Hisashi Urushihara: The Unreal Real World

2025/02/19

Pharmacoepidemiology is a field of public health that aims to improve people's health through the development and practice of methodologies for studying the use, safety, and efficacy of pharmaceuticals. Experimental research methods that artificially assign the administration or non-administration of a drug to be evaluated into two groups to compare efficacy and safety are called interventional studies, and are primarily used to prove the efficacy of drugs. In interventional studies, results with less bias are obtained by managing patient visits, collecting new data, and aligning characteristics between comparison groups. Data obtained under such experimental environments differ from general clinical practice and are "unrealistic," lacking generalizability, and are referred to as "champion data" because the effects of the drug tend to appear more easily. On the other hand, pharmacoepidemiology primarily uses observational studies that aim for more "realistic" evaluations using medical data generated in daily clinical practice, rather than in an experimental environment.

The advent of the information society has enabled the research use of massive databases consisting of vast amounts of electronic medical record information and insurance claims data, which are called "real-world data" as an antithesis to interventional studies. While it is one of the information sources traditionally handled in observational studies and can reduce the resources and costs for collecting new data from a vast number of patients, it is not always possible to obtain the information necessary for research because there are only test values already measured in actual clinical practice, and there is always bias in comparison because the characteristics of patients who use a certain drug and those who do not are different. In hospital medical record information, the occurrence of medical events can only be fully grasped during hospitalization, and follow-up during other periods is incomplete. Claims data may cause inconvenience because clinical information unrelated to insurance claims is likely to be missing.

However, by not simply leaving these basic shortcomings of "real-world data" as limitations but addressing them based on epidemiological theory, it is possible to lead to scientifically valid research results. Research using Danish "real-world data" proved the safety of antipyretic analgesic use in COVID-19 patients and had a major impact on global health authorities. I am also constantly aware of the limitations of data and how to deal with bias, but I continue my research while thinking that the feeling of "taming a wild horse" is also quite interesting.