Biomarkers of Nutritional Exposure and Nutritional Status

Transcription

1 Biomarkers of Nutritional Exposure an Nutritional Status Laboratory Issues: Use of Nutritional Biomarkers 1 Heii Michels Blanck,* 2 Barbara A. Bowman, y Geral R. Cooper, z Gary L. Myers z an Dayton T. Miller z *Division of Nutrition an Physical Activity, National Center for Chronic Disease Prevention an Health Promotion an Epiemic Intelligence Service, Division of Applie Public Health Training, Epiemiology Program Office, Centers for Disease Control an Prevention, y Division of Diabetes Translation, National Center for Chronic Disease Prevention an Health Promotion, Centers for Disease Control an Prevention an z Division of Environmental Health Laboratory Sciences, National Center for Environmental Health, Centers for Disease Control an Prevention, Atlanta, GA ABSTRACT Biomarkers of nutritional status provie alternative measures of ietary intake. Like the error an variation associate with ietary intake measures, the magnitue an impact of both biological (preanalytical) an laboratory (analytical) variability nee to be consiere when one is using biomarkers. When choosing a biomarker, it is important to unerstan how it relates to nutritional intake an the specific time frame of exposure it reflects as well as how it is affecte by sampling an laboratory proceures. Biological sources of variation that arise from genetic an isease states of an iniviual affect biomarkers, but they are also affecte by nonbiological sources of variation arising from specimen collection an storage, seasonality, time of ay, contamination, stability an laboratory quality assurance. When choosing a laboratory for biomarker assessment, researchers shoul try to make sure ranom an systematic error is minimize by inclusion of certain techniques such as blining of laboratory staff to isease status an incluing external poole stanars to which laboratory staff are bline. In aition analytic quality control shoul be ensure by use of internal stanars or certifie materials over the entire range of possible values to control metho accuracy. One must consier the effect of ranom laboratory error on measurement precision an also unerstan the metho s limit of etection an the laboratory cutpoints. Choosing appropriate cutpoints an reucing error is extremely important in nutritional epiemiology where weak associations are frequent. As part of this review, serum lipis are inclue as an example of a biomarker whereby collaborative efforts have been put forth to both unerstan biological sources of variation an stanarize laboratory results. J. Nutr. 133: 888S 894S, KEY WORDS: biomarkers iet assessment epiemiology nutrition methoology Nutritional biomarkers can serve as measures of nutritional exposure, or to use the nomenclature of environmental biomarkers, internal ietary ose. It is in this latter context that nutritional biomarkers go beyon being inicators of ietary intake an ai in our unerstaning of causal mechanisms between iet an isease. In 1983 Solomons an Allen (1) escribe the funamental role of biochemical parameters, which toay we call nutritional biomarkers, in assessing nutritional status. Their classic paper focuse on choosing appropriate biochemical parameters (irect measures versus functional assays) an confouning factors an assigning iagnostic cutoff values. Since 1983 there have been many technical 1 Publishe as part of The Journal of Nutrition supplement publication Biomarkers of Nutritional Exposure an Nutritional Status. This series of articles was commissione an financially supporte by International Life Sciences Institute, North America s Technical Committee on Foo Components for Health Promotion. For more information about the committee or ILSI N.A., call or ilsina@ilsi.org. The opinions expresse herein are those of the authors an o not necessarily represent the views of ILSI N.A. The guest eitor for this supplement publication was Jo Freuenheim, University at Buffalo, State University of New York, Buffalo, NY To whom corresponence shoul be aresse. hblanck@cc.gov. avances in the area of biomarkers as well as breakthroughs in the areas of genetics an metabolism. Avances in these fiels have been critical to nutritional researchers because biomarker levels can vary with absorption, metabolism, genetics an isease status. In aition to these avances, there has also been much iscussion about issues relate to proper sample collection (biological meium, fasting, contamination, stability), transport an storage (stability, temperature, light, oxygen) an analytical laboratory technique (precision, trueness, etection limit, recovery, stanarization an quality assurance). Some of these issues are also iscusse in this series by Potischman (2). Chapters by Hunter (3), Bates et al. (4), Sauberlich (5) an Myers et al. (6) an comprehensive articles by van t Veer (7) an Hermus et al. (8) provie further reaing on laboratory issues associate with nutritional biomarkers, which is the focus of this article. When choosing a biomarker, it is important for the researcher to unerstan how a nutritional biomarker relates to both ietary intake an the chronology of exposure. This inclues iscussion of whether the biomarker will be use to evaluate long-term nutritional status, recent ietary intake, effectiveness of ietary manipulation or the efficacy of an intervention. These issues have been reviewe elsewhere (7) Downloae from jn.nutrition.org by guest on October 13, /03 $3.00 Ó 2003 American Society for Nutritional Sciences. 888S

