C. Burgess

 for analytical procedure validation and development mean for a regulated Good Manufacturing Practice (GMP) laboratory? Are they consistent with the approach taken by 

United States Pharmacopoeia (USP) <1220> on Analytical Procedure Lifecycle

 two documents to describe what USP can do in one?

? Why does it take two documents to describe what USP can do in one?

Good analytical science and ensuring data integrity requires that analytical procedures should be validated for their intended use (1,2). In addition, the applicable US Federal Drug Administration (FDA) Good Manufacturing Practice (GMP) clause is Title 21 of 

 211.194(a), which has been in the regulation since 1978:

The suitability of all testing methods used shall be verified under actual conditions of use (3).

 Chapter 6.15 on quality control (QC) requires validation of test methods:

Testing methods should be validated. A laboratory that is using a testing method and which did not perform the original validation, should verify the appropriateness of the testing method. All testing operations described in the marketing authorisation or technical dossier should be carried out according to the approved methods (4).

Like equipment (analytical instruments and systems) that must be fit for intended use, so must analytical procedures. However, how many readers define what the intended use of their analytical procedure must be?

These regulations apply equally to in-house development as well as pharmacopoeial methods. However, there was no regulatory guidance on how validation or verification was to be achieved. In this column, we take you on a journey covering guidance for the validation of analytical procedures over the past 30 years. The starting point for this journey is to look in the sin bin of non-compliances related to analytical results.

Does Your Analytical Procedure Generate Out of Specification (OOS) Results?

One of the continuing sources of schadenfreude with FDA warning letters and 483 observations is the invalidation of out of specification (OOS) results for a variety of unscientific reasons, such as “human error,” often without a scientific investigation. This results in the invalidation of the result and has resulted in the FDA generating a quality metric for any QC laboratory of percentage of invalidated OOS results (5) as well as a revision of the Guidance on investigating OOS results in May 2022 (6).

Why are there many OOS results in some laboratories? After all, pharmacopoeial procedures always work as written, right? Wrong! Could a laboratory have rushed to validate a hastily developed procedure with the result that a poor procedure most probably is the root cause of OOS results? Therefore, we need a robust method development phase of any analytical procedure to have a sound foundation for validation and ensure that OOS results are rare.

ICH and USP Approaches to Procedure Validation

We focus mainly on publications from two organizations—United States Pharmacopoeia (USP) and International Conference on Harmonization (ICH) for analytical procedure validation. For those who are unaware, 

 was initially called International Conference on Harmonization of Technical Requirements for Registration of Pharmaceuticals for Human Use from its inception in 1990. It was a consortium of the regulatory authorities and pharmaceutical industry bodies from the United States, Japan, and Europe. The aim was to develop unified approaches to key regulatory topics so that the pharmaceutical industry only had to perform one activity and it was acceptable in the three regions. Since then, additional regulatory authorities have joined (for example, Brazil and South Korea), and the name of the organization has changed. Conference has been replaced by Council, but it is still known as ICH.

As you can see in Figure 1, the FDA have also published two draft and one final guidance document on validation for analytical procedures and methods, which brings us to an interesting question: procedure or method?

FIGURE 1: Timeline of Analytical Procedure Validation Guidance from USP, 

Analytical Method or Analytical Procedure?

It is important to understand what is meant by analytical procedures and if there is a difference with analytical methods:

Analytical procedures refer to the wholeprocess from sampling through to reporting the result.

Analytical methods usually only refer to the instrument portion of the analytical procedure or analytical principle (separation technique).

Therefore, analytical procedure is the preferred term because there may be problems when sampling (for example, light-sensitive analytes), during transport or storage (for example, unstable analyte), or during preparation before analysis, all of which will not be considered if you are focused on the analytical method unless you are the FDA who have bet on both terms.

In The Beginning 1: Initial Validation Guidance

 211 was published in 1978, there was no official guidance available on how to validate an analytical procedure; eventually, the USP issued an informational general chapter <1225> entitled “Validation of Compendial Methods” in 1989 (7).

In addition to general chapter <1225>, USP has also issued two more general chapters:

 <1226> Verification of Compendial Procedures (8), and

 <1224> Transfer of Analytical Procedures (9).

 <1225> served as the foundation for the development of 

 Guidance on Validation of Analytical Procedures, which was published in two parts: 

ICH Q2A Validation of Analytical Methods: Definitions and Terminology

 was published in 1994 and is the parent guideline, and

ICH Q2B Step 4 Validation of Analytical Procedures: Methodology

ICH Q2(R1) Validation of Analytical Procedures: Text and Methodology

 with no changes to the content of the combined document (10) shown in Figure 1.

Method development is ignored, which is problematic because this is vital. Cobbling a procedure together can result in variability contributing to OOS results as mentioned earlier.

 can be followed slavishly (for example, an analytical procedure for measuring assay between 90–110%) and often have LOQ and LOD limits that are determined in the validation. The reason? It is in 

! This is stupid as sound science has not been applied as required by 21 

 211.160(b) (3). We bet you didn’t know that a regulation required you to use sound science.

