Recently the controversy over long-term outcomes of those who use psychotropic medication has flared up again. Some people argue depression/bipolar/mental illness patients do the same, or better, when they don’t take psychiatric medications long-term. However, the statistics they use to assert this claim are often faulty.

A study from Calgary, Alberta, Canada (yes, we do research up here too, select Science Direct to see the study) has attempted to fix some of the bias seen in other long-term depression treatment outcome statistics. I’ll cut to the chase for you:

Over the course of eight years people with depression who took antidepressants had better outcomes.

Depressed People Do Not Do Better by Taking Antidepressants

The reason people say those with depression who take antidepressants do not do better, or do worse, goes something like this:

We looked at 100 people with depression and over five years those who didn’t take antidepressants were less depressed.

Put that in the middle of some persuasive text and the villagers gather with pitchforks at the doors of psychiatrists everywhere.

Why are they Wrong? What’s the Problem with This Data?

Treatment Outcomes in Depression

Um, OK. Anyone see the problem here? Anyone?

It should be obvious. Those who don’t take antidepressants (or who go off antidepressants) are typically less sick than those who do take antidepressants. If you’re sicker, you’re more likely to get treatment. It’s not rocket science.

For some reason everyone wants to gloss over that bit.

This is what you call sampling bias.

Measuring Depression Treatment Outcomes Long-Term

Trouble is, you can’t randomly assign people people with a mental illness to treatment/no treatment over the course of a year or more. Sure, the depression treatment outcome data would be better, but the people might not fair so well.

We have to think smarter. If you can’t assign people at random, can you account for variables like duration of mental illness and severity of depression symptoms?

The association between antidepressant use and depression eight years later: A national cohort study by Colman et al. tried to take these variables into account.

Confounding Depression Variables

As I’ve mentioned in the past, depression is not one thing, depression is a spectrum disorder. You may have a variety of symptoms with a variety of severities and still be “depressed” (bipolar is the same). Somehow, we have to quantify that.

We do know some variables that correlate to depression treatment outcomes. Colman et al. measured:

  • Severity of depression
  • History of depression
  • Duration of past depression
  • Suicidal behavior
  • Physical health
  • Demographics (gender, education, etc.)

Depression variables were measured based on scientific scales derived from various methods including personal interviews. You can read all about it here.

Population of People with Depression

For this study, patient population data used was collected by a Canadian agency in the National Population Health Survey in 1998/99 through 2006/2007 every two years. 486 people (of 17,276) were identified as having depression in 98/99 and were followed, 66% of which provided complete data in 06/07 (321 people). Population was all ages, treatments and representative of general population.

Colman et al. analyzed the above variables for all 486 so they could ascertain who was most likely to drop out of the study (or give incomplete information) based on their variables; however, obviously only 321 were used to determine treatment outcome measures.

Depression Treatment Outcomes and Adjusting for Variables

As I said above, if you do not adjust for variables that lead to treatment likelihood, you get statistics that aren’t valid (they’re too biased). Sicker people get treatment.

So, the numbers when you do not account for bias look like this in the depressed population over eight years:

  • Those who took antidepressants were as likely to be depressed as those who didn’t take antidepressants.

But when you adjust for the mentioned variables, that changes to:

  • Those who took antidepressants in 98/99 were less likely to be depressed eight years later (OR = 0.36, 95% CI: 0.15–0.88)

(There are other findings and statistics (not overly significant and in scientist-eze) here.)

Long-Term Depression Outcomes Better with Antidepressants

All that is a fancy way of proving the people who took antidepressants were less likely to be depressed eight years later.

In part two, discussion about the strength and limitations of this study and other interesting findings and citations.