A friend of mine passed on a new study published this month in JAMA Psychiatry he thought I’d find interesting. It’s one of those really big studies they can only do in Scandinavian countries like Denmark where healthcare is truly socialised and researchers can have access to every single person’s medical file (before anyone freaks out, this data is deidentified. I think.). The result? N’s bigger than your mom.
For example, this study found that out of 3.56 million people, over 91,000 had hospital contact for a mood disorder (depression, bipolar, etc). Previous hospital contact for an autoimmune disease increased risk for later contact due to a mood disorder by 45%.
The justification for this study was that a) medical illness and mood disorders are associated (true), b) immune processes affect the brain and could influence mood and mood disorders (also plenty of evidence for this, especially from animal studies, but human ones as well), and c) longitudinal studies are needed to demonstrate this model and there aren’t enough of them (also true, but there have been a few).
Unfortunately this study doesn’t contribute to c). Why? We’re pretty sure that inflammation affects mood and can increase depressive symptoms in some people. But depressive symptoms are different to a diagnosed depressive disorder – especially one that would require someone to make hospital contact. People with clinical depression don’t just wake up one day and have it, despite feeling fine the day before. And this study had no way of measuring, and therefore controlling for, the level of depressive symptoms at the time that they came in to hospital with their autoimmune disease problems. So how can we say that the autoimmune disease occurred before depression, let alone suggest that the autoimmune disease caused the depression?
It’s not to say that these large scale national studies with a bajillion people don’t have their place. They are a good justification for conducting further research with smaller sample sizes that might be able to more closely examine the specific mechanisms in the inflammation-mood relationship. But the thing is, we’ve already had lots of these large scale studies looking at other medical illnesses that involve inflammatory processes like heart disease, arthritis, and diabetes, and a lot of them show that depression may actually precede medical illness. Furthermore, there are smaller longitudinal studies looking at actual levels of inflammation and depressive symptoms, although results on which precedes which are still conflicting.
So the next thing I’d like to say about longitudinal research is that it can really only tell us about temporal relationships – it can only say that x happened before y. And that’s a very good thing, don’t get me wrong. It helps us to identify people early on that might be a risk for developing certain disorders and diseases and get them treatment or prevention as early as possible. But it can’t necessarily pin down causality. For example, even if the above study had controlled for mood symptoms at baseline, and found that autoimmune diseases truly manifested well before any signs of depression, that’s all it can tell us – about when each disease manifested. It doesn’t rule out the possibility that autoimmune diseases and mood disorders are both attributable to some other heritable phenotype that causes each to emerge at different points over the lifetime.
So what’s the best study? A randomised control trial testing an intervention with a longitudinal follow-up. For example, you take two groups and reduce inflammation in one, and follow them up to see who develops depression.
Expensive? Yes. A nightmare to manage? Yes. Somewhat difficult to get through ethics? I would think so. But you don’t need 3.56 million people to take part in it.