The drugs don't work?

In March I was writing about my reluctance to prescribe or recommend treatment that is unsupported by the available evidence. Soon after my blog was posted, a study purporting to prove the ineffectiveness of commonly used antidepressants was all over the news headlines. So have I responded by taking dozens of patients off their antidepressants? Not exactly, because what the study actually proved was somewhat different to what its authors, Kirsch et al, have been reported as claiming.

What certainly has been shown, by combining unpublished with published data, is that these pills are not as effective as their manufacturers have made out. Comparisons with placebo pills show that a difference can be demonstrated that is 'statistically significant', and this is undisputed. What has been claimed, though, is that the difference is too small to be 'clinically significant' i.e. that if doctors were to prescribe placebo pills in place of all their antidepressant prescriptions for mildly and moderately depressed patients, the difference would be too small to be of any real importance.

Unsurprisingly, this study has caused as much discussion within the medical profession as outside it, and many of the comments made have been along the lines of: 'who is to say how much of a difference needs to be proven in order for it to be deemed significant?'

When it comes to statistical significance, we fortunately have a generally accepted cut-off, namely a ‘p’ value of below 0.05 (corresponding to a less than 1 in 20 probability of the observed finding coming about by pure chance), but for clinical significance no such cut-off exists; what we do have is a scale that is more widely accepted in some circles than others. What's interesting is that if we use this scale to assess these new figures for antidepressants they come roughly midway between the markers for ‘low’ and ‘medium’ clinical effectiveness. Therefore, if one takes the ‘medium’ marker as the cut-off for a level of effectiveness that justifies use, then our widely-used antidepressants fail the test by a comfortable margin.  This is what our authors chose to do. Their rationale for doing so is that this is the cut-off chosen by the UK’s National Institute for Clinical Excellence (NICE) in assessing effectiveness for treatments.

What has taken place then is a judgment that, although not difficult to defend, has to be recognised as something more or less arbitrary. Many clinicians, myself included, will deem that a low-to-medium level of clinical effectiveness over that of placebo for a pill such as fluoxetine is quite acceptable, since the incidence and severity of side-effects, including dependence, is really quite low and its cost to the NHS nowadays amounts to a few pence per day.

What Kirsch et al are quite right to point out is that these tablets do not in fact have the medium-to-high level of effectiveness sold to us on the basis of trials sponsored and selected for publication by their manufacturers. This indicates some fairly dubious ethics within the drug industry but, to be honest, comes as no great surprise to me. I am in firm agreement with the commentator who has suggested that the latest results should make us more circumspect in our use of these tablets, without dismissing them out of hand on the basis of a criterion of clinical effectiveness that comes down to a matter of opinion.

Let's not forget that all this analysis is done in comparison to the placebo effect, an effect that for depression Kirsch himself describes as 'whopping'. Although I would consider it unethical to prescribe medication that was no better at all than placebo, to make use of a sizeable placebo effect alongside a pharmacological one that is known to be genuine (albeit fairly modest) seems quite appropriate to me, given that the costs, in terms of money and adverse effects, is so small.

One could easily argue that the cost to society in terms of what has been termed 'the medicalisation of mood' is far more significant, and some of my colleagues in medicine may see fit to fight a rearguard action against this, and its pharmaceutical perpetrators.

Of course, modern antidepressant medications are by no means the only tools available, and my next post will look at some of the alternatives.  But that doesn't mean we should be throwing the pharmaceutical baby out with the bathwater - even if the baby is smaller, and the bath rather larger, than we were previously led to believe.