Discovering Peak Performance at Supelco with Minitab

Supelco, Bellefonte, PA USA

You can download a PDF of this magazine article. You must have Adobe Reader 4.0 or better to view this PDF file.

When the scientists at chromatography product manufacturer Supelco developed the Discovery product line for the pharmaceutical industry, their goal was to produce columns with performance levels that surpassed all other products on the market. In order to meet that ambitious target, the company invested significant resources into research and development, building state-of-the-art facilities, and implementing an extensive quality assurance program.

Supelco's Discovery columns are generally used for high performance liquid chromatography (HPLC), the most common analytical method for separating substances into their individual components. The quality assurance protocol for Discovery columns involves producing and testing these columns under strict ISO 9001 documented processes. Several key characteristics are measured to make certain that each piece is designed and manufactured to strict customer and industry specifications. The company's manufacturing methods ensure column reproducibility, which means they've eliminated variations in column performance from batch to batch. Column performance and stability are critical for customers to be able to perform valid and reproducible analyses.

John Rumbaugh, the company's production manager, said, "At Supelco, we live by our ISO 9001 certification and practice continuous improvement methods in everything we do." When Supelco's Quality Assurance department discovered that the pass rate during product performance testing for one of their HPLC Columns was 80%, a rate far below the company's pass rate goal of 95%, Rumbaugh and a team of scientists and quality assurance professionals stepped in to find a solution. Their challenge was to increase the column's pass rate to the desired 95% by optimizing production processes and product performance. These improvements would ideally increase the throughput of the process and reduce costs associated with failing a column, which ranged from $40 to $85.

The product involved in the quality improvement project was a small, reversed phase, silica-based HPLC column used mainly for pharmaceutical research. Reversed-phase is the most common HPLC mode and involves the extraction of non-polar to moderately polar analytes from a polar solution using a non-polar sorbent. Typically, columns use hydrophobic packing bonded to silica or neutral polymeric beads. The mobile phase in the column is usually water and a water-miscible organic solvent such as methanol or acetonitrile. Various mobile phase additives are used to test for the presence of different components.

An experiment was designed to test the effects of key variables on efficiency and peak asymmetry, and to determine which combination would result in increased capability and pass rates, thus increasing throughput of the process. Once the team learned how the variables interacted to affect the asymmetry and efficiency of the columns, the settings could be improved until optimum quality and performance was achieved. The team also hoped to take what they learned from this experiment and apply their knowledge to improve or maintain performance/pass rate of other columns.

The dependent variables in this experiment were Efficiency with a specification greater than 3700 and Peak Asymmetry between .80 and 1.30. Asymmetry is the factor describing the shape of a chromatographic peak. Theory assumes a Gaussian shape and that peaks are symmetrical. The peak asymmetry factor is the ratio (at 10% of the peak height) of the distance between the peak apex and the backside of the chromatographic curve to the distance between the peak apex and the front side of the chromatographic curve. A value above one is a tailing peak, while a value below one is a fronting peak.

Supelco used a team approach to solving the problem and developed a cause and effect diagram (or fishbone diagram) to capture, categorize, and prioritize potential root causes. The team examined the processes within four distinct production areas for potential cause of column failure: column production, packing operations, testing, and final product packaging.

Several key variables were identified that influenced the throughput of the production of columns: difference in QA instruments, slurry, pressure profiles, and machinery.

Differences in Quality Assurance Instruments

Supelco used two types of column testing stations: multi-tester stations with the ability to test six columns at a time, and single column testing stations. In general, it was preferable to test all columns on the multi-tester because of the potential decrease in time spent on QA.

Slurry
Different combinations of solvents can be used to pack the columns with silica. Slurry packing is a technique that involves suspending the column packing in a slurry and rapidly pumping it into the empty column using a special high-pressure pump. Supelco used a 90:10 ratio of acetone to methanol in their slurry mixture. In the experiment, they compared the 90:10 slurry mixture to a 50:50 of Toluene and Acetone mixture to see if manipulation of this ratio could be used to increase the pass rate. For the experiment, the silica was blended in one "master batch" and the slurry was kept well-suspended to reduce batch-to-batch variation.

Pressure profiles
Columns are packed by machine at high pressure for a constant amount of time. The team decided to test whether increasing or decreasing these pressure profiles affected the pass rates of columns. They experimented with different levels of packing pressure (from 5000 to 7000 psi) and measured the results.

Machinery
The last variable tested was the influence on the pass rate caused by producing columns on two different types of machines: automatic and manual. The automatic machine is computerized and ramps up pressure to pack columns. The manual machine, on the other hand, is adjusted by an operator and uses an almost instantaneous pressure to pack the columns. The testing examined the differences in these two machine types and whether or not either produces more reliable parts.

