Construction of identification system for non-invasive analysis of macrolides tablets using near infrared diffuse reflectance spectroscopy.

A NIR identification system consisting of two proposed models (Model I and Model II) has been developed for the analysis of 10 different products of macrolides tablets from different manufacturers. A total of 253 batches of the 10 products from 93 manufacturers in China were used for the system construction. First, a universal classification model (termed Model I) was constructed for 10 products using all the samples with the objective to distinguish the macrolides homologues which have the similar molecular structures. Secondly, 10 models (termed Model II) were developed separately for each product by using their samples. For each type of macrolides products, the two qualitative models are used in tandem as an identification system in mobile labs. Only when Model I and Model II both show acceptance results can an unknown sample be identified as "genuine". Internal and external validation showed almost 100% correct identity. Our study has shown that the analytical accuracy can be greatly improved when using this identification system and it will be efficient for quickly prescreening the drug quality in the open market and distribution channels.