2 NUTRITIONAL BIOMARKERS: LABORATORY ISSUES 889S an are also aresse in this series (2). In aition, it is important to know about the various types of biochemical tests that may be useful for nutritional status assessment, e.g., irect or static measurements, metabolite measurements, functional tests, abnormal metabolites, proucts of the nutrient uner stuy, loa or saturation tests or other proceures such as using stable isotopes (5). Separate from essential knowlege about intake, chronology an biochemical tests is an unerstaning of how nutritional biomarkers are affecte by sampling an laboratory proceures. In this laboratory context, at least four methoologic consierations shoul be taken into account when choosing an appropriate nutritional biomarker: 1) valiity (how well the biomarker is measure in relation to its true value); 2) precision (how repeatable is the measure); 3) sensitivity (how well oes the biomarker ientify iniviuals with the conition); an 4) specificity (how well oes the biomarker ientify iniviuals without the conition) (9). The first two aspects assess how well the measurement can be mae an the latter two explain how well the result can be interprete. How well the measurement can be mae is partially epenent on the error associate with specimen collection an the analytic measurement of the biomarker. The following sections aress these issues. Measurement error: efinition Possessing valiity or trueness means that the biomarker measures the relevant exposure accurately. This measure of accuracy is also calle measurement error; it is the ifference between the true value of the biomarker an the measure biomarker. To test valiity, a gol stanar or reference metho that provies a goo measure of the true exposure is necessary. This gol stanar shoul reflect the true value without (or with minimal) laboratory or other sources of error. For a valiation stuy to be appropriate, both the measure biomarker an the reference metho must be relate to the relevant nutrient exposure (10). Valiation of biochemical measures has been reviewe by Hunter et al. (3) an Van t Veer (7). In most circumstances, measurement error is classifie as either biological (preanalytical) or laboratory (analytical) error. Briefly, preanalytical error usually inclues both biological an sampling errors, whereas analytical error focuses on the laboratory an inclues metho, instrument, reagent an/or matrix effects. Preanalytical measurement error: genetics, environment, behavior an health status. Common sources of withinsubject preanalytical biologic variability may arise from an iniviual s genetics, environment, behavior an health status an are therefore intrainiviual or within-subject sources of error. A goo review of boily processes that may affect the availability of nutrients is foun in the work of Bates et al. (4). Sources of variation inclue genetic polymorphisms; environmental an behavioral sources of variability can be from an iniviual s age, sex, iet, aiposity, weight loss, meication use, smoking status, physical exercise an alcohol intake. An example of variability from one s health status coul inclue the case where the concentration of a specific nutrient affects the levels of another nutrient of interest, i.e., nutrient interaction. Health-status variability can also occur when one s hormonal status or isease status influences a nutritional biomarker. This latter issue makes it important for researchers to be able to iscriminate between the impact of iet on isease an the influence of isease on nutritional biomarkers. The effect of isease on biomarkers of micronutrient status has been reviewe by Thurnham (11). Examples in the literature inclue reuction of plasma retinol, a-tocopherol an ascorbate by trauma; reuction of retinol an vitamin E ue to changes in their transport proteins or lipoproteins uring infection or other isease; an increase in vitamin C levels relate to the leukocytosis common in trauma an changes in measures of iron status incluing serum ferritin, transferrin receptor an retinol bining protein as part of the acute phase response of chronic isease. The impact of isease on biomarkers of macronutrient status (e.g., protein, energy) has been reviewe by Shenkin (9). Because of the possible effects of isease on biomarker level, researchers shoul etermine a priori whether biomarkers of interest shoul be measure concurrently with measures of infection or isease status. This coul inclue measures of malnutrition or protein status (total boy or specific acute phase proteins), presence of inflammation or infection (malaria smears, erythrocyte seimentation rate, C-reactive protein, white bloo cell count, transaminase levels) an presence of isease such as hypertension, iabetes, nephrosis or myocarial infarction. Having this information allows the researcher to better use an interpret the biomarker ata. Preanalytical measurement error: sampling error. In aition to these potential sources of error, measurement error ue to within-subject preanalytical variability can be ue to ifferences relate to sampling time. This variability within subjects can vary from hour to hour or year to year. Short-term variability may be hour to hour or ay to ay an can be cause by hormonal changes (e.g, menstrual cycle phase), fasting status or iurnal variation. An example of iurnal variation occurs with fasting glucose. If the accepte iabetes iagnostic cutpoint of 7.0 mmol glucose/l serum (126 mg glucose/l serum) is use with afternoon samples instea of bloo samples obtaine in the morning after an overnight fast, approximately half of all cases of iabetes will be misse (12). Meium-term variability can be month to month an be ue to seasonal changes in the iet such as those seen with cholesterol (13). Long-term variability may be year to year an ue to intentional ieting patterns or changes in health status. Analytical measurement error. Common sources of analytical or laboratory variability arise from errors in specimen collection an storage, errors uring specimen analysis an inherent matrix effects from reagents, instruments an interfering substances (14). In aition repeate freeze-thaw cycles of serum samples can lea to increase variability (15). In an article by White (16), a number of potential laboratory sources of variability are reviewe. These inclue errors or omissions