Estimate of Biological Variation of Laboratory Analytes Based on the Third National Health and Nutrition Examination Survey (original) (raw)

Journal Article

,

address correspondence to this author at: National Center for Health Statistics, 3311 Toledo Rd, Room 4215, Hyattsville, MD 20782; fax 301-458-4028, e-mail dol2@cdc.gov

Search for other works by this author on:

,

Division of Health and Nutritional Examination Surveys

, National Center for Health Statistics, Centers for Disease Control and Prevention, Hyattsville, MD

Search for other works by this author on:

Division of Health and Nutritional Examination Surveys

, National Center for Health Statistics, Centers for Disease Control and Prevention, Hyattsville, MD

Search for other works by this author on:

Published:

01 February 2005

• PDF
• Split View
• • Article contents
  • Figures & tables
  • Video
  • Audio
  • Supplementary Data
• Cite

Cite

David A Lacher, Jeffery P Hughes, Margaret D Carroll, Estimate of Biological Variation of Laboratory Analytes Based on the Third National Health and Nutrition Examination Survey, Clinical Chemistry, Volume 51, Issue 2, 1 February 2005, Pages 450–452, https://doi.org/10.1373/clinchem.2004.039354
Close

Navbar Search Filter Mobile Enter search term Search

Laboratory analytes for individuals are subject to several sources of variation, including biological variation, preanalytical variation (specimen collection), analytical variation (bias and imprecision), and postanalytical variation (reporting of results). Biological variation consists of within-person (WP) and between-person (BP) variation. These components of biological variation are used to set analytical quality specifications for bias and imprecision, evaluate serial changes in individual analytes, and assess the clinical utility of population-based reference intervals.

Desirable quality specifications for imprecision (I), bias (B), and total error have been related to the WP CV (CVw) and the BP CV (CVg) of laboratory analytes (1)(2)(3). Imprecision should be ideally less than one half of the CVW, and bias should be <0.25[(CVw)2 + (CVg)2]1/2. The quality specification for total error is to be less than _k_I + B, where k = 1.65 at α = 0.05. The total CV (CVt) can be estimated assuming that the CVs of all sources are measured at the same analyte mean and that pre- and postanalytical sources of variation are negligible. The CVt = [(CVa)2 + (CVw)2]1/2, where the analytical CV (CVa) equals the laboratory method imprecision (CVi) if there is no bias present.

Estimates of CVw and CVg for laboratory analytes were derived from the Third National Health and Nutrition Examination Survey (NHANES III) conducted from 1988 to 1994 (4)(5). NHANES III was a cross-sectional survey that collected data on the civilian noninstitutionalized US population through questionnaires and medical examinations, including laboratory analytes. NHANES III used a stratified, multistage probability design to collect a nationally representative sample. The laboratory methods including imprecision (CVi) for NHANES III have been described (6).

The BP and WP means, SDs, and CVs for 42 general biochemical, nutritional, immunologic, environmental, and hematologic analytes are listed in Table 11 . The BP and WP variations were estimated on 24 978 and 2426 sample persons, respectively. The WP sample, ∼10% of the sample persons, was recruited for a second analyte measurement. The WP sample was not selected randomly, but with the goal for obtaining approximately equal proportions of males and females with one half between 20 and 39 years of age and one half over 40 years of age. When possible, the second examinations were scheduled at the same time of day as the first examinations. Compared with the BP sample, the WP sample was older (mean age, 42.9 vs 30.8 years), had more non-Hispanic whites (42.2% vs 34.7%), and had fewer Mexican Americans (25.4% vs 30.6%). There was no statistically significant difference in gender proportions between the BP and WP samples.

The CVg values were estimated using a weighted, complex sample design by Taylor series linearization (7). The BP standard deviation was calculated as (SE2 + SDsrs2)1/2, where SE is the standard error of the mean obtained with the complex design and SDsrs2 is the square of the standard deviation assuming a simple random sample. The WP variations were estimated from a nonrandom, unweighted sample with a mean (SD) of 17 (8) days between two analyte measurements. The CVw was calculated as [(CVt)2 − (CVa)2]1/2. The distributions of several analytes were nongaussian, and extreme outliers were excluded to obtain an approximately gaussian distribution with more stable estimates of variation. Outliers were eliminated by use of Tukey’s method, which defines outliers as three interquartile ranges below the 25th percentile or above the 75th percentile (8). Statistical analyses were carried out with SAS for Windows software (SAS Institute) and SUDAAN software (Research Triangle Institute).

