U.S. flag An official website of the United States government

Extract Preparation for Chemical Characterization Studies – Liquid-liquid Extraction

Catalog of Regulatory Science Tools to Help Assess New Medical Devices 


This regulatory science tool is a model for establishing analyte recovery in chemical characterization studies following liquid-liquid extraction.


Technical Description

The liquid-liquid extraction tool is a model which establishes the impact of liquid-liquid extraction method parameters on recovery of extractables in chemical characterization studies. In the model, the recovery of an established universe of extractables is calculated from the combination of a modified Nernst Distribution Law equation and validated quantitative structure-property relationship equations for determining distribution coefficients (Kd) and pKa. The model’s input parameters are pH, extract volume (Ve), organic solvent volume (Vo), and extraction iterations (n), pKa, and Kd. A framework is also presented in the Procedure for Applying Framework, which provides instructions on how to select reference standards for evaluation of liquid-liquid extraction and apply the model to a universe of chemicals [1]. The model and framework for estimating recovery of a universe of chemicals is described in additional detail in a peer-reviewed publication [2].

The model has been verified for the following conditions:

  • Aqueous solutions – water and 0.9% saline
  • Organic solvents – hexanes, dichloromethane, ethyl acetate
  • pH – from 2 to 12
  • Ve/Vo ratio – from 0.2 to 10
  • n – from 1 to 5

Using the input parameters and models described, the framework outputs an estimation of the recovery of extractable chemicals. One can then examine the recovery for all compounds in the universe and obtain an estimate of total coverage based on the experimental parameters selected.

Intended Purpose

The liquid-liquid extraction tool allows the user to determine the impact of experimental parameter selection on recovery of extractables when liquid-liquid extraction is used in a chemical characterization study, and compare the impact of parameter selection across different laboratories. This information and finding can then be applied to the universe of extractables and predetermine their recovery behavior in a liquid-liquid extraction process. The presented model applies to aqueous (water or 0.9% saline) medical device extracts, which are generated as part of a chemical characterization study. The model may assist device manufacturers and test laboratories to establish a broadly applicable rationale for selecting the most appropriate experimental conditions and reference standards based on their needs.


The model and framework described has been tested and verified by demonstrating accurate prediction of recovery for a variety of extractables and a range of experimental parameters. These tests included a range of values for pH, Ve, Vo, and n-value for nine analytes. The nine analytes were chosen to demonstrate a range of Kd and pKa values to range in recovery from 0% to 100% under various pH conditions. Each test was performed in triplicate, with a total root mean square error of <20%. The instructions provided in the Procedure for Applying Framework document incorporate a requirement for each laboratory intending to use the tool to demonstrate the model is effective for their internal processes. This is accomplished by establishing the laboratory-specific recoveries empirically [1].

The details of testing performed to verify the model and framework performance are described by Duelge et al as well as in the supporting documents [1][2].


The tool is limited to liquid-liquid extraction using dichloromethane, hexanes, and ethyl acetate. The tool does not address other organic solvents or extract preparation methods such as evaporation or solid phase extraction. Additionally, the tool is limited based on the applicability domain of the prediction models used. Specifically, inorganic solutes, most organometallic solutes, and any solutes, which contained no carbon atoms, were removed and excluded from the training set for predicting Kd as they were considered to have near 0% recovery in most circumstances. Similarly, the workflow used to generate the pKa excluded inorganic chemicals and mixtures and removed salts, solvents, and counterions.

Supporting Documentation

  1. Procedure for Applying Framework
  2. Duelge, Kaleb & Young, Joshua. (2023). Estimating Recovery in the Liquid–Liquid Extraction Chemical Space. Biomedical Materials & Devices. 2. 10.1007/s44174-023-00123-7. https://link.springer.com/article/10.1007/s44174-023-00123-7


For more information