Implementation of a discrimination strategy for eight monoclonal antibodies by FIA-UV using a confusion matrix: application to pembrolizumab
1 Service de Pharmacie, Hôpital européen Georges Pompidou, APHP Centre Université de Paris, France
2 Equipe Lip(Sys)² Chimie analytique pharmaceutique, Faculté de Pharmacie, Université Paris Saclay, Chatenay-Malabry, France
In order to secure the production activity of chemotherapies, various approaches based on visual or analytical control can be set up. Analytical methods most often use the spectral properties of molecules in order to their quantification and/or discrimination. Due to the structural similarities of some macromolecules, performing discriminating analysis by spectral methods of an ever-increasing number of specialties is a challenge. The referencing of a new specialty in the spectral library can weaken the discrimination of other molecules with a risk of confusion. This work proposes a strategy for the development of a robust spectral discrimination model applied to pembrolizumab (PBZ) among seven other monoclonal antibodies (mAbs).
Implementation of an analytical control of PBZ preparations by FIA (Flow Injection Analysis) combined with UV detection.
PBZ analysis is performed by FIA-UV (3 µL injected in 0.9% NaCl solution as carrier fluid at 1 mL/min). Quantitative analysis was performed at 280 nm and validated according to SFSTP guidelines. Discrimination analysis was performed by comparing spectra acquired between 275 and 300 nm after a second derivative with the pre-registered spectral library including 7 mAbs. In order to guarantee the robustness of the discrimination, a confusion matrix (CMX) was developed for mAbs presenting critical spectral recognitions with PBZ (tocilizumab (TCZ) and trastuzumab (TRZ)). Thus, 47 real-life samples (TCZ n=7, TRZ n=6, and PBZ n=34) were analyzed by the 3 analytical methods.
With a maximum repeatability and intermediate precision of 0.4% and 0.7% respectively, a correlation coefficient (R2) of 0.9999, the quantitative method was validated for a concentration range from 1.5 mg/mL to 9.625 mg/mL. Of the 34 PBZ samples, 100% showed consistent recognition according to the developed method. CMX confirmed the robustness of the recognition with no cross-spectral recognition observed for TCZ and TRZ.
Within one month of the method implementation, 162 PBZ preparations were analyzed with a non-compliance rate of 0% and an average relative error of -2.91% on the quantitative analysis. 100% of the samples showed consistent spectral recognition and no spectral overlap was observed with the other seven antibodies. Because of the spectral similarities of these macromolecules in UV, setting up a CMX can be an interesting tool to guarantee the spectral recognition. This step should be performed systematically to study the robustness of the discriminant analysis.