The RAMbo acronym that Archi first chattered on about five years ago relied on the key suspected elements of risk of bias – the need for randomisation, accurate ascertainment of outcomes in the groups to which people were randomised, and being careful of measurement – thinking about what was actually measured and if it was blinded, and if not blinded, objective.
As with so many elements of EBM, the idea of checking if blinding of subjective outcomes really does make a difference has been subject to intense, methodologically robust, systematic review. A paper from the Cochrane centre undertook a review of trials where the same outcome measurement scale had been used in situations where both blinded and unblinded observers had made assessments. They showed that, on average, the unblinded observers noted a 65% greater effect than that recorded by blinded observers (but this average might be as small as about 15% or as great as 240%).
Can this lead to our interpretations of studies being different? We could, quite reasonably, use this to reduce all estimates of effect by about two-thirds if we are worried about unblinded interpretations. We’d be guessing, extrapolating and making some assumptions … but that’s nothing new for paediatricians, is it?