1689

Clinical Chemistry 42:10

1689-1694 (1996)

Results of an international round robin for serumand whole-blood folate

ELAINE W. GuNTER,l* BARBARA A. BOWMAN,’ SAMUEL P. CAUDILL,’ DELLA B. TWITE,1

MYRON J. ADAMS,2 and ERIC J. SAMpsoN’

Because of the increasing significance of folate nutriture to

public health, a “round robin” interlaboratory comparisonstudy was conducted to assess differences among methods.

Twenty research laboratories participated in a 3-day anal-

ysis of six serum and six whole-blood pools. Overall means,

SDs, and CVs derived from these results were compared

within and across method types. Results reported for serumand whole-blood folate demonstrated overall CVs of 2 7.6%

and 35.7%, respectively, across pools and two- to ninefold

differences in concentrations between methods, with thegreatest variation occurring at critical low folate concentra-tions. Although results for senun pools were less variable

than those for whole-blood pools, substantial intermethodvariation still occurred. The overall results underscore the

urgent need for developing and validating reference meth-ods for serum and whole-blood folate and for properly

characterized reference materials. For evaluating study orclinical data, method-specific reference ranges (established

with clinical confirmation of values for truly folate-deficient

individuals) must be used.

INDEXING TERMS: intermethod comparison #{149}radioassay #{149}chro-matography. microbiological assay #{149}ion capture #{149}chemilumi-

nescence

In addition to its requirement as an essential nutrient for

preventing megaloblastic anemia and its role as a cofactor in

one-carbon transfers required for DNA replication, the water-

soluble vitamin folate has assumed even greater epidemiological

importance in recent years because of its relation to the preven-tion of neural tube defects [1, 2] and to homocysteine metabo-

Divisions of’ Environmental Health Laboratory Sciences and 2 Birth Defects

and Developmental Disabilities, National Center for Environmental Health,

Centers for Disease Control and Prevention, Public Health Service, US Depart-ment of Health and Human Services, Atlanta, GA 30341-3724.

Author for correspondence. Fax 770-488-4609; e-mail ewgl@cehehll.em.cdc.gov.

Use of trade names is for the purpose of identification only and does notconstitute endorsement by the Public Health Service or the US Department ofHealth and Human Services.

Received January 19, 1996; accepted May 16, 1996.

lism in prevention of cardiovascular disease [3, 4]. Serum or

plasma folate concentrations are considered to reflect recent

dietary intake, whereas erythrocyte folate concentrations areindicative of body stores [5].

Typical study designs rely on retrospective evaluation of

banked specimens for folate analyses in serum or erythrocytes,necessitating the comparison of biochemical folate results be-

tween several large-scale studies (and introducing another vari-

able, long-term stability of folate in stored specimens). How-ever, results generated by various methods may not be

comparable, as can be discerned by examining external profi-

ciency testing survey results, such as those for serum folate fromthe College of American Pathologists (CAP) [6]. Such surveys,however, generally include only serum samples and do not

address variability among analytical methodologies for erythro-cyte folate.3 The UK National External Quality Assessment

Scheme (NEQAS) is unusual in having an erythrocyte folate

testing material but uses expired blood-bank materials ratherthan materials prepared as typical clinical specimens (personalcommunication, Mj. Lewis, NEQAS hematology coordinator,

October 1995). Comparability of folate assays among Europeanlaboratories was an objective of the European Community

FLAIR Concerted Action No. 10 Folate Intercomparison Study1992-1993 [7]. Specimens in this study also were not typical

clinical specimens but rather aliquots of Bio-Rad Labs. (Her-cules, CA) Lyphochek lyophilized whole-blood control pools.

Additional major problems are the lack of accepted Refer-

ence or Definitive Methods for analysis in serum or wholeblood, as well as Standard Reference Materials in these matrices.

The UK National Institute of Biologics, Standard, and Controls(NLBSC), in response to requests of the International Commit-

tee for Standardization in Haematology (ICSH) Folate andVitamin B 12 Committee, is currently undertaking the prepara-

tion of lyophilized whole-blood pools as a tentative reference

material; target values, however, are being determined as a

‘At the March 1996 meeting of the Ligand Survey Committee of CAP,

however, the Committee agreed to consider the implementation of severalerythrocyte folate challenges in 1997 (personal communication, avid Witte,chairman).

