A Quality by Design Approach to Rapid Biopharmaceutical HPLC

Method Development

Introduction
This work describes the use of Fusion AE™ for Galaxie™ - an integrated method development
software application that enables a true Quality by Design approach to develop and optimize a
biopharmaceutical HPLC separation. This method required the separation a single large
product peak from several co- or closely eluting impurity peaks in the shortest time possible.
The legacy method used to separate these components involved a very long (over 60 minute)
two-step gradient method that did not meet all method performance requirements, including
peak resolution and robustness.

Materials and Methods

Instrument – Gradient HPLC with Diode Array detector (Agilent Corp., Palo Alto, CA).
Column – 250 x 4.6 mm reversed phase column.
Buffer System – Water/TFA aqueous phase, Acetonitrile/TFA organic modifier
System Parameters Included as Experiment variables – Flow Rate, Gradient Slope,
Gradient Time, Column Temperature

Rapid Method Development Platform – Instrument control, chromatogram generation, peak

processing: Varian Galaxie chromatography data system (CDS), (Varian Inc., Palo Alto, CA.)
Statistical experimental design, data analysis, modeling, optimization: Fusion AE for Galaxie
(S-Matrix Corp., Eureka, CA.)

Experimental Method – Study factors were varied according to a model-robust screening

design generated by Fusion AE, which constructed the 27-run design as a set of ready-to-run
methods and the corresponding sequence in the CDS. The experiment was run overnight on
the HPLC under Galaxie CDS control. Peak results were imported from the CDS into Fusion
AE, using a file-less data exchange module, for automated data analysis. Optimization solution
searches were conducted with Fusion AE numerical and graphical optimizers using the
following goals:

USP Resolution: 1.5 — 2.5, USP Tailing: 0.8 — 1.5, Robustness Cp: ≥ 1.0.
Figure 1. Experimental Design Template -
experimental parameters (study variables)
including Flow Rate, Gradient Slope,
Gradient Time and Column Temperature
were entered into a standardized template.

Figure 2. A Model-robust Screening design

type was used to efficiently screen large
ranges of the variables and quantify their
effects on method performance.

Figure 3. The Fusion AE Software generated a

statistical experimental design. The variables included
process and mixture types (Gradient Slope is a two-
component mixture). Fusion AE therefore selected
mixture process algorithm design (1).
 Figure 4. A Galaxie sequence, comprising 27 individual automatically generated methods
based on the experimental design parameters was built by the Fusion AE software within the
Galaxie CDS.

Fusion AE automatically carried out a

rigorous statistical analysis on all the
responses for all named peaks. The data
analysis strategy included the following
inter-related analyses (2) :

• Experimental Error Analysis

• Transformation Analysis (nonlinear error)
• Regression Analysis (including Backward
Elimination modeling)
Figure 5. Peak results data were automatically
• Outlier Analysis
imported from the Galaxie CDS using file-less
• Residuals Analysis and Plotting data transfer.
• Model Effects Pareto Ranking and Plotting
 Pump Flow Rate
0.5 1.0 1.5
Column
Oven
Temp.

40.0

52.5

65.0

Figure 6. Single Response Surface plots for Peak 3 data showing the effect of
increasing flow rate and column temperature on resolution
 Size of unshaded
region around
optimizer answer
indicates very
stable method in
terms of meeting
goals.

Figure 7. Overlay Graphic of all responses showing the region (unshaded)

where all goals are met or exceeded and the regions (shaded) where goals are
not achieved
 Results

1. Responses were evaluated and modeled for all four critical peaks. All models fitted the
data.
(all coefficients are statistically significant, model prediction error ≈ experimental error). (3)

2. Cp Definition (4): UAL – LAL (Upper minus Lower Acceptance Limits)

6σ Variation (Response variation computed from
simulation)

A Resolution Cp value was computed for each run using a modified Monte Carlo simulation
approach (8) that employed (a) the Resolution models obtained from analysis of the
experimental results, (b) user-defined tolerance limits on the instrument parameters, and
(c) the user-defined acceptance limits for the response.

3. Optimization search goal settings:

USP Resolution – target = 2.0. range = 1.5 — 2.5.

(higher resolutions may cause problems with later adjacent
peaks)

USP Tailing – target = 1.0. range = 0.8 — 1.5.

Retention Time – minimize, upper bound = 30 minutes.

Robustness Cp – ≥1.0.
(means ±3σ of Resolution variation is within acceptance limits)

4. Software Automated Optimizer answer was a simple (one-step) 30 minute gradient, flow
rate of 1.4ml.min, initial organic modifier concentration of 20% and column temperature
o
setting of 46.7 C.
 a)
Figure 8. Chromatograms
showing increasing degree of
resolution between product
(Peak 2) and impurity peaks.
Method parameters for each
run were:
Chromatogram 1, 2
a) 1 ml.min, 10 minute
gradient, 32% initial organic,
o
52.5 C
b) 0.5 ml.min, 50 minute
gradient, 20% initial organic,
52.5 oC
c) 1.5ml.min, 30 minute b)
gradient, 32% initial organic,
52.5 oC

c)
Figure 9. Chromatogram resulting from a partially optimized method showing baseline
resolution of impurity peaks from product peak.
Conclusions
Fusion AE for Galaxie was able to successfully develop a HPLC method, optimized for
flow rate, gradient time, percent organic modifier and column temperature, that met all
the initial biopharmaceutical product purification separation goal settings.

Single response surface plots indicated that the major peak resolution response
effectors were a combination of high flow rate and elevated column temperature.

The Overlay Graphics plot showed that, at the predicted optimum method settings, the
method was extremely robust with respect to gradient time and column temperature
settings.

The 27 chromatograms, combined with the rigorous experimental design and

subsequent data analysis completed during this experiment, provide a statistically
defendable data set for justifying the optimized method parameter settings.
Literature Cited

1) ICH-Q2B - Guideline for Industry. Q2B Validation of Analytical Procedures:

Methodology. November, 1996.
2) Method Validation by Phase of Development. An Acceptable Analytical Practice,
Pharmaceutical Technology, November, 2004.
3) Montgomery, D., Design and Analysis of Experiments, Fourth Edition, John Wiley &
Sons, New York, New York, 1996.
4) Myers, Raymond H. and Montgomery, Douglas C., Response Surface Methodology,
John Wiley and Sons, New York, New York, 1995.
Authors
Huqun Liu,, Varian Inc., Lake Forest, CA, Graham Shelver, Varian Inc., Walnut Creek,
CA, Richard Verseput, S-Matrix Corporation, Eureka CA, USA.

Acknowledgements
The Authors would like to thank Ms Shan Lin for her assistance with this project and
George Cooney, S-Matrix Corp., Eureka, CA, for his help with data analysis.