The CVw and CVg exceeded the CVi for laboratory analytes (Table 11 ). For most laboratory analytes, the mean BP and WP analyte values were similar despite some demographic differences between the two groups. The analytical quality specifications for imprecision and bias can be judged by use of the CVw and CVg. For example, the total cholesterol imprecision should be less than one half of CVw (8.2%), or 4.1%. The method imprecision was 2.3%. The bias for total cholesterol should be <0.25[(CVw)2 + (CVg)2]1/2, or 0.25[(0.079)2 + (0.226)2]1/2, or ∼6.0%. The quality specification for total error is estimated as B + 1.65(I), or 6.0% + 1.65(4.1%), or 12.8%. This compares with the National Cholesterol Education Program (NCEP) performance criteria of <3% for imprecision and bias and 9% for total error (9).

Serum sodium had the lowest CVg (1.6%) and lowest CVw (1.3%). This reflects the narrow homeostatic range for sodium that the body maintains. High CVw and CVg values were seen for several analytes. High CV values could result from natural population or individual variations, diurnal variations, disease, outlying analyte values, and relatively lower analyte values.

The ratio of CVw to CVg, also known as the index of individuality, is important in determining the use of population-based reference (normal) intervals in detecting changes of disease status in individuals (10)(11). When the index of individuality is low (<0.6), the individual results stay within a narrow range compared with the population reference interval. Hence, a low index suggests the utility of evaluating serial changes in analyte values in an individual, and population-based reference intervals would be of limited use. A high index (>1.4) suggests that the reference interval is appropriate. The index of individuality ranged from 0.13 for serum alkaline phosphatase to 0.83 for serum bicarbonate (Table 11 ).

In this study, BP and WP estimates of CV were obtained for some selective environmental and nutritional analytes. NHANES III provides a better estimate of CVg than do other short-term local studies because the NHANES III sample was nationally representative (1)(12)(13). The CVw estimate was limited by the nonrandom, self-selected design and reflected a mean of 17 days between two measurements. The estimate of CVw could be improved by use of a stratified, multistage probability design over different time periods. Differences for CVw and CVg among subpopulations (gender, age, race, and ethnicity) can be further investigated by use of NHANES III data.

Table 1.

BP, WP, and method CV and index of individuality for laboratory analytes from NHANES III.

1

Specimen type is serum unless noted.

2

Index of individuality is the ratio of CVw to CVg.

3

Outliers were deleted by the Tukey method.

Table 1.

BP, WP, and method CV and index of individuality for laboratory analytes from NHANES III.

1

Specimen type is serum unless noted.

2

Index of individuality is the ratio of CVw to CVg.

3

Outliers were deleted by the Tukey method.

1

Ricos C, Alvarez V, Cava F, Garcia-Lario JV, Hernandez A, Jimenez CV, et al. Current databases on biological variation: pros, cons and progress.

Scand J Clin Lab Invest

1999

;

59

:

491

-500.

2

Fraser CG.

Biological variation: from principles to practice

2001

:

151pp

AACC Press Washington DC. .

3

Fraser CG, Harris EK. Generation and application of data on biological variation in clinical chemistry.

Crit Rev Clin Lab Sci

1989

;

27

:

409

-437.

4

. National Center for Health Statistics. Plan and operation of the third National Health and Nutrition Examination Survey, 1988–94.

Vital Health Stat

1994

;

1

(32):

415pp

.

5

. US Department of Health and Human Services. Third National Health and Nutrition Examination Survey (NHANES III), 1988–94.

NHANES III second examination file documentation. Series 11, No. 3A

1999

National Center for Health Statistics Hyattsville, MD. http://www.cdc.gov/nchs/about/major/nhanes/nh3data.htm#NHANES%20III%20Series%2011,%20No.%203a (accessed September 2004)..

7

Wolter KM.

Introduction to variance estimation

1985

:

221

-247 Springer-Verlag New York. .

8

Tukey JW.

Exploratory data analysis

1977

:

44

Addison-Wesley Publishing Co. Reading, MA. .

9

Current status of blood cholesterol measurement in clinical laboratories in the United States: a report from the laboratory standardization panel of the National Cholesterol Education Program.

Clin Chem

1988

;

34

:

193

-201.

10

Fraser CG. Inherent biological variation and reference values.

Clin Chem Lab Med

2004

;

42

:

758

-764.

11

Harris EK. Effects of intra- and interindividual variation on the appropriate use of normal intervals.

Clin Chem

1974

;

20

:

1535

-1542.

12

Fraser CG. Biological variation in clinical chemistry, an update: collated data, 1988–1991.

Arch Pathol Lab Med

1992

;

116

:

916

-923.

13

Ricos C, Garcia-Lario JV, Alvarez V, Caval F, Domenech M, Hernandez A, et al. Biological variation database, and quality specifications for imprecision, bias and total error (desirable and minimum). The 2004 update. http://www.westgard.com/guest26.htm (accessed September 2004)..

© 2005 The American Association for Clinical Chemistry

Citations

Views

Altmetric

Metrics

Total Views 1,135

841 Pageviews

294 PDF Downloads

Since 1/1/2020

Citations

96 Web of Science

×

Email alerts

Citing articles via

• Latest

• Most Read

• Most Cited

More from Oxford Academic