Debunking The Latest Gene Study

The headlines scream it excitedly. At last, we have found a genetic link between mental health conditions! A new study finds a shared genetic pattern between five different “disorders!”

Could this actually be true?

The researchers are just as excited—after all, their analysis was statistically significant, which is the only metric journal editors and peer reviewers care about.

Here’s what they write:

“In the first large scale meta-analysis of CNVs (copy number variations) across multiple neurodevelopmental/psychiatric diseases, we uncovered novel significant associations of structural variants in the locus of DOCK8/KANK1 shared by five diseases, suggesting common etiology of these clinically distinct neurodevelopmental conditions.” (See definition of terms.)

That’s right. In the first page of their manuscript, the researchers suggest that their finding implies a common genetic cause behind depression, schizophrenia, bipolar disorder, ADHD, and even autism. This is big news! If true, it validates the biomedical view of mental “illness” and suggests that future medical treatments could “cure” these conditions.

However, that grand conclusion is not supported by the data. In fact, an incredibly small proportion of people diagnosed with a psychiatric disorder actually have this “CNV at DOCK8” genetic pattern—99.5% of people with a diagnosis didn’t have it. And some of the “healthy” people did.

The Data Tells A Different Story

If you can read through the obfuscating genetics jargon for 6 pages, you’ll come to this table:

Even this table is pretty confusing, so I’m going to break it down. There are five different studies being analyzed here:

• “Janssen SCZ and BD” is a single study including people diagnosed with either schizophrenia or bipolar disorder
• “CHOP SCZ” is a study including only people diagnosed with schizophrenia
• “CHOP ASD” includes only people with an autism spectrum diagnosis
• “CHOP ADHD” includes only people with ADHD diagnosis
• Finally, there’s a study including only people with a depression diagnosis

The column “Number of Case CNVs” shows how many people who had the mental health diagnosis also had the genetic pattern being studied. The column “Number of Control CNVs” shows how many people without a diagnosis had the genetic pattern.

Of course, you have to look elsewhere in the article to find out how many total people were included in the study. There were 7,849 total people with a mental health diagnosis, and 10,799 “healthy” people in the control group.

If you add up the number of people under the “Number of Case CNVs,” you find that 32 people had both a mental health diagnosis and had the “CNV at DOCK8” genetic pattern. Thirty-two people out of 7,849. That means that out of 7,849 people with a diagnosis, more than 99.5% of them did not have the genetic pattern.

So how does this get reported as a “genetic link” between these mental health diagnoses? This is a prime example of how statistics can create a result that is utterly meaningless. Those 32 people out of 7,849 are compared to the “healthy control” group—of whom 9 (out of 10,799) had the genetic pattern. When those ratios are compared, the result is “statistically significant,” meaning that it is unlikely that this tiny difference happened by chance.

But that difference is meaningless if the number of people who share this pattern is so infinitesimally small. What clinical use could that tiny difference have? You could get tested for the “CNV at DOCK8” pattern. But if you have it, are you one of the 32 people in the “neurodevelopmental disorder” group, or one of the 9 people in the “healthy control” group?

And, if you don’t have the CNV at DOCK8 pattern, what does that signify?  Are you part of the 99.5% of people who don’t have that genetic makeup but have a “neurodevelopmental disorder?” Or, does this mean you don’t have a mental disorder? A test for this genetic variation would be totally useless.

Misleading the Public

Although it’s easy to see that the study results are meaningless, the researchers nevertheless wrote that their findings suggest a “common etiology of these clinically distinct neurodevelopmental conditions.”

In that sentence, you can see the misleading framing of their conclusions. Implicit in their statement is the idea that these mental health diagnoses are “neurodevelopmental conditions”—a statement which is not supported by their evidence. Their study is just another in a long line of studies that actually fail to find any specific abnormality in people diagnosed with a mental disorder.

Ties to the Pharmaceutical Industry

So what purpose does this study serve? It’s important to note that the study was conducted by the Janssen-CHOP Neuropsychiatric Genomics Working Group, meaning that the authors received their funding and worked under the command of a pharmaceutical company with a vested interest in biological conceptions of mental disorders. Several of the authors are current and/or former employees of Janssen. Janssen is the maker of Concerta (branded as a “treatment” for ADHD), Haldol (an old-school neuroleptic tranquilizer branded as an “antipsychotic”), and Risperdal (a drug branded as an “atypical antipsychotic”). If Janssen can convince the public that mental disorders are biological illnesses, this will help it continue to sell its drugs.

Final Thoughts

There are plenty of other concerns with this study. The researchers conducted tests on thousands of DNA patterns, for instance, which exponentially increases the possibility of errors. The more times you test, the more likely you are to find a result by chance. If you run a statistical test comparing people on thousands of variables, the likelihood is actually very high that you will find some differences just by chance, even if the two groups of people had no discernable differences.

Indeed, testing for CNVs is itself an unreliable process. It would be incredibly rare for two people to have the same duplication or deletion in the same area. Even those ranked as “having” the CNV at DOCK8 probably had very different protein lengths that were duplicated.

Also, the researchers combined five very different studies into this meta-analysis. In fact, because of differences in the way data was collected, and the inherent unreliability of the tests they used, they actually discarded 20% of their participants. One wonders what the results would have been had they included all the participants.

But ultimately, the biggest concern is that even if their finding is “true,” it’s completely meaningless. It doesn’t tell us anything about these supposed “disorders” or ways of “identifying” or “treating” them. It is a telling example of using genetics jargon and statistical significance to mislead the public.