By Agnes Shanley, Editor in ChiefNovartis began a pilot PAT program two years ago at its facility in Suffern, N.Y. Leading that effort has been Jim Cheney, the companys new global PAT program manager. Cheneys career represents a break with the current tradition of pharmaceutical PAT. Cheney, a process engineer with over 25 years experience, gained his expertise on the job, rising through the ranks of the organization, and combining process expertise with a thorough grounding in quality assurance. He is giving Novartis PAT program an edge, and a grounding in manufacturing realities, that could be the envy of many. Editor in Chief Agnes Shanley spoke recently with Cheney to get his perspectives. (For the entire interview, read Decembers issue of Pharmaceutical Manufacturing.)P.M.: When did Novartis decide to launch PAT at Suffern?J.C.: The idea for the project initially came from our quality assurance group in Basel, when Dr. Erni, a corporate quality executive, visited our facility and asked if anyone would be interested doing any PAT-related work. The company had not yet implemented PAT in Europe.P.M.: How large a team did you dedicate to the project originally?J.C.: Theres never been anyone full time dedicated to this project. Its something people have done on top of their regular workloads. I was only made permanent on PAT for Novartis in March. When the pilot was launched, I was a process engineer, responsible for the manufacturing operation from a processing standpoint. I developed a small team that also involved quality compliance people and IT group which was pivotal.P.M.: Where is the PAT installation at this point?J.C.: Were in the process of validating the application and plan to turn it on during the first quarter of next year. At this point, we are determining who should have access to the data and what level and kind of training each person will require at different positions.P.M.: How did you decide which technologies to use?J.C.: On the sensor side, we had not installed any PAT sensors at Suffern. We had not installed any NIR detectors available whatsoever on the production floor. However, we felt strongly that we had a lot of data available on our products and our processes. Our facility has been an automated plant since 1998, all the data has been going into databases. We just werent doing anything to analyze that data, and truly characterize our processes. I started to look around for ways to do this, using different statistical tools. Thats when I met Dr. Erni who talked about doing MVA.P.M.: How did you evaluate all the options?J.C.: We looked at a number of MVA packages, but since were a production facility we are so focused on manufacturing, we needed something very graphically interfaced, something that would not simply show a lot of data to the operators. I can explain a graph or a chart much more easily to an operator and once they are used to seeing it, they will know right away what to look for.P.M.: Did the level of automation at the facility aid the project?J.C.: It helped quite a bit to have the high level of plant automation, because it made a lot more data accessible. But data were sitting in multiple databases in multiple database tables, and we needed a way to pull the data together in a specific way to make the data accessible for MVA. Its not a trivial task to do that.P.M.: How did you go about it?J.C.: Early in the project, Novartis saw these issues and hired a consultant from TCGIvega to come in and write code for us to help us do this. He wrote the code and developed a hybrid that could combine the strengths of a relational database like Oracles with a time series type database, like PI.P.M.: Are similar projects being contemplated outside the U.S.?J.C.: Im starting a project to roll this out in Switzerland. It may take a bit longer because there they dont have some of the fundamental data collection technologies that Suffern had. The interesting thing about the project, though, is that it will not just look at drug product but drug substance.