Laserfiche WebLink
DEPARTMENT OF HEALTH SERVICES <br /> TOXIC SUBSTANCES CONTROL PROGRAM <br /> TOXICOLOGY AND RISK ASSESSMENT UNIT <br /> SCIENTIFIC GUIDANCE MEMO 89-2 <br /> August 4 , 1989 <br /> Page 2 of 4 <br /> Use of Non-Detect and Limit of Detection Values <br /> in Exposure Assessments <br /> Guidance In The Use Of Non-Detect And Limit Of Detection Values <br /> 1 . Professional judgement should be exercised to distinguish <br /> those chemicals reported as ND which are actually likely to be <br /> oresent at levels below the limit of detection, from those <br /> chemicals for which there is a high probability that the ND value <br /> truly represents a concentration of zero. <br /> 2 . Within either the characterization report or the risk <br /> assessment report, the analytical data on the concentration of <br /> chemicals in various media should be statistically analyzed to <br /> determine whether the frequencies of concentrations fit either a <br /> lognormal or a normal distribution. If these data are not <br /> available, consider requesting that the report be revised to <br /> include such an analysis. The default alternative is to assume <br /> that the data are lognormally distributed. <br /> 3 . As pointed out by Hornung & Reed (1987) and EPA (1988) , when <br /> chemical concentrations in a medium are lognormally distributed, <br /> as is generally the case fcr sclvent plumes, the best estimate, <br /> for statistical reasons, of a reported ND value is the limit of_ <br /> detection (LOD) divided by the square root of 2 (1. 414) . When <br /> determining the mean of several values, some greater than LOD and <br /> some ND, this method should be used to assign a non-zero value to <br /> ND values , unless the data justify an alternate treatment. <br /> 4 . The geometric mean, rather than the arithmetic mean, should <br /> be used to estimate the average concentration of a chemical in a <br /> defined volume of a medium which contains the entire plume. The <br /> geometric mean assumes that the concentration of the chemical in <br /> the medium is lognormally distributed, including NDs. Values <br /> reported as ND should be treated as defined in point 3 above. <br /> Geometric mean = (Cl * C2 * C3 * . . . * Cn) l/n <br /> C1 = Concentration of a specific chemical in sample 1 <br /> C2 = Concentration of a specific chemical in sample 2 <br /> C3 = Concentration of a specific chemical in sample 3 <br /> n = number of samples <br />