A radical note to end the conference: forget your standalone RCT, just publish the updated meta-analysis. Or, less provocatively, place the meta-analysis center-stage in the study publication and include the RCT results later or publish them on the internet.

Alex Sutton, a medical statistician from the University of Leicester, argued that greater attention should be paid to previous research, notably meta-analyses, when designing a new RCT. The statistics in his argument are complex, but they involve working out, for example, the sample size required in the new study if the findings are to add meaningfully to existing knowledge. One possible, and controversial, consequence of this argument could be that several smaller studies prove more useful than one or two large studies.

Sutton also advocated making better use of existing data before deciding to undertake a new RCT. ‘Mixed treatment comparisons’ are the new fashion, he explained. This method enables researchers to compare and rank the effectiveness of several interventions that have been evaluated singly or together in separate studies. So, A might be better than B, which is not as good as C, which in turn is better than A: CAB.

These powerful arguments boil down to building knowledge about effectiveness more coherently and efficiently. When it comes to RCTs in children’s services we are still learning to walk – doing good if modest RCTs where possible and appropriate. But we can learn from our colleagues in medicine who have traveled further than us, and, encouragingly, it sounds as if our faltering steps may get us further than we thought.