1690 Gunter et al.: Serum and whole-blood folate

consensus of different methods (Susan Thorpe, NIBSC, per-sonal communication).4

Serum and erythrocyte folate data from the Third National

Health and Nutrition Examination Survey (NT-lANES III,

1988-1994) will be released in 1996 (VVright J, Gunter EW,Johnson CL, et al., ms. in preparation). These data will be useful

for establishing US normative ranges and for assessing the

prevalence of inadequate folate status in the US population.

Because the folate analyses from NT-lANES III were performed

with the Bio-Rad QuantaPhase II radioassay (RA) kit [8, 9],

which was produced in response to the need to correct the

calibration of the original QuantaPhase kit [JO], it is importantto be able to compare these data with those generated by other

types of analyses [11]. Currently, the major analytical ap-

proaches are microbiological assay, RA, chemiluminescence,magnetic separation, HPLC, and ion capture. A few laboratories

use GC-MS for clinical research. Except for microbiological

assay, HPLC, and GC-MS, most methods are based on com-

petitive binding by a specific folate-binding protein.

Materials and MethodsTo assess the potential variability among methods, the Centers

for Disease Control and Prevention (CDC) invited clinicians

and laboratorians actively involved in folate research, as well as

manufacturers of frequently used assay kits, to participate in a

round-robin interlaboratory comparison study. To prepare a

series of serum and whole-blood pools with concentrations

spanning the ranges of values seen in healthy as well as deficient

subjects, we screened serum and EDTA-anticoagulated blood

specimens collected by the Emory University Hospital BloodCollection Services under an agreement with CDC (including

an omnibus informed consent and Human Subjects Review

protocol). The screening specimens were analyzed in the

NHANES laboratory at CDC by the Bio-Rad QuantaPhase II

RA kit, used exactly as it was for NIL-lANES III, with serum

quality-control pools at four concentrations (“anemic,” low,

medium, and high folate concentrations) and three additionalconcentrations of whole blood pools, all of which had been in

use for several years.

SPECIMENS

Of the 20 normal, noninfectious, apparently healthy adult men

and women donors screened, 12 were selected as having appro-

priate serum or whole blood folate concentrations. No effort

was made to add external pteroylglutamic acid (PGA) or

5-methyltetrahydrofolate (5-MTHF) to the pools or to char-

coal-treat any of the collected materials to contrive folate-

deficient samples. We also did not specifically request or exclude

donors who were taking folic acid supplements. Because multi-

vitamin supplement use is prevalent in the US population,

serumlplasma concentrations exceeding 45-68 nmolIL (20-30

‘ Nonstandard abbreviations: N1BSC, National Institute of Biologics, Stan-dards, and Controls; ICSH, International Committee for Standardization inHaematology; RA, radioassay; PGA, pteroylglutamic acid; 5-MTHF, 5-methyl-tetrahydrofolate; NHANES III, Third National Health and Nutrition Examina-tion Survey; and CDC, Centers for Disease Control and Prevention.

ng/mL) are frequently observed clinically. Our aim was to

prepare pools that would resemble samples typical of those

analyzed in most clinical laboratories.

Blood was collected in 250-mL bottles by phlebotomy from

the fasting donors as either EDTA-anticoagulated or nonanti-

coagulated whole blood. The nonanticoagulated blood was

allowed to stand for 2 h, covered with aluminum foil, at ambient

temperature to allow maximum serum yield. We determined

hematocrits of the screening specimens to ensure that no

potential donor was anemic. After serum was separated from the

clotted blood by centrifugation, 1 mg of L-ascorbic acid (Sigma

Chemical Co., St. Louis, MO) was added per milliliter of serumto enhance the stability of the folic acid during storage. Whole-

blood materials were allowed to stand at least 90 mm after

collection at ambient temperature to allow maximum hydrolysis

of the polyglutamyl folates by endogenous folate conjugase.

Whole-blood pools were diluted 10-fold, i.e., 50 mL of whole

blood with 450 mL of 10 g/L ascorbic acid diluent. This

combination of ascorbic acid concentration and the 1:10 dilution

to prepare the hemolysate was chosen because of its traditional

use in microbiological folate methods [12]. Users of all other

methods (e.g., RAs with a 1:11 initial dilution) were instructed to

correct their calculations accordingly.