When the team chose a statistical software package to help with data analysis, they turned to Minitab. Rumbaugh said, "I had first been exposed to Minitab when I attended Penn State's Quality and Manufacturing Management (QMM) Program. I feel that the software is easy to use, has clear graphics, and the data is credible. It has the tools we need for the job. At Supelco, our production supervisors and QA people use Minitab to create charts and reports used for review in weekly department meetings. The data is collected and fed directly into Minitab for statistical analysis. This provides us with vital information used to improve pass rates, decrease back-orders, and much more."

Minitab's Design of Experiments (DOE) capabilities allowed Supelco to screen the factors to determine which were critical for explaining process variation. The team was able to use Minitab to understand how the factors interacted and to identify areas for intervention and process improvement opportunities.

In a full factorial experiment, all combinations of the factors and the resulting responses are measured. The cost and time associated with running the large number of runs required in a full factorial model is often prohibitive. So, in order to reduce the number of runs to a manageable size, Rumbaugh and his team elected to perform a fractional factorial design experiment. Fractional factorial designs minimize expense while still providing valuable information regarding the factors to the experimenters.

Using Minitab, the team was able to determine optimum conditions and quickly identify which factors were most important and most likely to yield improvements in the process being studied. Two key variables were proven to have significant impact in this designed experiment: pressure, and the type of QA instrument used to test the columns. Both the type of machine/pump and the ratio of solvents in the slurry mix were insignificant at a 95% confidence level. There were no statistically significant interactions of the four factors (at the a=.05 level).

This insight gave the experimenters a new level of understanding of the factors that affect asymmetry in the 4.6mm x 3.3cm LC column. By understanding what significantly impacts its asymmetry, the pass rate would increase if the experiment's conclusion would be used in production. Twelve more columns were run using the proposed conditions, and the Cpk improved from -0.01 to 1.7. This significant increase in the pass rate of this column, to greater than 99%, translated into cost savings in materials and man-hours, particularly within the QA phase of production.

The team was able to make adjustments to the production and testing processes. An increase in the pressure used during the column packing process drove down asymmetry. They also changed the type of QA testing instrument used from a normal to an optimized instrument for shorter columns, resulting in the use of a more accurate measurement system. A single-column testing station was used versus a multi-column testing station to minimize the dead volume present. The ratio of volume in the column being tested to the dead volume in the test unit is critical.

An improvement in the process was demonstrated by using Minitab to perform an initial process capability study and then a final study using the results of the experiment. (Fig. 1) The quality of the product also improved as a result of the adjustments made to the process. The chromatograms produced by the column produced using the new process have a more ideal peak shape that is closer to the 1.0 target. Chromatography products that can produce the separation with the tightest peak shape are most desirable.

Based on the results of this experiment, Supelco revised their controlled documents to reflect the new packing conditions. The column is now being packed at 7000psi and tested on an optimized QA measuring system with all other factors remaining as they were prior to the experiment. Pass rates for the column have increased beyond the goal of 95%.

Mr. Rumbaugh emphasized the role of Minitab in the project: "Without Minitab, we probably would have changed one thing at a time and the whole process would have taken significantly longer." On a corporate level, Supelco hopes to increase sales and profit via increasing the throughput of the column production process. As the company analyzes future opportunities to design and launch new products and to make improvements on existing products, Minitab will be part of Supelco's strategy to develop production methods that ensure high pass rates.

Figure 1: Minitab's Capability Analysis Sixpack commands display the values of Cpk and Ppk, which are measures of how capable a process is of meeting specifications. At Supelco, they strive for a cpk of 1.2. Symmetry was 1.30 (-.01 Cpk prior to the adjustments). After the adjustments, asymmetry was 1.06 (1.74 Cpk). Asymmetry is a key measure of column performance.

Process Capability, Asymmetry After

About Supelco

Supelco manufactures chromatography products for analysis and purification. The company offers more than 10,000 stock and custom chromatography products and related tools for environmental, government, medical, pharmaceutical, biotechnology, food and beverage, pharmaceutical, biotechnology, and chemical laboratories. Supelco is a part of Sigma-Aldrich Corporation, which also includes Sigma Chemical Co., Aldrich Chemical Co., Fluka Chemie AG, Riedel-de Haen, and Genosys. The corporation has subsidiaries in 26 countries and more than 50 distributors worldwide.

This article was originally published in the September 2001 issue of Scientific Computing and Instrumentation).

Looking for other information or materials? For help, please contact Minitab's Marketing Department at PublicRelations@minitab.com.