in the collection or analytical protocol. This type of error can occur as a failure to comply with the proper protocol for specimen collection (e.g., using the appropriate type of anticoagulant or preservative container, allowing the esire whole bloo to clot, allowing sufficient clotting for serum, centrifuging in a timely fashion, keeping the specimen away from light or oxygen or at correct temperature), transportation or storage issues (keeping the specimen frozen or away from light or oxygen). Every effort shoul be mae to stanarize the specimen-collection protocol; however in a stuy that covers a wie geographic area or a long perio of time, it may not be possible to treat all samples exactly the same (i.e., some will travel further an some will be store for longer perios of time). Measurement errors ue to variations in analysis of the biomarker may also result from variability in technique an motivation between laboratory technicians, use of ifferent or contaminate reagents an failure to maintain stanarization of the instrument uring the course of ata collection. Once stanarization is accomplishe however, it is important to maintain quality control, but it is not necessary to frequently Downloae from jn.nutrition.org by guest on October 13, 2017

3 890S SUPPLEMENT restanarize. See the work of Petersen et al. (17) for a iscussion of propose guielines for internal quality control in the meical laboratory. Measurement error: assessment Precision or reproucibility implies reliability an expresses the variability of results obtaine from a single sample measure many times (4). Most laboratory researchers express the ranom analytical error of measurement as the stanar eviation etermine by repeate measurements performe on the same subject sample. Usually a mean an stanar eviation (SD) are calculate from several measurements on the sample an are then use to calculate the coefficient of variation (CV); CV 5 SD/mean The CV is ieally calculate for samples at the bottom, mile an top of the reference concentration range that was etermine on healthy people. Laboratories can provie both within-run an between-run CV to the researchers. The within-run CV is etermine by iviing the samples into two or more aliquots an analyzing them together. The between-run CV is obtaine when aliquots are analyze in ifferent runs, which is usually on ifferent ays. Multiple publications have presente approaches for estimation of reference intervals from ifferent numbers of specimens an numbers of analyses (18). Measurement error: minimization A major approach to minimize laboratory measurement errors involves blining the laboratory analyst to the isease status of the specimens or other pertinent characteristics an eliminating systematic ifferences in the way specimens are hanle. As reviewe by Hunter (3), ieally all specimens shoul be analyze in a single run to reuce between-assay variation (laboratory rift). In aition, if possible, paire-case an control samples shoul be analyze consecutively (with the within-pair orer ranomize) an with the pairs orere ranomly with regar to other variables of interest so that effects of orer (within-run laboratory error) are not attribute to another variable. If all samples cannot be assesse in a single run, then cases an an appropriate number of controls shoul be analyze together in the same batch to ensure the valiity of the paire comparison. For some nutrients, certifie quality controls an reference samples are available to help ensure comparability of results obtaine by ifferent laboratories (Table 1). Some stanars are certifie by the National Institute of Stanars an Technology an are mae available as Stanar Reference Materials. If suitable reference stanars are not available, researchers shoul test a laboratory by having the staff measure uplicate samples with laboratory personnel bline. Such samples shoul be inclue within the sample batches to monitor rift an reliability. Also it is essential for each laboratory to establish its own internal quality controls, for example, by incluing aliquots of known well-characterize eficient an normal controls for which mean an stanar eviation values have been establishe. These aliquots can also be exchange with other laboratories for comparisons of results. Routine inclusion of at least two samples from two quality control or reference samples in every batch of specimens helps to maintain precision an prevent or reuce between-assay 3 Abbreviations use: apo, apolipoprotein; CRMLN, Cholesterol Reference Metho Laboratory Network; CV, coefficient of variation; HDL, high-ensity lipoprotein; LDL, low-ensity lipoprotein; NCEP, National Cholesterol Eucation Program; NHANES, National Health an Nutrition Examination Surveys. TABLE 1 Examples of nutritional biomarker reference sample an certifie quality control sources National Institute of Stanars an Technology Stanar Reference Materials (U.S.A.) National Institute of Biological Stanars an Controls (U.S.A.) Worl Health Organization, Bloo Safety an Clinical Technology International Feeration of Clinical Chemistry, Scientific Division Centers for Disease Control an Prevention, Division of Laboratory Sciences Northwest Lipi Research Laboratories, University of Washington, Seattle, WA Solomons Park Research Laboratories, Kirklan, WA Commercial companies (primary stanars) Proficiency testing programs National reference material institutions variation, i.e., rift. In aition use of these reference materials over the entire range of possible values helps to assess the accuracy of the metho. A list of available stanars an seconary reference materials for lipis an lipoproteins has been publishe (6,19). Measurement error: effect in epiemiologic stuies Measurement error can lea to bias in measuring the association between nutritional exposure an outcome. This bias is known as either information bias or misclassification bias. In epiemiologic stuies, it is important