Statistical Data Evaluation

Repeatability is a measure of the method’s precision obtained under the ideal measurement conditions (single instrument, single analyst, a short period of time) and can be evaluated in two ways in EffiChem:
– From multiple measurements on a level
– From parallel measurements

Intermediate precision is a precision measure between repeatability and reproducibility. It can be evaluated from multiple measurements of parallel measurements per level.

Reproducibility is a precision measure obtained under the worst measurement conditions (multiple instruments, multiple analysts, a long period of time) and can be evaluated from multiple measurements or from parallel measurements per level.

Method accuracy describes the degree of agreement of a measurement with a reference value/method. The accuracy can be evaluated in several ways, depending on the concentration range and the data availability:
– Limited concentration range – sample reconstitution possible
– Limited concentration range – reference material available
– Large concentration range – blank available: regression
– Large concentration range – blank not available: t-test
– Large concentration range – blank not available : regression
– Comparison of two methods/laboratories: t-test by levels
– Comparison of two methods/labs: t-test for difference in results

Linearity describes the degree of linear dependence between the concentration and the measured signal. It is evaluated in different ways depending on the availability, type, and design of the validation data:
– Correlation and QC coefficient
– ANOVA for Lack of fit
– Significance of the quadratic term
– Sign test

The detection and quantification limits are concentration levels at which the signal can be distinguished from the noise or the concentration of the analyte being measured can be reliably quantified. The limits can be evaluated in several ways:
– 3s – IUPAC
– 3s – blank correction
– 3s – continuously measured blank
– From the calibration line

Robustness is an ability of the method to cope with small changes in the method setup or in the adherence to the prescribed conditions, such as pH of the mobile phase in HPLC or detector temperature in GC, without having impact on result. Robustness can be evaluated in EffiChem in two ways:
– Dong’s algorithm
– AOAC with evaluation of

Selectivity is the ability of a method to selectively measure the concentration of the analyte of interest in the presence of potent interferents. Selectivity can be evaluated in several ways:
– By comparing calibration lines
– By comparing the results with a standard
– By comparing the results with measurements without interferents

Sensitivity is the measure of change in the measurand (signal) per unit change in the concentration. It is calculated as the slope of the calibration line between the concentration and the measurand.

A blank experiment measures the magnitude of the measurand in a sample with zero concentration. It is the basis for calculating the limit of detection and the limit of quantification.

System suitability check is the confirmation that the system meets the prescribed parameters, namely PhEur requirements or USP/FDA requirements:
– Repeatability of injection according USP
– Capacity factor according to USP
– Resolution according to USP
– Tailing according USP
– Theoretical plate number according USP
– Theoretical plate number according PhEur
– Resolution according PhEur
– Symmetry factor according PhEur

Control charts are developed to monitor and evaluate the degree of variability in the method or results over time, to evaluate trends and to set warning and action limits. In EffiChem, there are several types of control charts to select from, depending on data availability, design, and range:
– Individual values – control characteristics determined
– Individual values – control characteristics not determined
– Repeated measurements – control characteristics determined
– Repeated measurements – control characteristics not determined
– Multivariate-variate control chart
– X-diagram (Control chart for Average)
– R-diagram (Control chart for Range)
– Western-Electric rules for evaluation of control charts

Quantifying uncertainty in analytical measurement is one of the key requirements of ISO17025. Several approaches or methods can be used for this purpose, depending on the situation and data availability:
– From Precision data – multiple measurements
– From Precision data – parallel measurements
– From Control charts – individual measurements
– From Control chart – multiple measurements

A linear or quadratic calibration model allows to estimate the concentration or the property of interest of the unknown sample from the value of the measurand. In addition, the calibration model is used for other purposes, e.g. to determine linearity, limit of detection, limit of quantification, and to determine the method sensitivity.

Inter-laboratory comparisons is used to assess whether a laboratory participating in an interlaboratory comparison is reporting results that are significantly different from those of other laboratories, or not. Furthermore, the inter-laboratory comparison can be used to evaluate the method’s repeatability and reproducibility. The following algorithms are included:
– Mandel statistics H
– Mandel statistics K
– Collaborative studies

The audit trail and history are kept for all Statistical Data Evaluations conducted. All records are 100% traceable and cannot be falsified or deleted. Undesired records can be moved to a bin with a rationale provided.

Statistical Data Evaluation can be either linked to the Methods LIMS module, or configured separately, within one single validated system, meeting the data integrity requirements defined.