Identifying and setting specifications to establish the quality dimensions of a drug substance represents an important aspect of project management. With significant cost and time-to-market pressures, a well thought out strategy regarding a product's specifications can have huge implications on ease of manufacturing and, ultimately, on the product's financial return. The cost of measuring numerous specifications on a commercial product is significant. An evolving specification "laundry list" generated during the development process may become "engraved in stone," incurring unnecessary costs year after year during the product's life cycle. Surely, this is an undesired situation in today's cost-conscious marketplace. Therefore, an important project management goal is to establish material acceptance specifications that clearly assess the API's quality as it is released from a validated manufacturing process.
From exploratory medicinal chemistry to clinical trial material and ultimately on to commercial production, the specifications that define the expected quality of the target molecule should be an evolutionary process. For instance, the specifications for a promising lead candidate must ensure proper molecular characterization and identification of the target. Typically, structure elucidation techniques of proton and carbon NMR, crystallography, mass spectroscopy and elemental analysis may be used. Further, the impurity profile throughout the entire development process should receive close attention from the project's stakeholders. The goal of this exercise is to insure the conclusions drawn from costly toxicological, safety and clinical studies are attributable to the target drug's characteristics and not to impurities. However, as the new drug candidate progresses toward commercialization, many of these specifications can be exchanged for those indicating manufacturing capability. Of course, the evolving specifications must adequately insure that the chemical intent and integrity of the API is not compromised.
For pharmaceutical companies wishing to outsource API production of clinical trial material or to initiate validation batches for commercial manufacturing, a specification and accompanying test method review is warranted. This review may have a number of implications. For a mature and formulated drug product, a Pre-Approval Supplement or a Change Being Effected in 30 days (CBE-30) filing must be submitted 1 to the FDA, but the upside potential from cost reductions in API manufacturing warrants the activity.
Conversely, for a clinical trial material, a specification assessment before scale-up can have a substantial net present value when the accrual of analytical costs over years of production are included in the manufacturing total costs. The analytical demands on the contract manufacturer can be contained and the client receive long-term cost benefit if the specifications representing molecular characterization are separated and removed from those suitable as manufacturing specifications. Ideally, manufacturing specifications should be based on batch-to-batch variability using well documented standards for the API and impurities. Obviously, using documented standards and a validated analytical method allows avoidance of costly analytical techniques during manufacture release testing merely to reassert structure confirmation—an activity defined and well covered during the development cycle 2,3 . The benefits accrued from separating structure elucidation specifications from manufacturing specs allows the contract manufacturer's plant management to focus on process robustness and predictability 4,5 (basic essentials of process validation) and to maintain these features throughout the API's lifecycle. Clearly, manufacturing API specifications directly related to consistent, high quality API delivery is everyone's goal. To that end, let's examine some observations and aspects of specification development, its impact on process validation and how these issues relate to the contract manufacturer.
Specifications and Natural Variation
It is very important to recognize the impact of natural (random) variation on a long list of specifications 6. Simply stated, nature tends to work against you as the cumulative variation from testing each specification is combined (a stacked tolerance). Unfortunately, many can testify to this observation by their participation in the well-intended, but often tedious Out-of-Specification (OOS) investigation process. The OOS investigation and follow-up can be cumbersome and costly. Clearly, OOS situations contribute to variable cost variances and have a negative impact on the financial health of the manufacturing process (as discussed in the cited article). Often, an OOS situation represents natural variation that could have been avoided by the thoughtful and responsible selection of specifications, and the conclusions drawn from the subsequent process validation study. An understanding of a process's natural variation as measured against the product specifications should be one of the desired outcomes from the process validation activity. This understanding can ultimately have a respectable financial payback7.
It is rumored that setting product specifications is the leading cause of project managers seeking professional psychiatric treatment. Few other topics will bring the armchair philosophers to their feet in tirades of dogma as they wave a map to the moral high ground. However, adopting a rational approach to setting manufacturing specifications can provide more predictable results and provide a data driven basis to process control.