to minimize misclassification bias from measurement error an to etermine whether this bias is ifferential or nonifferential. The effects of measurement error in epiemiologic stuies are iscusse in recent reviews by Saracci (20) an White (16), in the text by Willett (21) as well as in this series (22). Differential error occurs when the measurement error iffers between those with an those without the outcome or isease of interest. The influence of this type of error on the measure of association can only be evaluate if information about the error is known for all groups. However, if the presence of isease is in fact the cause of changes in a biomarker, it may falsely lea one to conclue that there is an association between biomarker an isease. To aress this problem, stratification by isease stage can be a way to etermine whether isease-inuce ifferential error exists (20). Nonifferential error occurs when the error oes not iffer between the comparison groups. This type of error is generally tolerable but can cause the measure of association to be attenuate towar the null, which results in a lack of effect between the nutritional biomarker an the outcome of interest. This can occur in a stuy where the overall CV for the nutritional biomarker of interest is high. In general, as the analytical precision ecreases an therefore the CV increases, the os ratio is attenuate towar the null. This can lea to an incorrect interpretation, i.e., the lack of an association between biomarker an outcome. It has been suggeste that for epiemiological stuies ieally the CV shoul not be. 5% (3). This level of accuracy is very ifficult to attain for many nutrients; often it is not possible to attain a CV, 10%. At levels. 10% there may be concern about the utility of the assay. In all cases, inclusion of the CV for the assay in the report of the stuy results is important. The nationally accepte CV cutpoint for total cholesterol is 3%; low-ensity lipoprotein (LDL) cholesterol, 4%; high-ensity lipoprotein (HDL) cholesterol, 4%; triglycerie, 5% (6). Downloae from jn.nutrition.org by guest on October 13, 2017

4 NUTRITIONAL BIOMARKERS: LABORATORY ISSUES 891S Misclassification bias can also come from the use of inappropriate biomarker cutoff values. One must unerstan the metho s limit of etection an laboratory cutpoints to maximize the biomarker s sensitivity an specificity. Overlap of persons with low or eficient nutritional biomarker values an those normal values can lea to misclassification of iniviuals thereby affecting the sensitivity an specificity of the biomarker. As reviewe by van en Berg (23), misclassification can result from confouning factors but can also be cause by both biological an laboratory variability; therefore minimization of sampling an laboratory errors are important in this aspect of analysis as well. To ai in the choice of a biologically relevant cutoff value, one can use serial or multiple measurements or a combination of nutritional exposure measurements, which is calle a composite value. This latter choice, although affecte by iniviual measurement error, may overcome some of the imperfection of nutritional biomarkers in epiemiology. As a final note, if the biological variability is such that nutrient levels show higher within-subject (intrainiviual) variability as compare to between-subject (interiniviual) variability, the researcher must be aware of the iminishe power of this biomarker an consier increasing the sample size of the stuy or, as mentione, consier using a combination of nutritional exposure measurements. Separate from issues of measurement error, use of a percentile cutoff value as etermine from a healthy population istribution may lack any real biological or functional basis. Similarly using percentile cutpoints from an even istribution of the survey s participant values may lea to null results. Therefore functional tests or the use of specific subpopulation ata such as elerly, young, aolescent or pregnant women may be of value in etermining the optimal cutoff value (23). The National Health an Nutrition Examination Surveys (NHANES) are an example of a reference ata source that allows for etermination of subpopulation reference ranges an cutoff values for a healthy U.S. population by sex, racial/ethnic group an age group. It shoul also be note that many epiemiologic analyses use continuous-exposure ata. Appropriate regression iagnostic approaches shoul be use to check the assumption that use of a continuous epenent variable is warrante, an one also nees to etermine whether a one-unit change in the epenent variable is a meaningful way to interpret the ata. Graphing of the ata shoul be use to ensure that a linear association is appropriate. It may be necessary to transform the nutritional values, i.e., log transformation Y9 5 log Y, square transformation Y9 5 Y 2, to normalize the epenent variable, to stabilize the variance of the epenent variable or to linearize the regression moel if the original ata suggests that the relation is nonlinear. Another problem inherent in nutritional biomarker research is the possibility that biologic processes such as genetic polymorphisms can alter biomarker values. This phenomenon is terme metabolic confouning by Saracci (20). If it is known that iniviuals may have ifferent forms of an enzyme that can ultimately affect biomarker metabolism, then the enzyme activity shoul be consiere as a potential confouner an controlle for. One way to control for this is by stratifying on the enzyme status. An example of a genetic polymorphism of this type is human apolipoprotein (apo)e, which has three common alleles (E2, E3 an E4). These alleles coe for three isoforms that are associate with ifferent levels of serum cholesterol (24). Similarly an inverse relationship has been foun between apo(a) isoform size an plasma lipoprotein(a) levels (25). Quality control an calibration The aim of quality control is to