This article introduces an example of using JMP to perform design of experiments (DOE) in the development of an analytical method for a pharmaceutical.
In recent years, the ICH-Q14 guideline for the development of analytical methods has been developed, and "analytical methods utilizing more advanced techniques" are attracting attention.
The method of examining quality based on scientific evidence in the development of analytical methods is called AQbD (Analytical Quality by Design), and the use of design of experiments (DOE) is proposed as one way to achieve this.
For example, determining the optimal operating conditions for the following HPLC configuration is an example of analytical method development.
Analysis example: Optimization of HPLC operating conditions
Objective: To develop an analytical method that achieves complete peak separation (resolution ≧ 1.5) as specified by the Japanese Pharmacopoeia.
To achieve this goal, we will conduct experiments based on the concept of experimental design and use the data obtained to determine the operating parameters and conditions that will achieve complete separation.
In experimental design, the characteristic values being optimized or modeled are called " responses ," which in method development would be resolution or number of peaks, etc. On the other hand, the manipulable parameters that are thought to affect the response are called " factors ," which in HPLC would be column temperature or mobile phase flow rate, etc.
In this example, the study is conducted in two steps: "An experiment (screening experiment) to narrow down important factors from many factors," and "An experiment (optimization experiment) to find optimal conditions by fitting a more accurate model (response surface model) using the factors narrowed down by screening."
- Screening experiment : Extracting effective factors from a large number of factors that are thought to affect response
- Optimization experiments (response surfaces) : Finding factor settings that satisfy a response goal (maximization, minimization, etc.)
In JMP, we use " Custom Design ," a typical design creation method, and proceed with the analysis using the following steps. 
1. Screening Experiment
Objective: To screen factors that affect resolution.
As a result of the risk assessment, five factors thought to affect separation performance (shown in yellow in the figure below) were selected.
The conditions for the factors are as follows:
| factor |
role |
Lower limit |
Upper limit |
| pH |
continuous |
7 |
9 |
| Column Temp（℃） |
continuous |
35 |
45 |
| ACN（%） |
continuous |
30 |
50 |
| Flow Rate (mL/min) |
continuous |
1.4 |
1.6 |
| Column Type |
Categorical |
C1,C2,C3,C4 |
pH: pH of the mobile phase buffer;Column Temp: Column temperature.ACN(%): Percentage of organic solvent (acetonitrile) in the mobile phase
Flow Rate: Flow rate of the mobile phase.Column Type: Column type (4 types: C1 to C4)
*This is an example of isocratic elution (mobile phase composition is constant).
Create a plan
- Launch "Custom Design" and set "Response", "Factors" and "Model".
- Here, the model (factors to be screened) includes main effects and interactions ("pH*Column Temp", "pH*ACN", "pH*Flow Rate", "Column Temp*ACN", "Column Temp*Flow Rate", and "ACN*Flow Rate").
- Under Create Design, accept the default number of runs (24).
Entering response values and fitting the model
- You will be presented with 24 experiments based on an optimal design, so you will conduct the experiments and enter the response (degree of separation) values.
- After entering the response values, run the "Model" script in the upper left of the data table (click the green ▷ button). This will fit the model specified in the custom design.
Fitting the model and screening important factors
Check the "Summary of Effects" in the displayed report and identify important factors by "pooling for error," which removes factors with large P values in order.
In this example, the important factors were selected as "pH", "ACN", "Column Temp" and "pH*Column Temp".
Therefore, in the subsequent optimization experiments, we use "pH", "ACN", and "Column Temp" as factors.
*Here, we used "pooling for error" to identify important factors, but other methods such as stepwise variable selection methods can also be used.
2. Optimization Experiments (Response Surface)
In this step, in addition to Resolution**, N of peaks is also set as a response, and a response surface is created and optimized.
The responses and factors considered are:
| response |
specification |
| Resolution |
1.5 or higher |
| N of Peaks |
6 or more |
| factor |
role |
Lower limit |
Upper limit |
| pH |
continuous |
7 |
9 |
| Column Temp(℃) |
continuous |
35 |
45 |
| ACN(%) |
continuous |
30 |
50 |
Create a plan
- Launch "Custom Design" and set "Response" and "Factors".
- Specify the response surface model by clicking the [RSM] button as the "Model".
- Under Create Design, accept the default number of runs (16).
*The response surface model is a model that includes main effects, second-order interactions, and squared terms as effects.
*The design created with 16 experiments is consistent with the "Central Composite Design (CCD)."
Entering response values and fitting response surface models
- Enter the response values as in the screening experiment and run the "Model" script to fit the response surface model.
Optimization, building design space
- The "Prediction Profiler" report allows you to visually check the relationship between factors (horizontal axis) and responses (vertical axis) in the prediction equation.
- We search for a combination of factors that meet our goals (Resolution ≧ 1.5, N of peaks ≧ 6) by selecting [Optimization and Satisfaction] > [Maximize Satisfaction] ( optimization ).
- The bottom right figure shows the optimized settings. When pH=8.38, Column Temp=40.98℃, and ACN=30%, the predicted Resolution value is 2.45 and N of peaks is 7.39, which meets the response specifications.
The JMP Contour Profiler allows you to visually identify the region of factor space that meets the specifications ( design space ).
Below are the contours of the optimization conditions.If Column Temp is fixed at 40.98℃, pH is plotted on the horizontal axis, and ACN is plotted on the vertical axis, the white contour area is the range where the specifications are met. This area also includes the optimal value obtained by [Maximize Satisfaction].
When constructing a design space, a confirmation experiment is conducted near the optimal conditions (an experiment is conducted under the optimal conditions to confirm whether the obtained response value meets the specifications). If there are no problems as a result of the confirmation experiment, a rectangular area within the white area at the bottom right of the above contour plot, taking into account the specific technical requirements, is considered to be set as the design space.
Reference Information
■New User Case: Ono Pharmaceutical Co., Ltd. (Example of JMP usage for analytical method development)
Ono Pharmaceutical Co., Ltd. | JMP
■Basics of Design of Experiments (DOE) using JMP and examples (YouTube)
by Naohiro Masukawa (JMP Japan)
Naohiro Masukawa - JMP User Community