All pools were dispensed in l.0-mL aliquots into 2.0-mL

high-density polypropylene cryovials (Nalge, Rochester, NY)

labeled with appropriate pool identifiers (e.g., SFOL-0l, RBCF-

04). The aliquots were promptly frozen at -70 #{176}Cuntil ship-

ment. Extra aliquots were retained by CDC. Each laboratory

was sent a shipment on dry ice containing six vials of each of the

12 pools and instructed to perform two determinations per vial

on two vials per day for 3 days, for a total of 12 measurementsper pool. No effort was made to ensure that the participating

laboratories did not know the identity of the pools; rather, the

objective was to assess the degree of comparability among

leading clinical research laboratories. Laboratorians were in-

structed to report their actual results and key information

describing their assay method such as calibration material,

throughput, and expected reference ranges. Although we had

originally measured the hematocrit of each donor pool, partic-

ipating laboratorians were requested to report results as whole-

blood folate, in nanograms per milliliter.

PARTICIPANTS AND METHODS

Of the 20 participating laboratories, 11 were in the US and 9

were international. The participants included two manufactur-

ers, five government laboratories, six academic laboratories, and

seven clinical research facilities.Eight laboratories used the Lactobacillas casei microbiological

assay, some following the traditional assay set-up in culture

tubes and others using microtiter plates and readers. Eight

laboratories used RAs: Bio-Rad QuantaPhase II; Simultrac S#{174}

and Simultrac S No-Boil#{174}(ICN/Becton Dickinson, Orange-

burg, NY); and Dual Count Solid Phase No-Boil#{174}(Diagnostic

Products Corp., Los Angeles, CA). Two laboratories used

chemiluminescence methods: Corning MagicLite#{174} and ACS:

180#{174}(Ciba-Corning, E. Walpole, MA). Two newly developed

Lab (method) SFOL.O1 SFOL-02 SFOL-03 SFOL-04 SFOL-O5 SFOL-08

1 (M) 69.91 (1.68) 5.48 (0.48) 27.61 (0.86) 38.28 (0.90) 19.91 (0.82) 0.90 (0.12)

2 (M) N/A N/A N/A N/A N/A N/A

3 (M) 59.51 (12.42) 4.92 (2.26) 29.24 (3.96) 38.50 (3.85) 15.41 (2.54) 3.36 (2.80)

4 (M) 43.68 (1.65) 3.79 (0.45) 27.18 (3.23) 34.64 (3.40) 12.31 (1.47) 1.59 (0.39)5 (M) 55.30 (3.81) 4.69 (0.26) 27.60 (2.17) 42.28 (3.21) 15.55 (1.15) 2.32 (0.52)6 (M) 67.95 (5.78) 5.21 (0.63) 29.65 (2.34) 42.70 (1.72) 16.35 (0.87) 3.08 (0.15)

7 (M) 53.64 (3.46) 4.92 (0.81) 24.56 (1.94) 37.92 (3.83) 13.67 (1.38) 3.45 (0.28)8 (M) 58.75 (3.30) 4.53 (0.19) 26.65 (1.88) 37.69 (2.74) 15.18 (1.49) 2.72 (0.56)9 (B-R) >458 3.44 (0.11) 22.29 (3.16) 39.70 (6.16) 12.89 (1.48) 3.50 (0.29)

10 (B.R) >45#{176} 3.81 (0.29) 21.93 (1.09) 38.13 (2.53) 13.46 (0.55) 4.15 (0.20)11 (B-R) 65.05 (12.01) 3.53 (0.28) 22.26 (1.34) 35.75 (2.59) 13.05 (0.37) 3.67 (0.12)12 (B-R) 51.66 (5.21) 3.68 (0.11) 19.73 (0.66) 30.69 (1.38) 12.07 (0.28) 4.05 (0.11)13 (B-D) 56.91 (3.10) 5.47 (0.29) 22.25 (0.86) 34.47 (1.75) 14.95 (0.48) 5.55 (0.15)14 (B-D) 52.36 (10.19) 5.27 (0.32) 22.46 (2.35) 30.97 (4.33) 14.36 (1.15) 5.27 (0.26)15 (B-D NB) 71.23 (3.64) 7.56 (1.76) 28.26 (2.58) 35.63 (2.97) 19.55 (2.82) 5.26 (0.57)16 (DPC) 78.03 (8.00) 7.44 (0.74) 34.28 (3.65) 48.57 (3.54) 21.03 (1.64) 7.69 (0.64)17 (ACS) 81.20 (9.72) 9.17 (1.82) 30.71 (3.31) 47.62 (7.94) 22.23 (2.93) 8.36 (1.69)

