“Clinical data is the gold standard” for setting manufacturing specifications, said Patrick Swann, PhD, acting deputy director of the Division of Monoclonal Antibodies at FDA, at a session on specification setting at the AAPS National Biotechnology Conference that was held June 19–21 in Boston. The challenge, as Swann and other speakers pointed out, is that clinical data that shows relationships to manufacturing specifications is generally limited. The additional challenge in setting specifications for biologics is that these molecules are structurally complex and hard to characterize.
Wassim Nashabeh, PhD, director of quality control for clinical development at Genentech (South San Francisco, CA, www.gene.com), said that for biotech drugs, purity and impurity specifications are often the hardest to set. “With biotech products, it’s difficult to provide a definitive purity analysis,” he said. Adding to that, he said, it’s rare to have conclusive scientific data that links safety to individual variants or impurities, which means that statistical methods are needed.
Part of the difficulty, said Nashabeh, is that process consistency and capability generally are established in the late phases of the clinical program or after trials are complete. To create a closer link between manufacturing data and clinical results, Nashabeh suggested performing many manufacturing runs and campaigns when producing material for clinical trials. “That will give you more assurance that your clinical trial material covers the range of variability likely to arise in commercial manufacturing,” he said. “Don’t plan to manufacture just enough material [for the trials]; manufacture more to test variability.”
Setting Specifications for Vaccines
Potency specifications for live virus vaccines, in contrast, can be clearly linked to clinical results, because dose determines efficacy, said Robert Sitrin, PhD, executive director of bioprocess and bioanalytical research at Merck & Co.’s research labs (West Point, PA, www.merck.com). For live virus vaccines, the upper potency specification is set at the highest dose shown to be safe, and the lower specification is set at the level of the lowest efficacious dose. The minimum release specification is derived by correcting for loss of potency over shelf life.
In contrast, a different strategy is used for setting specifications for another class of vaccines, which are made from recombinant proteins. In this case the dose is set by protein level, and the potency is an independent parameter. Sitrin illustrated his discussion with a case study of the recently approved Gardasil vaccine against cervical cancer and genital warts. Once the protein dose was determined, the company developed an in vitro relative potency (IVRP) assay, based on ELISA technology, that measures specific antigenicity. They then used this assay to compare the relative potency of commercial lots to clinical lots. To set specifications in the final container, they incorporated a propagation of error model to account for potential variation in bulk manufacturing, bulk stability, transfer and weighing, formulation and fill, and assay variability, all of which are assumed to affect the error rate multiplicatively.
Because the potency of all vaccines decays over time, Sitrin displayed a graph to show how stability data were used to set lot release potency specifications high enough to ensure that the required potency is maintained at the end of the vaccine’s shelf life. “You must allow for assay variability,” he said, “then work within the window of safety.”
Sitrin agreed with the other speakers on the importance of clinical data. “You need clinical data to confirm that the final specification is not near the edge of clinical performance,” he said. For the Gardasil vaccine, even though the company conducted large conducted trials, they still wanted to set wide safety margins around manufacturing specifications to allow for patient variability. “50,000 women do not represent the entire world,” he said. “So for women’s protection, we want to be safely away from limits. [With a vaccine] Safety includes lack of efficacy.” To confirm this, an additional clinical study was conducted to demonstrate that a vaccine lot at the minimum potency specification would still be effective.
Managing Specifications and Limits
Several speakers also reminded the audience not to confuse manufacturing specifications and limits. “There is no connection between control limits and specifications,” said Nashabeh. He pointed out that specifications are determined externally, from clinical data, whereas limits are determined by the natural variability of process. “Just because a manufacturing process is naturally capable of producing x quality doesn’t mean those are acceptable specifications for the product,” he said.
Stephen Liebowitz, PhD, group director of global regulatory science and CMC at Bristol-Meyers Squibb’s pharmaceutical research institute (Princeton, NJ, www.bms.com), emphasized the importance of managing specifications and limits during a product’s lifecycle—from early development through commercial manufacturing. “Significant manufacturing changes during the development phase must be assessed for comparability of product to see if any additional preclinical or clinical testing is required,” he said. Limits and specifications must be supported by a strong rationale, he added, and then verified periodically, after a predetermined number of lots or in response to a known change or empirical observations. “Changes to operating limits are subject to the same considerations as the original limits,” he noted.