That was the repeated phrase of my middle child’s obsessive bedtime reading for a while. Picture of police bikes, fire engines, ambulances, mountain rescue 4×4 and lifeboats.
In not one frame was the rescued individual entered into a clinical trial of therapy or diagnostics.
I guess that might have been asking a bit much, but is it also a bit much to ask for signed, informed consent with an appropriate time to reflect between information delivery and accession? If we worry about risk of bias in non-randomised trials, should the acuity of emergency studies make this even more important to get right?
And yet … the practical challenges are immense, so many emergency trials are quasi-randomised at best. But might this not be OK – after all, there isn’t really time to understand the likely outcomes of the patient and make a biased choice. Surely that makes them OK?
How can we tell if quasi-randomised are going to be valid enough to make a decision upon?
The question was addressed in a recent review that analysed a series of systematic reviews from the Cochrane Library. The team took seven systematic reviews, including 27 trials that had both quasi~ and truly randomised studies in them. The then looked at the baseline characteristics for important imbalances that could signify biases in selection and found it in 2 of the 11 quasi-randomised trials (18%) and 4 of the 16 trials using true randomisation (25%) – but 3/4 did have small sample sizes that could well have produced these effects. The team concluded that, in the emergency setting, the difference between true and quasi~ randomisation is likely to be minimal — but the problem of small sample sizes appears to be a greater one.
We can take the results of this to let us relatively happily look at both types of study and be equally confident. Work like this is essential in keeping up the flow of trial data that makes every step of the treatment pathway evidence-based.