O’Keeffe et al (2016) are critical (disapproving) of the current binary approach to classifying process parameters and quality attributes as being critical or noncritical. It is too simplistic to adequately reflect current science and risk-based approaches to product quality. The authors note how embedded the term “critical” has become in GMPs and guidances, but question the difference between criticality and importance. They advocate a more realistic “spectrum of importance” with respect to process parameters and material attributes. This is of particular relevance for excipients, which in Europe are subject to formalized excipient risk assessment, as of March 2016 (European Commission, 2015).

Criticality
In English the term “critical” has several meanings, which may cause confusion, even before translation into other languages. Merriam Webster (2016) lists several relevant definitions:

• Crucial, decisive, a critical test
• Indispensable, vital, a component critical to the operation of a machine
• Being in or approaching a state of crisis, a critical shortage, a critical situation
• Of, relating to, or being a turning point or specially important juncture, a critical phase.– a state in which or a measurement or point at which some quality, property, or phenomenon suffers a definite change, a critical temperature

All excipients are critical (indispensable, to the formulation) but not all are critical (crucial, to the design, or finished product performance). The ICH (2009) requirement that “at a minimum, those aspects of drug substances, excipients, container closure systems, and manufacturing processes that are critical to product quality should be determined and control strategies justified,” begs the question, how important to be decisive? Inappropriate selection and control of excipients will lead to a critical situation (state of crisis).

MIL-STD-1629A (US Dept Defense, 1980) uses the word ‘critical’ as a severity classification, one stop short of catastrophic, and defines criticality as a relative measure of the consequences of a failure mode and its frequency of occurrence. A similar approach was adopted by ICH (2009)together with the additional criterion of detectability. Mcfarland & Waldron (2015) criticise reliance on detectability because, “in the absence of knowledge management and a science- and risk-based rationale supporting monitoring programs, undue emphasis may be placed on implementing detection controls rather than directing resources towards preventing failure in the first place”. The authors concern was to avoid Quality by Inspection, but it could be argued that Process Analytical Technologies (PAT) enable process control and even Real Time Release (RTR) because of increased detectability.

ICH (2009) did not include the last definition of critical, as relating to, or denoting, a point of transition from one state to another. The criticiality (importance) of an excipient may suddenly change if a finished product criticality (critical transition) is encountered during the lifecycle.

“Excipients are often divided into critical and non-critical categories, the latter receiving less attention. Such arbitrary classification runs the risk of surprises if subsequent experience invalidates the assumption of non-criticality.”

If the choice of materials is critical (of the essence) for product safety, how may excipients be successfully factored into product design. Determination of the criticality (importance) of an excipient is critical (essential) but, as suggested by O’Keeffe et al (2016) this will be a continuum rather than a binary classification. If every excipient is treated as critical then there is the ‘potential for Quality Risk Management (QRM) to degenerate into a non-value added exercise of identifying noncritical, improbable, low risk scenarios indefinitely.’

How Critical?
Kano analysis (Matzler & Hinterhuber, 1998) can be applied to provide a more structured approach to excipient incorporation in formulation design:

• Must-be requirements
• Performance requirements
• Surprises

Must-be requirements
Compliance with Pharmacopoeial specification, manufacture under GMP and a secure supply chain are all examples of basic or must-have requirements, critical (essential) to excipient selection, regardless of the finished product design. Such requirements are also known as “entry tickets”, or minimum standards, to operate in the Pharmaceutical market. Unfortunately, compliance with specification by itself ensures neither GMP nor supply chain security, hence the European guidelines on formalized risk assessment for ascertaining the appropriate GMP for excipients (European Commission, 2015), and corresponding US requirements under FDASIA (FDA, 2012). Reliance on specification alone opens the door to economically motivated adulteration. Non-compliant materials should be rejected by the quality system. Compliant materials are actually a greater and more insidious threat. Melamine was added to pet foods and infant formulae to bolster the apparent protein content, as specified by a test for nitrogen content in lieu of an actual protein determination (FDA, 2009).

There may be additional basic requirements dependent on route of administration, such as absence of endotoxins for injection use, which in turn imposes additional excipient GMP requirements. Chemical compatability with the API is another basic requirement, regardless of the finished product design. If an incompatability is dependent on the source of the excipient, it is likely that the reactants may be trace components, which are not always specified.

In terms of these basic requirements all excipients are critical because non-compliance, contamination, or incompatability can render the finished product unacceptable. Such requirements tend to be taken for granted by users, with extreme dissatisfaction if not met, but with little reward for quality beyond compliance with specification.