18 (ML) 63.19 (4.57) 9.40 (4.72) 27.92 (2.53) 42.49 (4.74) 19.32 (2.18) 7.74 (1.66)19 (IC) 39.79 (1.18) 5.87 (0.58) 23.80 (0.71) 30.11 (1.09) 16.91 (0.74) 5.91 (0.34)20 (HPLC) 24.12 (1.37) 2.70 (0.42) 14.59 (1.35) 19.53 (1.62) 7.76 (0.71) 3.16 (0.33)All lab mean 58.37 5.31 25.42 37.14 15.58 4.30

All lab SD 14.18 1.88 4.54 6.65 3.60 2.06All lab CV, % 24.29 35.41 17.86 17.89 23.10 48.00Range of means 24.12-81.20 2.70-9.40 14.59-34.28 19.53-48.57 7.76-22.23 0.90-8.36Range of CVs, % 2.4-20.9 2.9-50.2 3.0-16.2 2.3-16.7 2.3-16.5 2.7-83.3

Clinical Chemistry 42, No. 10, 1996 1691

assays were also used, ion-capture IMX#{174}(Abbott Labs., Abbott

Park, IL) [13] and HPLC with fluorescence detection (unpub-

lished observations, CM Pfeiffer andJF Gregory, 1995), by one

laboratory each.

For calibration materials, folinic acid was used in two of the

microbiological assays, 5-MTHF in one chemiluminescence

assay and the HPLC assay, and PGA in the remaining assays.

STATISTICS

For each laboratory, the Inean, SD, and CV were computed for

each pool. No attempt was made to separate out among-day,

among-vial, or pure within-run analytical variance when com-

puting SD and CV. To obtain an all-laboratory mean, SD, and

CV, we first computed the mean and SD of the laboratory

means for each pool. Results from any laboratory with a mean

lower than this preliminary all-laboratory mean, minus three

times the corresponding all-laboratory SD, were then excluded,

and a final all-laboratory mean, SD, and CV were computed.

Although the HPLC laboratory provided data from two separate

runs, only one set of results was used so that accuracy and

precision would be determined on the same basis as for the other

laboratories.

Results and DiscussionIndividual serum and whole-blood folate results reported by

each laboratory, as well as the overall results, are presented in

Tables 1 and 2. Results are grouped by similar method type and

are reported in nmol/L (1 nmol/L = 2.265 ng/mL). Laborato-

ries 6, 7, and 8 are grouped to facilitate comparison because they

all used the same version of the L. casei assay. Similarly,

laboratories 9-12 used the Bio-Rad RA, and 13 and 14 used the

Becton Dickinson boil RA.

For pool SFOL-01, two laboratories reported only “>45

nmol/L” and not a diluted specimen result for the highest-

concentration sample; thus, their results for this sample were not

included in the final calculations. Laboratory 14 reported serum

results only, and laboratory 2 reported whole blood results only.

Mid-range folate concentrations appeared to have the best

agreement. Variation at low and high concentrations tended to

be greater than mid-range concentrations, as seen for high

whole blood pool RBCF-01, low whole blood pool RBCF-02,

high serum pool SFOL-04, and low serum pool SFOL-06.

Unlike proficiency-testing challenges, which are essentially aone-time snapshot of method performance, the objective of this

round robin was to present performance averaged over 3 days,

N/A. results not available.

Table 1. Serum folate, mean (SD), by pool, nmol/LSerum pool

Methodcodes: M, L. casei microbiological assay: B-R, Bio-Rad QuantaPhase II 25l radioassay; B-D, ICN/BectonDickinson SimultracS 125 radloassay; B-D NB,

ICN/Becton Dickinson Simultrac S No-boil 1251 radioassay; DPC, Diagnostic Products SPNB Dual Count 1251 radioassay; ACS, Ciba-Coming ACS:180 chemilumines-cence assay; ML, Ciba-Corning Magic Lite chemiluminescence assay; HPLC, in-house assay [16); IC. Abbott Mx ion-capture assay.

b Specimen not diluted for high value (45 nmol/L high standard).

