VP, Global Head of Data Science Odysseus Data Services Inc. New York, United States
Background: Prevalence estimates of rare malignancies are typically associated with a high degree of uncertainty when submissions to the EMA Committee for Orphan Medicinal Products are being assessed.
Objectives: To develop standardized analytic approaches for estimating prevalence of rare malignancies in the Data Analysis and Real World Interrogation Network (DARWIN EU®).
Methods: Included databases are chosen based on having a representative population among whom diagnoses of rare malignancies can be identified. All persons in a database are eligible to contribute to the denominator in the prevalence calculation once they satisfy a prior history requirement. To allow estimation of partial and complete prevalence, a person is considered to have the outcome of interest only within a prespecified time after their initial diagnosis; n years after for n-year partial prevalence and all time after for complete prevalence. Point prevalence is estimated on January 1st of each calendar year. For annual period prevalence, a requirement for participants to be observed for the full period can be required. The first DARWIN EU® study estimated the prevalence of rare blood malignancies: follicular lymphoma, diffuse Large B-Cell Lymphoma, multiple myeloma, chronic lymphocytic leukemia, acute myeloid leukemia, and acute lymphocytic leukemia. Data came from Integrated Primary Care Information Project (IPCI), The Netherlands, the Information System for Research in Primary Care (SIDIAP) linked with hospital discharge data, Spain, and Clinical Practice Research Datalink (CPRD) GOLD, UK, IQVIA Belgium LPD, and IQVIA Germany DA. In the study the impact of partial versus complete prevalence, different requirements for prior history and full periods for period prevalence were considered. Where available, results were compared with available prevalence estimates from cancer registries.
Results: Prevalence estimates were generally in-line with estimates from other resources including cancer registries and followed similar age and sex distribution. The choice between partial and complete prevalence had the largest effect on estimates, with complete prevalence up to double that of 5-year partial prevalence. An upward trend in prevalence over time was also more pronounced for complete prevalence. While they did have an impact, requirements for prior history and full periods for period prevalence had a less marked effect.
Conclusions: Our findings highlight the approach to prevalence estimation for rare malignancies in DARWIN EU®. The proposed methodology offers fast, reliable, and systematic assessment of prevalence of rare malignancies in Europe. However, methodological considerations, particularly around appropriate outcome duration, must be kept in mind.
