Experts weigh in on the right approach to validate a liquid chromatography method.
With instrumentation becoming easier to use, method development and validation are two of the few remaining “expert realms” in liquid chromatography.
Robust methods demand a balance between speed and resolution. A service lab may accept relatively poor resolution provided that ingredients baseline-separate from impurities, while in-process analytics demand resolution at the expense of speed.
Improving resolution is achieved by increasing retention time, selectivity, or efficiency (number of theoretical plates). In practice, selectivity is the parameter over which chromatographers exert the most control through varying solvent strength/type, temperature, column, pH, and additives.
“It’s not rocket science as long as the samples aren’t too complex,” notes Tom Jupille, a trainer and consultant with LC Resources (Walnut Creek, Calif.). “You line up the parameters in preferred order and start changing them, systematically.”
Speeding things up
One key to accelerating method development is to speed up its individual components. “Faster chromatography methods lead to more rapid method development through higher throughput and greater productivity,” says Bill Hedgepeth, HPLC applications chemist at Shimadzu Scientific Instruments (Columbia, Md.). Yet, investing in rapid LC can be expensive, particularly if a company is not committed to adopting it outside of specialized situations, because methods are not generally transferrable from rapid LC to standard systems.
“Purchasing a specialized system simply to conduct ultra-fast, high-pressure analysis is costly, and must be carefully considered,” Hedgepeth notes. Methods are not simply transferrable between conventional and rapid systems, and usually need to be re-validated. Mobile phases are generally interchangeable, even when transferring a method to a sub-two-micron particle size column. Users should be prepared, however, to tweak flow rate, gradient time program, and injection volumes to match the new column dimensions.
Abu Rustum, PhD, senior director of global analytical sciences at Schering Plough (Union, N.J.) and his department of about 50 doctoral-level scientists, use a variety of methods software packages, including ChromSword (Iris Technologies, Olathe, Kan.), DryLab (Molnar-Institut, Berlin, Germany), and Method Development and LC Simulator (ACD Labs, Toronto, Ontario, Canada).
But Rustum’s group relies primarily on ChromSword, which has an automated mode that allows an analyst to specify a mobile phase and sample—even an unknown—and obtain the best separation conditions for that system. ChromSword employs natural intelligence algorithms to arrive at the mobile phases most likely to succeed. The system does not always provide the absolute best answer, Rustum says, but “the data it generates becomes invaluable for deciding on what to try next.”
ChromSword’s strengths are its simulation functions, which arrive at a method without the need to run many physical experiments. The package works with reverse phase, normal phase, and ion exchange LC. Users can build a method from the ground up, beginning with either two physical runs of unknown samples at different organic solvent concentrations, or by inputting a known chemical structure and proposing solvents through ChromSword’s “virtual chromatography” function.
Rustum’s group optimizes stationary phase conditions with a column selection product, the LC Spiderling column selection selector from Chiralizer Services (Newtown, Pa.). Column selectors are essential for taking full advantage of method development software. Chiralizer’s four models handle up to nine columns, interface with various manufacturers’ LC instruments, and provide heating and cooling modes. “Screening these conditions manually would take ten times as long as with the automated system,” Rustum notes.
Validation: the finishing touch
Method validation is critical in regulated industries. Analytical scientists at US drug firms follow International Committee on Harmonization (ICH) guidelines, with refinements from the US Pharmacopoeia and FDA. “You can deviate, but you must provide scientific justification for doing so,” notes Herman Lam, PhD, president of the standards-setting organization Calibration and Validation Group (Toronto, Ontario, Canada). A former analytical chemist at GlaxoSmithKline, Lam is co-editor of Analytical Method Validation and Instrument Performance Verification (Wiley).
Lam warns of being seduced by modern automated method development tools. It is easy to devise a method that is scientifically-
satisfying, but not validatable. The method may not be sufficiently robust, or investigators may not have collected enough information for validation. “It is critical to determine if a method is applicable to day-to-day operations,” he adds.
Alessandro Baldi, product manager for chromatography systems at PerkinElmer (Waltham, Mass.), notes that validation settles issues that plague poorly-developed methods, namely sample stability, repeatability/reproducibility, method robustness, and performance of the method or system over time. All hinges on robustness, a quality that distinguishes a true method from a one-off analysis. “Methods that lack robustness will not provide reproducible data over time, and will be difficult to export outside the group.”
Knowing what you know
Defining analytic goals poorly or incompletely is the most serious mistake method developers make during validation, says John Waraska, director of strategic marketing for life sciences at ESA Biosciences (Chelmsford, Mass.), a developer of LC detection systems. “It is possible to validate a method that doesn’t quite give you the information you need.”
Narrowly defining parameters, for example, by specifying one vendor’s column, is a common error, as is misunderstanding parameters that impinge on the analysis. “Make sure your assay is measuring what’s important, and not boxing you in,” Waraska says.
Two pitfalls affecting development/validation are over-validation and over-reliance on published methods. Over-validation—an extreme example would be examining the effect of temperature on temperature—is more an issue of managing people than of managing the method. Published methods are a tremendous resource, but only when instrumentation systems cited in the literature match one’s own. When they do not, constructing a method from first principles could save time.
Jupille, of LC Resources, recommends treating methods development and validation organically rather than separately. “If you develop methods correctly, formal validation will not be a problem,” he says.
That means not changing more than one variable at once by playing hunches. Instead of constructing a grid of conditions and working through it, some analysts will change columns, pH, solvent, etc., in the hope that a method will pop out. “When they work, hunches are a source of great emotional satisfaction,” Jupille says. “You get this ‘yes!’ moment.” When they fail, developers are back to square one, with no real understanding of what to do next.
About the Author
Angelo DePalma, Newton, N.J., covers the pharmaceutical, biotechnology, chemical, and healthcare industries.
This article was published in Drug Discovery & Development magazine: Vol. 12, No. 1, January, 2009, pp. 29-31.
Filed Under: Drug Discovery