Comparison of different linear calibration approaches for LC–MS bioanalysis

Many different calibration approaches are used for linear calibration in LC–MS bioanalysis, such as different numbers of concentration levels and replicates. However, direct comparison of these approaches is rare, particularly using experimental results. The purpose of this research is to compare different linear calibration approaches (existing and new ones) through simulations and experiments. Both simulation and experimental results demonstrate that linear calibration using two concentrations (two true concentrations, not forced through zero) is as good as or even better than that using multiple concentrations (e.g. 8 or 10) in terms of accuracy. Additionally, two-concentration calibration not only significantly saves time and cost, but is also more robust. Furthermore, it has been demonstrated that the extrapolation of a linear curve at the high concentration end to a linearity-known region is acceptable. When multi-concentration calibration is used, the difference between the two commonly used approaches, i.e. singlet (one curve) or duplicate (two curves) standards per concentration level is small when a method is very precise. Otherwise, one curve approach can result in larger variation at the low concentration end and higher batch failure rate. To reduce the variation and unnecessary reassays due to batch failure or possible rejection of the lowest and/or highest calibration standards, a partially duplicate-standard approach is proposed, which has duplicate-standard-like performance but still saves time and cost as singlet-standard approach does. Finally, the maximum allowable degrees of quadratic (non-linear) response in linear calibration are determined for different scenarios. Because of its multiple advantages and potential application in regulated bioanalysis, recommendations as how to implement two-concentration linear calibration in practice are given and some typical “concerns” regarding linear calibration using only two concentrations are addressed, e.g. how does one know if the response is truly linear over a given range when only two concentrations are used?.