A number of scenarios can occur when specifications are not developed as an integral whole or in a self-consistent fashion. For instance, a combination of analytical techniques used to measure a given parameter should be shown to yield comparable results with an expected variation. Perhaps an obvious conclusion would be to eliminate one of the methods, but often specifications added to maintain current GMP compliance makes it difficult to remove the original spec and its accompanying test methodology. To illustrate this, a drug substance specification list containing both an (original) titrimetric assay and an (updated) HPLC assay with analyte specificity is probably a non-value added redundancy. Similarly, a TLC impurities test remaining on a drug substance specification sheet after a validated HPLC impurity analysis has been implemented should be avoided if possible. It's well known (okay, so I'm exaggerating here) that performing the TLC test with each batch is significantly cheaper than filing a Supplement!
Another example is the situation where related specifications are not centered to one another. If two (or more) related specs are not centered (or aligned), the "approval" range or overlap artificially reduces the specification range of the individual specs. An example of this is the amount of allowable enantiomer in a HPLC chirality assay and the relationship to an optical rotation specification. In this case it may be possible for the process to deliver material that meets one specification but not the other(s). This situation necessitates that the process include a search and sort mechanism that "inspects quality into the product." Clearly, such an approach to meeting specifications is unacceptable and would not represent a validated process 8.
Conversely, well-characterized standards and their validated analytical methods are intended to, and easily obviate, the need for lengthy specification sheets. Establishing the identity, assay, impurities and confirming that the process itself did not contribute to undesired materials should be the goal of consistent manufacture of the drug substance. For manufacturing purposes, a logical set of specifications that represent the API and are based on validated methods and standards should be sought. Again, the steps of molecular characterization activities should be separated from manufacturing specifications.
One rule of thumb for setting specifications is to set the spec range at ± four sigma for any given, independent specification. The steps to this are to: 1) determine the natural variation of the process (including measurement systems) to ± three sigma, 2) question and challenge whether the product would meet other acceptance criteria if material was produced in the ± three-to-four sigma range, and 3) if so, establish an accepted specification range at four sigma. One criteria to this approach is that the measurement systems must be "in control and capable," as determined using a traceable standard. Secondly, it must be remembered that specifications do not drive performance requirements—just the opposite is true. However, setting specs that reflect an honest assessment of a process's natural variation, while simultaneously supporting the uncompromising delivery of a drug substance's quality aspects, has long-term financial payback to all parties.
Specification Goals and Variation
Simply stated, API specifications provide a quantitative means to evaluate product quality. Operationally, the manufacturing management desires a process sufficiently robust that they can produce material predictably and consistently within specification. Hence, the attention drawn to manufacturing six-sigma programs is justified for an effective means to demonstrate processes are in control. If process development activities addressed control factors sufficiently to move process averages (by design) and to understand the factors contributing to random variation, then the process and the specifications can be "centered." Perhaps the comments vide supra about establishing specification ranges is more apparent. A ± four sigma specification range based on ± three sigma process control limits provides a manufacturing system statistically capable of producing the drug substance correctly—every batch. This scenario represents a plant manager's utopia, but unfortunately, life is usually not this simple.
Often, product specifications and manufacturing control limits are not centered—at least not for every specification. At times, the statistical process control limits are equal to, or greater than the specifications. Instead of utopia, this keeps plant managers awake at night or on the phone arranging "technical coverage" to see the batch through. If a company culture rewards "fire-fighting" or the appearance of heroic efforts, it will also inadvertently support the "arsonist." Ultimately, higher productivity and good employee morale are achievable with manufacturing processes that are in control and capable. Amazingly, it all starts with the product specifications!
For the purists who are crying foul over this manufacturing-driven approach to product quality, it should be firmly stated that product performance in the application should drive product specifications; the reciprocal is not necessarily true. Phrased another way, there is no relationship between product specifications and process control limits. However, the financial return from the described approach is significant when off-quality costs are considered—OOS investigations, lost plant time or opportunity costs, lost raw materials, stock-out situations, etc.
From a project management perspective, it is important to understand when and how to set manufacturing specifications versus how to handle information obtained for molecular characterization purposes. Molecular characterization is a scientific exercise to confirm a material's structure. There is essentially no end to the tests available to "explore" the molecule's structure. In nearly all cases, molecular characterization is an R&D function. In contrast, manufacturing specifications indicate the acceptable tolerances of a material for its intended use when produced by a defined manufacturing process. Lastly, specifications do not represent a scientific investigation process but are used to confirm what is already known and expected from routine manufacture of the substance.
