(T1530-10-66) A Regression Approach for Assessing Large Molecular Drug Concentration in Breast Milk

Purpose: The development of an effective method for predicting the transfer of biologics from plasma into breast milk is important to ensure the safe use of medications during lactation. The aim of this study was to develop a regression model that could predict the transfer of monoclonal antibodies (mAbs) and Fc-fusion proteins from plasma into breast milk.

Methods: A preliminary list of approved and under-reviewed monoclonal antibodies (mAbs) was first obtained from the Antibody Society’s website. Then more drugs were included through search results from published papers from Pubmed database and Drugs and Lactation Database. We compiled a list of eleven mAbs and Fc-fusion proteins with available information of presence in the breast milk, and conducted multiple correlation analyses using their physiochemical properties. Physicochemical properties such as the isoelectric point (pI), molecular weight (MW), dissociation constant (Kd), and pharmacokinetic (PK) parameters such as clearance (CL), volume of distribution (Vd), and half-life (T1/2) were collected or calculated. A two-variable regression analysis was used to evaluate the correlation of each physicochemical property with M/P ratios of the drugs first. Then a multivariate regression analysis was employed to establish correlation of M/P ratios with different combinations of two physicochemical properties. The 3D isoelectric point (pI) of the Fv region and the buried surface area (BSA) between the light and heavy chains (LC_HC) are two factors that emerged as a promising predictor of the milk-to-plasma concentration ratio (M/P).

Results: The outcomes of the regression analyses between M/P and physiochemical properties can be seen in Figure 1. No statistically significant correlation was observed through these bivariate regressions, indicates that M/P ratio is not solely dependent on a single physiochemical property. Based on the two factors regression analyses result, multivariate regression was conducted for selected factors. The correlation between M/P ratio, 3D p1 of Fv region, and BSA_LC_HC was found to be good with R² of 0.9058. The 3-D regression plot is shown in Figure 2. Other combinations of the physicochemical properties did not show a statistically significant correlation. The multivariate regression model was used to predict the MP ratios for 79 different mAbs and the results were plotted in Figure 3.

Conclusion: We believe that this regression model could serve as a valuable tool to estimate the M/P ratios of mAbs and Fc-fusion proteins. Further model validation is necessary when the M/P ratios of additional biologics are available. This could inform clinical decision-making and improve the safety of drug use during lactation.

Acknowledgements: All authors declare no conflicts of interest relevant to the content of this article.

Fig. 1 Linear relationships between M/P ration and physiochemical properties: pI (A), Fab pI (B), 3D pI of Fv region (C), molecular weight (D), BSA_LC_HC (E), PatchRatio (F), DM.HM (G), Avg_HI (H), and Kd (I). Correlation equations are shown in the plot. R2 means the percentage of information contained in this regression model.

Fig. 2 The 3D plot of correlation between observed M/P ratio, 3D p1 of Fv region, and BSA_LC_HC of 11 large molecule drugs

Fig 3. The 3D plot of correlation between predicted M/P ratio, 3D p1 of Fv region, and BSA_LC_HC of 79 large molecule drugs