Hardly a week goes by without us being bombarded with alarmist headlines of doom relating to our diet or lifestyle. ‘Alcohol doubles your risk of y’, or, ‘Red meat causes x,’. Moreover, many don’t make any sense. Like the recent one that warned of the dangers of coconut oil – as if a modern disease could be related to an old-fashioned food.
If you read enough health-related headlines, you would be forgiven for not wanting to eat anything at all, or worse, wanting to eat everything with reckless abandon, in protest to the apparent inevitability of disease or death.
Nutrition Headlines
In our quest to navigate these headlines and to avoid being hoodwinked, we must first understand how these research studies are designed. Some studies are designed to establish cause and effect (to prove that x causes y), and others to generate a hypothesis (a proposed explanation of a particular finding).
Nutritional epidemiology is a branch of science that observes how diet affects health and disease. These studies tend to employ food-frequency questionnaires that are given to participants. A typical question could be…’’ how many beef burgers did you eat, on average, over the last 12 months’’. The researchers collect this data and then try to work out what types of foods cause what types of diseases.
Correlation does not imply causation.
There are innumerable flaws with this methodology; the most obvious being that participant recall is notoriously unreliable. Most of us can scarcely remember what we had for lunch last Wednesday, let alone what we ate in January. The outcome of any study is largely dependent on the quality of data collected. Nonsense in = nonsense out.Entire papers have been written about how inadequate such questionnaires are as a way of capturing what people actually eat.
The typical modern diet contains hundreds of ingredients. It is near enough impossible to construct a questionnaire capable of capturing this level of complexity.
Observation Studies – The Challenges
These observational studies – where the investigator observes individuals without manipulation or intervention- are not designed to establish cause and effect. At best, they can be used to generate hypotheses (good guesses) about which foods may be implicated in which diseases. They are limited to the realm of correlation – not causation. In other words, they can inform us that eating X is linked to, or associated with, developing Y. But they cannot assume, or prove, that eating X causes Y.
In addition, any associations found must be strong enough to warrant further investigation in a controlled experiment. Frustratingly, when these ‘statistically sound’ hypotheses are tested in clinical trials, they are wrong at least 80% of the time. You’d literally have better odds in flipping a coin.
In science, if observations do not fit the hypothesis, it is the hypothesis that needs to change, not the facts.
Back to the beef burger. Did the researchers adjust their data in acknowledgment of the influence that the soda, chips, burger bun and tomato sauce would have had on their results? Hardly anyone eats the meat in isolation. In fairness, researchers do use statistics to adjust for some of these biases, but this seems like a near-impossible task to get right; because you can’t adjust for an entire lifestyle or individual.
Neither can we lock people up in a research centre and study them like we do mice. Researching free-living adults is very challenging. Scientists cannot expose humans to extremely rigid or unpalatable diets for long periods of time. They cannot expect people to eat nothing but red meat for years on end.
Correlation does not imply Causation
Even if red meat is found to be associated with an increased risk of colon cancer, it can be very difficult to determine whether it is the red meat or, in fact, the chips that are affecting disease risk. Or perhaps it is less about the diet and more about the lack of exercise. How one disentangles all these competing exposures to a particular outcome is proving to be something of a Gordian Knot.
In spite of all these limitations, observational studies form the lion’s share of nutrition studies that dominate the literature, and consequently, the headlines. Nutrition epidemiology has major setbacks, but there is no doubt that it has made major contributions to the body of science. The question is whether these types of studies should be used to form public policy.
The next time you read another sensationalist headline, take it with a pinch of salt, never stop asking why and consider the challenges our well-meaning scientists have in trying to study humans.
This article was published in print and online by The Star, Kenya on 23rd September 2019.
