Home > Technical Articles > Detection Limits
Introduction
Detection Limits (DLs) are estimates of concentrations at which we can be fairly certain that the compound is present. Concentrations below this limit may not be detected. Concentrations above this limit are almost certainly detected in the analysis. Using statistics, the certainty of detection can be quantitated as 99%. Therefore, Not Detected (ND) indicates that the analyte may be present below the DL. Some labs report DLs without actually measuring them. If you don't believe their claims, ask to see their data. DLs are usually determined on "clean" samples. In reality, each sample will have its own DL that is determined by the matrix of the sample. The "dirtier" the sample, the higher the DL. What does "clean" and "dirty" mean? A clean sample is one that does not have an interference that affects the analyte of concern.
Here are answers to five of the most common questions about DLs:
What is the difference between DL, IDL, MDL, PQL, etc.? An IDL (Instrument Detection Limit) is the lowest limit that the instrument can detect. It is determined on samples which have not gone through any sample preparation steps. An MDL (Method Detection Limit) is similar to an IDL, but is based on samples which have gone through the entire sample preparation scheme prior to analysis. A PQL (Practical Quantitation Limit) is normally 3 to 10 times the MDL and is considered the lowest concentration that can be accurately measured, as opposed to just detected. DLs are actually determined by analysis of replicate low-level samples or blanks. This information gives the variation in instrument response at levels near the detection limit, from which 99% confidence limits are calculated from the standard deviation.
Why can't you raise the DL in a report so that the hits are no longer reported? Ethics. We must report what we find. Audits of the data would reflect detection. You may want to discuss DL requirements with us before we analyze the sample.
Why can't you just use more sample to get a lower DL? On a clean sample this can be done, and in fact there are many methods that require large sample volumes or concentrating steps. You must remember that the key to this is clean samples. A larger amount of a dirty sample will just increase the interference.
Why is the DL in one sample different from another? This problem usually occurs when either the sample batch is not homogenous or the site is not similar from location to location, requiring some samples to be diluted or because of varying amounts of interferences.
When a result is reported at the DL, how sure are you that its there? We will not report an analyte if we are not sure that it is present. We may not be as confident in the amount of an analyte that is present. Since near the detection limit the signal:noise ratio is low, it is difficult to accurately measure the signal, leading to increased uncertainty near the detection limit.
Estimating Detection Limits
One option that the FDA allows in estimating detection limits is a measurement of signal to noise (S/N) ratio. The detection limit is then determined as the concentration (or amount) of an analyte which will give a signal three times the background noise. This method may be adequate for techniques which have an almost instantaneous change in signal from background to sample (e.g. flame AA) but we believe the S/N ratio is misleading for most other techniques.
For example, in any chromatographic method the response of an analyte gradually changes over time. It is obvious that the software used to detect this change from background must also be considered in the determination of detection limits. For this reason we recommend that detection limits be determined by the analysis of low level standards or samples. By doing so, the precision of the entire response measurement system is measured, and a statistically valid 99% confidence limit can be calculated from the standard deviation of the response. This method of determining a detection limit is another option allowed by FDA and the one that we prefer.
In any case, these measurements only provide an estimate of the detection limit and certainly no better than one significant digit. And remember, anything that changes the sensitivity of the method will change the detection limit.
We are sure that there are many more questions which cannot be covered in a newsletter of this size.
9240 Santa Fe Springs Rd
Santa Fe Springs, CA 90670562.948.2225 Fax 562.948.5850
Formerly West Coast Analytical Service (WCAS) and Bodycote Testing Group
