DETECTION AND QUANTIFICATION LIMITS
The detection limit of an assay is the lowest concentration that can be detected but necessarily quantified; the quantification
limit is the lowest concentration that can be quantified with acceptable precision. The quantification limit is the lowest
level of analyte that can be reported. The ICH guidelines suggest three different methods for determining the detection and
quantification limits. These are: visual determination, signal-to-noise determination, and standard deviation and slope method.
Each method will give different results. The signal-to-noise method is the most logical, because it is based on comparing
low levels of the analyte to a blank or background sample.
Determination of the signal-to-noise ratio is performed by comparing measured signals from samples with known low concentrations
of analyte with those of blank samples, and subsequently establishing the minimum concentration at which the analyte can be
reliably detected. A signal-to-noise ratio between 3:1 and 2:1 is generally considered acceptable for estimating the detection
Using a well-designed experiment can reduce the total testing time. A well-designed experiment can also improve the quality
of the analysis by improving the statistical power. Certain characteristics need to be tested individually because the sample
preparation is unique for the test. Assuming one needed to show acceptable results for all characteristics in Table 1, a testing
scheme could be developed to meet the minimum requirements of the guidance. Three experiments would cover all the characteristics
and their minimum sample requirements.
The first experiment is for specificity alone. A minimum of three different levels for each potential interfering substance,
plus a neat sample, is repeated six times for a total of 24 repeats for the experiment. A one-way analysis of variance is
used to test the level that can be shown to be statistically different from the neat sample. The last level of the interfering
substance that is shown to be statistically equivalent to the neat sample is the level of specificity. Specificity testing
should be conducted for all potential interfering substances. It is not necessary to spike levels of interfering substances
that cannot reasonably be expected to be present in test samples.
The second experiment establishes the detection limit and quantification limit. This experiment is run only for those assays
that require such characterization. If one does not expect the range of the assay to test samples near the quantification
level, this experiment can be eliminated. As stated above, there are several ways to demonstrate detection and quantification
limits. Here we propose a test scheme using the signal-to-noise method of a blank sample. Testing eight repeats of the blank
sample gives a sufficient estimate of the error. Using the standard deviation of the blank sample, the detection and quantification
limits are set. The mean of the eight repeats, plus three times the standard deviation, is the detection limit, while 10 times
the standard deviation is the quantification limit. If additional precision around these estimates is necessary, the number
of repeats is increased. The signal-to-noise method can be used to assess the detection limit, and one of the other methods
can be used to assess quantification limit. Regardless of the method used for assessing quantification limit, a sample at
that level should be incorporated into experiment three.
Table 2. An overview of the three-experiment protocol
The third experiment covers the majority of characteristics required by the guidelines. After the data from experiment three
is analyzed, this data should be compared to the levels of interfering substances and detection limit to show acceptable precision
of the test method at these levels. A minimum of five samples spanning the expecting range of the assay are each tested a
minimum of 12 times (two analysts over two days testing three replicates per day). Usually, multiple analysts and days are
used to estimate intermediate precision. A minimum of two analysts will perform the assay on a minimum of two days, with three
repeats on each day (a total of 12 observations per sample). To meet the linearity requirement, a minimum of five samples
will be used. Table 2 gives an overview of the three-experiment protocol discussed here. More samples and repeats can be added
to the study to gain additional information about the test method.
Table 3. The results of the specificity analysis. Using four samples each with six repeats, an equivocal zone of 0.375 to
0.465 would be used to determine if there are differences among the four levels of analyte. For each level, one would compute
the 95% confidence interval and compare it to the equivocal zone.