This year I took a poster to the Society for Clinical Trials meeting in Philadelphia.
The Society for Clinical Trials is an international organisation dedicated to the development and dissemination of knowledge about the design, conduct and analysis of clinical trial methodologies - rather like HTAi but with a focus on clinical trials rather than systematic review. Despite its claim to be international its meetings are almost always in North America - of those 32 meetings one has been in London and one in Brussels. The attendance this year reflected a similar pattern with 480 attendees, of whom less than 100 were not from either the USA or Canada. Of those 100 however, there were several HTA grant holders, and at least one Commissioning Board member.
I took home three things from this conference:
1) A stepped wedge trial is a special case of a cluster randomised trial, which should have been self-evident, but I'd never thought of it like this before. The difference between a stepped wedge and a traditional cluster design is the timing of when clusters receive an intervention.
2) A stepped wedge design is almost never the right choice. At the end of a conventional stepped wedge trial all the clusters will be on one intervention. This is usually the intervention that the investigator wants to push, and it can be quite difficult to withdraw the intervention should it not be shown to be superior to the alternative.
There isn't an efficacy argument, as a conventional stepped wedge design is not much more efficient than a cluster design with even allocation, and is subject to confounding by time. Therefor my conclusion is only to use a full stepped wedge design when it's impossible to remove the intervention.
Interestingly, the most efficient option is to have a number of clusters which only receive control, a number which only receive the intervention, and the remainder adopting a stepped wedge approach.
3) There's scope for exploring adaptive interventions in pragmatic trials.
An adaptive intervention (not an adaptive trial design) is when the interventions are clarified by rules such as
- try treatment A, if the patient responds continue else randomise between treatments B and C.
If large parts of the patient pathway need to be explored this is a potentially efficient option to do so.
As I wrote after my previous trip to this conference in 2011, for people interested in clinical trials this is a high value event with great content, and an oportunity network with the wider trials community.
While usually it's held in North America (Arlington next year, and Montreal in 2016), there is currently a plan to hold a joint meeting with the MRC Methodology Hubs in May 2017 in the UK - with Edinburgh being the current favourite venue.