ESTABLISHING PROCESS TRANSFER CRITERIA
After sharing client process history with the CMO, discussions to establish mutually agreeable transfer criteria can begin.
Examples of cell culture process transfer criteria jointly agreed to by Millennium and a CMO are shown in Tables 1 and 2.
Table 1. Process transfer criteria for transferring Millennium (MPI) bench-top cell culture bioreactor process to a CMO
These tables demonstrate a few key features. First, the transfer criteria should be a direct comparison: compare bench-top
data with bench-top data and pilot-scale data with pilot-scale data. If undesirable results are attained, a comparison of
CMO pilot data to client bench-top data alone or to client pilot data alone cannot distinguish between generic transfer issues
(e.g., a misunderstanding of the process by the CMO) or a scale-up issue. Thus, it is prudent to first evaluate the efficacy
of the transfer with a laboratory-scale to laboratory-scale comparison as illustrated in Table 1. There typically is much
more client laboratory-scale data than pilot-scale data available to enable statistically valid comparisons. Only when the
efficacy of this laboratory-scale transfer is confirmed should a pilot-scale to pilot-scale comparison be completed (Table
2), although these activities can overlap to save time.
Table 2. Process transfer criteria for transfer of pilot-scale cell culture bioreactor process to CMO
Figures 1 and 2 present actual data generated by the transfer activities represented in Tables 1 and 2, and illustrate the
wisdom of this approach. Figure 1 demonstrates similar titer performance between the CMO's 2-L bench-scale production-phase
bioreactors and Millennium 5-L bench-scale production-phase bioreactors. However, Figure 2 shows that the CMO's 300-L pilot-scale
bioreactor generated approximately 20% lower titers than the Millennium 100-L pilot-scale bioreactor. In light of the close
agreement between bench-scale systems in Figure 1, the CMO quickly agreed that there was an implementation problem with the
300-L process that required investigation.
This evaluation procedure quickly identified a serious issue well before current good manufacturing practices (cGMP) manufacturing
was scheduled to begin. Moreover, this procedure pointed to the most fruitful area for investigation. A careful comparison
between the operation of the CMO's 300-L bioreactor and the Millennium 100-L bioreactor and both sites' bench-scale bioreactors
identified three operating variables that were unique to the 300-L bioreactor: the age of the bioreactor feed; the percentage
carryover of medium from the inoculum bioreactor; and bioreactor pH. Regarding the pH, the set points were identical for all
systems. Unfortunately, the actual pH of the 300-L bioreactor was approximately 0.1 pH units lower than all the other systems,
because of differences in control algorithms and the reduced frequency of re-calibration of the pH probe in the 300-L suite.
Based on this investigation, a factorial-designed experiment with three variables as inputs was executed in the Millennium
5-L bioreactors (Table 3). The data clearly demonstrated that only the pH had a significant impact on the titer. Consequently,
the CMO changed its bioreactor pH control and recalibration procedures for the large-scale GMP runs, which were successfully
implemented without an engineering run.
Table 3. Bench-scale bioreactor experiment to troubleshoot a lower- than-expected titer from a CMO pilot-scale bioreactor
This example highlights another benefit of establishing a process history and transfer criteria before transferring a process
to a CMO. Most CMOs have established their own tightly controlled procedures over many years. These procedures have been proven
to meet practical, regulatory, and business needs, making CMOs reluctant to change standard practices. However, when scientifically
sound data suggest that deviation from standard practices is warranted, most CMOs will be quite willing to cooperate.
Another essential feature of process transfer criteria illustrated in Tables 1 and 2 is the comparison of key product quality
attributes. The acceptance criteria for product quality presented in this example were intentionally vague. Unless there is
significant process history approaching the level attained nearing a formal process validation or a biologics license application
filing, there is usually insufficient process knowledge either to establish definitive product quality acceptance criteria
or to guide process troubleshooting efforts.