Always check for changes in the Laboratory Procedure Manual for the environmental chemical of interest when considering NHANES data analysis, especially when combining data across survey cycles. The Laboratory Procedure Manual is released for each NHANES cycle. This example is the 2003–2004 version of the NHANES Laboratory Procedure Manual [PDF - 19.9 MB].
Changes over time in laboratory methods, laboratory locations, or other laboratory aspects may influence temporal patterns in the data or when combining data across survey cycles. The Laboratory Procedure Manual will help to identify factors that may need to be considered when analyzing the NHANES laboratory results or interpreting the findings. An example of this would be the improved 2003–2004 results for urinary 2,4-dichlorophenol (Appendix B: Fourth National Report on Human Exposure to Environmental Chemicals).
Changes may also be noted in the data documentation. For example in the 2005–2006 release for urine mercury and iodine the documentation states “There were no changes to the site or laboratory from the previous two-year cycle. Beginning in 2005 urinary iodine and mercury were tested from the same instrument.”
Scientific consensus standards may be developed regarding individual chemicals in an effort to provide more valid national and international standards for their laboratory measurement. This permits valid comparisons across the various scientific studies being published in the literature. When such standards are developed, NHANES may update its datasets and/or documentation to provide guidance.
In addition to possible laboratory changes, check laboratory documentation specifically for an environmental chemical, for possible changes in the subsample design, or for targeted demographic groups used to obtain specimens for measurement of that environmental. For example, in the 1999–2002 NHANES, urine mercury concentrations were measured in all women ages 16-49 years. In 2003–2004, urine mercury concentrations were measured in a one-third subsample of participants ages 6 years and older.
While data files may be renamed, updated or changed infrequently, it is useful to look carefully at the location of laboratory data files on the NHANES website. Some data files may be included in larger data and documentation files for one data release, but may be in a separate or independent data for the next two-year cycle. Before 2003–2004, some environmental chemical documentation and data files were contained within the Nutritional Biochemistries file (Lab 06) on the NHANES website. In more recent years, separate data files and documentation were released for each set of related environmental chemicals. Module 4 of the Continuous NHANES Tutorial may assist in learning to identify variable names.
Also, it is strongly suggested that you check for updated data files and documentation when accessing older data files, even if you have used that data or documentation previously. On occasion, laboratory results are modified, new laboratory findings are added, or further documentation may be provided. When this occurs previous data files or documentation that you may have downloaded may not be the most current data set available. For example, the data documentation for 2003–2004 Volatile Organic Compounds in Blood and Water (Lab04) was originally released in August, 2008, and updated in June, 2009 due to a change in the elevated detection limits reported for a small subset of the data. A notice for such updates can usually be found either as a comment next to the Environmental Chemical data file name on the NHANES Webpage, in the new documentation itself, or listed in the “What's New” link in the “Contents at a Glance” section of the NCHS Webpage. When an update is released, the previous data and documentation are removed from the NHANES Website.
Estimates based on pooled specimen measures have important limitations that should be considered. Most environmental chemical measurements have been shown to have a log-normal distribution. However, the measured value for a pooled sample is comparable to an arithmetic average of log-normal results and thus represents a biased estimate of the central tendency of the samples making up the pool. Also, results from pooled specimens do allow an assessment of the variability of the environmental chemical concentrations in the population. Methods have been proposed for correcting the bias associated with using data from pooled samples with a log-normal distribution. One potential approach has been published by Caudill et al.
Reference: Caudill S.P., Turner W.E., and Patterson Jr. D.G. Geometric mean estimation from pooled samples. Chemosphere 2007: 69: 371–380.).
Detailed discussion of the uses and limitations of pooled samples is beyond the scope of this tutorial, but further discussion of this topic is available in the following references:
In a cross sectional study because both exposure and disease are determined simultaneously for each subject: it can be considered a snap shot of the population at a certain time point. It is generally not possible to know what came first, the exposure to the potential risk factor or the disease (Szklo and Nieto, 159).
Causality cannot be readily established in a cross sectional study design like NHANES. Analysts need to be cautious when interpreting results.
Szklo, Moyses and F. Javier Nieto. 2000. Epidemiology Beyond the Basics. Gaithersburg: Aspen Publishers, Inc.
Many factors that can influence the results and may be of interest in your analysis may not be available in every NHANES data release cycle as the content of the survey changes periodically. Other variables of interest may be available in more detailed data files maintained within NCHS or by other agencies, (e.g., EPA), While NHANES provides for many additional demographic and medical conditions, there may be other desired factors of interest that NHANES has not been able to measure.
Examples of potential analytic covariates available in NHANES are demographics (age, sex, or socioeconomic status) or lifestyle or behaviors factors (such as smoking status and diet). Other important variables may have been collected and are only available in the NCHS Research Data Center, including detailed occupation data, survey location, or season at the time of examination.