There is always confusion on how low a lab can go when it comes to what their instruments can detect.  The EPA and TNI came together and changed how Method Detection Limits (MDL) are established for an instrument or group of instruments. They wanted to account for the inherent contamination that occurs when samples are prepared. They also wanted, by this change, to stop labs from creating non-supportable low detection limits.

What does MDL mean?

So what does MDL really mean, and why do we care about how it is prepared?  Under the new definition, it means a lab can be 99% confident that a result is a detectable hit above the background contamination found in prepared blanks.  This is probably not clear to you because it is not clear to those in the laboratory industry.  Let me break down a simplified explanation of what is going on.

Every instrument has an electronic signal that is influenced with how much “noise” is created when all the moving parts and electronics combine together (the output of which is then interpreted by a signal detector).  For a lab to be able to give you legally defensible results, those results have to be somewhere above the background signal (the noise) that were determined by an MDL study.  The study is undertaken to find the beginning of the method’s confidence level, this level is where 99% of the recoveries are discernable/detectable above the background signal.  Therefore, 1% of those results that come in above the background signal are statistically not valid.  To establish the confidence level where we can account for the 1% error rate we establishing reporting limits that are greater than the MDL baseline. The area between the MDL and the reporting level is a no man’s land and does not have legally defensible confidence. Results in this area are qualified as estimated to account for the potential for error.  

How low can you go?

Here is the answer to question posed above – only at or above the reporting limit can a lab report to you results with confidence and without qualifications – but with qualifications they can report down to the MDL level. 

The “old” way of looking at MDLs never accounted for the inherent bias of the noise associated with blank analysis; this and other practices, led to artificially low MDL values that were then marketed to capture or retain business.  The “new” way to look at MDLs takes into account blank results versus spiked results.  The MDL value is statistically derived based upon a minimum of seven blanks versus seven spiked samples taken over a two year period.  After the compilation and calculations of these results, the larger number between the two becomes the MDL baseline. 

Statistically determined MDL results should be cut and dried right? Its science… what could go wrong… plenty.  Laboratories can manipulate study parameters, adjusting spiking levels and how samples and blanks are prepared in order control variability.  Variability has a direct influence on how a constituent of the calculation (the standard deviation) is established, which can lead to a lower (and more marketable) MDL baseline.  This means that limits may be forced lower than they should be – funny, I just mentioned this twice it must be important.  Which is why the EPA and TNI decided to look at blanks and spikes quarterly over a two year period to account for a more natural variance which is a little more difficult to manipulate.

The big take away

So what can you take from this blog? Well if you only want to be 99% confident that the potential contaminant that you are looking for is there, then yes, the MDL is as low as you can go.  You can do this by asking the lab to report down to the MDL. If you ask a lab to report data between the MDL and the reporting limit then that data should be qualified as estimated and more than likely it will not be acceptable for most government permit requirements.  But if you want to do legally defensible regulatory work where the 1% error rate is not acceptable then you must accept nothing lower than the lab’s reporting limit.