Beating Cancer with Better Data
Pharmaceutical Manufacturing is one of the most important yet most challenging of industries. Yield variations, particularly with early-stage experimental medicine, can lead to difficulties for both business and patient treatment plans. The complexity of these processes cannot be understated, and companies have traditionally relied on individual heroics and long processes. These workings can be mastered, accelerated, and brought to market by leveraging a strong data and decision-making strategy.
The specific issue was straightforward; a breast cancer detection assay suffered a 14.5% yield loss, which was completely unacceptable. Three factors made this challenge exceptionally complex: Regulated Environment, Novel Chemistry, and Experimental Manufacturing Process. The convergence of these factors caused the traditional consulting approach to become ineffective, as the problem spanned both management and technology.
The start of the partnership involved reviewing how the organization was attempting to address the issue. Quality engineers had already been collecting data for 21CFR11 compliance. Furthermore, technicians had theories and tribal knowledge of the root cause of the issue. Even though data was being collected, the knowledge was not being shared. The information was retained in files, and the tribal knowledge was not being combined with the data collection systems.
The pathway to reducing the loss of yield started with making the already collected data available to everyone. This involved several technical steps, from making production dashboards on the screens of the manufacturing floor to better reports for the quality assurance team. Over the course of the investigation, a piece of knowledge arose, and in the winter the process was more reliable than in the summer. Although previously dismissed as a root cause, it was combined with another key finding that the temperature monitoring graphs were showing the humidity was not stable at all times.
With these findings, we assembled a cross-functional team of technicians, quality engineers, management, and facilities. The days when the humidity changed by >10% were also days in which the yield dropped. Facilities cross-referenced this to the maintenance logs and found that these also corresponded with days of HVAC outages or scheduled maintenance. The chemists confirmed that the oils used could change viscosity in response to the humidity, and the machine logs confirmed this.
A cross-functional fix formed part of the solution and was subsequently implemented. Quality engineers would leverage the new dashboards, and collaborate with facilities and technicians to estimate if the humidity would vary during the next day. This approach led to a significant reduction in yield loss from 14.5% to 1.6%, without requiring any capital investment or retooling.