Validation of liquid chromatography mass spectrometry (LC-MS) methods
10.3 Different ways to evaluate robustness
In the lecture 10.1 Robustness and ruggedness relation to LC-MS method development we saw different LC-MS parameters that influence and , as well as what the influence of these parameters. On the basis of this information it is possible to plan changes to the method. As a reminder, there was a list of LC parameters, sample and sample preparation parameters and mass spectrometry parameters.
Because of the very large number of potentially variable parameters it is reasonable to divide assessment of ruggedness into separate parts. A very logical division would be to test ruggedness separately for the sample preparation and for the LC-MS analytical part.
Before starting the investigation of robustness it is crucial to find out what are the most important performance characteristics of the method.
For example, if the method’s is very close to the LoQ required by legislation, then the changes in the LoQ value have to be monitored against small changes in the method parameters. The most influential method parameters impacting the LoQ could be MS parameters, mobile phase pH and sample preparation parameters.
The main criteria for choosing parameters are (a) how much a given method parameter can influence the critical characteristic and (b) how likely it is that this parameter will change uncontrollably.
If those parameters are chosen, then we should use one of two options to evaluate the method robustness and ruggedness – to use the experimental design or the One Factor At a Time approach.
Different ways to evaluate robustness
http://www.uttv.ee/naita?id=23735
https://www.youtube.com/watch?v=U1nchnq8TZE&feature=youtu.be
Example of the One Factor at the time approach
An Example of how an OFAT can be used is described in the table below. O represents here the optimal value, + and – are values larger and smaller than the optimal value. All three are chosen by the chemist in charge of the method development. For robustness study we change, as the name suggest, one factor at a time and monitor the response variable – the retention time – in our example. The actual experiments need to be performed in random order to prevent a on the response variable. After finishing the OFAT the responses can be used to calculate e.g. % deviation from the “optimal response” (experiment 7) and thereby deduct if the factor change has a significant influence on the response.
|
Experiments |
Actual |
A |
B |
C |
Response |
|---|---|---|---|---|---|
|
1 |
3 |
O |
O |
+ |
7.95 |
|
2 |
6 |
O |
O |
– |
8.13 |
|
3 |
5 |
O |
+ |
O |
8.12 |
|
4 |
1 |
O |
– |
O |
7.72 |
|
5 |
4 |
+ |
O |
O |
8.32 |
|
6 |
2 |
– |
O |
O |
9.82 |
|
7 |
7 |
O |
O |
O |
8.03 |
| *O Optimal factor setting | |||||
One Factor at a Time
https://www.uttv.ee/naita?id=32134
https://youtu.be/kBR703IK08A
In most cases experiments with one-by-one variations (One Factor At a Time approach) of the most important parameters are carried out.
Based on the common practice and for the sake of simplicity, we recommend the following if using OFAT:
- Change parameters one by one (One Factor At a Time approach) in both directions from the nominal (optimal) value. Changes in the parameters should be realistic in the context of normal use of the method.
- “Do not stop there!” Often parameters may be mutually unrelated (uncorrelated), but in some cases this does not hold. For example: change in mobile phase pH can decrease resolution between two adjacent peaks. Likewise, increase of the mobile phase flow rate can also lead to a decrease of resolution. While separately either of these two changes can still lead to insignificant loss of resolution, their occurrence together may lead to peak overlap. Whether this is the case, can often be determined by educated inspection of the effects of the changes (without additional experiments) and noting potential problems.
- Effects from the change of parameters should be recorded and if necessary, graphical or statistical analysis of the effects should be done.
- Regarding the robustness tests results, if necessary, measures to improve the performance of the method should be taken.
Alternative to OFAT, Design of Experiments (DoE), is somewhat less used, especially at routine laboratories, because these approaches require knowledge and experience with mathematical statistics. Furthermore, programming skills in R or Python are beneficial if the chemist is not willing to buy a software tool for the evaluation of the experimental designs. In this course we will give an overview of both – One Factor At a Time approach, which we covered here, and the Experimental Design (DoE) approach. For DoE we offer two approaches:
1. Shorter introduction in chapter 10.4.
2. Longer introduction in chapter 10.5 (this is bonus content, not mandatory to pass the course).