Lab (method) RBCF-O1 RBCF.02 RBCF.03 RBCF-04 RBCF.05 RBCF-06

1 (M) 561.81 (28.53) 151.09 (6.80) 303.79 (10.33) 644.77 (52.32) 451.77 (132.44) 228.48 (16.58)

2 (M) 592.01 (44.73) 161.64 (24.06) 357.64 (42.17) 830.23 (99.32) 546.71 (81.24) 340.32 (51.54)3 (M) 645.62 (130.58) 180.65 (30.81) 456.98 (72.73) 999.92 (212.11) 638.99 (133.63) 381.45 (84.49)4 (M) 399.55 (30.07) 99.85 (14.60) 308.23 (52.68) 452.25 (55.10) 392.98 (21.52) 262.17 (58.43)5 (M) 541.11 (63.25) 150.95 (14.04) 345.75 (47.44) 812.17 (62.51) 510.49 (35.37) 341.73 (42.57)6 (M) 555.87 (29.87) 132.50 (6.07) 326.16 (15.15) 763.87 (38.10) 465.27 (19.50) 281.99 (8.17)

7 (M) 441.30 (46.56) 127.78 (11.36) 318.04 (16.15) 732.92 (29.21) 426.76 (20.21) 263.12 (17.91)

8 (M) 499.62 (40.50) 122.69 (9.10) 332.77 (24.10) 737.64 (62.63) 475.46 (61.88) 270.67 (29.67)

9 (B-R) 470.44 (55.54) 64.74 (3.35) 170.40 (15.21) 349.23 (41.61) 238.88 (26.37) 124.65 (7.01)

10 (B-R) 465.46 (22.43) 61.53 (2.33) 160.82 (9.99) 326.92 (8.82) 215.36 (6.44) 116.84 (2.64)

11 (B-R) 489.99 (44.72) 63.98 (4.48) 165.57 (11.78) 317.84 (20.74) 222.94 (19.41) 121.57 (8.85)

12 (B-R) 403.85 (23.40) 73.12 (2.76) 168.63 (5.83) 332.92 (14.28) 226.24 (13.66) 130.20 (4.58)13 (B-D) 595.32 (25.95) 140.81 (6.07) 283.50 (11.17) 684.03 (30.17) 411.85 (12.29) 244.62 (11.10)

14 (B.D) N/A N/A N/A N/A N/A N/A

15 (B-D NB) 666.61 (34.58) 149.26 (22.18) 285.09 (42.04) 629.20 (53.36) 469.70 (54.28) 229.63 (25.96)

16 (DPC) 916.76 (61.30) 141.60 (11.49) 352.59 (38.91) 521.52 (46.43) 415.06 (39.39) 241.41 (21.96)

17 (ACS) 264.06 (37.67) 63.99 (14.13) 172.52 (28.38) 207.63 (33.54) 168.52 (29.09) 132.88 (29.04)

18 (ML) 436.01 (59.89) 79.84 (21.69) 214.04 (44.53) 366.93 (43.76) 256.89 (38.59) 156.85 (25.88)19 (IC) 464.70 (33.87) 120.23 (5.75) 250.09 (9.97) 309.17 (10.86) 297.47 (11.32) 226.50 (11.26)20 (HPLC) 101.68 (14.39) 72.54 (8.80) 165.96 (15.14) 411.53 (35.19) 240.82 (22.12) 130.04 (24.53)

Al) lab mean 500.62 113.58 270.45 548.98 372.22 222.38

All lab SD 165.57 39.23 86.44 226.90 134.65 82.91

All lab CV, % 33.07 34.54 31.96 41.33 36.17 37.28

Range of means 101.68-916.76 61.53-180.65 160.82-456.98 207.63-999.92 168.52-638.99 116.84-381.45Range of CVs, % 4.4-20.2 2.2-27.2 3.4-20.8 2.7-21.2 3.0-29.3 2.3-22.3

N/A, results not available.a Method codes as in Table 1.

1692 Gunter et al.: Serum and whole-blood folate

Table 2. Whole-blood folate, mean (SD), by pool, nmol/L.Whole-blood pool

with the average of 12 replicate measurements for each of the 12

pools. The variation among laboratories was especially striking

for the whole-blood pools, which varied from threefold (pool

RBCF-06) to ninefold (pool RBCF-01). Results for serum pools

varied from twofold (pool SFOL-03) to ninefold (pool SFOL-

06). Within pools, accuracy also differed markedly, as seen in the

extreme results reported for pool RBCF-0 1.

As stated earlier, the higher-concentration pools were not

achieved by adding purified PGA or 5-MTHF, but instead

reflect naturally occurring folate exposure. Clearly, some char-acteristic of pool RBCF-01 affected the HPLC method quanti-

fication more than did any of the other five whole-blood pools.

