Run Number

Run Number

Figure 4 Use of sample controls for trend analysis on variability. Low, middle, and high sample controls of a biomarker were monitored in Levey Jennings control charts. There was a noticeable shift in trend in all the sample control levels after run 27. The average, upper and lower control limits of analytical runs up to run 27 (B) were compared to the overall parameters (A). (See insert for color reproduction of the figure.)

LCL =11.80

The Levey Jennings control charts showed a noticeable shift in trend at all the SC levels after run 27, indicating a systemic bias. The average, upper, and lower control limits of analytical runs up to 27 (right panels) were different from the overall values (left panels). This shift was traced to a change in commercial kit lot.

Method Validation on Performance Parameters

The foremost assay performance parameters to be established for quantitative methods are precision and accuracy. In addition to these two parameters, sensitivity, selectivity, specificity, stability, and reproducibility should also be demonstrated. These parameters are established during prestudy method validation to demonstrate reliably that the method is fit for the purpose of biomarker characterization over multiple clinical studies. If more than one bioanalytical laboratory is used, each additional laboratory should also demonstrate its ability to perform by conducting a full or partial method validation.

Accuracy and Precision Method performance is consistently better understood and validated through the appropriate use of control samples. To ensure data quality, assay performance is evaluated during method validation with validation samples (VSs) and monitored during sample analysis with quality control samples (QCs). VSs are used in method qualification or validation to define intra- and interrun accuracy and precision, providing data to demonstrate the robustness and suitability of the assay for its intended application. On the other hand, QCs are essential in determining run acceptability during specimen analysis. However, many laboratories have not made the distinction, using the term QC for both VS (for prestudy method validation) and QC (for in - study run acceptance).

For a well - defined biomarker assay, at least six accuracy-and-precision validation runs should be performed to provide statistical data to calculate these assay parameters (DeSilva et al., 2003; Lee et al., 2008). Each run should consist of standards prepared by spiking the reference standard into blank biological matrix or an alternative blank matrix. QC/VS at low, middle, and high concentrations should be prepared in the biological matrix or alternative matrix by spiking the reference standard. In addition, sample controls (SCs) should be prepared by pooling authentic samples at the low and high levels of the biomarker to reflect the performance of the endogenous biomarker in assay precision, stability, and reproducibility. At least two sets of QC/VS and SC should be run with the standards in each accuracy and precision experiment.

Accuracy and precision can be evaluated from the total error of VS data from the validation runs in a way similar to that of the macromolecular protein drug as analyte (DeSilva et al., 2003). However, given biological variability and other factors in biomarker research, more lenient acceptance criteria may be used for biomarker PD than that for PK studies. Still, it should be recognized that accuracy and precision data of VS in buffer provide only a relative quantification, which may be quite different from measurements in the authentic matrix.

Concentrations of the endogenous SCs will be determined by multiple validation runs. The true values are then defined after sufficient data collection from pre- and in-study validation. For example, the mean of 30 runs and 2 standard deviations can be used to define the target concentration and acceptable range of the SCs. Because the reference material may not fully represent the endogenous biomarkers, the SCs should be used for stability tests (such as minimum tests of exposure to ambient temperature and freeze-thaw cycle). In addition, the SC can be used for in-study, long-term storage stability analysis and for assessment of lot-to-lot variability of key assay reagents.

Regression models are essential for data calculation from sigmoid curves like those in LBAs. The most commonly used four- or five - parameter logistic regression should be evaluated in conjunction with weighting factors during method development. Final decisions on which curve-fitting model to use should rest on which offers the best fit for all the standards in the precision profile. In some cases, a less than optimal fit will suffice to allow for greater assay sensitivity.

Sensitivity To provide data for PK/PD studies, assay sensitivity is usually defined by the assay lower limit of quantification (LLOQ), which is the lowest concentration that has been demonstrated to be measurable with acceptable levels of bias and precision and total error (the sum of bias and precision). However, low-concentration clinical samples may fall below the LLOQ (i.e., the method lacks the required sensitivity). In some instances, the investigator may want to use values below the LLOQ but above the limit of detection (LOD), to obtain a numerical estimate of the changes while recognizing the high variability below the LLOQ region.

LOD is often used as the analytical "sensitivity" of the assay in a commercial diagnostic kit. A common practice of diagnostic kits to determine the LOD is the use of extrapolated concentrations from a response signal of +3SD (or -3SD for a competitive assay) of the mean background signal from 30 or more blank samples. The National Committee for Clinical Laboratory Standards (CLSI) has recommended an approach for determining LOD (National Committee for Clinical Laboratory Standards, 1999; Tholen et al., 2004). This statistically sound approach evaluates the limit of blank (LOB) as type I error and LOD as type II error from sufficient numbers of blank and low concentration samples with normally distributed analyte concentrations. One approach of calculation is described briefly as follows: LOB is estimated from the blank result at the 5th percentile position. LOD is calculated as

SDs is the estimated standard deviation of the low-concentration sample measurement, and cp is a correction factor for the population bias:

Project Management Made Easy

Project Management Made Easy

What you need to know about… Project Management Made Easy! Project management consists of more than just a large building project and can encompass small projects as well. No matter what the size of your project, you need to have some sort of project management. How you manage your project has everything to do with its outcome.

Get My Free Ebook

Post a comment