82.6 years of life expectancy. Iceland, Australia, and Sweden also rank as the highest antidepressant users, ranking one, two, and four. (Denmark is number three)

Among the lowest antidepressant users (where data for life expectancy and antidepressant use can be accurately determined) are Estonia, Turkey and Slovakia.The life expectancy for the people of those countries averages 76. So by simple association we find that the longest lived people in the world consume the greatest number of antidepressant pills per person.

Applying my own meta analysis to this data I can arrive at the conclusion that high average consumption of antidepressants prolongs (oops, is associated with an increase in) life expectancy by 6.6 years, or almost 9%. The headline this could generate would be: Prozac increases life expectancy by 9%

But, academics have an ethical duty to explain the limitations of associations found in population studies and meta-analysis, and the true meaning of various statistical analyses in real life terms.

Reporters should have an ethical duty to avoid golly gee whiz headlines in health matters. (probably in a few other matters as well)

And medical historians should point out the dramatic change in the number of home and hospital beds utilized by moribund patients suffering from severe depression pre 1960 and today.

A curious side note: On the same Google page for Health news there is a report of a British teen dying from “eating her own hair”. They go on to discuss Rapunzel syndrome, and trichophagia. But such a compulsive behaviour is just that. A compulsion. A serious symptom of OCD. And easily treated today with one of those antidepressants maligned in the other article, along with some counseling of course.



2 thoughts on “Lies, damned lies, and statistics

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 )

Twitter picture

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

Facebook photo

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

Google+ photo

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

Connecting to %s