A Statistical Review of ICH Q10 Pharmaceutical Quality System

The International Conference on Harmonization ICH Q10 guideline, Pharmaceutical Quality System, and its two companion guidelines Q8 Pharmaceutical Development and Q9 Quality Risk Management, have been readily accepted if not fully implemented by the pharmaceutical industry over the past few years (1–3). Discussions of the statistical implications of Q8 and Q9 have appeared since theguidelines were harmonized (4). Little has been said, however, about the statistical content of the Q10 model, probably because it is perceived to be focused only on the management of the quality system. There are many Q10 recommendations that affect statistical issues facing the pharmaceutical industry, however, the guideline states that it is not "intended to create any new expectations beyond current regulatory requirements" (1).

Lynn D. Torbeck

Although no new statistics or sampling plans are explicitly required by Q10, it goes without saying that current regulatory requirements are, in fact, mandatory. In addition, cGMPs continue to improve over time and according to Q10, "Implementation of ICH Q10 throughout the product lifecycle should facilitate innovation and continual improvement and strengthen the link between pharmaceutical development and manufacturing activities" (1). That link should include the results of statistically designed experiments and related statistical and risk analysis.

While not explicitly requesting these approaches, ICH clearly implies that companies need to be proactive when it comes to corrective and preventive action (CAPA) programs. In today's environment, it is not sufficient to be reactive alone when problems occur. The Quality department must routinely seek out potential problems and prevent them before they result in rejects or recalls. For example, Q10 notes that companies should "Establish and Maintain a State of Control. To develop and use effective monitoring and control systems for process performance and product quality, thereby providing assurance of continued suitability and capability of processes" (1).

Having control over one's product and process is not a new expectation, although there is still confusion as to what a proper "state of control" means (4). It is not enough to ask for a state of control; the industry must provide and define additional modifiers. There are several ways in which a process can be in a state of control or, conversely, in a "state of out of control."

A process can be in control, for instance, for financial and accounting, for regulatory compliance, and for organizational and managerial control. These forms of control are usually assumed to be in place. There are two other states of control that are germane to statistics: engineering and statistical.

A process is said to be in a state of engineering control when the process can be changed and adjusted using control knobs and/or by setting the critical process parameters (independent variables) that affect the dependent responses (5). When in control, the product always meets its specifications even if inconsistent and erratic. Time plots with specification lines are used to monitor the process. A process is said to be out of engineering control when it fails to meet its specifications.

A process is said to be in a state of statistical control when the process has been designed, developed, and adjusted to produce product that, while still containing some variability in the critical quality attributes (dependent variables), is predictable in that variability over time. Statistical control charts are used to monitor the process. A process is said to be out of statistical control when it fails one or more of the eight Western Electric control chart rules (6). As Q10 notes, "The pharmaceutical quality system should include the following elements, process performance and product quality monitoring, corrective and preventive action, change management, and management review" (1).

Product quality monitoring can be interpreted as trending the critical quality attributes. Again, proactive CAPA is preferred to reactive CAPA. As Q10 highlights: "Advocate continual improvement" (1). This continual improvement should include proactive variability reduction.

Also recommended in Q10 is: "... a written agreement between the contract giver and contract acceptor." This agreement should include the acceptable quality limit (AQL) and limited quantity (LQ) limits for incoming sampling plans as well as the usual specification methods and acceptance criteria. Data collected in-coming and in-process can be used to determine compliance with a contract agreement. Per Q10, "Throughout the product lifecycle, companies are encouraged to evaluate opportunities for innovative approached to improve product quality" (1).

There are many ways, statistically, to achieve this goal. Trending, designed experiments, variability reduction, and design space are just some of the tools that can be used to make process improvements.

Many of the terms in ICH Q10 imply trending of critical parameters and attributes. It is a given that this must be done. Q10 states: "An effective monitoring system provides assurance of the continued capability of processes and controls to produce a product of desired quality and to identify areas for continual improvement" (1).

Process capability is measured by comparing the variability of the product/process to the width of the specification range. This comparison can best be achieved using statistical tolerance intervals because they take into account the sample size where Cpk and Ppk do not. Per Q10, "Identify sources of variation affecting process performance and product quality for potential continual improvement activity to reduce or control variation" (1).

Some Six Sigma programs have gotten a poor reputation in certain circles because of a single-minded focus on saving money as opposed to giving equal consideration to improving quality and reducing variation. It is the author's opinion that management needs to give equal attention and resources to both. As Q10 calls for, "Proposed changes should be evaluated by expert teams contributing the appropriate expertise and knowledge from relevant areas (e.g., Pharmaceutical Development, Manufacturing, Quality, Regulatory Affairs and Medical), to ensure the change is technically justified" (1). The company's statistics department or a statistician should be included in the team.

Many statements in ICH Q10 have important implications for the correct and consistent use of statistics in the day-to-day implementation of pharmaceutical quality systems. Addressing these harmonized recommendations proactively and in context can help to strengthen one's quality system and thereby reduce rejects and recalls.

Lynn D. Torbeck is a statistician at PharmStat Consulting, 2000 Dempster, Evanston, IL 60202, tel. 847.424.1314, LDTorbeck@PharmStat.com, www.PharmStat.com.

References and notes

1. ICH, Q10 Pharmaceutical Quality System (2008).

2. ICH, Q8 Pharmaceutical Development (2009).

3. ICH, Q9 Quality Risk Management (2005).

4. L. Torbeck, Pharm. Technol. 35 (10) 46–47 (2011).

5. Note: Other definitions of Engineering Control exist in other industries.

6. Note: It is common practice to use only one to three of the eight Western Electric rules for a given control chart. It is counterproductive to use more than three rules at a time.