RESUMEN
Biologic reference drugs and their copies, biosimilars, have a complex structure. Biosimilars need to demonstrate their biosimilarity during development but unpredictable variations can remain, such as micro-heterogeneity. The healthcare community may raise questions regarding the clinical outcomes induced by this micro-heterogeneity. Indeed, unwanted immune reactions may be induced for numerous reasons, including product variations. However, it is challenging to assess these unwanted immune reactions because of the multiplicity of causes and potential delays before any reaction. Moreover, safety assessments as part of preclinical studies and clinical trials may be of limited value with respect to immunogenicity assessments because they are performed on a standardised population during a limited period. Real-life data could therefore supplement the assessments of clinical trials by including data on the real-life use of biosimilars, such as switches. Furthermore, real-life data also include any economic incentives to prescribe or use biosimilars. This article raises the question of relevance of automating real life data processing regarding Biosimilars. The objective is to initiate a discussion about different approaches involving Machine Learning. So, the discussion is established regarding implementation of Neural Network model to ensure safety of biosimilars subject to economic incentives. Nevertheless, the application of Machine Learning in the healthcare field raises ethical, legal and technical issues that require further discussion.