We have been taking a different approach to press relations in recent months. With the help of a journalist colleague, Tom Parkhill, we are trying to be more proactive in shaping how relevant science is reported. This could sound like an exercise in propaganda, but actually it is an essential service to science journalists, who are usually only reporting what people tell them and trying to deliver a good story. Of course, we are keen to be positive when stories have merit. However, providing balance can have the fortunate side effect of killing a misleading story, so that is also what we try to do. Providing balance is rarely rocket science; it often consists merely of pointing out methodological limitations that referees should already have picked up before publication. Unfortunately, bad news is best for hunter gatherers like ourselves, so stories without balance do well in newsrooms.
Accordingly a major concern is any story that misleadingly attributes bad outcomes to exposure to psychotropic drugs. Perhaps the most damaging example has been the story that antidepressants cause suicide, especially in children. A decade on from that fiasco, it is clear that the black box warnings that regulators were spooked into putting on antidepressants for young people achieved the exact opposite of what would have been expected if antidepressants really did cause deliberate self harm. The rates of self harm went up, rather than down as the rate of antidepressant prescribing went down (http://www.bmj.com/content/bmj/348/bmj.g3596).
The reason why this story always seemed likely to be a turkey, is the phenomenon of ascertainment bias. Patients on antidepressants have more adverse effects of all kinds than those on placebo. That will be very likely to increase the rate at which all adverse effects are reported and the minor increases in rates of self harm claimed for the positive studies were all derived from side effect reports. Prospectively determined suicide ratings in the same studies showed a decrease in antidepressant treated groups proportionate to the effect on depression. However, the suicide story was driven by other factors: ambitious doctors with an axe to grind, lawyers with punitive damage claims in mind and grieving parents seeking an explanation for why a loved one had died. Maybe balance was never going to be possible.
Unquestionably, balance continues often to be lacking in the reporting of new alleged adverse effects of psychotropic drugs. The basis for such claims is usually a simple case control study. In such studies, cases, who have the adverse outcome, are compared with controls, who do not, and the measure is the percentage exposed to a particular risk. The choice of exposure can be anything from cornflakes to environmental pollution. The ‘health’ pages of newspapers and magazines abound with stories reporting the results of just such case-control studies. A perennial favourite is to publish scare stories about exposure to drugs taken by the mother while a child is in utero. The thalidomide story stalks behind less substantial findings. Certainly such stories tap into a fear that almost all the mothers I have ever treated express – that the drugs they are obliged to take to stay well will harm their children. It is not surprising that it has high news value when the news is bad.
Unfortunately, case control studies have a major problem and that is confounding by indication. This simply means that if you look for an outcome of exposure to any drug, you risk identifying problems that are associated with the diagnosis for which the drug is indicated. Death is for sure going to be associated with the use of anti-cancer drugs because only cancer patients get them and cancer patients have high mortality.
In the case of mothers receiving antidepressants, for example, the key issue will be whether any effect observed is due to increased genetic or other risks of a particular outcome in the offspring of the women prescribed antidepressants, rather than the effects of the drugs themselves. And experience (plus common sense) suggests that particularly where the bad outcome is behavioural/psychiatric (autism, ADHD etc.), special care should be exercised not to over-interpret any association. This is simply because depression itself has the potential for such a wide range of associations, independent of antidepressant treatment. Thus, depression is both moderately heritable and common in the probands of effectively all other psychiatric disorders, so will have shared predisposing genetic risks. Depression is very likely to be associated with epigenetic factors we do not yet understand. Depression is associated with misuse of alcohol and smoking. Depression is common in the physically ill so may indirectly imply vulnerability to a spectrum of physical illness. And finally, maternal depression has independent effects on child development. None of this (well, not much) will be estimated in large data bases so attempts to control for demographic differences between cases and controls may be reassuring but are likely futile. Even in the case of more certain diagnoses (of birth defects), diagnosis rather than drug exposure appeared to explain the association that had previously been confidently attributed to drugs (Jimenez-Solem et al BMJ Open 2012; 2; e001148. doi:10. 1136/bmjopen-2012-001148).
By maintaining awareness of what press releases are current we hope to continue to provide balance to public debate in the coming years. Fair reporting of science is important. Ultimately the public pays for what we do and it is critical that their understanding of biomedical science increasingly reflects the nuances of doubt and certainty, which we may take for granted.
Guy Goodwin, ECNP President