Holy carp! Autistic fish!

New Scientist is normally a shining beacon of high standards of scientific journalism, so I must confess I was rather surprised when I saw an article with the headline "Antidepressants in water may contribute to autism", based on a study that showed nothing of the kind. To be fair, New Scientist have since realised that their headline was over the top and have changed it to "Antidepressants in water trigger autism genes in fish", for which they deserve some credit. They also deserve credit for acknowledging that the headline was changed in a footnote to the article.

However, the article still fills me with despair. For a start, the idea that fish have autism genes seems a little bit silly. Although I should start by acknowledging that I'm not a great expert in the finer points of fish behaviour (although I do much enjoy watching the antics of the goldfish in my garden pond), to the best of my knowledge autism is not a recognised diagnosis in fish. So even the revised headline is not great. The genes in question are similar to genes that appear to play a role in autism in humans, but that's not the same as being "autism genes". Especially in fish.

But that's a minor point in comparison with what comes next. The article talks about antidepressant drugs in drinking water. Urine from patients taking antidepressants contains antidepressants, urine ends up in sewage, which is treated to purify the water and then ends up in drinking water. But the purification process is not 100% perfect, so some antidepressants get through the process. That all seems reasonable.

But the article then says that "at most, they are present at levels 100 times lower than the prescription doses". That statistic struck me as odd. OK, so water purification isn't perfect, but the idea that as much as 1% of a prescription dose of antidepressant could be present in tap water seemed suspiciously high to me.

So, where did this number come from? I don't know, but the article is based on a paper in PLoS One, which does give some figures for antidepressant concentrations in tap water. Table 1 of that paper states that the highest observed concentration of fluoxetine in drinking water is 0.014 μg/L. So someone drinking 2 L of tap water a day, in a system with the highest observed concentrations, would get a daily dose of fluoxetine of about 0.03 μg. The standard dose of fluoxetine is 20 mg per day. Now, you don't have to be a mathematical genius to know that 0.03 μg is rather less than 100 times lower than 20 mg. It's actually closer to a million times lower than 20 mg. So that "100 times lower" statistic seems to be way off.

But anyway, should we be worried that fish gene expression was affected by antidepressant drugs? Well, I'm not. The study used a combination of 3 drugs: fluoxetine, venlafaxine, and carbemazapine (the last of which is not actually an antidepressant, but that's not really important), and found that it did change patterns of gene expression in the fish, particularly in genes supposedly related to autism. Interestingly, the researchers tested other genes as well, not all of which had their expression significantly changed. It's not clear to me why the other genes were tested. It's not stated in the paper whether the specific effect on autism was a pre-specified primary hypothesis, or if many different genes were investigated in what is generally known as a "fishing expedition" (sorry about that!) If the latter, then the significant result for the autism related genes becomes much less impressive.

There may be other weaknesses in the study for all I know, but I'll leave it to those more expert in fish genetics than I to comment on them.

But here's the really, really, big problem I have with the study. The concentrations of the drugs used in the study were massively higher than the concentrations observed in drinking water. Recall that the highest concentration of fluoxetine in drinking water was 0.014 μg/L. The concentration used in the experiments was 10 μg/L, or in other words almost 1000 times higher. Venlafaxine was used at 50 μg/L, and that hasn't even been observed in drinking water according to the data in the paper. Carbemazepine was used at 100 μg/L, compared with a highest level in drinking water of 0.25 μg/L, so about 400 times higher than humans are likely to be exposed to.

I'm guessing that fish are exposed to greater amounts than humans would be (at least on a per-bodyweight basis) as well, for the same concentration in water, given that fish have a rather more intimate connection with water than humans do.

So what this study tells us is that a combination of fluoxetine, venlafaxine, and carbemazepine may affect gene expression in fish. But it does so at doses massively higher than any human would ever be exposed to from even the most pharmaceutically contaminated drinking water. So the evidence this study provides that drinking water can affect gene expression, let alone a tendency to develop autism, in humans is precisely zero.

New Scientist really should have known better than to write headlines like that, and certainly should have known better than to suggest that the concentrations of drugs used in the fish experiments were in any way comparable to what comes out of our taps.