When results for this pool are excluded, the HPLC assay results

(laboratory 20) for pools RBCF-02 though RBCF-06 are very

similar to those for the Bio-Rad RA laboratories (labs 9-12).

However, the serum results for laboratory 20 (I-IPLC) are only

about half of those obtained by all other methods. This HPLC

method is specific for 5-MTHF, which constitutes ‘-90% of

circulating plasma folate, but the somewhat mixed results may

indicate a possible loss of folate during extraction.

Similarly, the results for pool RBCF-01 by laboratory 16,which used the Diagnostic Products RA, were extremely high

(917 nmol/L), but results for all other whole-blood pools by thismethod agreed more closely with the results by laboratory 13,

which used the Becton Dickinson boil RA. Serum results from

the Diagnostic Products assay, however, were more comparable

with the ACS:180 results (laboratory 17) than with those from

the Becton Dickinson assays.Within-laboratory precision also varied considerably within

and across methods, as seen in the summary entries for Tables I

and 2. The best overall precision results, as reflected by the

smallest CVs, were those of laboratories 6 (M), 12 (B-R), 13(B-D), and 19 (IC). The ion-capture (IMx) assay is the newest

addition to the folate assay methodology and, as can occur withmany new products, the results in the hands of laboratorians

using the newly adopted method may initially tend to be

somewhat less precise. Overall, the microbiological assay results

tended to have higher CVs than the RA group.

Traditionally, RA has been used clinically because of its

relative ease and greater throughput compared with the L. caseiassay. The RAs are calibrated, as are most of the microbiologicalassays, with PGA and take advantage of the fact that at pH 9.2

PGA and 5-MTHF have equal affinity for the folate-binding

protein. The best microbiological vs RA comparison can be

drawn from laboratories 6-8 (all of which used the same version

of the L. casei assay) and laboratories 9-12 (all of which used the

Bio-Rad RA). Within groups, each assay compares well withitself, but the two assays are not comparable across the range of

serum and whole-blood concentrations examined in the survey.The Bio-Rad RA was initially marketed with calibrator concen-

Lab/co.

B-D S

3.4-38.3

M-Lab 5

Clinical Chemisiiy 42, No. 10, 1996 1693

#{176}Codesas in Table 1.b Includescorrection for hematocrit in final calculation: NOT whole-blood folate as used in round robin.

trations to match the performance of the L. casei assay, but in

response to questions raised by Levine [10], the QuantaPhase II

assay was introduced in late 1993 with spectrophotometrically

verified PGA calibrator concentrations, resulting in a 30% shift

downward in measured folate concentrations. Except for pool

RBCF-0 I, results for the other five erythrocyte pools by this RA

are -50% lower, on average, than those from laboratories 6-8(all using the same L. casei assay). Serum results for the Bio-Rad

RA and laboratories 6-8, however, agree much more closely.

The results for the lower-concentration serum pools,

SFOL-02 and SFOL-06, were especially interesting. Too often,

clinical diagnoses or population assessments are made from

single measurements of serum or plasma folate. Such measure-ments are a problem, particularly because normal subjects and

patients with definite deficiency may have a substantial overlap

in values; accordingly, test results often include “borderline” or“indeterminate” ranges [14]. Clearly, it would be inappropriate

to use one set of reference ranges for all methods. Examples of

the variation in normal ranges cited by the some of the

participating laboratorians, or suggested by the manufacturer of

kits used, vary widely and are compared in Table 3.

In the FLAIR Folate Intercomparison, which preceded the

recalibration of the Bio-Rad QuantaPhase RA., the authors

concluded that most RAs tended to overestimate serum folate

content because standards were improperly calibrated and PGA

rather than 5-MTHF was used as the calibration material /7]. In

our study, however, the RAs as a class clearly did not overesti-

mate the serum folate. The chemiluminescence assays and the

Diagnostic Products RA had the highest overall values for serum

folate and demonstrated substantial bias with the lowest-con-

centration serum pools, SFOL-02 and SFOL-06.Using three different calibrator source materials did not have

a clear-cut effect but may have added to the overall variability.

