Approaches for harmonization of biomarker data from multiple studies: a narrative methodological review

Abstract Research consortia frequently attempt co-analysis of biomarker data from multiple sources. However, limited guidance exists on the harmonization of biomarker data, which are prone to various sources of variation. This narrative review aimed to address this gap by providing an overview of the existing methods for harmonizing biomarker data and their application. A systematic search of literature published by March 2024 was conducted using PubMed and Web of Science. Articles were included if co-analysis of individual participants data with at least one biomarker from multiple studies were reported and the biomarker harmonization methods were described. A total of 23 articles were included in the narrative review. Overall, 5 methods used for biomarker harmonization were identified: (1) calibration, (2) round robin test, (3) variable rescaling, (4) categorization, and (5) logistic regression-based method. Among these, calibration appears to be superior to other methods. To facilitate researchers in their choice of method, we suggest a roadmap for a step-wise biomarker data harmonization approach. There is a pending demand for developing standard definitions, methods, and quality criteria to assure proper harmonization practices in biomarker-based pooling initiatives.