About TCAW - Subscription Info
February 2001
Vol. 10, No. 02, pp. 42–48.
Focus: Separation Sciences

FEATURE

“Committeeing” to Calibrate

opening artA case history of one company’s effort to expedite regulatory compliance

What is HPLC system “calibration”? Hint: Neither internal adjustment nor response curve for quantitation is intended here. “Calibration”, in good manufacturing practices (GMP) terminology, refers to instrument “qualification” or HPLC performance verification. In most pharmaceutical laboratories, calibration is performed on each HPLC module usually every 6–12 months, according to the company’s standard operating procedures (SOPs). If results are within the pre-established criteria, a calibration sticker is placed on the instrument to signify its readiness for generating GMP data. This periodic calibration, together with initial qualification (installation qualification and operation qualification or IQ/OQ) and daily system suitability testing, forms the backbone validation program to ensure data validity in the pharmaceutical laboratory. An analyst, a metrologist, or a qualified contractor can perform calibration. Because regulatory compliance is necessary, expensive, and time-consuming, the challenge to make validation more efficient becomes a universal priority. Here is the case history of one company’s effort to expedite a segment of this process.

The Problem
In our Department of Pharmaceutical Analysis, we perform precise and accurate analyses of drug substances and formulations to support new drug development. Until recently, we had a 4-year-old calibration procedure that was thorough but time-consuming. Using that procedure, we checked the wavelength accuracy of the detector; the flow and compositional accuracy of the pump; and the precision, linearity, carryover, and sampling accuracy of the autosampler.

It typically took an analyst 2–3 days to track down the myriad apparatus, prepare the standards, and perform the experiments, plus additional hours to document results in the laboratory notebook. This calibration time, when multiplied by more than 50 HPLC systems twice a year, translated into thousands of nonproductive hours. Other complaints included test procedures with insufficient details, acceptance criteria that were too wide or difficult to pass on the first trial, and the use of an antiquated HPLC column. Most analysts agreed that an update was needed, but no one volunteered; that is, until this summer, when we decided to tackle the problem.

The Calibration Team
Like most creative folks, our first step was to form a committee or an HPLC calibration team. We went for diversity. Our team consisted of a leader experienced in writing SOPs, an instrument specialist well-versed in specifications, a consultant with expertise in regulations, and our very own departmental calibration administrator. Since everyone was overwhelmed with primary project responsibilities, a summer intern was hired for lab work. Our team’s mission was to create a 1-day calibration procedure applicable to all on-site HPLCs, as well as a set of meaningful acceptance criteria that conform to the industry norm. The revised procedure would have sufficient detail to eliminate references to instrument manuals, yet be simple enough that new analysts could follow it with little training.

Our early euphoria from achieving these lofty goals quickly gave way to the reality of biweekly meetings, where new ideas were proposed, debated, rejected or embraced, amended, and adopted. Experiments progressed steadily. Considering the personalities of its members, the team worked together surprisingly well. We went public early during departmental meetings, sharing our goals and strategies to solicit feedback and support. A draft procedure emerged in August, and an independent scientist performed a final check. HPLC Calibration Procedure Rev. 1 was in effect on September 11.

The Expedited SOPs
Table 1 summarizes the tests and criteria of our revised SOPs. It is the result of numerous debates, compromises, and rewrites. On the surface, this list looks remarkably similar to that of our previous procedures. The difference, of course, is in the details. First, let’s identify some concepts and philosophies that guided our decisions along the way.

  • Calibration frequency remains every 6 months, because we are concerned that lengthening the period might compromise the integrity of our data.
  • Annual preventive maintenance (when the seals, lamp, and filters are replaced) is to be scheduled before calibration when possible.
  • Before calibration, each module is shut down and powered up to initiate built-in diagnostics.
  • The calibration order is detector right arrow pump right arrow autosampler (as the detector is used to calibrate other modules).
  • To enhance productivity, the number of apparatus, reagents, or mobile phases is reduced to a minimum. Procedures are updated for speed and robustness. Automated instrumental methods and calculations are used whenever possible.
  • Acceptance criteria generally mirror manufacturers’ specifications, although many are necessarily relaxed to accommodate diversified models and aging components.
  • Developing template forms facilitates documentation. Raw data are entered directly on the forms to which data reports are attached and subsequently approved. No laboratory notebook entries are required.