In R&D activities, routine characterization of complex molecules would probably include high field NMR techniques and the like—used to characterize reference markers and analytical standards. Using the analytical standards, validated analytical testing procedures can be established 9 to confirm identity, assay and purity of the API produced in manufacturing. Further, the validated analytical methods are an integral part to establishing process validation and are often employed during the cleaning validation activities.
Obviously, the importance of analytical standards to manufacturing operations is significant. In fact, two financial benefits arise from well documented standards employed in manufacturing QC labs:
1) standards obviate the need for capital intensive, high maintenance, high budget,
sophisticated analytical laboratories attached to manufacturing facilities, and
2) the variable cost contribution per API batch from expensive molecular
characterization instrumentation is eliminated.
If these reasons still seem insufficient to the cautious project manager, consider the financial burden of retaining specifications emanating from the discovery laboratory (as opposed to manufacturing process development).
As a product matures and larger quantities are produced, actual specifications per batch must be tested on a statistical sampling of the batch. Often, this sampling is the square root of the number of containers required to package the batch, plus one. The analytical cost for release testing then becomes the summation of the number of tests times their associated costs, times the statistical sampling. Sure, there are ways to reduce this cost: One method is to validate testing on a composite sample of the statistical sampling; another way is to truncate the specification list to eliminate structure confirmation tests. A careful selection of specifications for API release testing essentially generates a financial annuity for the life of the product.
A few recommendations for establishing API specifications have been offered that can substantially reduce the manufacturing cost of a drug substance. Without question, choose specifications that represent the API's quality when produced by a validated manufacturing process using documented analytical standards. These specifications should, if possible, be divorced from molecular confirmation techniques since structure determination was accomplished long before the product reached manufacturing. Secondly, use a statistical basis to establishing the specification range based on the process and measurement capabilities of the production facility. This approach is dependent upon the caveat that performance requirements drive specifications, not what the plant can, or wants to, do.
Lastly, an emphasis on analytical reference markers and standards is well warranted. For instance, in the case of contract manufacturing, the sponsor anticipates a cost benefit to the outsourcing activity. For a new drug substance, standards may not have been prepared; for an older drug, updating documentation to cGMP guidelines may be needed. In either case, careful attention to preparing, characterizing and documenting analytical standards is an investment paying a dividend on every API batch produced. The project manager who effectively addresses these issues without compromising API quality contributes significantly to the financial well-being of their organization 10.
1. Guidance for Industry; Changes to an Approved NDA or ANDA; Food and Drug
Administration, November 1999. Back
2. Chemical Engineering; June 6, 1977; pp. 168 – 182: "Design and Analysis of
Industrial Experiments"; Thomas D. Murphy, Jr. Back
3. Chemical Engineering; November 1995; pp. 142 - 147: "Design of Experiments";
Kymberly K. Hockman and David Berengut. Back
4. Quality, June 1996, pp. 28 – 30: "Put ‘Process' Back in SPC"; Edmund S. Fine.
5. Chemical Processing's 1997 Project Engineering Annual, pp. 56 – 59: "Basic
Elements of Chemical SPC"; Jodi Kay. Back
6. Contract Pharma, October 2000, pp. 54 – 64: "Improve Financial Performance from
Statistical Process Control: Six Sigma Programs Are a Valuable Addition to a
Manager's Toolbox"; Cliff R. King Back
7. Chemical Market Reporter; July 16, 2001; pp. fr10 – 12; "Six Sigma and the
Bottom Line"; Pamela Sauer. Back
8. BioPharm, August 2001, p. 18: "Documentation: It's Not Just Paper Work"; Dian
Feuerhelm and Gregory S. Blank. Back
9. Pharmaceutical Technology, October 1999, pp. 166 – 129: "A Statistically
Integrated Approach to Analytical Method Validation"; Donald H. Weed, Jr. Back
10. Chemical Market Reporter; September 3, 2001, pp. 27 – 30: "EVA (Economic
Value Added) and the Chemical Industry"; John Ballow, Henri Perrson and Red