If we were to want to work out if a drug (or another treatment) is any good, we’d do well to conduct some ‘randomised controlled trials’. Basically, this means randomly assigned individuals to the drug being tested or placebo (or another drug). Neither the researchers nor the study participants know what’s being taken. Symptoms or other markers of health (e.g. biochemical markers) are monitored, as well as side-effects. After a period of time the ‘code is cracked’, and we can learn who was taking what. Now we can know if the drug being tested was more effective at helping people than the placebo, and how safe it is. A few studies showing favourable results will usually get a drug passed as ‘fit for purpose’. And this is how, by and large, drugs end up getting licenses so that doctors can prescribe them to their patients.
Not so fast, though. Because while some claim (often the manufacturers) that there is a stack of research supporting a drug, it can sometimes be the case that the published research does not tell the full story. You see, sometimes there exists other research that is not so supportive of the drug being tested, that may not have seen the light of day. It’s natural (though not ethical) for drug companies to generally want to publish supportive studies, and ‘shelve’ more negative findings. There is an expression for this practice: ‘publication bias’, and it’s not the first time I’ve written about it. See here and here for other relevant posts.
Publication has been known to go on for decades, but only relatively recently has some members of the scientific community taken steps to stamp it out. One huge step forward has been a decision in the US for trials to be registered on a central database at some point before or during the undertaking of a trial. That way the study has been logged, and if the results mysteriously fail to appear, then questions can be asked.
Not so long ago, trial results had to be forced out of the drug companies making the cholesterol-reducing agents simvastatin and ezetimibe. It took two years for the manufacturers to cough up their data after the conclusion of the study. Once the data came out, we learned why: this drug combo didn’t work to reduce signs of cardiovascular disease. Subsequent studies have also proved negative. Now, not that long ago, such data would have been relatively easily hidden, even from prying eyes. These days, drug companies don’t have things so easy (and a good thing too).
Mind you, now we have these new rules, not all researchers are keen to play by them. Only last week I wrote about how researchers have decided to shift the goalposts in terms of the outcomes assessed for the cholesterol-reducing combo referred to above. See here for more about this.
Things are better now, however, than they were. And a natural question to ask is how many drugs earned their stripes on the basis of publication bias? Well, it seems some principled researchers are keen to answer this question, by re-assessing drugs using not only published data, but unpublished data too.
An example of such a piece of work appeared on-line this week in the British Medical Journal . German researchers decided to assess unpublished and published data on the anti-depressant reboxetine. Reboxetine is a relatively new type of antidepressant – it is a selective noradrenaline reuptake inhibitor’. Essentially, what this means is that it helps maintain levels of the brain chemical noradrenaline, which is believed to enhance mood. It’s similar, though not the same, as the more commonly-prescribed anti-depressants known as the ‘selective serotonin reuptake inhibitors’ or ‘SSRIs’.
Previous published evidence showed that reboxetine was more effective in treating depression than placebo, and was about as effective as the commonly-prescribed SSRIs. It’s been licensed for use for depression in many European countries (including the UK and Germany) since 1997.
The researchers found that almost three-quarters of data on reboxetine was unpublished. This is disquieting in itself. And when they put this data into the mix, an altogether different picture emerged.
When the totality of the evidence was assessed, it turns out that reboxetine was not better than placebo. Most researchers would summarise this by concluding that ‘it doesn’t work’.
And perhaps not surprisingly, reboxetine turns out to cause more harm than placebo, plus more adverse effects than the most commonly prescribed SSRI (fluoxetine). The authors conclude that reboxetine is `an ineffective and potentially harmful antidepressant.’ Yet, here in Europe it’s been licensed for more than a decade.
The authors of this re-analysis mention that here in the UK, the National Institute of Health and Clinical Excellence (NICE) describes reboxetine as “superior to placebo and as effective as other antidepressants in the treatment of depression.” As the authors point out, “this conclusion can no longer be upheld.”
It’s interesting to note that in the US, reboxetine was originally licenced but then this was revoked. While it appears the US authorities made quite the right decision, the discrepancy with some European countries suggest that the US authorities had more data to go on, or perhaps set different licensing criteria.
Such discrepancies and this flagrant example of publication bias does not instil confidence. The good news is that, more and more, at least some researchers are not content on doing what paymasters in industry tell them to, and genuinely appear to be seekers of the truth. I’d say we can expect many more of skeletons to emerge from the cupboard over the coming years.
1. Eyding D, et al. Reboxetine for acute treatment of major depression” systematic review and meta-analysis of published and unpublished placebo and selective serotonin reuptake inhibitor controlled trials. BMJ 2010;341:c4737