The Decline of Six Sigma

Published on: 

Pharmaceutical Technology Europe

Pharmaceutical Technology Europe, Pharmaceutical Technology Europe-12-01-2003, Volume 15, Issue 12

The Six Sigma approach would appear to be ideally suited to pharmaceutical processes, yet the industry has been very slow to adopt it. This article looks at the possible reasons for this, and suggests an alternative methodology that takes advantage of Six Sigma tools and techniques, backed by good statistical principles.

It has recently been suggested that the adoption rate of Six Sigma has started to slow. The term "Six Sigma" is used to describe processes that operate at the highest quality levels with minimum variation and producing less than four defects per million opportunities. This low rate is achieved through applying statistical techniques and other quality improvement tools in a formalized way.

The "Sigma" referred to is the standard deviation of a quality characteristic (such as hardness, impurity or friability) against a target value. Figure 1 shows a Three Sigma process that is on target and producing approximately three defects per thousand. The accepted statistic that companies quote to reflect their process performance is Cp, a process capability index, which is defined as

for normally distributed data. In Figure 1, Cp equals 1.

As processes are rarely centred on their target value, a second statistic, Cpk, is used to measure the degree to which the process is off target. For normally distributed data the Cpk statistic is defined as



whichever is the smaller.

Six Sigma processes achieve a Cp of at least 2 and a Cpk of at least 1.5. This means that the specification range represents 12 times the standard deviation of the process. A Six Sigma process is illustrated in Figure 2. In the pharmaceutical industry the upper and lower specification limits are often defined in the relevant pharmacopoeia and have to allow for lifetime degradation.

Motorola coined the term Six Sigma more than a decade ago, defining the target capabilities for processes, the standard set of quality tools to use in pursuit of the goal and the way in which they should be applied. Many manufacturing companies use Cp and Cpk as key measures in production, but few pharmaceutical companies apply the Six Sigma methodology during development to fully realize Motorola's vision.

It is interesting to speculate why Six Sigma has not been whole-heartedly endorsed by the industry, particularly in research and development (R&D). Motorola's goals - to reduce product variation to a bare minimum and achieve specific quality targets - also apply to pharmaceutical companies. The tools and techniques used in Six Sigma are ideally suited to manufacturing, and with its emphasis on strictly meeting targets and specifications, the approach seems particularly relevant to the pharmaceutical industry.

Figure 1: Data from a process with a standard deviation of approximately 0.1. Cp is approximately 1, resulting in 3 defects per thousand opportunities.

Six Sigma has now spread far and wide, and is used in many service industries as well as in manufacturing. Some large manufacturers, such as General Electric, were quick to adopt the technique, but now service giants such as American Express and Starwood Hotels also have major programmes. So why does the pharma- ceutical industry remain sceptical?

In the author's opinion, the teaching of applied statistics to science students in higher education is often inadequate, so it is left to employers to give their bench scientists and technicians training in basic statistical skills. And although employees of service companies may be expected to have less experience in using an analytical approach to process improvement, they seem willing to adopt the methods.

Regulatory agencies have been increasingly pressurizing drug companies to understand their manufacturing processes better. A recent report from the US Food and Drug Administration (FDA) shows that the inability to validate processes is the major reason for withholding GMP approval. A greater emphasis on statistical techniques, such as design of experiments - key to Six Sigma - would improve process characterization.

Another reason why Six Sigma receives little attention could be that supplies of active pharmaceutical ingredients (APIs) are often low, early in development, and batch production costs are high during scale-up, so there may not be sufficient data to measure Cp and Cpk. However, all this may have to change as regulatory authorities apply more pressure on drug companies to understand their processes better before new drug application (NDA).

The problem may lie in the DMAIC methodology that is most commonly used by companies implementing Six Sigma. The DMAIC acronym comes from the five main steps involved - define, measure, analyse, improve and control. The names given to the steps in the methodology highlight the main goal of each. Within each step there are a number of objectives, activities and deliverables, along with a toolbox of statistical and quality techniques to help achieve the goal. The model is not rigid, giving improvement teams room to adapt it to their own situations.

Figure 2: Data from a process with a standard deviation of approximately 0.05. The Cp is approximately 2 and the Cpk approximately 1.5. Despite being off target this process will produce less than 4 defects per million opportunities.

The DMAIC model begins by identifying a problem - such as poor uniformity or an unwanted polymorph - and then sets about fixing it. Once the problem is resolved controls are put in place to make sure it does not happen again. The approach is useful for well-defined processes that are non-conforming. For example, it could be applied to a plant which has been producing a generic drug for some time but where improvements can still be made. Similarly the approach could be used to improve efficiency and rectify problems in an accounts department where invoice processing is well understood.

Drug development in the pharmaceutical industry, however, is an evolving process. As a drug moves from early formulation through scale-up and the production of pivotal batches to submission, the processes involved often change - sometimes dramatically. This is equally true for pharmaceutical development, chemical development and analytical development. So managing change and moving towards the ideal process is more important than correcting static processes, and the DMAIC model is less useful.

"Process," for the purposes of this article, refers to a processing step or small group of steps during which the raw materials are chemically or physically changed. Examples are a blending stage for dry powders, or a chemical stage consisting of a number of operations including crystallization, filtration and drying.

Figure 3 outlines a valuable approach, which takes a life cycle view of process development and deployment but has at its core the achievement of Six Sigma performance. It is more prescriptive than DMAIC, but at the same time takes advantage of all of the Six Sigma tools and techniques, and is backed by good statistical principles.

Figure 3: Using Six Sigma techniques for pharmaceutical development.

This methodology owes its existence to work done at IBM in the early 1990s, when process engineers were tackling similar manufacturing problems. The approach is to map a course for applying tools and techniques, beginning at formulation and ending in volume production. Despite its origin in the semiconductor industry, the ideas have been used in several pharmaceutical companies.

As with the DMAIC model, there are clear goals for each of the 12 steps and a set of quality tools available. Process targets are set early on, and by the third step key process measures, targets and measurement capabilities defined. From the fourth step, the process is being modelled and decisions made regarding which process quality characteristics and parameters are most influential. Potential problems in manufacturing are being predicted and addressed from the fifth step, and from step 6 the process is being assessed for control. Initial measurements of Cp and Cpk are being made within confidence limits determined by the volume of data available.

Control chart design begins once technology transfer is approved. The charts are finalized and implemented with a full statistical process control (SPC) strategy once sufficient data are available. Documentation of all activities is key to the long-term success of this approach, and ensures against warning letters from the regulators.

Unlike the DMAIC model, this methodology allows for several iterations during the development and scale-up of the process, as indicated in Figure 3. Outputs from each step provide inputs for the next step, allowing data-driven decisions regarding the process to be made. The methodology provides for the most efficient and consistent use of the tools and techniques of Six Sigma whilst ensuring the process under development takes the critical path from discovery through to manufacturing.

The methodology outlined here can be readily integrated with in-house development procedures to help embed the Six Sigma and design for manufacture culture in the organization. The technique is equally applicable to equipment development and manufacturing as it is to drug development.

An increase in statistical awareness within the scientific and management communities may still be needed to get the most from this way of working. A "just-in-time" training programme built around the application of the methodology to real processes can deliver dividends in terms of successful development, training effectiveness and risk management. The latest software packages allow scientists to use statistics in a way they never could 25 years ago, and a few are now providing support for the application of methodologies.

The potential return for a relatively small investment in training is very high - an important consideration while pressure on companies from external bodies continues to grow.