What is Epidemiology?
How do regulators make decisions on what were not allowed to do, like smoking in public places, or driving gas-guzzling, air-polluting cars? And how do health organizations decide what healthy eating recommendations to make? Like having five servings of fruit and vegetables a day?
These are examples of public health decisions based on epidemiology – the science of understanding how what we’re exposed to, or what we do, may affect the overall health of society. But to make sense of epidemiology, we need to dig a little deeper.
Let’s start by going back to basics and ask: How do we work out whether something we do, or are exposed to, is harmful; and how harmful is it? The only reliable way is to use science to develop evidence that shows something is bad for health. But this isn’t as easy as it sounds.
One way is to do research on animals, called “in vivo” studies, or on cells in test tubes and petri dishes. These are called “in vitro” studies. But such research can only tell us so much. For example, many substances can affect us differently than they affect animals. So while chocolate is harmful to dogs, and aspirin is toxic to cats, both are safe for humans, when used appropriately.
As for in vitro studies, cells that are isolated in the lab behave differently than cells inside the body. So we have to be very careful applying the results of such studies directly to human health.
Of course, the most straightforward approach is to do experiments on real people. But with a few exceptions, this is highly unethical! You can’t go around exposing people to potentially harmful substances, just to see what happens. This leaves us with observing the real world – measuring what a lot of people are exposed to – and by a lot we mean tens and hundreds of thousands of people – and trying to work out what the connections are between these exposures and their health. Studies like this form a big part of epidemiology. And it is a powerful way to understand how exposures and behaviors potentially affect large groups of people.
But it’s also sometimes difficult to make sense of. To start with, just because someone was exposed to something and they got sick doesn’t mean the two events are related. The exposure may not have caused the sickness. For example, lots of people eat ice cream when it’s hot. And lots of people get sunburned when it’s hot. But ice cream clearly does not cause sunburn. These events are instead correlated, meaning that they often happen together. But they’re not causative, meaning that eating ice cream doesn’t directly cause sunburn.
And while this example is pretty obvious, making sense of epidemiology data is often really hard! And care needs to be taken that we don’t jump to the wrong conclusions.
When examining relationships between exposures and health outcomes, there are a number of reasons why we might see an association. These include an actual cause, pure chance, bias, and what epidemiologists call “confounding” – other things interfering with what we observe. Bias can come about because of errors in how a study is designed. For example, if we observe only people who eat a lot of ice cream, and live in really hot sunny areas, and don’t include anyone else who doesn’t eat so much, or lives elsewhere, the association between ice-cream eating and getting sunburned will seem very strong. In other words, the result will be misleading; there will be bias toward a particular – and in this case, wrong – conclusion.
Confounding on the other hand is where other factors confuse our interpretation of exposure and outcome. For example, imagine a study that suggests people who drink more coffee are more likely to develop heart disease. It would be tempting to conclude that coffee causes heart disease. But people who drink coffee also tend to smoke. And in this case smoking is a confounder. Since it is associated with both drinking coffee and heart disease, it can make it seem that coffee causes the condition, if we don’t take smoking into account.
In real life, there are many confounders – some more obvious than others. Because of this, epidemiological investigations take quite a bit of detective work to figure out what exposures really lead to the health outcomes, and which ones only appear to. To tease out what is relevant, and what is not, epidemiologists use statistics.
One standard practice in analyzing data is to look at the probability, or “p” value, to determine if the findings are likely to be true, or are simply due to chance. The lower the p-value, the more likely it is that the results of the study represent reality, and did not just happen because of chance or random variation.
Usually epidemiologists consider a p-value of 0.05 or lower as indicating that the study results are statistically significant, which is just a fancy way of saying that there’s less than 5% chance of these results being due to random variations. However, the p-value only helps you get a sense of whether study outcomes are due to chance or not. It does not help us examine how strong the association is, or how important the health implications are. And if statistics aren’t done well, even low p-values can be misleading.
Consider cancer risk. Epidemiological work shows that both smoking and processed and red meat, have a statistically significant association with increased cancer risk – meaning they both have p-values below 0.05. But, while smoking increases your chances of getting cancer by around 20x, or 2,000%, eating red and processed meats increases it by only 20%, or 0.2x. This is extremely low when you consider all the other things you’re exposed to that potentially impact your health – especially if the chances of getting cancer aren’t high to start with.
And this is why looking at the p-value on its own is not enough. And researchers also need to consider how large of an effect an exposure has – called “effect size” – not just whether there’s likely to be an effect or not.
Lastly, when making sense of epidemiological studies, it’s important to remember that this science deals with the collective health of thousands and millions of people – and not individuals. Because we’re all different, and live under different conditions, it’s very hard to apply broad conclusions from such studies to single people. But they are good at indicating what whole communities should do to stay healthy. And just to complicate things further, epidemiology studies may not include people like you; meaning that they may be less important to you than to others.
The bottom line is that it takes a lot of work to conduct epidemiology studies well, and it takes a lot of work to interpret them correctly. Despite these difficulties, epidemiology is crucial to making public health decisions, and improving the well-being of people – so that, on balance, we all live healthier lives. These decisions do not only mean better healthy lifestyle recommendations and programs; they’re also important for reducing health care bills, and increasing productivity over tens of millions of people, which is why, even though it’s complicated, epidemiology is so important.
