By Julia O’Neill and Richard Burdick
If you have ever had a family member diagnosed with a disease such as cancer or a degenerative muscle or cognitive disorder, you likely have encountered barriers to accessing treatments — high cost, limited supply, delayed commercial launch, or restrictions on clinical trial enrollment. Time is the enemy. Delays in accessing treatment can result in life-altering disease progression or even death.
Many factors contribute to the delay in accessing treatment, but one factor is within our power to change — the way we set specifications for pharmaceuticals.
2 WAYS TO SET SPECIFICATIONS
The International Council for Harmonisation of Technical Requirements for Pharmaceuticals for Human Use (ICH) Q6A and Q6B outline the expectations for specification setting in broad terms. The ICH guidance states that a product conforming to specifications will be “acceptable for its intended use.” In practice, a set of critical quality attributes (CQAs) is selected and specification ranges are established based on what has been produced and tested in clinical trials. These ranges define typical performance for initial manufacturing. The implicit purpose of these ranges is to ensure that future production is consistent with initial production. These limits rest on an empirical assumption: What was good enough for patients during clinical studies should be good enough for future commercial material.
There is another way to set specifications. With Quality by Design (QbD), specifications represent the true requirements of a product to be used for its intended purpose. Think of jet fuel — the specifications on its chemical and physical characteristics should be set to ensure it can power an engine as designed, rather than to ensure future production of jet fuel is consistent with its initial production. If we find a way to improve the characteristics of the fuel in the future, our ability to update production methods will not be hampered by specifications set on what we achieved in initial production runs. This allows fuel manufacturers to make continuous improvements aimed at improving quality, lowering production costs, and increasing the volume and reliability of supply.
The QbD framework relies on two sets of limits, not just one. Specifications are firm limits on the product characteristics that ensure it is fit for its intended use. A product failing to meet the specifications should never be released commercially. There is also a second tighter set of limits that falls within the specification limits — statistical process control (SPC) limits. SPC limits are used to monitor process and product consistency. This monitoring is carried out in real time, while there is the best opportunity to identify trends, catch issues quickly, and track the results of process improvements.
Pharma developers and regulators have adopted a QbD approach and established specifications representing true requirements in some cases. The WHO and the International Alliance for Biological Standardization (IABS) have published an upper limit of no more than 10 nanograms per dose for the common CQA residual host cell DNA. These published limits define boundaries for how the product will perform (in this case regarding its safety) independently of its manufacturing history.
Specifications may be based on dose-ranging studies, which if specific to a product, could be an ideal basis for QbD specifications. However, the power of their information is typically limited by small study size. Also, these studies are unlikely to push the limits of safety and efficacy for ethical reasons, so the true boundaries remain undetermined.
DECOUPLING SPECIFICATIONS AND CONTROL LIMITS
Why are specifications and control limits so muddled, and is it possible to decouple them? Historically, our understanding of product characteristics and therapeutic action has been incomplete and unable to provide science-based specifications, such as those that define the characteristics of fuel needed to power an engine. Without this knowledge, we typically establish specifications as limits consistent with past process performance instead of true requirements from the patient’s perspective. We have confused the purpose of specifications and SPC limits by combining them into one set of limits, and we’ve lost the best features of both.
Today, through advances in analytical methods and precision medicine, new gene therapies may be very precisely targeted to replace or correct a specific defective gene. Although potency may remain challenging to quantify precisely, the action of the treatment is well understood, and the genetic material can be precisely identified.
Until recently, this depth of knowledge was out of reach, and pharmaceutical specification setting practice relied on experience gained during clinical trials and initial manufacturing. But this experience-based approach hits a wall when the limits of safety and efficacy are not explicitly challenged during clinical studies. Knowledge about true boundaries is imprecise. This empirical approach fosters a “tighter is better” mindset. Regulatory review of a new product submission usually involves a negotiation, with the manufacturer proposing wider limits and the regulators pushing for narrower limits — with limited knowledge of true efficacy or safety requirements.
This blurring of purpose for specifications has negative impacts that translate directly to patient experience. Specifications that are set too tightly based on initial manufacturing create regulatory hurdles to process improvement. Tight specifications may be challenging for manufacturers to meet, and they may cause supply disruptions and launch delays. Specifications based on an initial manufacturing site may impede the manufacturer’s ability to transfer product to additional sites, interfering with efforts to increase supply volume and reduce costs.
MOVING TOWARD PATIENT-CENTERED SPECIFICATIONS
Developers, regulators, and manufacturers should move toward more patient-centered specifications in two steps. With full understanding of therapeutic action and the related product characteristics, specifications should be determined based on the true requirements for safety and efficacy in patient use. As this deeper knowledge is being gained, we should challenge the risk-benefit assessment of specifications for new products.
Patients of breakthrough products have an even greater probability of negative impact because limits must be set with very limited process experience. Standard statistical approaches for setting specifications, such as using three-sigma limits, provide specifications that are much too tight when small data sets are used. Alternatively, a phased approach using limits that are updated to reflect the growing database of process experience can be used to manage this risk.
The Goldilocks principle is a useful paradigm for setting early and interim limits. Limits should not be too tight and not too wide, but just right. Statistical procedures such as tolerance intervals enable developers and reviewers to quantify probabilities of negative impact associated with choices of specification limits for critical quality attributes.
One helpful approach during the transition to QbD limits is to set initial specifications based on experience with other products in the same product family (e.g., monoclonal antibodies). When products are made on the same technology platform or share a product composition that is very similar except for perhaps the identity of an active ingredient, it may be possible to draw on prior experience with similar products to set limits for nonspecific attributes. Limits based on a family of products are more reliable since they draw on a larger body of experience.
The revolution in scientific knowledge that is bringing hope to patients with devastating diseases is also setting the stage for modernizing our approach to specifications and process control. Updating our approach to specification setting will accelerate the delivery of breakthroughs to patients.
JULIA O’NEILL is the founder and principal consultant of Direxa Consulting, LLC. She works with companies developing novel treatments including vaccines, gene therapies, oncolytic viruses, microbiome-based products, and other therapies derived from biological starting materials.
RICHARD BURDICK is an emeritus professor of statistics, Arizona State University, and former quality engineering director at Amgen for 10 years. He presently serves as principal for Burdick Statistical Consulting, located in Colorado.