Beware of uninformed warnings about risk

By: Dan Gardner

The Ottawa Citizen, 01 August 2007

Perhaps you are reading this column while sitting at the kitchen table, or
in a booth at the local diner, or bouncing along in bus. Wherever you may be, you are at risk. Right now — at this very instant — there is a chance that a large rock will blaze down from the sky and smash you to a thousand unrecognizable bits.

Most people are not terribly concerned by this mortal threat for the very
sensible reason that while the risk is real, it is tiny. Very, very tiny.

But what if we were to open the newspaper tomorrow and read that scientists had concluded the number of asteroids smacking the earth is 200 per cent higher than previously believed? Just like that, the chance of being crushed while watering the delphiniums would triple.

A 200-per-cent increase in the danger. Triple the risk. Would it be rational to respond to this news by carrying a cast-iron umbrella and replacing the delphiniums with a backyard bomb shelter? Of course not. Three times zero is zero. And a tripling of an itty-bitty risk produces a slightly larger itty-bitty risk. The delphiniums can stay.

At this point, the reader may think a flying rock has hit Gardner’s head.
This is so obvious. Why is he telling us this?

Because sometimes it is not so obvious. Take a look at any newspaper or news broadcast. Journalists, activists and politicians routinely warn the public about increases in relative risk without saying anything about the absolute risk. And people just as often get the impression that the danger is greater than it actually is.

Case in point: A paper published in the British medical journal The Lancet
last week found that users of marijuana have “an increase in risk of
psychosis of about 40 per cent” compared to those who had never used
marijuana. The most frequent users were at 50-per-cent to 200-per-cent
greater risk.

There are many things to note about this study. First, it’s not new. It’s a
meta-study of existing work that supports earlier conclusions. There are
doubts about the methodologies of some of those papers, a subject I’ve
written about before. Second, as the authors were careful to note, the study does not prove marijuana causes schizophrenia and other forms of psychosis. It shows they are associated. The researchers tried to control for various factors but they can’t rule out the possibility that something else may be at work.

But let’s leave all that aside. The issue at hand is risk, not marijuana. So
let’s assume that the Lancet paper really does show that marijuana causes psychosis. And let’s assume the increased risk really is as high as 200 per cent. What does that mean?

Nothing. Or rather, it means nothing by itself.

If the lifetime risk of being crushed by an asteroid were to triple, we
would ignore it because the original risk is so tiny. But a tripling of the
lifetime risk of getting cancer is serious because the existing risk is big.
So to make sense of the increased risk of psychosis, we have to know what the existing risk of psychosis is. Without that, these stats are scary but meaningless.

And yet, many journalists did not provide that key piece of information,
while others buried it as if it were a trivial detail. Even an editorial
that accompanied the paper in The Lancet only mentioned the relative risk.

An Agence France-Presse story that appeared in this newspaper did modestly better: “The report stresses that the risk of schizophrenia and other chronic psychotic disorders, even in people who use cannabis regularly, is statistically low, with a less than one-in-33 possibility in the course of a lifetime.”

That’s not a lot of information, but it’s enough to work out the basic
numbers: According to this paper, someone who never uses marijuana faces a lifetime risk of around one per cent; a light user’s risk is about 1.4 per cent; and a regular user’s risk is between 1.5 and three per cent.

These are significant numbers, but they’re not nearly as scary as the
numbers that got the headlines. And as far as I can tell, not one news story anywhere in the world reported them in full.

And that, I’m sorry to say, is all too typical.

Last year, a study found, in the words of one newspaper article, that women who use the Ortho Evra birth-control patch “were twice as likely to have blood clots in their legs or lungs than those who used oral contraceptives.” Twice the risk: In newspapers across North America, that was the only information readers got. It sounds big. No doubt some women found it quite frightening.

An Associated Press story provided the missing piece of the puzzle: “The
risk of clots in women using either the patch or pill is small. Even if it
doubled for those on the patch, perhaps just six women out of 10,000 would develop clots in any given year …” The AP story was carried widely across North America but many newspapers that ran it, including The New York Times, actually cut that crucial sentence.

We face lots of risks in life, including the risk of getting bad information
about risk. Reader beware.

Republish