Figure 1. Spectrum of anthracene solution (1 mg/mL in acetonitrile) from Waters 996 PDA detector shows the annotations of lmax.
Figure 1. Spectrum of anthracene solution (1 mg/mL in acetonitrile) from Waters 996 PDA detector shows the annotations of lmax. The inset shows an expanded view of the bands centered between 320 and 380 nm.

Figure 2. Calibrations of a UV?vis detector (Waters 2487) near 340 nm by incremental scanning from 336 to 344 nm.
Figure 2. Calibrations of a UV–vis detector (Waters 2487) near 340 nm by incremental scanning from 336 to 344 nm. The lmax of anthracene is determined to be at 339 nm for this detector. The autozero on wavelength function in the detector must be inactivated for this test.
Calibrating Each Module
Photodiode Array (PDA) or Variable UV–vis Detector. A number of reference chemicals with well-defined UV spectra have been used for detector wavelength calibration, including uracil, erbium perchlorate, holmium oxide, caffeine, and anthracene. We selected anthracene because of its sharp absorbance bands and good solution stability. We use a single anthracene solution at 1 µg/mL in acetonitrile to be flushed into the flow cell using a disposable syringe. Wavelength accuracy of the detector is verified by measuring the absorption bands at 251 nm and 340 nm. In photodiode detectors, a spectrum annotated with lambdamax is used (Figure 1). In UV–vis detectors, two scans from 247 to 255 nm and from 336 to 344 nm are used to locate the actual lambdamax (Figure 2). Detector absorbance linearity is tested simultaneously with autosampler linearity. Detector noise is not tested, as it can be covered by system-suitability tests in methods for trace impurities, if required.

Pump. Flow accuracy is checked at 1 mL/min, with the column in place, by measuring the time required to fill a 10-mL volumetric flask from the detector outlet. Other flow rates can be tested if the lab routinely performs analyses using narrowbore or Fast LC. Compositional accuracy is determined by making 5-min step gradients from water in one solvent line to 0.1% acetone in water in a second solvent line. The absorbance of each step against that of the 100% step is measured. A higher flow rate of 2 mL/min is used (vs. 1 mL/min in the previous method) to sharpen the step definition and increase test robustness. All four solvent lines are checked in a 1-h solvent program (see Figure 3). Calculations of absorbance ratio are reported automatically by the data system (Waters Millennium32).

Figure 3. Typical step gradient profile of a four-solvent pump (Alliance 2690 module) using a solvent program at 2 mL/min with water and 0.1% acetone in water.
Figure 3. Typical step gradient profile of a four-solvent pump (Alliance 2690 module) using a solvent program at 2 mL/min with water and 0.1% acetone in water. Absorbance ratio of each step against the 100% step is used to determine compositional accuracy. All four solvent lines are tested in a 1-h gradient program.

Autosampler. For autosampler precision, 10 consecutive 10-µL injections of an ethylparaben solution (20 µg/mL) are used. A faster symmetry column packed with 5-µm particles replaces the older and longer 10-µm µ-Bondapak C18 column. The acceptance criterion for peak area precision remains at 0.5% relative standard deviation. A suggestion to relax this stringent criterion was voted down because an assay precision of <1% is needed for drug substances. Single injections of 5, 10, 40, and 80 µL of the ethylparaben solution perform the linearity test. The largest volume is set at 80 µL because the default sample loop is 100 µL in many autosamplers. This test actually checks the combined linearity of the injector, the detector, and the data system. Carryover is determined by measuring the carryover peak area of an 80-mL injection of the ethylparaben standard in a blank injection.

