We would like to thank Professor Burns for his thoughtful reply to our recent editorial and we are grateful for this opportunity to respond. To clarify: we would be happy to see more RCTs in psychiatry, but only as one form of evidence among others. Interestingly, the same work of Karl Popper referred to in the reply is drawn on by a leading proponent of realism to support such a position.Reference Pawson¹

Professor Burns gives two examples of RCTs of complex interventions to demonstrate their value. Our view of the implications of these trials is, unsurprisingly, different. We find it hard to believe that assertive community treatment teams and community treatment orders are not effective for anyone, anywhere, or in any way. And although we agree with Professor Burns that the scarcity of trials evidence is problematic – in the case of community treatment orders, there have only been three RCTs with a total sample size of 749 patientsReference Kisely and Hall² – we also believe that RCTs alone will never be the whole answer.

Rather than privileging a method designed to estimate singular ‘average treatment effects’ and whether a treatment does or does not ‘work’, we would argue that a more sensible way to proceed is to develop approaches intrinsically attuned to detecting variation and difference and, most importantly, understanding what gives rise to it.Reference Subramanian, Kim and Christakis³ Where RCTs design out the effects of context, realist approaches see this as key.

We agree that other medical and healthcare specialities rely on evidence for the effectiveness of complex interventions. But what distinguishes mental health is the preponderance of interventions that require human agency, and factors such as therapeutic alliance, empathic communication and motivation: the relationship between community treatment orders and readmission rates is of a different complexity than that between chemotherapy and cancer remission, or between digitalis and cardiac function.

We acknowledge, and celebrate, the contribution of RCTs to evidence-based healthcare. But there remains a need for a plurality of methods. However astute and research-literate the clinician, RCTs select participants in ways that can make generalisation to real-world settings difficult. Realist approaches that help bridge the gap between the ‘what’ and the ‘how’ of clinical outcomes can only be a good thing. And the more complex the intervention – and the more context dependent – the more important this is. For us, RCTs alone are unlikely to be sufficient.

Parity of esteem for psychiatry is undoubtedly worthwhile, but this does not mean we have to imitate other specialities; as so often in the past, we can lead the way instead. Primus inter pares.