Authors
Dr. Frank Deruyck
University College Ghent
Dr. Volker Kraft
JMP
Muralidhara A
JMP
Objective
Use Measurement Systems Analysis (MSA) to assess the precision, consistency and bias of a measurement system.
Background
FVM Pharmaceuticals is an international drug manufacturer, specialized in manufacturing finished formulations that cater to the most demanding global needs. FVM delivers contract manufacturing of tablets, capsules, and liquids.
A typical manufacturing process involves milling an active pharmaceutical ingredient (API) into a powder of uniform particle size. The milled material is then blended with other ingredients to bulk up and evenly distribute the API. This blended material is then compressed into tablets, which are finally coated to aid shelf life, taste, and other properties. At the end of the process, various quality parameters including critical to quality (CTQ) metrics are populated, which further drive batch acceptance.
The process starts with a raw material that is a concentrated emulsion containing two organic compounds. The raw material is supplied by two vendors, and the incoming quality is monitored by measuring the concentration of the compounds in milligram per liter (mg/l). Each day, a quality lab operator takes raw input material in two batches from each supplier into the process.
The process of chromatography is a laboratory technique for the separation of a mixture. The company is currently leveraging gas chromatography (GC), a common type of chromatography used in analytical chemistry for separating and analyzing compounds that can be vaporized without decomposition.
The Task
Recently, the Quality Control Team observed a significant variability in one of the critical product properties of the drug delivered. To address this issue, a cross-functional improvement team was formed to identify the root cause of the problem and then solve it. The head of the Cross Functional Team is Lawrence, a quality engineer who is a firm believer of data-enabled decision making. He knows that building a strong quality culture into the process demands application of statistical techniques to the data to discover the actionable insights. He also knows from his experience that bringing operational excellence into the manufacturing process is a sequential and multistage process starting from raw material to final inspection. At the same time, Quality by Design (QbD) involves ensuring quality throughout the production process (starting from raw material to finished product), while giving flexibility to the manufacturing system.
Lawrence and his team have a huge challenge ahead of them to identify the reasons for variation and find ways not only to minimize them, but also to identify the optimized process parameters to meet the quality standards.
In Part 1 of this case study, Lawrence decided to start the investigation with the raw materials, where he found that Vendor A’s processes for Compound 1 and Compound 2 are stable but incapable. However, for Vendor B, Compound 2 is stable but incapable and Compound 2 is both incapable and unstable.
Based on their initial findings from raw material investigation, Lawrence and his team started exploring a UHPLC measurement system to enable faster and closer raw material monitoring. While raw material monitoring is still offline, it should be possible with UHPLC to measure all batches. They set faster inline analysis as a stretch goal for the future.
The dataset UHPLC_GC.jmp shows the measurements of GC and UHPLC for both Compound 1 and Compound 2 for each of the operators: John, Laura, and Sarah. A simple data visualization using Graph Builder help explain the precision.
As shown in Exhibit 1 (see PDF), UHPLC is still in the preliminary exploration phase; it shows poor precision compared to the actual standard GC method.
Since the process involved three operators, the team wanted to examine how much of the variability is due to operator variation (reproducibility) and measurement variation (repeatability) when different raw material batches are analyzed, so they launched a Gauge R&R study. The main goal of the study is checking the extent to which the actual UHPLC method can discriminate between batches so that quality deviations can be detected. Poor Gauge repeatability and/or poor reproducibility are signs that the discrimination power is too low.