Deconstructing bias in clinical trials

‘The way medical journals publish the results of clinical trials has become a serious threat to public health’. ^{Reference Smith and Roberts1} With this arresting sentence Richard Smith, former editor of the BMJ, began an article published 4 years ago. Why did Richard get so aerated? Well, let him explain. ‘Although most editors would like to meet expectant researchers shortly after a clinical trial's conception (or even before), to find out who the parents are and to ensure that the trial receives high-quality antenatal care, more often than not labouring researchers arrive at their offices heavily pregnant with results that require immediate, fast-track delivery. Some trials are deposited on the editor's doorstep, so that it is hard to tell who the parents are’. ^{Reference Smith and Roberts1} So Richard is asking the awkward question: ‘do authors manipulate editors in order to get trials published in journals?’ Unfortunately I think he is right. Medical editors are certainly midwives in this conspiracy, for we all reject papers that do not follow the rules for publication of trials, so isn't it to be expected that authors will fillet, finesse and finagle their papers to fit a journal's requirements and preferably produce positive results? Like most editors I like to feel I am publishing giant strides in science rather than shuffling backwards and forwards in centimetres, so I connive quietly in this deception. We publish two randomised trials (Kellner et al, pp. ; Yip et al, pp. ) and one systematic review in this issue and another paper that shows the importance of attention to methodology in clinical trials (Bethell et al, pp. ). Cuijpers et al (pp. ) do a valuable service in pointing out that publication bias is not restricted to drug trials supported by the pharmaceutical industry; psychological treatments suffer in the same way and, if their findings with cognitive–behavioural therapy in depression were replicated with other disorders, the effect size of this treatment would be reduced by 30% compared with equivalent controls. The defence that failure to find benefit may be a consequence of poor therapeutic competence ^{Reference Davidson, Scott, Schmidt, Tata, Thornton and Tyrer2} may be valid but such trials still need to be published. Seeing the review by Cuijpers et al had me rushing back to recent issues to see to what extent we could be castigated for publication bias. I was convinced our high-standing journal must have learnt something from Richard Smith's admonishments – sadly I was wrong. In 2009 we published five randomised controlled trials of psychological treatments; four reported clearly positive results ^{Reference Haddock, Barrowclough, Shaw, Dunn, Novaco and Tarrier3–Reference Furmark, Carlbring, Hedman, Sonnenstein, Clevberger and Bohman6} with only one being unequivocally negative. ^{Reference de Graaf, Gerhards, Arntz, Riper, Metsemakers and Evers7} The valuable editorial by Scott (pp. ) shows that much can still be learnt from trials with equivocal results and I hope that solid honest studies illustrated by the careful diligence of Kellner et al will still find a place between these covers.

We also must not feel that randomised trials should be automatically at the top of our publishing agenda. We still need reminding of this ^{Reference Slade and Priebe8} and Rawlins has recently reinforced it in a magisterial account of different forms of scientific evidence. ^{Reference Rawlins9} The notion of randomised controlled trials as the gold standard of evidence of efficacy should be revised. The standard is only silver or bronze; in the High School of Evidence the randomised controlled trial is at the top of class B but no one has yet passed the examination allowing access to class A. So when you read the results of randomised trials in the British Journal of Psychiatry look at them carefully and be prepared to cavil as much as to marvel, to read reviews and commentaries ^{Reference Parker10} that put them into perspective, and spare a tear for the many manuscripts that gather dust on shelves forlornly waiting for a call to see their negative findings.

A global tribute

Mental health initiatives rarely get the same accolades as those in other parts of medicine, so they are worth celebrating when they arrive, especially when they concern the plight of the forgotten needy in so many countries of the world. The Health Service and Population Research Group, led by Professor Graham Thornicroft at King's College London, has been recognised with a Queen's Anniversary Award, the first to be given in mental health, for ‘its commitment to enabling the recovery and improving the quality of life of people with mental health problems throughout the world’. We report often on the mental ill-effects of neglect, abuse and deprivation; indeed we do so in this issue (Oladeji et al, pp. ; Nayak et al, pp. ), and collectively this evidence is a call for action to correct preventable disease. Thornicroft has identified one of the main reasons for inaction in one word, ‘shunned’, ^{Reference Thornicroft11} and his group are being rightly acknowledged for their courage in taking this task on. It may even call for a small piece of doggerel: