2023 年 40 巻 4 号 p. 450-453
RWD is mainly used in clinical research and Post Marketing Surveillance (PMS), and recently its use in clinical trials is also being actively discussed in pharmaceutical companies. However, there are still issues regarding the variety and quantity of RWD and the reliability of outcomes, so further development is expected.
On the other hand, in overseas where the development and utilization of RWD has preceded, the evolution is still continuing.
Some trends in RWD development include, for example, “accumulation of outcome validation results” “structuring of unstructured data” “conversion to common data models” and “anonymization of risk-based approaches”
“Accumulation of outcome validation” is to validate the outcome defined in RWD, and to calculate indicators such as positive predictive value and sensitivity compared to the gold standard.
Although it is also required for PMS in Japan, there are still issues in implementing validation and accumulating results.
In “structuring of unstructured data” the text portion of an electronic medical record is structured using Natural Language Processing (NLP) to enable analysis of patient background information and efficacy of drug treatment. This will lead to more efficient recruitment for clinical trials at each hospital in terms of primary use, and will enable the generation of RWD that can be used for research and clinical trials in terms of secondary use.
“Conversion to a common data model” is an effort to integrate and analyze different RWD such as medical claims and electronic medical records, and a representative common data model is OMOP CDM (Observational Medical Outcomes Partnership Common Data Model). OMOP conversion enables coaxial analysis of multiple RWDs using standard programs that have already been accumulated, leading to more efficient research.
“Anonymization of risk-based approaches” is a common approach in the U.S. that addresses the dilemma that the more information in RWD is processed by deleting or masking, the less risk there is of re–identification, while the more useful information for research is lost at the same time. Efforts based on this concept have been started in Japan as well.
