Is convalescent plasma futile in COVID-19? A Bayesian re-analysis of the RECOVERY randomised controlled trial

Abstract

Introduction: Randomised trials are generally performed from a frequentist perspective reporting point estimates and 95% confidence intervals. This approach can confuse evidence of no effect with no evidence of an effect and does not allow for contextual knowledge. The RECOVERY trial evaluated convalescent plasma for patients hospitalised with COVID-19, the interaction test for the primary outcome was not statistically significant, and the trial concluded no evidence of an effect. From the clinical immunology perspective, there is strong justification to expect differential responses to convalescent plasma in patients who already have their own antibodies to SARS-CoV2 (seropositive) versus those who do not (seronegative). Methods: Outcome data was extracted from the RECOVERY trial both overall and for seronegative participants. A Bayesian re-analysis with a wide variety of priors (vague, optimistic, skeptical and pessimistic) was performed calculating the posterior probability for both any benefit or a modest benefit (number needed to treat of 100). Results: Across all patients, when analysed with a vague prior the likelihood of any benefit or a modest benefit was estimated to be 64% and 18% respectively. In contrast, in the seronegative subgroup, the likelihood of any benefit or a modest benefit was estimated to be 90% and 74%. Results were broadly consistent across all prior distributions. Conclusion: Performing clinical trials during a pandemic is challenging, and RECOVERY has provided high quality evidence for numerous therapies. However, the use of frequentist hypothesis testing in this trial has led to the trialists and governing bodies to conclude a strong evidence of no effect. Based on this trial, and other prior knowledge there remains a strong probability that convalescent plasma provides at least a modest benefit in seronegative patients.