Multivariate thinking for optical microfluidic analytical devices – A tutorial review

Abstract Microfluidic analytical devices (µFADs) made from glass, polymeric materials or paper are considered as a powerful tool in several fields such as chemistry, engineering, pharmaceutics, forensics, medicine, environment, food, biology and biotechnology, or security and biodefence. Optical-based µFADs measuring molecular spectroscopic properties such as transmission, absorption, reflection, fluorescence, or Raman are not alien to this framework and most of the developments currently underway are still focused on the use of univariate models. Multivariate models could be used to: (i) optimize concurrently the experimental variables (factors) in order to find the best operative conditions for chemical or technical processes, (ii) classify or discriminate material systems regarding one or more interesting features, and/or (iii) determine simultaneously the value of multiple physical-chemical properties directly or indirectly related to the chemical composition. In addition, multivariate models overlook problems arising from the interaction between experimental variables, or biases caused by the lack of measuring device selectivity. However, this strategy is still poorly applied and is a clear focus of progress in the development of µFADs. This tutorial aims to describe the potential and tools available to carry out these sorts of applications focused on the use of multivariate models and to critically review the published literature on this topic. Technical aspects related to the design and development of optical µFADs will not be considered.