When the SPIRIT moves you: protocol changes can introduce bias in non-inferiority trials

In recent weeks, the UK has seen masked students queueing for emergency meningitis jabs and the publication of the module 3 of the UK covid-19 inquiry,1 which criticised the government’s reliance on flawed advice that the virus did not spread through the air.2One aspect of our lack of readiness for the next pandemic is the current policy on protecting healthcare workers from airborne respiratory pathogens. Specifically, whether respirators should be used instead of ordinary medical masks.

Current policy in the UK and many other countries,3456 based on non-inferiority randomised controlled trials,78 is that respirators are needed only for aerosol-generating medical procedures such as intubation. But such trials are inherently predisposed to produce null results and mislead policymakers and potentially cause harm.Randomisation reduces some forms of bias, but it does not abolish bias altogether.

Post-randomisation biases can arise once a trial is underway, for example in how interventions are delivered, how...