It's the
world's most popular legal drug. Coffee!
Yes, it is a drug, and so we should remember the words of
Paracelsus, 'all drugs are poisons, what matters is the dose.' Based on the fact that coffee has been used widely for over
a millennium, we would expect its bad side-effects would be fairly minimal—so
long as we don't take too much.
If several of the studies are flawed,
then the result is suspect. A study result is only as good as the data it’s built on.
But in a surprising (and lucky) break for coffee drinkers, there
is a body of evidence that some of the side effects of coffee actually seem to be good. And interestingly, most of these
good side effects appear to have nothing to do with caffeine—there are a bunch
of other chemicals in coffee that seem to be involved.
Taken with a grain of salt (which may, or may not,
be good for your health), the main positive effects seem to be on life
expectancy, Type II diabetes, prostate, skin and oral cancers, and heart
conditions.
However, coffee sometimes has bad outcomes in the case of lung
cancer.
But there's a niggling problem—the
statistics. Most of the data on these good effects from coffee drinking have
popped up incidentally.
This is because most of the data comes from observational
studies that were set up to look at the actions and activities of huge numbers
of people—and over many years, sometimes decades. The studies weren't looking for anything specific—instead, they were
just observing, or fishing, to see what would pop up incidentally.
They were looking for 'correlations'. A 'correlation' is
a mutual link between two things. But a correlation does not mean that one
thing caused another. In some cases, a seemingly obvious correlation can be
wrong.
For example, the original Nurses' Health Study was set up in
1976 in the USA. It was an observational study and looked at various factors
that could be related to the health of over 120,000 female registered nurses in
the USA. By 1990, some 14 years later, a clear correlation had popped up. The
nurses who were going through combined hormone replacement (CHR) were having
fewer cases of coronary heart disease. It seemed quite clear; combined hormone
replacement would definitely protect women from
coronary heart disease.
However, this was not true.
You see, the women who took CHR were also wealthier, and also took more care of their health, such as eating well
and doing exercise. Once these factors were corrected for, it turned out that
combined hormone replacement had either no effect on coronary heart disease—or
maybe a small increase.
For another example, consider the excellent correlation between
'how much margarine each person in the USA consumes each day' and 'the divorce
rate in the state of Maine'.
Is the correlation excellent over a decade? Yes.
Does margarine cause divorce? No. Is there a 'causation'? No.
This is what statisticians mean with the short phrase, 'correlation
is not causation'.
That's the
first thing to recognise when looking at observational studies.
The second factor to keep in mind is that much of the data on
coffee (and coffee consumption) comes from grouping several observational
studies—the so-called 'meta-analysis'.
A meta-analysis tries to be a statistical version of the 'wisdom
of the commons'. This is where, in some cases, the
'many' can be smarter than the 'few'.
Ideally, a meta-analysis looks at studies that are all observing
the same factors. If all the individual studies are done to the same high
standard, the final meta-analysis should be able to arrive at better
results—e.g., provide better estimates of the result, and lower uncertainty.
But sometimes the individual studies are flawed. And it's not always possible to see the problems (and source of
bias) in a study, just by reading the paper related to that study. If several
of the studies are flawed, then the result is suspect. A study result is only
as good as the data it's built on.
The old adage still holds: 'GIGO', or 'garbage
in, garbage out'.So now
we've got that rider out of the way, it seems that with regard to life
expectancy, coffee drinkers live longer.
One meta-study reviewed 20 other studies that included over
970,000 people, while another looked at 17 studies with over a million
participants. They compared those who drank the most coffee, with those who
drank the least. The heavy coffee drinkers had a 14 per cent lower risk of
dying prematurely from any cause. Even having just one or two cups each
day drops the risk of premature death by eight per cent.
And drinking decaffeinated coffee gave the same advantage.
Drinking two to four cups of decaf per day still kept
their risk of premature death at 14 per cent lower.
So
bottoms up (if coffee is your tipple), and hang in there for the next
episode with even more 'life-enhancing' effects of coffee to come.