## The Standard Curve Calibration Curve

The relationship between drug concentration and the response of the analytical system needs to be determined. This mathematical relationship will allow us to later determine analyte concentrations of unknown clinical

20,000

ULOQ Standard Curve

100 200 300 400 500 600

20,000

Upper Bound A

ULOQ Standard Curve

Lower Bound A

100 200 300 400 500 600

Range

FIGURE 4 Standard Curve. The detector responses to a drug are plotted against six duplicate concentrations of drug ranging from 5 to 500ng/mL (•). The upper level of acceptable error in the drug concentrations is represented by the triangles, and the lower level by the open circles. The ULOQ is 500 ng/ml, and the LLOQ is 5 ng/ ml. The solid line through the actual data was linearly regressed, and generated an equation for a straight line with the form Y=AX+B, where Yis the machine response, A is the slope of the curve, X is the drug concentration and B is the intercept on the y-axis. With the values of A and 8, the value Y for unknown samples is determined by analysis, and the corresponding concentration is then back-calculated.

samples from the response obtained from the analytical method. The standard curve of the method is specific for each drug in a specific matrix (e.g., blood, plasma, urine, cerebrospinal fluid, etc.). If the drug will be measured in plasma during the clinical study, the standard curve should be constructed by spiking drug into plasma, and then extracting and analyzing the concentrations. The use of different solvents such as water or methanol is not recommended because there may be differing solvent characteristics (such as solubility, protein binding, etc.), and this could complicate the interpretation of the data. The drug stock solution must be made in a solvent, but all subsequent dilutions should be in sample matrix.

If samples will be taken from more than one matrix (e.g., plasma and urine), then standard curves must be constructed for each. The same is also true if more than one analyte is to be measured (e.g., parent drug and metabolite). Although parent and metabolite may be simultaneously quantified from the same sample, a standard curve for each specific analyte must be constructed. It is also advisable to incorporate the use of an internal or external standard in sample preparation, although this step is not a requirement for method validation. Standard curves should be constructed with a minimum of six drug or analyte concentrations spiked in the appropriate matrix (see Fig. 4). Once these standards are measured, the data should be plotted (response vs. analyte concentration), and the simplest curve which best fits the data should be generated to describe the relationship. Zero or blank samples should not be included in the curvefitting procedure because the assay is characterized by a lower limit of quantification which is higher than "zero" or no drug, and inclusion of this point might alter the fit of the curve. Curves generated without weighting of the data are preferred, but weighting the data is permitted. Usually, weighting is used in cases where the range in drug concentrations spans several orders of magnitude, and weighting helps account for the heterocedasticity in the data. The relationship that is derived is then used to back-calculate drug concentrations from clinical study samples. The slope of the curve indicates the sensitivity of the assay; small changes in concentration that induce large changes in response indicate a sensitive method [9].