ensure that the analytical values prouce by a laboratory are sufficiently reliable for their intene purpose (26). A goo quality control program monitors the following: 1) preanalytical variation; 2) clerical error proper labeling an logging in of specimens as well as maintenance of appropriate recors for all specimens for future reference; 3) technique assurance that all laboratory analysts unerstan the principles that unerlie a specific assay an that all personnel use the same technique an have reay access to a etaile an current metho manual; 4) reagents an materials proper labeling, evaluation for matrix effects in analytical system, confirmation of absence of interfering substances, substitution of new reagents oes not change level of values of quality control materials; 5) calibration using a purifie primary stanar or a commercially available serum seconary calibrator traceable to an accepte reference metho or reference material, confirmation that each newly acquire calibrator possesses an accurate target value; 6) bench performance using controls an stanars for each assay, ocumentation of aily bench performance for etection of slight error over time, an 7) instrumentation performance of perioic preventive maintenance for all instruments with appropriate maintenance ocumentation. Some methoological problems associate with certain laboratory assays are liste in the appenices of the Dietary Reference Intake reports (27), where tables summarize available ata on ifferent methos: accuracy an precision, analytic sensitivity an specificity, interlaboratory agreement an how changes in methos have affecte estimates over time. However, iniviual laboratories shoul be able to provie internal an external quality control ata for biomarker assessment. Gunter et al. (28) have publishe an article on the results on an international roun-robin for serum folate an whole bloo folate. This interlaboratory comparison stuy was conucte to assess ifferences among methos. Twenty research laboratories participate in a 3- analysis of six serum an six whole bloo pools. Overall mean, stanar eviation an CV values erive from these results were compare within an across metho types. Results reporte for serum an whole bloo folate resulte in overall CV of 27.6 an 35.7%, respectively, across pools an two- to ninefol ifferences in concentrations between methos with the greatest variation occurring at low folate concentrations. Although results for serum pools were less variable than those for whole bloo pools, substantial intermetho variation occurre. These results emphasize the urgent nee for eveloping an valiating reference methos for biomarker measurements an for properly characterize reference materials. In aition these finings highlight that when one is evaluating stuy or clinical ata, metho-specific reference ranges (establishe with clinical confirmation of values for truly eficient iniviuals) must be use. In aition to the interlaboratory comparison stuy performe on serum an whole bloo folate, Pfeiffer et al. (29) compare plasma total homocysteine measurements in 14 laboratories. These 14 laboratories across the worl use eight ifferent analytical methos. The laboratories participate in a 2- analysis of 46 plasma samples an plasma quality control pools. The mean among-laboratory an within-laboratory (among-run) CV were 9.3 an 5.6%, respectively, for plasma samples. Specific techniques ha systemically higher or lower values when compare to gas chromatography/mass spectrometry. This analysis showe that among-laboratory variations within one homocysteine metho coul excee among-metho Downloae from jn.nutrition.org by guest on October 13, 2017

5 892S SUPPLEMENT variations. Although some of the methos appeare to be interchangeable, these finings further raw attention to the nee for improve analytical precision. Quality control: long-term planning For researchers involve in long-term prospective stuies, consieration of long-term quality control is important. A review of long-term laboratory quality control planning has been one by researchers at the NHANES nutritional laboratory (30,31). As reviewe, long-term quality control not only encompasses methoologic accuracy an precision but also must ensure positive specimen ientification (incluing ahesion of labels after liqui nitrogen or boiling-water bath immersion) an aress time-associate trens such as specimen stability (incluing rate of eterioration over time an freeze-thaw effects), changes in analysts, reagents or instrumentation an seasonal/geographical variation. For example, at the NHANES lab, both bench-quality controls prepare at low, normal an elevate concentration levels are run two to four times in each analytical run, an blin-quality controls an low an moerately elevate normal levels are incorporate into every run. To assist the reaer, Tables 2A C list consierations for choosing the appropriate nutritional biomarker, choosing a laboratory, an unerstaning the necessities of specimen collection an processing. For an example of a etaile review of choosing an appropriate assay an accompanying laboratory, see the article by Nexo et al. (32), which escribes this type of evaluation for homocysteine. Unerstaning sources of variation: results from the National Cholesterol Eucation Program Serum lipis are an example of important biomarkers for which collaborative efforts have le to both unerstaning biological sources of variation an stanarization of laboratory measurements (6). Cooper et al. (13) reviewe the magnitue an impact of the major biological an analytical sources of TABLE 2A General consierations in choosing a nutritional biomarker for an epiemiologic research stuy 1. Timing relative to ietary exposure: recent intake versus usual intake, acute versus chronic exposure. 2. Type of measurement: irect measure (static inicator) versus functional assay. 