A QbD with Design-of-Experiments Approach to the Development of a Chromatographic Method for the Separation of Impurities in Vancomycin

This paper describes a novel method development approach using Quality by Design (QbD) with Design of Experiments to develop a UPLC method for separating 39 impurities in vancomycin resulting in an optimally performing analytical method while simultaneously applying robustness limits to ensure success in final method validation and ultimately in method transfer.

Benefits

Using specialized software in conjunction with UPLC Technology, an optimized QbD method for the impurities in vancomycin can be developed that will be robust for method validation and transfer.

Analytical methods are developed at various stages of the drug develoment process for samples of varying complexity. Due to the inherent nature of the method development process, redundant efforts take place across an organization, resulting in a very costly and time-consuming activities. If we can streamline the process by which we develop methods, products can be brought to market faster and in a more cost effective manner.

Many different approaches are typically used to develop chromatographic methods today including trial and error, method/column scouting, and software approaches such as first principles approaches and simplex optimization procedures. All these approaches suffer from the inability to determine complex interactions effects between method variables or measurably consider method robustness during the method development process.

Vancomycin is a tricyclic glycopeptide antibiotic derived from Amycolatopasis orientalis (formerly Nocardia orientalis) and is indicated for the treatment of serious or severe infections caused by susceptible strains of methicillinresistant (beta-lactam-resistant) Staphylococci. Vancomycin is a large molecule (MW 1485.71 daltons) and contains many impurities that are difficult if not impossible to separate. Traditional HPLC gradient methods have shown the ability to separate out as many as 13 of these impurities, while the use of sub-2-μm ACQUITY UPLC Column chromatography has demonstrated the separation of as many as 26 impurities.

This paper describes a novel method development approach using Quality by Design (QbD) with Design of Experiments to develop a UPLC method for separating 39 impurities in vancomycin resulting in an optimally performing analytical method while simultaneously applying robustness limits to ensure success in final method validation and ultimately in method transfer.

Analytical instrumentation

The vancomycin studies described here were carried out using an automated integrated system consisting of Fusion Method Development Software, Empower 2 Chromatography Data Software (CDS), and an ACQUITY UPLC System with PDA, Column Manager, and Solvent Select Valve allowing for the screening of up to four different column chemistries, six different aqueous buffers/pHs, and two different organic mobile phases in one run.

Data management

Fusion Method Development Software (S-Matrix Corporation, Eureka, CA) is a Quality by Design based LC Method Development software package with built-in robustness metrics. Fusion includes a built-in interface with the Empower 2 CDS Software that controls the ACQUITY UPLC System. Using the chromatographic results collected from Empower 2 CDS, Fusion manages complex statistics and models for method optimization. Fusion builds experiments, analyzes data, and presents results as visual and numerical method predictions.

Phase 1: Rapid screening

Experiment design

The first phase of the method development involves the screening of the major effectors of selectivity, primarily the column chemistry, buffer pH, and organic mobile phase. The variables and ranges screened along with the constant conditions are listed in Figure 1’s tables.

Overlay graphics

The experimental design is run and data processed on the chromatographic system and the results are imported back into Fusion. The software predicts the optimum LC method after modeling all significant effects – linear, interaction, and complex – on each critical method performance characteristic. The unshaded (white) area of the overlay graph shown for the BEH C₈ column with methanol as the organic mobile phase (Figure 2) highlights the experimental region where the mean performance goals are obtained.

Optimization

The Automated Optimizer wizard defines the LC method performance goals and ranks them in order of importance. The software searches for the LC method that meets all the performance goals simultaneously. The best result(s) are reported along with predicted results for an experimental run (Figure 3). These conditions are used for the next stage, Method Optimization.

Phase 2: Method optimization

Experiment design

Phase 2 experiments use the column (ACQUITY UPLC BEH C₈, 2.1 x 100 mm, 1.7 μm) and mobile phase (pH 5.0 buffer, methanol B solvent) results from Phase 1 plus additional variables with tighter ranges to determine the optimum LC method. The experimental design is created using pump flow rate, gradient time, final percent organic, and column temperature as final optimization variables in the ranges shown (Figure 4).

Fusion Software creates the experimental design and exports it to Empower 2, automatically creating all the necessary instrument methods, method sets, and sample sets. The experimental design is run and data processed on the chromatographic system and the results are imported back into Fusion.

In addition to the data analysis for method optimization, Fusion applies a combination of Monte Carlo Simulation and Process Capability statistics to evaluate method robustness without running additional experiments.

Multiple response surface plots

Visualizing the results with Fusion’s 3D Response Surface Plots demonstrates the combined effects of variables on key chromatographic responses such as resolution, peak tailing, and retention time. Colors represent the magnitude of interaction and the curvature indicates the type of interaction (Figure 5).

Multiple response effects plots

Another option for visualizing the interactions between variables is through Multiple Response Effects Plots. These plots clearly identify simple linear additive effects (1), complex interactions of variables (2), and other types of response effects (Figure 6).

Final results of method optimization

Next, an overlay graph is created using a number of critical chromatographic responses. Of primary importance for the vancomycin separation is maximizing the number of peaks observed and the number of peaks exceeding different levels of resolution while minimizing the area of the vancomycin peak, which equates to separating out the most impurities. The overlay graph shows the QbD Design Space (unshaded region) where the method meets the mean performance goals and robustness criteria (Figure 7).

Using ranked response variables, the Method Optimizer determines the optimum method to best meet the performance and robustness goals specified. The final method conditions are listed along with predicted response results with confidence limits for this method (Figure 8).

This method was exported to Empower 2 and the vancomycin sample was run to evaluate the prediction accuracy. The chromatogram in Figure 9 shows the separation of vancomycin impurities obtained with the optimized method.

The experimental results compare favorably with Fusion Software’s predictions.

The QbD-based Fusion Software method improved the separation of impurities in vancomycin from 26, obtained previously with UPLC methods developed manually, to 39 impurities observed with the method shown.

• Fusion Method Development Software, used with the ACQUITY UPLC System, generated an optimized method for the analysis of vancomycin and its impurities in two business days.
  
• The use of UPLC data managed and processed by Fusion and Empower 2 software established a valid design space with both mean performance (set point optimization) and robustness (operating space).
  
• The QbD method’s resolution improved from 26 peaks in previous method to 39 peaks.
  
• Integrated robustness calculations ensure a reproducible method, which increases confidence in the ability to validate and transfer that method.