More antidepressants please…and bump up the Ritalin too. That seems to be the suggestion from this study published by the National Bureau of Economic Research in the US, making a link between new drug treatments for depression and ADHD and falling crime rates (Hat tip: The Economist).

Since the 1990s violent crime rates have declined markedly (no, it’s true, really), especially in the US. Criminologists have struggled to explain the trends as the usual suspects seem to have had little impact. So enter the economists with their alternative explanations, such as Steven Levitt (he of Freakonomics fame) presenting evidence that legalizing abortion in the 1970s led to a decline in the number of young people at risk of criminality, thus reducing crime rates.

In their provocative paper, A Cure for Crime?, Dave Marcotte and Sara Markowitz use data on international drug sales and crime rates as well as more detailed US data from the National Comorbidity Study (showing that in those with a mental disorder the percentage receiving treatment has increased from 20 to 33%) and national prescribing rates to show that ‘the countries with the largest declines in crime rates in the 1990s were almost exclusively those with the fastest growth in SSRI sales’. Details of the analysis are in an ungated preliminary version of the paper here. To control for overall improvements in health care they also looked at the impact of the non-psychotropic medicines, statins and COX-2 inhibitors which have also seen a rapid growth in prescriptions, but here there was no effect. They found that that increased prescribing of psychiatric drugs, notably SSRIs and stimulants (Ritalin), were associated with a reduction in violent crime and go on to conclude:

“Our evidence suggests that, in particular, sales of new generation antidepressants and stimulants used to treat ADHD are associated with rates of violent crime, with weaker evidence that anti-psychotic medications played a role in declining crime rates. The magnitude of the elasticities estimated here are clearly small. We estimate that a one percent increase in the total prescription rate is associated with a 0.051 percent decrease in violent crimes. To put this in perspective, doubling the prescription rate would reduce violent crimes by 5 percent, or by about 27 crimes per 100,000, at the average rate of 518 crimes per 100,000 population. While doubling the prescription rate seems like a large change, it has been estimated that 28 percent of the U.S. adult population in any year has a diagnosable mental or addictive disorder, yet only 8 percent seeks treatment (USDHHS 1999). Doubling the treatment rate would still leave a substantial portion of the ill untreated.

From the beginning to end of our panel, prescriptions per visit increased by 41 percent. Our elasticity estimates imply that this would reduce the total number of violent crimes committed by about 35,000. In fact, the total number of violent crimes reported to police declined by 300,000 during the period. Our estimates imply that just under 12 percent was due to expanded mental health treatment.”

Medical journals tend be dismissive of natural experiments and ecological studies such as this, considering them pretty weak evidence. But given that questions like this are never going to be answered by randomized controlled trials, if the methods are robust some freakonomic epidemiology may be just what we need.

Can you believe what you read in a medical journal? Probably not, as many if not most research findings turn out to be false. Poor research design and underpowered studies are part of the problem but looming large in the background is the spectre of publication bias.

No one doubts that negative studies should be published, yet it remains the case that they struggle to get into journals. By negative studies I mean studies that don’t show a statistically or clinically significant effect, or where a new treatment is more effective than standard treatment or placebo but has intolerable or dangerous adverse effects. Much of the blame has been heaped upon those unscrupulous drug companies callously suppressing unfavourable data. But before we all climb up on our collective high horse you should read this month’s Editor’s Choice (free to access) in Evidence–Based Mental Health.

In his personal account of his struggle to publish negative data on the drug lamotrigine, Nassir Ghaemi points the finger at not just the pharma industry, but at the FDA, journal editors and the peer review process itself. A Boston psychiatrist, he speaks as an insider having sat on an advisory board for GlaxoSmithKline as well as the editorial board of the journal Bipolar Disorders (he also writes an entertaining blog, Mood Swings). The contemptuous tone of the rejection letters will be familiar to anyone who’s submitted a paper, as will the contradictory reasons for refusal.

We now have clinical trials registration – requiring that all results end up somewhere in the public domain - which is clearly a good thing. There are also journals such as BMC Research Notes and the Journal of Negative Results in BioMedicine that are specifically aiming to publish negative studies. However the fact that a paper attempting to address publication bias should itself fall victim to that bias indicates that this is a problem that won’t go away.

That drug companies play fast and loose with study data is hardly news. It’s a widely-acknowledged problem that has been highlighted particularly with antidepressant trials where the advantage over placebo is often equivocal. Not before time, someone has decided to draw a line: “Deception through concealment is no trivial offence”, says the director of IQWiG – the German version of NICE – as he accuses Pfizer of concealing data about the antidepressant, reboxetine.

IQWiG (The Institute for Quality and Efficiency in Health Care) has ruled that reboxetine has no proof of benefit. They report that the drug has been tested in at least 16 trials but in 9 of them key information is not reported so they are unable to evaluate the effect of the antidepressant on 3000 of the 4600 patients enrolled. What are the implications of their ruling? The watchdog provides an assessment that is used to inform which medical treatments can be reimbursed through the public health insurance system in Germany – Europe’s largest market for drugs.

A rather sniffy spokesman for Pfizer told the BMJ that IQWiG is not a ‘regulatory authority’ but a ‘private institute’ and that there was no obligation to provide them with information. This seems at odds with Pfizer’s own policy on disclosure: “in all cases study results are reported by Pfizer in an objective, accurate, balanced and complete manner and are reported regardless of the outcome of the study or the country in which the study was conducted.”

Reboxetine itself has had a bit of a mauling of late. The Lancet’s recent comparative meta-analysis of 12 newer antidepressants put reboxetine at the bottom of the pile both in terms of efficacy and tolerability. It’s also not available in the US as the FDA turned down the application for a license for reasons which are still unclear. Paradoxically, if this license had been granted current US legislation would have required full disclosure of all the trial data. IQWiG is now calling for a similar European Union-wide legal obligation to publish all trial results as clearly self-regulation by the industry is not working. Corrado Barbui, writing in EBMH, made the same point back in 2007. Isn’t it time to get on with it?