According to the manufacturer (personal communication, ICN/

Becton Dickinson, December 1995), the chemistries for theAbbott Magic-Lite (using PGA calibrators) and ACS: 180 (using5-MTHF) are similar; however, we found that, although they

gave comparable results for the serum pools, their results for the

erythrocyte pools were quite different. The manufacturers’

suggested reference ranges for these assays vary slightly as well,which may be partially due to the different populations surveyedfor this purpose or may reflect true assay differences. Labora-

tories 1 and 5 used folinic acid as calibrator material, but these

laboratories are not necessarily more comparable with each

other than with other laboratories using the L. casei assay.Lacking a “gold standard” method or material, a definition of

which results are “correct” is impossible, in that all of thesemethods have inherent errors. We examined weighted linear

regressions of each laboratory’s mean results against the meansfor the same method type, separately for the serum and whole-

blood pools. For the eight microbiological assays, the slopes

varied by 0.8 1-1.22 for the serum pools and by 0.77-1.32 for the

whole-blood pools. For the eight RAs (collapsed together for

greater statistical power), the respective slope variances were

0.81-1.28 and 0.75-1.15, and for the two chemiluminescence

assays they were 0.91, 1.10 and 0.71, 1.29. (Results for the

ion-capture and HPLC assays are not presented because they

represent only one laboratory per method type.)

Table 3. VariatIon in reference ranges used by some participatIng laboratorIes or suggested by manufacturers of assay kits.Folate range

Lab 15(B-D S-NB)

DPC

ACSM-LB-R

M-Lab 1M-Lab 2M-Labs 6-8

“Low”“Indeterminate”“Normal”

“Low”

“Normal”

“Low”“Indeterminate”“Normal”“Normal”“Normal”“Deficient”“Normal”

“Deficient”“Normal”“Normal”“Normal”“Deficient”“Possible deficiency”“Probable normal”“Normal”

In serum, nmol/L<3.6

3.6-5.0

>5.0

<3.4

6.8-38.55.9-39.2

>8.2<3.4

3.4-46.7<6.3

7.0-28,0>6.85.00-50.0

<4.5

4.5-6.1

>6.1

4.5-22.7

In erythrocytes,nmol/L RBCb

<283

�283

<272

�272-1948

<227227-453

396-1586328-1110

272-974<136

215-1291

337-1461

>317500-1300

<227227-340

>340317-770

1694 Gunter et al.: Serum and whole-blood folate

In summary, our comparisons of serum and whole-blood folate

concentrations measured in 20 research and clinical laboratories

by 7 different types of assays showed considerable intra- and

intermethod variation, These results demonstrate that folate

concentrations measured in one laboratory cannot be reliably

compared with those assayed in another laboratory without an

evaluation of interlaboratory differences. Hence, comparing

folate data for different study populations obtained by use of

different analytical methods is difficult. Excellent discussions ofmethod comparisons and possible sources of variability have

been previously presented by Shane et a!. [15] and Brown et al.

[16]. The reasons they cite for potential methodological differ-

ences are still valid but have become even more complex with

the emergence of new analytical technologies. Using method-

specific reference ranges in which “deficient” status results areclinically confirmed by other hematological markers is essential,

Developing Definitive Methods and Certified Reference Mate-

rials is urgently needed for serum and erythrocyte folate analy-sis. In 1994, at the Sixth Conference for Federally Funded

Human Nutrition Research Centers in Bethesda, MD, a strong

argument was presented for a collaborative development of

these methods and materials (Wayne Wolf, personal communi-

cation). Recent epidemiological evidence on the increasing

importance of folate underscores the need for analytical accu-

racy in its measurement. Improved standardization is needed

before researchers can collectively contribute to the evaluationof how well various folate cutoff values predict folate-prevent-

able conditions such as megaloblastic anemia, hyperhomocys-

teinemia, and neural tube defects. Joint efforts toward this goal

by the National Institute for Standards and Technology, CDC,

US Department of Agriculture, National Committee for Clin-

ical Laboratory Standards, Institute for Food Researchll.JK,NIBSC, and ICSH should be rigorously supported.

This work was supported in part by the National Center for

Environmental Health and by the National Center for Health

Statistics, CDC. We thank the following laboratorians for their

participation and support: Lynn Bailey, Jesse Gregory, and C.M.