The focus is mainly on chromatographic analysis with little coverage of spectroscopic analysis.

In The Beginning 3: USP Lifecycle of Analytical Procedures 

 general chapter <1220> on Analytical Procedure Lifecycle (ALM) (11) that became effective on May 1st this year. Bet you didn’t know that either! 

 <1220> has had a long gestation period, starting in 2012 with the first Stimulus to the Revision Process (12) with further Stimuli articles (13–15) and two draft versions of <1220> (16,17) to support the development of the general chapter.

The lifecycle approach is now a regulatory expectation as it features in 

ICH Q12 Technical and Regulatory Considerations for Pharmaceutical Product Lifecycle Management

 is to move from quality by testing to quality by design (QbD), using knowledge, science, and statistical design to achieve more robust pharmaceutical products. QbD is defined by 

A systematic approach to development that begins with predefined objectives and emphasizes product and process understanding and process control, based on sound science and quality risk management (18).

The key points are: predefined objectives, product and process understanding, sound science, and risk management.

These four points are incorporated in the three-stage lifecycle of <1220> analytical procedures (11) is shown in Figure 2 that comprises:

Stage #1: Procedure Design and Development (PPD). PPD, or method development, is based on the analytical target profile (ATP) for the procedure. Within this phase of the lifecycle, the critical parameters are identified and risk-based ways to control them are implemented. A design space is defined and tested using experimental design software.

Stage #2: Procedure performance qualification (or method validation) of the final design space against the parameters for intended use of the analytical procedure as defined in the ATP and elucidated during the PPD phase of the lifecycle.

Stage #3: Procedure performance verification, which is an ongoing assessment of the procedure’s performance during operational use (for example, trending data and results over time). Any changes made within the design space are considered validated and can be made without formal change control. Changes outside the design space return to either phase 2 for qualification, or if major to revise the ATP with further development. This is the longest portion of the lifecycle and it is imperative that it is controlled and monitored to ensure verification against intended use.

FIGURE 2: The USP <1220> Analytical Procedure Lifecycle, from reference (11).

The key to a lifecycle approach for any analytical procedure is to define the intended use as required by the GMP regulations. This is achieved by writing an ATP to provide the foundation for all other work in the lifecycle. Get the analytical target profile (ATP) wrong, and the procedure is not fit for intended use.

 established an Expert Working Group (EWG) to revise 

 on Analytical Procedure Development (20). The aims of the EWG were as follows: 

 required revision as it lacked guidance for near-infrared (NIR) and Raman spectroscopy, especially using multivariate models, resulting in inadequate validation reports in regulatory submissions resulting in multiple questions being asked by regulators.

 needed a new guideline outlining the principles and scientific approaches for analytical procedure development.

Specifically, the feasibility to combine both documents into one for simplification and clarity should be examined.

ICH guidances go through a five-step process:

Step 1: Define the scope and drafting by the EWG.

Step 2: Issue the draft for public comment (the stage that 

Step 3: Review the public comments and update of the document.

Step 4: The final approved version of the document is released.

Step 5: The guidance is incorporated into a member’s national regulatory framework. This can vary from updating regulations or issuing a guidance for industry.

Let us now look at the contents of these two draft guidance documents. Avid readers should also read a review of these two draft documents by Teasdale and others (21), who have been involved in analytical procedure lifecycle development for a number of years.

The first thing to realize is that the expert working group (EWG) have failed to combine both documents to define, simplify and clarify a unified approach to lifecycle management of analytical procedures. This leave analysts in regulated laboratories juggling between two regulatory documents with differing approaches on the same subject. It is not a great start.

 (22) with the continuing failure to mention the most critical part of the lifecycle—procedure development. Given the pharmaceutical industry’s inability to change, this means perpetuating poor science. Furthermore, the failure to consider a lifecycle approach breaks the principles outlined in 

 is a tweak, not a substantive advance—for example, the issues mentioned about lack of validation in regulatory submissions are discussed in 

To be fair to the EWG, an argument is that they wanted to maintain the status quo, but this brings its own problems. Will analysts read 

 and integrate them together? Probably not. In contrast, that has not stopped USP <1220> managing to incorporate all phases in a single general chapter (11).

Let’s start with the major flaws in the draft (22):

define an analytical procedure lifecycle? No.

Does it define an Analytical Target Profile to document intended use of the procedure? No. 

 does mention “performance characteristics” as the nearest to ATP.

Method development is still conspicuous by its absence. This is a critical component of the lifecycle that has been missing throughout the various versions of 

There is no mention of validating the analytical procedures against the intended use as defined by an ATP. This is disappointing because it is the linkage between 

, and it is only shown in the outline in Figure 1. If these documents are to remain separate, they should be linked via the procedure ATP.

 is in the glossary, which is copied from 

The draft guidance jumps straight into a validation study for the procedure, assuming that development is not done, undocumented, or cobbled together. Given the fact that 

 both focus on lifecycle and the proposed 

 does not, it leaves us with the presumption that the authors are happy to accept a suboptimal and unscientific approach.

