How does the EMC view analytical lab results that indicate
nondetectable or below the detection limit values?
Conceptually, the "limit of detection" is the smallest amount of a substance that an
analytical method can reliably distinguish from zero. More formally, it is the
minimum concentration or amount of a target analyte that produces a signal the
tester can distinguish, at a specified confidence level, from the signal produced by
a blank. Testers sometimes mistakenly use the term "limit of quantification" when
they mean "limit of detection." The limit of quantification is the minimum
concentration or amount of an analyte that a method can measure with a specified
degree of precision. One procedure for calculating the limit of detection is to
collect and analyze a series of at least seven blank or very low-level samples.
The tester then calculates the standard deviation of the results. If the tester
used seven samples multiplying the standard deviation by three will produce a limit
of detection at roughly the 99% confidence level (please see NOTE below).
Each laboratory should report its limit of detection, identify its procedure for
measuring the limit of detection, and label results below the detection limit as
"below detection limit (BDL)." As the measured results approach the estimated limit
of detection, it becomes critical for the tester to determine the limit of detection
for each test program. The tester must report the experimentally measured limit of
detection and the procedure used to measure it whether or not individual results are
above or below the detection limit. If the user averages results of samples from
the same "test," where some results are BDL and other results are above the
detection limit, then the user should substitute the estimated detection limit for
the BDL results. The user should then report the average as "equal to or less than"
the averaged result. If all results are BDL, the user should report the average as
BDL also.
When designing a test program for a particular purpose, you should not use a method
whose limit of detection will affect your decision. In other words, you should
design the test program so substituting any value from zero through the limit of
detection for BDL results will not affect the primary data user's ability to make
decisions based on the data. If your test program uses methods that measure multiple
pollutants, you must decide how to treat the results from all the potential
targets of the test method that are not detected. Some methods designed to measure
multiple pollutants can have many potential targets. Substituting a value other
than zero for all potential measurement targets that are not detected may result in
a total that exceeds regulatory limits even when the method has low limits of
detection. No general guidance can address all the possibilities, so you must
decide how to handle this on a case-by-case basis. In addition, when designing a
test program, the data user and his tester may want to consider potential secondary
uses of the data. Some examples of secondary uses are as inputs for risk
assessments, for characterizing emissions by nonregulatory groups and for
developing emission factors.
Using data gathered for other purposes or "secondary data" to support a decision may
present special problems related to detection limits. The secondary data user may
find published data sets where substituting a value from zero through the limit of
detection for BDL values would affect the user's decision. The potential effect of
the decision should dictate which value (zero through the detection limit) to
substitute. This is best shown with examples. The examples are for illustrative
purposes; they are not the only right ways to evaluate BDL data. In the end, the
data user must decide what is the most suitable way to treat results reported as
BDL.
If the data are used to perform a risk analysis for emissions from a particular
facility, the value of the limit of detection could be substituted for any BDL
results. This would give the highest (worst case) value for the emissions which
would protective of the public health. Some risk assessment evaluations use
uncertainly estimations to produce probabilistic estimations of emissions.
If the data are used to develop an emissions factor, half of the limit of
detection is typically substituted for BDL results. This is consistent with
developing emissions factors to represent average emissions per unit of
production across the industry category.
If the data are used to set emissions limits, different procedures may be used
for different source categories. However, the procedure used to determine
compliance with the emission limit should be consistent with the procedure
used during testing to set the emission limit. For example, when measuring
dioxin/furan emissions from municipal waste combustors, EPA substituted zero
for BDL results during the testing to gather data to support the emission limit.
To be consistent, the Agency allows testers to substitute zero for BDL results
when determining compliance with the municipal waste combustor dioxin/furan
emission limit provided the test run duration is four or more hours.
NOTE: This information was derived from the American Chemical Society Committee
on Environmental Improvement, "Guidelines for Data Acquisition and Data Quality
Evaluation in Environmental Chemistry"; Analytical Chemistry 1980, 52,
pp.2242-2248.
What are Guideline Documents?
The EMC's Guideline Documents
are developed to clarify undefined technical areas of methods
and performance specifications, to set policy for the methods
and monitoring where the method does not address the particular
application necessarily, and to describe alternatives to sampling
procedures within the methods or performance specifications.
Some of the guidance may result in changes to the methods
or performance specifications.
What are Information Documents?
The EMC's Information Documents
are developed to explain the derivation of factors, equations,
and procedures, and to provide technical information necessary
for regulatory agencies to make decisions about acceptance
of data and alternative methods and monitoring.