Just fancy that! Odds ratios and autism

Drug News: WDDTY misrepresents news about drugs. I'm shocked. Shocked!
Drug News: WDDTY misrepresents news about drugs. I’m shocked. Shocked!

In “Medical McCarthyism“, Lynne McTaggart seeks to defend her inaccurate use of statistics and her undermining of the proven benefits of chemotherapy thus:

It all has to do with absolute vs relative risk. Let’s say you have osteoporosis, the brittle-bone disease. Your condition may be at a stage where your risk of suffering a fracture is 4 per cent, but the drug can reduce that risk to 2 per cent.

There are two ways of expressing the same thing: as a relative risk, the drug has a 50 per cent rate of effectiveness – it’s reduced your risk from 4 to 2 – and that sounds attractive, but in absolute terms its effectiveness is just 2 per cent.

For years researchers have been presenting the effectiveness of chemotherapy in terms of relative risk, and this has influenced the way the media has reported on cancer too.

What the original researchers concluded:

Although the number of children exposed prenatally to selective serotonin reuptake inhibitors in this population was low, results suggest that exposure, especially during the first trimester, may modestly increase the risk of ASD. The potential risk associated with exposure must be balanced with the risk to the mother or fetus of untreated mental health disorders. Further studies are needed to replicate and extend these findings.

Statistics nerds call this the “odds ratio”.

And of course WDDTY would never do such a thing.

Or would they?

One environmental cause could be antidepressants that were taken by the mother, especially during the first trimester. Researchers from Kaiser Permanente Northern California discovered that the drug more than doubles the risk of autism, especially if the mother was taking the drugs before conception or early in the pregnancy.

Doubles the risk, eh? What does the cited source say?

A history of maternal (adjusted odds ratio 1.49, 95% confidence interval 1.08 to 2.08) […] was associated with an increased risk of autism spectrum disorders in offspring. […] Assuming an unconfounded, causal association, antidepressant use during pregnancy explained 0.6% of the cases of autism spectrum disorder.

Yes, that’s right. Your risk of autism increases by 0.6% not 50% if you avoid the odds ratio trick that WDDTY don’t want you to use when it makes chemo look good. That’s from the Stockholm study cited in WDDTY, by the way, not the source they didn’t cite and from which the odds ratio of 2 is drawn.

The odds ratio is between 1.2 and 4.3 in the original Kaiser study; a wide odds ratio indicates either a weak correlation or a small study. In this case it’s the latter: this study is based on 20 cases matched against 50  controls. And unlike the source WDDTY used, this has a specific correlation with SSRIs. For some unaccountable reason WDDTY forgot to include this rather important item: the conclusion, usually considered the most important thing in a scientific study, not the one that you ignore in order to make pretty much the opposite case:

Although the number of children exposed prenatally to selective serotonin reuptake inhibitors in this population was low, results suggest that exposure, especially during the first trimester, may modestly increase the risk of ASD. The potential risk associated with exposure must be balanced with the risk to the mother or fetus of untreated mental health disorders. Further studies are needed to replicate and extend these findings.

So we’ve gained a valuable insight into the WDDTY editorial policy on statistics.

  • Look at the relative and the absolute risk.
  • Look at whether the outcome is favourable (improved survival) or unfavourable (greater risk of something).
  • For favourable outcomes, use the lower figure and castigate anyone who uses the higher.
  • For an unfavourable outcome, use the higher figure.
  • Forget the conclusions, stick with the WDDTY agenda.

For WDDTY’s probably statistically unsophisticated audience, the message is the important thing. Torture the statistics until they give the worst possible appearance, and go with that. It’s what they want to hear (or at least what you’ve told them they want to hear).

But there’s more:

In the subsample with available data on drugs, this association was confined to women reporting antidepressant use during pregnancy (3.34, 1.50 to 7.47, P=0.003), irrespective of whether selective serotonin reuptake inhibitors (SSRIs) or non-selective monoamine reuptake inhibitors were reported. All associations were higher in cases of autism without intellectual disability, there being no evidence of an increased risk of autism with intellectual disability.

So: antidepressants, of any type (thus reducing the chances of a biochemical cause) are associated with a small increase in the diagnosis of autism using the subjective (behavioural) but not the objective (intellectual disability) criteria. The association is not proven to be causal, and the lack of a common factor between the drugs compared with the obvious common factor of depression, raises a very real possibility that the autism has a common cause with the mother’s depression.

Early days yet and there may well be more to this in the discussions in the literature, but would a responsible health journalist use the claim that antidepressants cause a 50% increase in autism diagnosis?

Thanks to Southwark Belle for the tipoff and for pointing out a minor error, now fixed above.

What doctors don't tell you
Why don’t doctors tell you that antidepressants in pregnancy double the risk of autism?

Because it’s misleading and alarmist claptrap.

One thought on “Just fancy that! Odds ratios and autism”

  1. Pingback: Daily Overload – News in short (08-11-2013) « The Skeptical Bear

Leave a Reply