3. Woul a ietary assessment metho such as a foo-frequency questionnaire or a 24-h recall provie aequate ietary information precluing the nee for biomarker assessment? 4. Has within-person (intra-) an between-person (inter-) variance been ocumente for the biomarker measurement metho of interest? If yes, is the between-person variance larger than the within-person variance? If no, it will be ifficult to assess associations without an extremely large sample size. a) Has the laboratory metho been stanarize against an accepte reference metho? b) Has the laboratory metho been shown to be reliable/reproucible? c) Have the specificity an sensitivity of the biomarker been establishe? ) Do appropriate cutoff values or reference ranges exist for the biomarker within the population you are stuying? e) Can you assume there is a sufficient range of concentrations in the biomarker concentration within your sample to warrant its use? f) Is it logistically possible to obtain the specimen as the protocol specifies or o collection an transportation conitions negate its use, i.e., store specimen away from light at 48C? TABLE 2B Consierations in choosing a laboratory for assessment of a nutritional biomarker 1. Does the laboratory have a written escription of the entire preanalytical, analytical an reporting proceures for the analytes to be etermine for the epiemiologic, clinical investigations an methoology research stuies or for collaborative services in the clinical laboratory? 2. Does the claime analytical performance (coefficient of variation an bias) of the laboratory meet the requirements necessary to accomplish the goals of the epiemiologic or clinical stuies? 3. Confirm with acceptable quality control serum pools that the laboratory meets the claime acceptable precision an bias requirements. 4. Confirm that the calibration process uses a purifie primary stanar of known purity or a commercially available serum calibrator that is traceable to an international or national accepte reference material. Inquire about recovery experiments an whether frequency of use of calibrator meets statistical requirements. 5. Confirm that the laboratory uses statistical quality control charts at the bench level to maintain accuracy among runs; alert of bias changes with new calibrators, new instruments or eteriorating reagents an provie analytical performance ata over long-term perios. 6. Determine whether a clinical laboratory participates successfully in an effective proficiency program an Clinical Laboratory Improvements Amenments evaluations or whether an epiemiologic laboratory has proven its performance is traceable to highly accurately labele concentration values of national or international reference materials. 7. Before analysis, ensure that the sample-collection process, sample preparation an storage conitions an patient meication or isease status o not cause matrix effects in the analytical instrument-reagent system, which can result in inaccurate results. Suspect matrix effects with frozen, lyophilize or eteriorating samples enature by unsuitable storage, interfering substances an inappropriate hanling techniques. 8. Confirm that highly accurate measuring instruments are being use for measurement of critical volumes, carry-over contamination oes not occur in the spectrophotometer or absorption measuring instrument, the sample-hanling proceure is reproucible an possible transcription errors are controlle. 9. What are the laboratory s analytical reference ranges of the analytes? Is the laboratory report suitable for use in the stuy of the population or subgroup of iniviuals being assesse? What is the lab throughput an turnaroun? Will the assays be complete in the time frame neee? 10. Is the cost of the assay appropriate an competitive for the emans on accuracy an precision, the necessity to use uplicate or more eterminations, the cost of highly qualifie personnel assigne to the project an the possibility that instrumentation an reagents are highly expensive? 11. Is the laboratory staff motivate towar high-quality analytical performance an oes the staff receive close supervision? variation in serum lipi an lipoprotein levels on risk of coronary heart isease. Because of the concern regaring serum cholesterol an isease risk, concerte efforts were mae to attain analytical accuracy an precision (e.g., CV, 3% by 1992) as well as to minimize biological an sampling sources of error. It was etermine that behavioral sources of variation in serum lipis arise from the lifestyle of the iniviual. Lifestyle factors inclue iet, weight status, smoking, physical activity an alcohol consumption. In aition to behavioral sources of variation, there are also a number of clinical sources of variation in serum lipis. These sources inclue metabolic states, illness an seconary or other isease states such as myocarial infarction, stroke, hypertension, nephrosis, iabetes an Downloae from jn.nutrition.org by guest on October 13, 2017

6 NUTRITIONAL BIOMARKERS: LABORATORY ISSUES 893S TABLE 2C Consierations in nutritional biomarker specimen collection an processing 1. What is the correct specimen type to collect (e.g., plasma, serum, urine, bloospots)? 2. What is the appropriate metho of collection (e.g., capillary sample, venipuncture, mistream urine, etc.)? 3. What is the appropriate vial? Vials must hol appropriate volume, be stable when frozen an hol their contents without leakage; label or markings must have long-term ahesion an have integrity at a range of temperatures. 4. Is a preservative or anticoagulant require? 5. What volume of sample is require? 6. Will ilution, preservation, aliquoting or hemolysate preparation of the sample be neee? 7. How large shoul the aliquots be to maximize the efficiency of use of the samples an to meet requirements for the laboratory assay? 8. Do the subjects nee to fast or follow other instructions before sample collection? 9. Are samples sensitive to light, oxygen or temperature? 