Method Validation vs Quality Control Flashcards
What is the difference between validation and quality control?
Validation is the process of showing that a method is good enough for intended use
Quality Control is the process of showing that the method works with real sample analysis
What are the categories of validation?
Selectivity Calibration Detectability Accuracy Precision
What is the matrix effect? What is interference? What do the lines on a graph look like?
These are components of selectivity FYI
The matrix effect is a proportional error due to an interaction with the analyte. (Lines have intercepts at 0 but different slopes)
Intererence is the predeterminatant error due to a FIXED amount of substance in the sample (lines are parallel)
How do you assess selectivity criterion for immunoassays? When do you accept the criterion?
By calculating the Percent Crossreactivity (%CR)
%CR = 100 x (response of a test compound with a given mass)/(response of the analyte at the same mass)
You accept when %CR < 0.1%
Differentiate between:
Limit of Detection (LOD)
Limit of Identification (LOI)
Limit of Quantitation (LOQ)
What are the standard deviation values compared to the noise?
LOD is the lowest concentration in which you can say that the desired substance is present in sample (3SD)
LOI is the lowest concentration in which you can say that the desired substance is present WITH certainty (6SD)
LOQ is the lowest concentration that can be reliably measured quantitatively (10SD)
In an analytical chromatogram, what is Npp? What is w?
Npp = peak to peak noise w = Peak elution width
What confidence interval is used for LOD and LOI?
99%
If you were to only use the LOD to argue the presence of a substance, how often will you be correct? What are the types of errors? What can you do to correct this?
You would only be right 50% of the time. To correct it, you can use the LOI.
Errors:
Beta: false negative
Alpha: false positive
What is total error equal to? (also called bias)
The sum of precision and accuracy errors
How do you reduce random (systematic) error?
Do more experiments (repeated experiments)
When looking at the graph with diagnostic values of lab tests, how do we modify the amount false positives/negatives? What if we move it too much to the right or left?
We can modify on the graph by moving the cuttoff line (dotted line). If placed too far to the right, false negatives will start to increase. If placed too far to the left, false positives will start to increase.
What are some examples that you would want to avoid alpha or beta error ?
Cancer patients - avoid beta error bc you dont want to send someone home that actually has cancer
Healthy patients - avoid alpha error bc you dont want to treat someone for heart failure that doesnt have it
