Control of parenteral nutrition preparations in ternary mixture : contribution of infrared spectroscopy for the discrimination of three lipid mixtures

Introduction
Ternary parenteral nutrition preparations are complex magistral preparations containing macronutrients (glucose, amino acids, lipids) and micronutrients (electrolytes, vitamins and trace elements). Depending on the needs of the patients, different types of lipid emulsions can be used such as fish oil, soybean oil or olive oil. Parenteral nutrition preparations present a risk of physico-chemical instability due to the large number of constituent elements. A change of ingredient can decrease the stability of the mixture. The aim of this work was to evaluate the contribution of infrared vibrational spectroscopy for the qualitative analysis of ternary parenteral nutrition preparations.

Material and method
Three types of lipid emulsions were prepared three times and analyzed at different concentrations of lipids (Smoflipid^®, Clinoleic^® or Omegaven^®) ranging from 0 to 9.7% in a glucose (14.5%) and amino acids (Vaminolact; 2.0%) constant matrix. 10 µL of each solution were deposited on the crystal and dried under a stream of nitrogen. Acquisitions were made using Perkin-Elmer Spectrum IR software between 450 and 4000 cm-1 for 30 minutes (1 acquisition per minute). The spectra acquired were divided into a calibration set for the development of the prediction models and an external validation set to determine the performance of the models. Spectral data were processed by chemometrics to develop prediction models to discriminate the three types of emulsion. Different multivariate analysis algorithms have been studied for the optimization of discrimination models.

Results
A total of 3090 spectra were acquired. The analyzes carried out on the emulsions before the deposit dried only allowed partial discrimination of the lipids; the water band between 3000 and 3800 cm-1 masking those of the lipids. The step of drying the samples with a flow of nitrogen was therefore necessary to limit the interference of water and maximize the spectral differences of the emulsions. After 30 min of drying, a prediction model by PLS-DA (10 principal components) was developed and optimized from the calibration set (n = 69) after a spectral preprocessing by vector and first derivative normalization. All spectra (n = 34) of the validation set, were predicted in the correct class of lipids, validating the discriminant analysis of lipid emulsions by IR spectroscopy for all the concentrations studied.

Discussion / Conclusion
Despite the relatively similar compositions of the three emulsions, IR spectroscopy offers an interesting possibility for securing production by discriminating the lipid emulsions contained in ternary parenteral nutrition preparations produced in hospital pharmacy.