The autosampler injection accuracy test was hotly debated by the committee and in group discussions. Many believe that it is not needed because HPLC is a relative quantitative technique. Also, this test is not performed by the FDA, allegedly because of the lack of an acceptable testing method. However, the committee felt that an independent verification was desirable because the sampling volume is controlled by the size of the sampling syringe, which can be changed easily in most autosamplers. The previous procedure stipulated the injection of 50 µL of a paraben solution and the collection of the entire peak into a volumetric flask. After dilution to volume, the concentration of paraben was then compared using a validated UV spectrophotometer with a similar solution prepared by pipetting the same paraben solution from a validated syringe. This arduous procedure took 2–3 h and had an acceptance criterion of ±15% that was judged too broad by the committee.

After many discussions, a procedure from a manufacturer’s operation qualification was adopted. It involves the gravimetric determination of the average volume of water per injection withdrawn from a tared vial after six 50-µL injections. The procedure takes <10 min and has an acceptance criterion of 50 ± 2 µL.

Column Oven. A temperature accuracy test of the column oven using a calibrated thermal probe is added because of the increased popularity of column ovens. An acceptance criterion of 35 ± 2 oC is adopted.

Overall Procedures
Minimizing sample preparation and the number of calibrated test apparatus saves significant time. In the revised procedure, only a balance, a 10-mL volumetric flask, a stopwatch, a thermal probe, two standards, one column, and one mobile phase are used. Sufficient details are included to eliminate any references to manuals. For instance, directions to inactivate the autozero function required to determine the maximum absorbance in a UV–vis detector are included, as well as the exact solvent program used to perform the compositional accuracy test. The development of template forms (see Figure 4) also improves the consistency of calibration records.

Epilogue
Six weeks after the revised procedure was in effect, the verdict was in, and almost everyone seemed happy. About 10 calibrations were performed, and most analysts successfully calibrated the system in 1 day. Most were impressed with the robustness of the compositional accuracy tests and the gravimetric autosampler accuracy test. They loved the template documentation but uncovered two minor procedural problems. The calibration administrator is happy with the clarity in documentation. Management is happy about the time savings. The committee is writing version 2 of the calibration procedure to fix bugs and to extend the flow rate accuracy test at both 0.3 and 1.5 mL/min. The instrumental specialist still grumbles about not including tests for instrumental bandwidth and dwell volumes to address the increasing use of narrowbore columns. The committee disagrees with this inclusion at the moment.

Life in the pharmaceutical laboratory is increasingly burdened with numerous SOPs and massive documentation. Perhaps a better way to create a more efficient system is by setting up small task forces to encourage ideas and solutions to “bubble up” instead of the traditional “top down” approaches. Our three-month sojourn taught us that through active participation and continuous refinement, there is light at the end of the dark tunnel. So here is a story with a happy ending, even though it involves a committee—or should we say—because of a committee.

Acknowledgments
Raphael Ornaf, Mark Loranger, and Clark Choi, members of the scientific staff of the Pharmaceutical Analysis Department of Purdue Pharma L.P. in Ardsley, NY, and Jaspal Mayell, consultant, were either members of the team or had significant contributions to this project. The authors would also like to thank Phil Palermo, Lane Gehrlein, Phil Goliber, Glenn Sweerus, Ashley Lister, and Katharina Jakaitis of Purdue Pharma for helpful advice and suggestions.

Further Reading

  1. Parriott, D. Performance Verification Testing of HPLC Equipment. LC-GC 1994, 12 (2), 134.
  2. Furman, W. B.; Layloff, T. P.; Tetzlaff, R. F. Validation of Computerized Liquid Chromatographic Systems. J. AOAC Int. 1994, 77 (5), 1314.
  3. Huber, L. Validation and Qualification in Analytical Laboratories; Interpharm Press: Englewood, CO, 1998.


Michael Dong and Roy Paul are principal scientists at Purdue Pharma, L.P., Ardsley, NY, and David Roos is a student at Fairfield University, CT. Send your comments and questions about the artical to the Editorial Office by e-mail at tcaw@acs.org, by fax at 202-776-8166 or by post at 1155 16th Street, NW; Washington, DC 20036.

Return to Top || Table of Contents


 CASChemPortChemCenterPubs Page