Performance requirements
Performance requirements are those excipient attributes which govern finished product quality in terms of performance. Patient-centric product performance is specified by the Critical Quality Attributes (CQAs), as formalized in the Quality Target Product Profile (QTPP). Rather than arbitrarily classifying the excipients in a formulation as critical or “non-critical” a better approach is to ask to what extent are they design critical?

Design-critical implies a direct cause-effect on finished product performance where the level of the excipient is titrated above a minimum, or to an optimal level, for specific performance in the finished product. Examples of design-critical excipients include modified release polymers and suspending agents, which control the CQAs of release and content uniformity respectively. At the other end of the spectrum a filler-diluent should have no impact on the CQAs, and therefore is not design- or performance critical. In practice the distinction is not absolute. Between the two extremes of a dominant functionality and no impact, other excipients in the formulation may interact and modulate finished product performance. Their design criticality will vary along a spectrum of importance.

A filler-diluent in a tablet formulation, affecting compactability and release, is more performance-critical than the same filler used to stop mini-tablets rattling inside a hard gelatin shell. Magnesium Stearate in tablets is more performance-critical because of side effects on dissolution rather than its function as a lubricant. (As a process aid it has no function in the finished product).

If design-critical excipients are titrated into the formulation to deliver specific performance it follows that excipient variability could impact performance. Two questions must be answered:-

1. What is the concentration-response relationship for that excipient in that formulation?
2. What attribute(s) of the excipient govern performance, and how variable are they?

If the excipient use level corresponds to a steep part of the concentration (x)-response (y) curve, then performance is sensitive (dy/dx) to precision of dosing and content uniformity of that excipient. Performance is also likely to be sensitive to variability in that excipient. Similarly, if there is a minimum effective level for that excipient, operating close to that minimum has a higher risk of impact from excipient variability. Near the minimum effective level the concentration-response curve may be a step function (dy/dx ® ¥) characteristic of percolation, or threshold, effects.

For example if a matrix sustained release polymer gives a desired release profile close to the minimum feasible level of usage it is better to look for a “faster” grade which can be used at a level higher than the minimum feasible but still give the desired release profile. For a given release profile, using the smaller amount of the “slower” polymer (with minimum feasibility) is likely to make the formulation more sensitive to variability in that polymer than using a higher amount of the “faster” grade.

Having selected an excipient to deliver a functionality, what attributes of the excipient govern its performance. What are the Critical Material Attributes (CMAs)? There may not be a CMA on the official specification, usually Pharmacopoeial. For example Avicel® RC591 is a structured vehicle former used in pharmaceutical suspensions to ensure content uniformity, a CQA for suspensions. As implied by its British Pharmacopoeial title (“Dispersible Cellulose”) the particle size is not a CMA. The specified viscosity also cannot be a CMA for two reasons. Firstly viscosity is a liquid property and therefore never a CQA in terms of suspending power. Secondly the specified (apparent) viscosity is measured on a de-structured sample and is therefore not predictive of structured vehicle performance. In the absence of a CMA an alternative design approach is to consider the Critical Formulation Attribute (CFA), in this case Rheology, and assess the extent to which variation in the CFA can impact on the quality of the drug product (FDA, 2009). In this case the CFA is both formulation and process dependent.

Excipient viscos parent dilute-solution viscosities of no relevance to applications beyond low level use as thickeners. Fu et al (2010) found viscoelastic properties of sodium alginate at higher concentrations to be more predictive of performance as a controlled release matrix former than the specified dilute solution viscosities, noting that “polymeric excipients are often the least wellcharacterized components of pharmaceutical formulations.” Performance requirements of an excipient for a particular application should be explicitly specified, both to avoid regulatory requests to justify reliance on official specifications, and to avoid surprises if the suppliers are unaware of the requirements. If possible the design should also utilize design-critical excipients at levels where the sensitivity of performance to concentration is mimimised.

Surprises
Surprises, also called exciters, are those product criteria which have the greatest influence on customer satisfaction and are neither explicitly expressed nor expected by the customer. When eventually specified they become performance requirements. Surprises can be beneficial where increasing raw material experience and insight drives process efficiency, higher yields or enhanced finished product performance. Unfortunately most excipient-related surprises are negative, reflecting design uncertainty due to the complexity of the raw materials and the products into which they are formulated (Carlin, 2016). Risk applies when the odds can be calculated, but with uncertainty there is no information to set the odds in the first place (Knight, 1921).

