Improving the transparency and reliability of observational studies through registration

From the use of booster doses against covid-191 to approval of novel cancer therapies,2 health authorities and health technology assessment bodies are increasingly relying on non-randomised trials and observational studies. The methodological problems of observational studies, such as residual confounding,3 are difficult to resolve, but concerns about transparency and reliability can be reduced through registration.

Registration is an open science practice whereby research hypotheses, elements of study design, and planned statistical analyses are prespecified, preferably in a centralised repository. Registration has become the norm for clinical trials (interventional studies),4 and some have called for it to be adopted for observational studies.56 However, opponents argue that systematic registration is unrealistic, unnecessary,78 and too complex.9 Some have even suggested that it may cause more harm than good as it could force researchers into scientific dishonesty9 and enable stakeholders with vested interests to use deviations from the initial protocol to discredit inconvenient findings.10

However, unreliable research findings may be just as misleading as unjustified criticism of valid evidence. Indeed, unclear methods and discrediting of evidence are often two sides of the same coin, as shown by nutritional research11 or the marketing of doubt by the tobacco industry.12 Given the stakes, if observational studies are to be used in regulatory or public policy decisions, they should be required to meet the same basic quality standards as interventional studies, which includes registration. This already happens for certain studies, such as non-interventional post-authorisation safety studies in Europe, but it can be further expanded to other observational research.