Task 2: How to Address Different Fill Values and the Limit of Detection (LOD) in NHANES Environmental Chemical Data

Step 1: Use Indicator Variable Included in Dataset to Determine LOD

The Limit of Detection may change for an environmental chemical across survey cycles. Examine the summary indicator variable to identify individual results at or below the Limit of Detection of the assay. This variable usually categorizes the results as “0” (at or above the LOD) or “1” (below the limit of detection). At time you will find a “2” to show that the result was above the upper measurement limit and exceeds the calibrated range of the laboratory assay.

warning iconWARNING

Be aware that some chemicals may have multiple limits of detection on the same data file (due to changes in the LOD during the NHANES cycle). For some chemicals (e.g.dioxins) there may be a different LOD in each sample test mostly because the sample volume used for analysis differed for each sample. So all records below the limit of detection will have an indicator value “1” to indicate that the result is below the limit of detection. However, the below LOD value recorded in the field for the laboratory finding may not all the same.

For earlier releases of NHANES environmental chemical data, an indicator variable may not have been included in the data file.  For more recent data releases LOD indicator variables are provided.  If no indicator variable is available, you can follow the next steps to determine the Limit of Detection.

Step 2: Use Laboratory Methods Documentation to Determine LODs

The limit of detection information is often provided in the Laboratory Methods documentation. In this example, you will be looking for LODs for iodine and mercury in urine for the 2005–2006 survey cycle. First, go to the NHANES 2005–2006 Laboratory Files page, located at: http://www.cdc.gov/nchs/nhanes/nhanes2005-2006/lab05_06.htm. Then, click the Laboratory Methods link. Scroll down and click the Urinary Iodine and Mercury link.

After the PDF file opens, search for 'limit of detection'. In the Limit of Detection section, you find that the Limits of Detection is based on three times the standard deviation of approximately 20 or more low concentration samples, each analyzed in a separate run.

If the actual limit of detection is not provided in the laboratory methods documentation, follow the next step to determine the LOD.

Step 3: Determine CDC Exposure Report Limit of Detection Values

The CDC National Report on Human Exposure to Environmental Chemicals includes the various chemical LODs . Users can use the information found here to create their own summary variable to indicate whether or not a chemical has a value above or below the Limit of Detection. This strategy can also be adopted  when there are multiple fill values for a particular analyte of interest.

For example, in the Phthalate dataset, there are multiple fill values noted for Mono-cyclohexyl phthalate in 1999–2000 and 2001–2002. You would like to create an indicator variable for Mono-cyclohexyl phthalate (URXMCP) values over the various LODs used. To do this, follow the conservative approach outlined below. 

Go to CDC's National Center for Environmental Health website, located at: http://www.cdc.gov/nceh/.

Click National Report on Human Exposure to Environmental Chemicals link under Laboratory Sciences.

Click the Data Tables by Chemical Group link in the Explore the Report box on the right.

Since your example is a phthalate, click the Phthalates link.

Scroll to the bottom of the page and click the Mono-cyclohexyl phthalate link under Data Tables.

Review the Limit of Detection (LOD) information for Mono-cyclohexyl phthalate. The table in Appendix D notes for 1999–2000 the maximum LOD is 0.9, for 2001–2002 it is 0.3, and for 2003–2004 it is 0.2. 


Step 4. Create an Indicator Variable

As previously stated, a conservative approach would be to use the highest LOD fill value for all years of data being analyzed. For example in 1999–2000, Mono-(2-ethyl)-hexyl phthalate, (URXMHP) had an LOD fill value of 0.8 ng/mL and in 2001–2002 the fill value was 0.7 ng/mL. To use the most conservative approach an analyst can change the LOD fill value to be less than or equal to 0.8 ng/mL for the four survey years (1999–2002). Therefore, for 1999–2002 all values between 0.7 ng/mL and 0.8 ng/mL would be considered below the LOD. An LOD indicator variable for all records in all of these years would then be created for any record a value of 0.8 ng/mL or below (excluding missing values). For the Mono-(2-ethyl)-hexylphthalate example, you might use the following SAS code:

Sample Code

if URXMHP>0.8 then MHP_aLOD=1;
else if URXMHP>. then MHP_aLOD=2;

Alternately, use the LOD fill value for each survey cycle to create an indicator variable for values over the LOD. For the Mono-(2-ethyl-hexyl phthalate example, you might use the following SAS code:

Sample Code

if (sddsrvyr=1 and URXMHP>0.8) or 
         (sddsrvyr=2 and URXMHP>0.7) or 
         (sddsrvyr=3 and URDMHPLC=0) 
         then MHP_aLOD=1;
      else if (sddsrvyr=1 and URXMHP=0.8) or 
              (sddsrvyr=2 and URXMHP=0.7) or 
              (sddsrvyr=3 and URDMHPLC=1)
              then MHP_aLOD=2;


Step 5: Distinguish Results that are Missing from Results that are below the LOD

In all NHANES laboratory data sets, results below the lower detection limit, (“non-detectable”) are assigned a “fill” value that is calculated as the lower limit of detection divided by the square root of 2. This distinguishes persons with any result at all from those with missing laboratory results due to absent or insufficient specimen volume or due to nonparticipation (i.e., did not provide blood or urine). 

To understand the use of the fill value for results below the LOD, go to the NHANES 2005–2006 Laboratory files page and click the “Docs” link next to Serum Cotinine. Scroll to the Codebook section and review the codes for LBXCOT.  In 2005–2006, the detection limit for cotinine was 0.015 ng/mL and the fill value assigned was 0.011 ng/mL.

Cotinine (ng/mL)
Code or Value Description Count Cumulative
0.015 to 1156

Range of Values




Below Limit of Detection








close window icon Close Window to return to module page.