On the plus side is the expansion of the scope of 

Spectroscopic methods are covered but not in much detail as we shall see later.

Given the large number of biological products and biosimilars that are available and in development, biological assays (for example, polymerase chain reaction (PCR) and cell based and binding assays) are included in an expanded 

 draft. This has resulted in an expansion of calibration methods to accommodate them.

Good results require a good calibration model and the draft guidance includes calibration models that are linear, non-linear, and multivariate. Use any calibration method but keep it appropriate and use simplest model for your data.

Linearity as a validation parameter is only applicable now for linear calibration models.

In Annex 1, there is a good process flow diagram leading from the objective of an analytical procedure that then defines which parameters need to be considered during a validation (not forgetting development as well): specificity, range, accuracy, and precision. For example, a procedure for identification would just focus on specificity in contrast to an assay where all four would be determined. This is better than the table in the current version of 

 in terms of a lifecycle approach to analytical procedures but with a big gap—validation of the procedure which is devolved to 

Minimal versus Enhanced Approaches to Analytical Procedure Development

 discusses minimal versus enhanced approaches. The use of the term minimal is a concern because it is often interpreted as this is all we need to do. Only in the enhanced option is the ATP mentioned and this is defined as:

a prospective summary of the performance characteristics describing the intended purpose and the anticipated performance criteria of an analytical measurement (23).

The ATP defines what is wanted from the procedure but not how it is achieved. However, from the ATP comes the outline of the analytical procedure with knowledge of the analyte and matrix:

Any sampling and sample storage requirements

Instrumental technique used to analyze the sample

Any data interpretation or calculation required

A good idea is to define the ATP for any analytical procedure including the minimal approach. However, without an ATP, is the minimal approach scientifically sound (3)? No. Outlines of the minimal and enhanced approaches are shown in Table I (23). Apart from defining the ATP, the critical component of the enhanced approach is the definition of an analytical control strategy. Here, time is taken to identify the parameters that are critical to ensuring a robust analytical procedure and how they must be controlled. Knowing how a component can influence the outcome is essential. What is not shown in Table I is the iterative nature of the development process between the analytical control strategy and the results from experiments. As more information and knowledge about the procedure are generated, the procedure and control strategy can be refined. Risk management is used to identify critical parameters that may need to be carefully controlled and other that have a lesser or zero impact on the outcome of the procedure. 

Experimental design software can be used either standalone of controlling some analytical instruments to identify and define a design space where the procedure is robust. Having this information before the validation of the procedure as well when it is used operationally is essential as changes within the design space can be made without the need for revalidation. 

 section 10 deals with the submission of analytical procedures. One question that arises is why is this not in 

 as this is where the bulk of the data are generated? This is a good argument for merging the two documents.

One of the remits for the EWG for revising 

 was to include multivariate analysis methods. We want to focus on the development of and validation of spectral libraries used widely for identification of raw materials and active pharmaceutical ingredients (API) by various IR methodologies (NIR, Fourier transform (FT)-IR, and Raman) and how they are covered by the draft 

There is no clear link or explanation of “fit for use” for these methods. A discussion of what fit for use means would help define some of the vague concepts such as “enough” and “small.” This is a general problem with 

. In the revision of these guidances a useful discussion of how to demonstrate fit for use would be very valuable.

Section 8 discusses the development of multivariate analytical procedures, noting that multivariate models are constructed from samples (the bullet points added by us):

...a careful strategy for sample selection is essential for obtaining the relevant in- formation from the analytical data and contributes to the robustness of the resulting model.

Based on the method and measurement principle, the sample population should encompass the sources of variability likely to occur during manufacture and analysis, such as raw material quality, manufacturing process variability, storage conditions, sample preparation, and testing.

Use of risk assessment tools can help to identify sources of variability with the potential to influence the measurements and resulting model outputs.

Obtaining samples with appropriate variability at commercial scale can be challenging. Therefore, development laboratory and pilot scale samples are often utilized to provide enough variability to improve accuracy and robustness of the model.

Inclusion of commercial scale samples is recommended to capture variability related to specific equipment or processing conditions.

Careful consideration should also be given to sample distribution in the calibration and validation sets, as this will influence the model predictive capability.

The key to building an identification library is selecting enough appropriate samples reflecting the expected variation a manufacturer would expect (as noted in the second bullet point above), which can be aided by reducing the number qualified suppliers for each compound. However, finding detailed guidance about how to develop an identification library is lacking, the closest being using Raman to identify incoming glucose samples in section 13.3 Annex C: Example of Multivariate Model Lifecycle Components. However, if you want to identify a material from similar structured compounds, there does not appear to be any advice on building a library. However, there is still the excitement of validating the spectral library in 

Articles

A look at two ICH draft publications reveals that the EWG has much revision work to do for these publications to fully cover the validation lifecycle process.

C. Burgess

Articles

Quo Vadis Analytical Procedure Development and Validation?