10. Do you have access to a local laboratory or can you transport collection materials to the fiel ( i.e., centrifuge, refrigerator or freezer, wet or ry ice, phlebotomy supplies incluing gloves, alcohol or alcohol pas, gauze, banages, Vacutainer tubes an accompanying neeles, barrels an biohazar isposal containers)? 11. Will you be able to collect fiel controls? 12. Do you have enough supplies to split the samples to prouce uplicates? 13. Is the stability of the sample known uner fiel conitions? At what optimal temperature shoul the samples be transporte (e.g., must the samples be shippe refrigerate within 24 h of collection or can they be frozen or refrigerate an then shippe or transporte on wet or ry ice)? 14. At what optimal temperature shoul the samples be store? What is the optimal size for storage aliquots to maximize utility an minimize loss of sample? infections. Sampling sources of variation in serum lipis also inclue fasting status, posture an choice of serum versus plasma. The accuracy of the etermination of risk for cariovascular isease of a cholesterol measurement in clinical trials an other epiemiological stuies epens on the availability of an accurate reference metho an cholesterol reference materials labele with accurate target values (33). The Centers for Disease Control an Prevention (CDC)-moifie Abell-Levy- Broie-Kenall cholesterol metho is accepte as the cholesterol reference metho of the National Reference System for Cholesterol. This metho is also the reference metho for HDL an LDL cholesterol (34). An investigation of the effect of systematic bias an ranom error, quality control an intraperson biological variation on the National Cholesterol Eucation Program (NCEP) clinical classifications for reporte lipi measurements has been performe. This investigation foun that the NCEP guielines are aequate to ensure correct patient classification at least 90% of the time if the laboratories are meeting the NCEP guielines for analytical precision an are using stanar quality control proceures (35). This probability is ensure regarless of the size of the systematic bias of the laboratory or increase ranom analytic error. In collaboration with manufacturers of cholesterol iagnostic proucts, valiity of trueness of serum lipi measurements in the clinical laboratory is being improve by manufacturers calibration of clinical laboratory equipment to the reference methos maintaine at the CDC through activities by the National Cholesterol Reference Metho Laboratory Network (CRMLN) (36). The CRMLN uses CDC serum lipi reference materials labele by the CDC reference metho to assist manufacturers of in vitro iagnostic lipi assays with the valiation of their lipi an lipoprotein assays before the assays are istribute to the clinical laboratories. This is one so that clinical laboratories across the U.S. can be confient that they can accurately measure total, HDL an LDL cholesterols. SUMMARY Measurement error is often classifie as either biological (preanalytical) or laboratory (analytical) error. Preanalytical error usually inclues both biological an sampling errors, whereas analytical error focuses on the laboratory an inclues metho, instrument, reagent an/or matrix effects. Common sources of laboratory variability arise from errors in specimen collection an storage, errors uring specimen analysis an from ifferences in reagents, instruments an interfering substances. Two major approaches to reucing laboratory measurement error are to blin the laboratory analyst to the case-control status of specimens an to eliminate systematic ifferences in the way case an control specimens are hanle. Reucing measurement error (either preanalytical or analytical) an choosing appropriate cutpoints is extremely important in nutritional epiemiology, where weak associations are frequent. To correctly interpret the association of nutritional exposure an outcome, it is also important to anticipate the nature of the errors. It is the investigator s responsibility to choose a vali biomarker an to interpret results with knowlege of the measure error. This error shoul be inclue in the reporting of the results to ai the reaers interpretation of the finings an comparison with other stuies. Although the use of nutritional biomarkers in screening was not specifically aresse in this paper, similar issues an concerns must be heee. In aition previous knowlege or estimation of sensitivity an specificity for the isease of interest is neee to ensure correct interpretation of the results. The interpretation is important for proper implementation of treatment an/or control measures. Future stuies on the impact of isease an genetics on nutritional biomarker levels will help researchers better interpret measure biomarkers an hence ai in the unerstaning of the causal role between iet, metabolism an genetics an isease. As assessment of nutritional biomarkers improves, they may be use to assess ietary exposure as well as to provie important information on early biological effects an isease characteristics. Determining an quantitating the sources of biological an laboratory error for all nutritional biomarkers (such as has been one for lipis) will therefore be an important step in the area of nutritional epiemiology. LITERATURE CITED 1. Solomons, N. W. & Allen, L. H. (1983) The functional assessment of nutritional status: principles, practice, an potential. Nutr. Rev. 41: Potischman, N. (2003) Biologic an methoologic issues for nutritional biomarkers. J. Nutr. 133: 875S 880S. 3. Hunter, D. (1998) Biochemical inicators of ietary intake. In: Nutritional Epiemiology, 2n e. (Willett, W., e.), pp Oxfor University Press, New York, NY. 4. Bates, C. J., Thurnham, D. I., Bingham, S. A., Margetts, B. M. & Nelson, M. (1997) Biochemical markers of nutrient status. In: Design Concepts in Nutritional Epiemiology, 2n e. (Margetts, B. M. & Nelson, M., es.), pp Oxfor University Press, Oxfor, U.K. 5. Sauberlich, H. E. (1999) Introuction. In: Laboratory Tests for the Assessment of Nutritional Status, 2n e. (Wolinsky, I., e.). CRC Press, Boca Raton, FL. Downloae from jn.nutrition.org by guest on October 13, 2017

