Answer to Question #305 Submitted to "Ask the Experts"

Q

When performing a dose assessment for the consumption of vegetables, how are isotopes which are reported as less than the minimum detectable activity (MDA) handled? Is there a national or international standard on how to handle this? I have heard some suggest that if the reported value is less than 1/2 the MDA, use 1/2 the MDA in the dose calculation. Likewise, if the reported value is > 1/2 the MDA but < the MDA, use the MDA in the dose calculation. Any thoughts?

A

This is a good question.There are a number of situations in which we would like to interpret, aggregate, average, or otherwise utilize data containing values that are reported as less than the minimum detectible activity (MDA). The situation is further cluttered by the various minimum detectible levels, lower limits of detection (LLD), etc., and their varying definitions. We are not aware of any national or international standards on how to deal with MDA values in evaluating data sets. This seems like a good topic for a discussion forum and a good assignment for a standards committee. Your question implies that you have values as reported by the laboratory and are concerned with how to treat those values that are less than the laboratory's stated MDA. Several of our experts feel that for averaging or other use in dose calculation, the values should be carried through as reported without artificially adjusting to some value, even if they are less than the method or sample MDA. Radioactivity measurements are variable quantities–results for a sample that is exactly at the MDA will report out as below, at, or above the MDA; conversely, a result that reports out as below the MDA could actually occur when the true activity is below, equal to, or above the MDA. The MDA is not necessarily a speculation for the true activity in an individual sample, but more properly an indication of the capabilities of a measurement system. The reported value is still our best estimate, even if it is below the method MDA. Naturally, use of these less-than-MDA values add uncertainty to the assessment, and the assessment should contain an uncertainty analysis to account for the possible variations in the factors that went into the assessment. Sometimes one needs a method of dealing with the fait accompli where the laboratory has already truncated the data and "less than MDA" is the only value surviving. Several methods are used: - Exclude less-than-MDA values–(this will bias results high and is not appropriate for dose assessment). - Use the MDA–(this will provide an upper bound but will bias the result high). - Assign a value of ½ MDA–(this is probably not a bad approximation in the long run but some individual values will be high or low). - Assign a value of some other fraction of MDA as being close to representing the average value under the truncated tail of the distribution curve. - Assign a value of zero–(this will provide a lower bound but will bias the results low). If only a part of the values in the data set are less than MDA, then various methods have been used or proposed for reconstructing or projecting into the truncated (or "censored") region from the noncensored data and the assumed frequency distribution. This answer was prepared with input from Brian Birky, Ph.D., and Tom Borak, Ph.D., CHP. C. E. Roessler, Ph.D., CHP