Often, it is variability in a so-called “non-critical” excipient, which causes a surprise when it suddenly correlates with an out-of-specification (OOS) or out-of-trend (OOT) excursion, a special cause variation. The variability may be within norms for a known attribute. The excipient variability may not have caused the problem but is now controlling the transition in and out of acceptable finished product quality. Two preconditions for such indirect excipient impact are a finished product criticality and product or process drift. Criticalities, a point of transition between two states, should always be anticipated in complex products such as pharmaceuticals. Such product weaknesses or latent defects may not be seen during development and come to light as a result of cumulative changes during the product lifecycle. Individual changes, usually subject to univariate change control, do not affect the CQAs but may allow system drift. Eventually, one change too many reveals the interaction of an excipient variability with a criticality. What was previously “non-critical” is now critical in controlling the critical transition between states.

Surprises are inevitable in complex systems with too many degrees of freedom. For excipients the answer is to anticipate application-specific failure modes and design the Control Strategy accordingly. Designing for failure will make the product more robust, especially when compensatory mechanisms are built in to address raw material variability. The fixed processes and formulae so common in pharmaceutical products are rigid systems through which raw material variability can feed forward to impact CQAs.

Manufacturability
Manufacturability should always be taken into account in product design (FDA, 2009) but treated separately from the CQAs in the QTPP. Critical (one step short of catastrophic) manufacturing problems are self-limiting. Only manufacturing with zero, minor or marginal severity defects will yield commercial product. If manufacturability is included as a CQA then the number of potential CPPs and CMAs increases and focus is lost on what is critical for performance in the hands of the patient.

The distinction is exemplified in the QbD Sampling Guide (IPEC Americas, 2016) where bracketing with multiple grades can be used to widen the range of material attributes in order to gain a higher degree of understanding of raw material impact. The criterion is not absence of impact but more specifically “no unanticipated effects of bracketing with readily available grades on the performance of the process or finished product”. For example if you substitute Avicel® PH102 with Avicel® PH101 in a high speed direct compression it should come as no surprise that you have to slow the tablet machine down. If it does not affect the CQAs, such as content uniformity and release rate, then the particle size distribution is not critical to the QTPP, even if critical to the flow of the compression mix (by design).

The quality of manufacturing, as opposed to the manufacturing of quality products, is better controlled through the Quality Metrics initiative (FDA, 2015).

This is currently specified in terms of lot rejection rate (FDA, 2016) but a quality culture will seek to control excipient-related special cause variation before an out of specification excursion. Most excipient-related impact on manufacturability will not be known at time of filing.

A critical look at excipient criticality
Applying the simple binary logic of critical vs non-critical to excipients can lead to inconsistencies. Can a “non-critical” excipient have a CMA? Logically a CMA governing finished product performance would make that excipient critical, but, as illustrated by the earlier discussion on excipient surprises, CMAs can arise during the product lifecycle due to drift and interactions. If what was originally considered non-critical suddenly becomes critical the question is a dynamic lifecycle-management issue, rather than an initial static assessment. Continuous monitoring of impact from all excipients throughout the lifecycle is more important than a one-off arbitrary binary classification during development.

Must a critical excipient have a CMA? The question assumes a linear one-to-one relationship between an excipient attribute and finished product performance which is not always the case given the multi-functional nature of many excipients and the number of interactions in complex systems. As originally envisaged by ICH (2009) would a CFA be a more appropriate level of control where there is not a simple correlation of product performance with an excipient CMA? The biggest problem in assigning a CMA is the tendency to pick an attribute on the official specification, without critically (careful judgement) assessing how relevant that attribute (and associated measurement methods) is to finished product performance (fitness-for-purpose). Failure to specify the material attribute actually controlling a CQA is akin to a skater using the coefficient of friction to measure the thickness of the ice (Consistency and compliance, up to point of failure without warning).

The development of a spectrum of importance advocated by O’Keeffe et al (2016) is a better way to handle quality attributes (including raw material attributes), instead of classifying the excipients as critical or non-critical. If the relative importance of the excipients (to both CQAs and manufacturability) is understood then qualification and validation efforts can be prioritized accordingly to reduce variation and to control the risks of producing a product that is not of the required quality. The authors note that this approach is consistent with the FDA Process Validation Guidance for Industry (2011), which saw criticality as a continuum rather than a binary state. The importance of all attributes and parameters should be evaluated for impact, and re-evaluated as new information becomes available.