Why cortisol sucks as a measurement of stress

Standing in the ward of a veterinary hospital, you see a dog jumping up and down in his run, barking. Is he distressed at being here? Or is he just barking to get attention? Obviously, you decide to perform a research study on dogs in the hospital, to measure their stress levels. How are you going to go about getting some sort of numeric measurement of stress, so that you can perform statistics on your data and publish it in a journal?

People have approached similar problems in a lot of ways. One of the most common answers is to measure the dog’s cortisol levels. (Or corticosterone, if it’s actually a rat, mouse, rabbit, or bird.) This is the approach I’m using; cortisol is in saliva, which is why I spent so much time over the last year trying to get dogs to drool more.

What is cortisol, actually? There was an excellent post on mindhacks.com recently about cortisol and how science journalists sometimes misrepresent it. I’m going to use dogs as my examples, but what I have to say is just as relevant to studies on humans, if that is your cup of tea. It might help you to understand some of the news stories floating around about various things which “raise cortisol levels.”

Cortisol is a hormone made by your adrenal glands. Your adrenals sit next to your kidneys, but they produce cortisol in response to hormones released from your brain in stressful situations. So we like to measure cortisol levels because they tend to increase when the brain is sending out “I’m stressed” messages.

OK, but what is stress? What I care about, and what many people who measure cortisol care about, is psychological distress — being yelled at, being scared you’re going to be eaten by a predator, being left in a loud veterinary hospital with no familiar faces around you. Stress is a lot of other things as well, however. It is hunger, illness, feeling too cold, having exercised recently. In fact, cortisol has a normal rise and fall over the course of the day to help your body know that it is time to be awake or to go to sleep. Your adrenals also produce it to help you deal with anything which requires some extra energy. You may need that extra energy for a good reason, such as competing in an athletic event. “Good” stressors like that are known as eustressors. So if you’re going to use cortisol to measure stress, you are going to be measuring both eustress and normal daily stress like hunger, in addition to whatever source of distress you may be interested in.

Knowing this, you’ll try to design your experiment to work around the problem. You’re interested in whether dogs find their time in a veterinary hospital to be distressing. So you will try to remove eustressors from the equation — you will make sure that none of our dogs have exciting things like getting fed or taken for walks happen while we’re studying them. You will also make sure that all of your study dogs are healthy, since illness can raise cortisol levels. And you will measure the dogs’ cortisol levels at exactly the same time of day, because of cortisol’s diurnal cycle. (There is some very interesting debate about whether dogs, unlike all other mammals which have been studied, actually don’t have a daily cycle of cortisol. One theory is that they don’t because they sleep most of the day.) Now you believe you're just measuring distress.

Cortisol is still an awfully bad way to measure distress! Males and females react to stressors in different ways. (This has mainly been reported in humans, but it’s been said that studies of stress in rats which are limited to males miss an important segment of the population.) Age has something to do with cortisol levels as well, though mainly just in the very young and very old. In dogs, it is an open question of whether breed matters, but I’m guessing it does, since personality affects cortisol responses to stress in humans.

So you control for that, too. You get a bunch of dogs of the exact same age, gender, and breed. They are all laboratory animals, so you can be reasonably sure their histories are the same, and you aren’t going to find out at the end that half of them have spent more time in a veterinary hospital than others. (This wasn’t the direction I chose, but some studies do give it a go, using laboratory beagles of similar ages and only one gender. There are obvious ethical implications here, but that’s a post for another time.) You put these dogs in a veterinary hospital and measure their cortisol levels. Now are you measuring their reaction to the hospital setting?

Maybe. The next problem is that all these animals have their baseline “unstressed” cortisol level set at a different point. We don't understand all the genetics having to do with how this system works, but we are learning. We do know that a cortisol level that indicates stress (good or bad) in one individual might indicate total relaxation in another. Many studies deal with this problem by looking only at changes in cortisol levels. They measure cortisol before and after the stressor, and look at the difference, rather than at absolute levels. So let’s assume you can do this in your hospitalized dogs. You keep them in one environment for a few weeks or months, until they have time to settle in and relax, and you keep track of their average cortisol levels there. Then you put them in the hospital and look for a difference. Now are you measuring their reaction to the hospital?

You probably are, but what exactly are they reacting to? Something which is a stressor for one individual isn’t necessarily a big deal for another. For example, the Trier Social Stress Test (TSST) is a test specifically designed to raise cortisol levels in humans. It’s used to study things like how gender affects responses to stress. You stress the person out by making them do some public speaking and public arithmetic. But only 70% of people who take the TSST actually have increased cortisol levels compared to just before they took the test. Doing arithmetic in front of a hostile audience just isn't alarming for some people.

In the case of your hospitalized dogs, some don’t like the noise, and some don’t like being in a cage, and some don’t like having other dogs around, and some don't like all of the above. But some think it’s awesome to be in such an exciting environment with so much going on. The hospital isn’t just one big stressor, it is a lot of different little ones.

If cortisol is such a bad way to measure distress, why do we use it? Unfortunately, it is still the best understood method we have. There are lots of other methods, but they all have their own problems. It’s a good idea to use at least two methods together, actually.

So what do you do? Give up? My approach has been to cross my fingers (maybe close my eyes) and just proceed. I think a lot of research involves just circling around a problem, picking away at it until it starts to give in. Studies of stress may not be able to give precise answers to questions about stressors. But if enough of them are done, our picture of how the stress response works will continue to get clearer and clearer. It’s really hard to know what is going on in the mind of a member of a different species; it can even be hard to know what’s going on inside the mind of a member of your own species. We just have to keep trying.

[ETA: See the follow-up post, “Why do other measurements of stress suck worse than cortisol?”]