Study-design in pandemics: From surveillance and performance-evaluation to licensing and pharmacovigilance

Abstract

Andy Grieve, the first pharmaceutical statistician to be President of the Royal Statistical Society, practiced in the regulated world of drug development. With reduction in drug development costs as his motivation, Grieve advanced Bayesian methods for developing predictive methods for efficacy and toxicity - to be used as early as possible in the drug development process; and his presidential address exhorted statisticians to weigh-in wherever data are used to make decisions. Diagnostic tests for infectious diseases are less regulated than drugs and vaccines unless the blood supply is at risk. Unlike in the HIV and HCV pandemics of the late 20th century, even well-designed surveys linked to a volunteered biological sample (to be tested for SARS-CoV-2 antigen or antibodies) have had modest or low consent rates. Record-linkage, statistical design and reporting standards have seen triumph and tragedy. Among the triumphs are: Liverpool's insistence on dual testing (lateral flow device; polymerase chain reaction [PCR]) of some 6000 asymptomatic citizens who attended for SARS-CoV-2-screening; two tricky randomized controlled public-policy trials on daily contact testing for close contacts of index cases of SARS-CoV-2 infection versus self-isolation (with or without initial PCR); and among already-consented participants in surveillance, over 80% secondary consent for linkage to their health records, including the Immunization Management Service. Before the next pandemic we need to entrench better regulation of diagnostic tests, better informed consent (not via weblinks), better feedback to participants, and transparency about basic safety data.