Analytical performance assays, calibration curves

An immunoassay was developed to determine the penicillinase stable isoxazolyl penicillins cloxacillin and dicloxacillin in milk by Usleber et alJ The assay detected lOpgkg" of cloxacillin and 30pgkg of dicloxacillin with recoveries of 102% and 84%, respectively. The calibration curve was prepared by fortifying skimmed milk powder (lOOgL ) with standards. Fortified samples were prepared in pasteurized milk and analyzed directly after decreaming by centrifugation. This immunoassay was performed with minimal sample preparation, probably because the extensive water solubility of the penicillins prevents problems associated with more lipid-soluble analytes. 

Matrix effect — To demonstrate that the assay performance was independent from the sample matrix, QC samples were prepared using two different lots of matrix. The QC samples were evaluated using the same calibration curve. With regard to analytical recovery, no significant difference was observed for the QCs prepared in two lots of plasma. 

With known analyte concentrations, the processed data provide calibration points, and Immusoft comprises fitting procedures to deduce the corresponding calibration curve that can be stored in the program for further experiments. Different calibration curves can be stored depending for instance on the assay protocol, on the specific features of the chip used or on the medium in which the assay is performed. In most cases, the calibration is performed with six independent chips of eight channels and cumulated in order to get a stable batch calibration. Then the results can be referred to this internal batch calibration. For routine control, one calibration each week is recommended to be sure that the chemistry is still in the specifications (e.g. + 10% of inter-assay standard deviation). 

Zarghi et al. [76] developed an HPLC method, using a monolithic column, for quantification of omeprazole in plasma. The method is specific and sensitive with a quantification limit of 10 ng/ml. Sample preparation involves simple, one-step extraction procedure, and analytical recovery was complete. The separation was carried out in reversed-phase conditions using a Chromolith Performance (RP-18e, 100 x 4.6 mm) column with an isocratic mobile phase consisting of 0.01 mol/1 disodium hydrogen phosphate buffer-acetonitrile (73 27) adjusted to pH 7.1. The wavelength was set at 302 nm. The calibration curve was linear over the concentration range 20-1500 ng/ml. The coefficients of variation for intra- and interday assay were found to be less than 7%. 

A series of unknown samples is usually measured together with two sets of calibration standard samples (covering the concentration range for the assay, usually two sets of six or more calibration standards) and two or more sets of quality control samples. The calibration standard samples will be used to establish the calibration for the unknowns. Quality control samples (usually at least 5 % of number of unknowns) are matrix samples of known concentration, which are equally distributed over the analytical run (usually two sets of three different concentration levels 2-3 times the LOQ, mid concentration range and close to the upper limit of quantification). They establish a set of control samples in order to verify the assay performance within the run. Typically, the calibration standards and quality control samples should be within +/-15% of the nominal value. However, in typical assays, it is considered to be acceptable, if 75 % of the standards are within the +/-15 % criteria. Outliers will not be used for the calculation of the calibration curve. Not all standards at one concentration should be excluded. A similar criteria is applied for the quality control samples 2 out of 3 of the quality control samples should be within +/-15 % of their nominal value. 

There are various ways to derive a calibration curve. Multi-point calibration curves, for example, include a minimum of three different concentrations of the analyte. For semiquantitative assays, a single-point calibration is common. The single point is usually the threshold concentration used to determine whether a specimen is positive or negative for the analyte of interest. Depending on the validation process and performance characteristics of the assay, a single-point calibration curve may also be used in quantitative applications over a limited range of linearity. A historical (pre-established) multi-point calibration curve may also be used, but only if the stability of the analytical method over time has been well established (Goldberger et al., 1997). 

Organophosphate pesticides studied in this work were the model low-toxic OPC trichlorfon, and some common organophosphate pesticides malathion, parathion, dichlorvos, and diazinon (Table I). Calibration curves for these pesticides (dependences of the sensor inhibition response on the analyte concentration) were obtained for all of these OPCs. These calibration curves were obtained under conditions (time of inhibition, pH and temperature) optimize with the model analyte trichlorfon. All of the pesticide calibration curves are similar and Fig. 4 illustrate the method by the example of malathion. The lowest concentration of pesticide samples assayed with 10 min. of incubation of the electrode in inhibitor containing solution was 5 ppb. This resulted in approximately 10 % of the relative inhibition signal. Fig. 4 predicts much better performance of our system compared with the literature data. For example, trichlorfon detection by means of ISFET had a reported limit of detection of ca 250 ppb (5), while conductometric sensor assay registered trichlorfon at ca. 25 ppb (5), still an order of magnitude higher than the described sensor. An amperometric sensor was used to detect dichlorvos with a limit of detection of 350 ppb (2J) and a potentiometric (pH-sensitive) sensor was shown to detect parathion at 39 ppm and diazinon at 35 ppb (9). 

Procedurally, dilutional linearity should not be confused with the MRD of an assay. The MRD is a designed integral part of an analytical method and involves a predefined dilution of test samples, QC samples and, often, calibrators usually with a buffer-based matrix. In contrast, dilution linearity is used only to support analysis of study samples that exceed the assay s ULOQ and involves dilution(s) intended to result in an analyte concentration within the standard curve s validated range. Another notable difference is that, while MRD is usually performed in buffer, dilutional linearity is performed in matrix, often the same one used to prepare the standard curve. 

Calibration standards can be included in the ELISA experiment to produce a sigmoidal standard curve, and can provide the means to quantify the extent of the binding event and hence, the amount of target analyte present in an unknown sample. ELISAs can be employed as quantitative screening assays, depending on the particular performance characteristics of the assay and assuming the use of adequate control samples. 

---