Pfeiffer, Department of Food Science and Human Nutrition,

University of Florida, Gainesville, FL; Selwyn DeSouza, Bio-

Rad Laboratories, Hercules, CA; Phillip Garry and Patricia

Stauber, School of Medicine, University of New Mexico, Albu-

querque, NM; Ralph Green, Michael Miller, and Lynn Man-

teuffel, Cleveland Clinic Foundation, Cleveland, OH; Paul

Finglas, Institute of Food Research, Norwich, UK; Victor

Herbert and Spencer Shaw, Bronx VA Medical Center, Bronx,

NY; Sean O’Broin and Brian Kelleher, St. James’ Hospital,

Dublin, Ireland; Robert Jacob, USDA Human Nutrition Re-

search Center, The Presidio, San Francisco, CA, John Linden-

baum, Columbia University Medical Center, New York, NY;

Klaus Pietrzik, Department of Pathophysiology of Human

Nutrition, University of Bonn, Bonn, Germany; Jacob Selhub,Marie Nadeau, and Irwin Rosenberg, USDA Human Nutrition

Research Center at Tufts University, Boston, MA; John Scott,

Zana Kelley, and Eileen Hallinan, Department of Biochemistry,

Trinity College, Dublin, Ireland; Nefertiti Sourial, St. Bar-

tholomew’s Hospital, London, UK; Tsunenobu Tamura and

Howerde Sauberlich, Nutrition Sciences Department, Univer-

sity of Alabama-Birmingham, Birmingham, AL; Chris Thomas

and Regine Steegers-Theunissen, University Hospital St. Rad-

boud, Nijmegen, The Netherlands; Johan Ubbink, University

of Pretoria, Pretoria, South Africa; and David Wilson, Abbott

Laboratories, Abbott Park, IL.We also thank the staff of the NHANES Laboratory, CDC,

for their contributions in preparing and shipping these materials

to the participating laboratories.

References1. MRC Vitamin Study Research Group. Prevention of neural tube

defects: results of the Medical Research Council vitamin study.Lancet 1991;338:131-7.

2. Czeizel AE, Dudas I, Metneki J. Pregnancy outcomes in a random-ized controlled clinical trial of periconceptional multivitamin sup-plementation. Final report. Arch Gynecol Obstet 1994;255(3):

131-9.3. Green R, Jacobsen DW. Clinical implications of hyperhomocys-

teinemia. In: Bailey LB, ed. Folate in health and disease. NewYork: Marcel Dekker, 1995:99-102.

4. Selhub J, Jacques PF, Wilson PWF, Rush 0, Rosenberg IH. Vitaminstatus and intake as primary determinants of homocysteine in anelderly population. JAMA 1993:270:2693-8.

5. Sauberlich HE. Folate status of U.S. population groups. In: BaileyLB, ed. Folate in health and disease. New York: Marcel Dekker,

1995:171-94.6. College of American Pathologists. 1996 CAP Surveys. Northfield,

IL, November 1995:36.7. Van den Berg H, Finglas PM, Bates CJ. FLAIR intercomparisons on

serum and red cell folate. Int J Vit Res 1994:64:288-93.8. Instruction Manual, Bio-Rad QuantaPhase folate radioassay kit.

Hercules, CA: Bio-Rad Labs., 1987.9. Gunter EW, Lewis BG, Koncikowski SM. Laboratory methods used

for the Third National Health and Nutrition Examination Survey(NHANES Ill), 1988-1994. Atlanta, GA: Centers for DiseaseControl and Prevention, 1996: in press.

10. Levine S. Analytical inaccuracy for folic acid with a popularcommercial vitamin B12/folate kit [Letter). Clin Chem 1993;39:2209-10.

11. Raiten DJ, Fisher KD, eds. Assessment of folate methodologyused in the Third National Health and Nutrition ExaminationSurvey (NHANES Ill, 1988-1994). J Nutr 1995;125:1371S-98S.

12. Baker Fl, Herbert V. Frank 0, Pasher I, Hutner SH, Wasserman LR,

Sobotka H. A microbiologicmethod for detectingfolicacid defi-ciency in man. Clin Chem 1959:5:275-82.

13. Wilson OH, Herrmann R, Hsu 5, Biegalski T, Sohn L, Forsythe C,et al. Ion capture assay for folate with the Abbott IMx#{174}Analyzer[Tech Brief]. Clin Chem 1995:41:1780-1.

14. Lindenbaum J, Allen RH. Clinical spectrum and diagnosis of folatedeficiency. In: Bailey LB, ed. Folate in health and disease. NewYork: Marcel Dekker, 1995:53.

15. Shane B, Tamura T, Stokstad ELR. Folate assay: a comparison ofradioassay and microbiological methods. Clin Chim Acta 1980;100:13-9.

16. Brown RD. Jun R, Hughes W, Watman R, Arnold B, Kronenberg H.Red cell folate assays: some answers to current problems withradioassay variability. Pathology 1990:22:82-7.