7 894S SUPPLEMENT 6. Myers, G. L., Cooper, G. R., Greenberg, N., Kimberly, M., Waymack, P. P. & Hassemer, D. J. (2000) Stanarization of lipi an lipoprotein measurements. In: Hanbook of Lipoprotein Testing, 2n e. (Rifai, N., Warnick, G. R. & Dominiczak, M. H., es.). American Association for Clinical Chemistry Press, Washington, D.C. 7. Van t Veer, P. (1994) Measuring nutritional exposures incluing biomarkers. Proc. Nutr. Soc. 53: Hermus, R. J., Verhagen, H. & van Poppel, G. (1994) Biomarkers in nutritional assessment. Bibl. Nutr. Dieta 51: Shenkin, A. (1997) Impact of isease on markers of macronutrient status. Proc. Nutr. Soc. 56: Margetts, B. M. (1994) Linking the fiel to the laboratory in nutrition research. Proc. Nutr. Soc. 53: Thurnham, D. I. (1997) Impact of isease on markers of micronutrient status. Proc. Nutr. Soc. 56: Troisi, R. J., Cowie, C. C. & Harris, M. I. (2000) Diurnal variation in fasting plasma glucose. Implications for iagnosis of iabetes in patients examine in the afternoon. JAMA 284: Cooper, G. R., Myers, G. L., Smith, S. J. & Schlant, R. C. (1992) Bloo lipi measurements. Variations an practical utility. JAMA 267: Miller, W. G. (2000) Matrix effects in the measurement an stanarization of lipis an lipoproteins. In: Hanbook of Lipoprotein Testing, 2n e. (Rifai, N., Warnick, G. R. & Dominiczak, M. H., es.), pp American Association for Clinical Chemistry Press, Washington, D.C. 15. Comstock, G. W., Burke, A. E., Norkus, E. P., Goron, G. B., Hoffman, S. C. & Helzlsouer, T. (2001) Effects of repeate freeze-thaw cycles on concentrations of cholesterol, micronutrients an hormones in human plasma an serum. Clin. Chem. 47: White, E. (1997) Effects of biomarker measurement error on epiemiological stuies. IARC Sci. Publ. 142: Petersen, P. H., Ricos, C., Stockl, D., Libeer, J. C., Baaenhulgsen, H., Fraser, C. & Thienpont, L. (1996) Propose guielines for the internal quality control of analytical results in the meical laboratory. Eur. J. Clin. Chem. Biochem. 34: Horn, P. S., Pesce, A. J. & Copelan, B. E. (1998) A robust approach to reference interval estimation an evaluation. Clin. Chem. 44: National Committee for Clinical Laboratory Stanars (NCCLS) (1995) How to Define an Determine Reference Intervals in the Clinical Laboratory. Document C28-A, ISBN , Villanova, PA. 20. Saracci, R. (1997) Comparing measurements of biomarkers with other measurements of exposure. IARC Sci. Publ. 142: Willett, W. (1998) Correction for the effects of measurement error. In: Nutritional Epiemiology, 2n e. (Willett, W., e.), pp Oxfor University Press, New York, NY. 22. Marshall, J. (2003) Methoologic an statistical consierations regaring use of biomarkers of nutritional exposure an status in epiemiology. J. Nutr. 133: 881S Van en Berg, H. (1994) Functional vitamin status assessment. Bibl. Nutr. Dieta 51: Siest, G., Pillot, T., Bailey, A. R., Muller, B., Steinmetz, J., Galteau, M. M. & Visvikis, S. (1995) Apolipoprotein E: an important gene an protein to follow in laboratory meicine. Clin. Chem. 41: Graw, A., Brown, E. A. & For, I. (1998) Impact of apo(a) length polymorphism an the control of plasma Lp(a) concentrations. Arterioscler. Thromb. Vasc. Biol. 18: Gunter, E. W., Lewis, B. L. & Koncikowski, S. M. (1996) Laboratory methos use for the Thir National Health an Nutrition Examination Survey (NHANES III), In: NCHS CD-ROM: NHANES III Reference Manuals an Reports, Centers for Disease Control an Prevention, Hyattsville, MD. 27. Institute of Meicine (2000) Appenix E: Methoological problems associate with laboratory values an foo composition ata for B vitamins. In: Dietary Reference Intakes for Thiamin, Riboflavin, Niacin, Vitamin B6, Folate, Vitamin B12, Pantothenic Aci, Biotin, an Choline. A Report of the Staning Committee on the Scientific Evaluation of Dietary Reference Intakes an Its Panel on Folate, Other B Vitamins, an Choline an Subcommittee on Upper Reference Levels of Nutrients, Foo, an Nutrition Boar, pp National Acaemy Press, Washington, D.C. 28. Gunter, E. W., Bowman, B. A., Cauill, S. P., Twite, D. B., Aams, M. J. & Sampson, E. J. (1996) Results of an international roun robin for serum an whole-bloo folate. Clin. Chem. 42: Pfeiffer, C. M., Huff, D. L., Smith, S. J., Miller, D. T. & Gunter, E. W. (1999) Comparison of plasma total homocysteine measurements in 14 laboratories: an international stuy. Clin. Chem. 45: Bartley, S. L. & Gunter, E. W. (1996) Laboratory practice an archival storage of biological specimens. Toxicol. In. Health 12: Gunter, E. W. & McQuillan, G. (1990) Quality control in planning an operating the laboratory component for the Thir National Health an Nutrition Examination Survey. J. Nutr. 120(suppl. 11) : Nexo, E., Engbaek, F. & Uelan, P. M. (2000) Evaluation of novel assays in clinical chemistry: quantitation of plasma homocysteine. Clin. Chem. 46: Myers, G. L., Cooper, G. R., Winn, C. L. & Smith, S. J. (1989) The Center for Disease Control National Heart, Lung, an Bloo Institute Lipi Stanarization Program: An approach to accurate an precise lipi measurements. Clin. Lab. Me. 9: Duncan, I. W., Mather, A. & Cooper, G. R. (1982) The Proceure of the Propose Cholesterol Reference Metho. Clinical Chemistry Division, Centers for Disease Control, Atlanta, GA. 35. Cauill, S. P., Cooper, G. R., Smith, S. J. & Myers, G. L. (1998) Assessment of current NCEP guielines for total cholesterol, triglycerie, HDLcholesterol an LDL-cholesterol measurements. Clin. Chem. 44: Myers, G. L., Kimberly, M. M., Waymack, P. P., Smith, S. J., Cooper, G. R. & Sampson, E. J. (2000) A reference metho laboratory network for cholesterol: a moel for stanarization an improvement of clinical laboratory measurements. Clin. Chem. 46: Downloae from jn.nutrition.org by guest on October 13, 2017