## You are here

View all publications

# Appendix F: Determination of Method Detection Limits

## Calculation of method detection limit (MDL)

Method detection limit (MDL)is the minimum concentration of a substance that can be measured and reported with 95% confidencethat the value is greater than zero. It is determined from analysis of a sample of a given matrix containing the analyte and considers all of the analytical operations on a sample (sub-sampling, extractions, digestions, dilutions, reagents, instrument parameters, etc.). It is the preferred term used by the US EPA and corresponds to the 'criterion of detection' of ASTM.

As part of a method validation, a laboratory should calculate a99% confidence MDL by analysing a low-level real matrix sample, containing the analyte at levels 2-10 times the expected detection limit. The analysis is performed in triplicate on three separate occasions at least one day apart. The MDL is then calculated using the formula:

MDL = 2.896 x (standard deviation of pooled nine results).

Practical quantitation limit (PQL) (US EPA) is the lowest level of quantitation that can be reliably achieved within specified limits of precision and accuracy during routine operations. The PQL may be 10 to 100,000 times or more greater than the MDL for a method. For most environmental samples the PQL is taken as 5-10 times the MDL. The converse of this is that if a desired standard level (such as a guideline value) is the target against which results will be compared, the laboratory MDL should be 5-10 times lower than the standard level.

## Comparing data from different laboratories

The following points should be checked when comparing data and detection limits between laboratories.

• Which detection limit is being quoted? An instrument detection limit (IDL) will always be much less than an MDL.
• What confidence level is used for calculating the MDL? A 95% MDL will be two-thirds of a 99% MDL.
• Is an MDL calculated from 'blanks' (ie, laboratory-grade deionised water) or a real matrix? Blanks will always give an MDL much less than a real matrix MDL.