Content uploaded by Nathan G Dodder
Author content
All content in this area was uploaded by Nathan G Dodder on Oct 16, 2015
Content may be subject to copyright.
Certification of Creatinine in a Human Serum
Reference Material by GC-MS and LC-MS
Nathan G. Dodder,
1*
Susan S.-C. Tai,
1
Lorna T. Sniegoski,
1
Nien F. Zhang,
2
and
Michael J. Welch
1
Background: To meet recommendations given by the
Laboratory Working Group of the National Kidney
Disease Education Program for improving serum creat-
inine measurements, NIST developed standard refer-
ence material (SRM) 967 Creatinine in Frozen Human
Serum. SRM 967 is intended for use by laboratories and
in vitro diagnostic equipment manufacturers for the
calibration and evaluation of routine clinical methods.
Methods: The SRM was produced from 2 serum pools
with different creatinine concentrations. The concentra-
tions were certified using a higher-order isotope-dilu-
tion GC-MS method and an isotope-dilution LC-MS
method. The LC-MS method is a potential higher-order
reference measurement procedure.
Results: The GC-MS mean (CV) concentrations were
67.0 (0.9%)
mol/L for serum pool 1 and 346.1 (0.45%)
mol/L for serum pool 2. The LC-MS results were 66.1
(0.2%)
mol/L and 346.3 (0.2%)
mol/L, respectively. For
serum pool 1, there was a 1.4% difference between the
mean GC-MS and LC-MS measurements, and a 0.10%
difference for serum pool 2. The results from the 2
methods were combined to give the certified concentra-
tions and expanded uncertainties.
Conclusions: The certified concentration (expanded un-
certainty) of SRM 967 was 66.5 (1.8)
mol/L for serum
pool 1 (a value close to the diagnostically important
concentration of 88.4
mol/L) and 346.2 (7.4)
mol/L for
serum pool 2 (a concentration corresponding to that
expected in a patient with chronic kidney disease).
© 2007 American Association for Clinical Chemistry
The concentration of creatinine in serum is a diagnostic
marker for chronic kidney disease (CKD).
3
An estimated
20 million Americans have kidney disease (1 ), and the
number is rising, primarily owing to the increasing inci-
dence of diabetes and high blood pressure. Between 1988
and 2003, the number of patients on dialysis nearly tripled
(2). Early detection of CKD, followed by drug treatments,
can prevent or postpone kidney failure. The simplest and
most widespread method of detecting kidney disease is
through measurement of blood creatinine concentrations.
Recognizing that more accurate blood creatinine measure-
ments will lead to better diagnosis of early stage kidney
disease, the Laboratory Working Group of the National
Kidney Disease Education Program (NKDEP) outlined a
series of recommendations, including the development of
a reference material (3 ).
The NKDEP recommended that a serum reference
material with a creatinine concentration of 88.4
mol/L
(1.00 mg/dL) be developed (3). This value corresponds to
a glomerular filtration rate (GFR) of approximately 60
mL 䡠 min
⫺1
䡠 1.73 m
⫺2
. CKD is defined as a GFR of ⬍60
mL 䡠 min
⫺1
䡠 1.73 m
⫺2
for ⱖ3 months (3 ). The GFR de-
scribes the ability of the kidneys to filter waste products
from the blood and is estimated based on the patient’s
serum creatinine concentration, age, sex, and race. There-
fore, 88.4
mol/L is near the critical concentration that
would determine a positive or negative diagnosis. Also,
compared with higher creatinine concentrations, errors
associated with the calibration or measurement precision
of creatinine at this relatively low concentration will have
a greater impact on the error of the estimated GFR (3).
To meet the recommendations set by the NKDEP, NIST
has developed standard reference material (SRM) 967,
Creatinine in Frozen Human Serum (4). This SRM con-
sists of 2 serum pools with target creatinine concentra-
1
Analytical Chemistry Division and
2
Statistical Engineering Division,
National Institute of Standards and Technology, Gaithersburg, MD.
* Address correspondence to this author at: National Institute of Standards
and Technology, 100 Bureau Dr., Stop 8392, Gaithersburg, MD 20899-8392. Fax
301-977-0685; e-mail nathan.dodder@nist.gov.
Received April 6, 2007; accepted July 3, 2007.
Previously published online at DOI: 10.1373/clinchem.2007.090027
3
Nonstandard abbreviations: CKD, chronic kidney disease; NKDEP, Lab-
oratory Working Group of the National Kidney Disease Education Program;
GFR, glomerular filtration rate; SRM, standard reference material; JCTLM,
Joint Committee on Traceability in Laboratory Medicine.
Clinical Chemistry 53:9
1694–1699 (2007)
General Clinical
Chemistry
1694
tions of approximately 88.4
mol/L (the diagnostically
critical concentration) and 354
mol/L (4 mg/dL, a
concentration corresponding to that in a patient with
CKD). The creatinine concentrations were certified using
2 independent methods. The isotope-dilution gas chroma-
tography–mass spectrometry (GC-MS) method (5) is con-
sidered to be a higher-order reference measurement pro-
cedure by the NCCLS (6 ) and the Joint Committee on
Traceability in Laboratory Medicine (JCTLM) (7 ). The
isotope-dilution LC-MS method is similar to a procedure
developed at the Laboratory of the Government Chemist
that was approved by the JCTLM as a higher-order
reference measurement procedure (7, 8). The LC-MS
method has fewer sample preparation steps and is a
higher-throughput method than the GC-MS method. The
commutability of SRM 967 was then verified in a study
organized by the NKDEP.
SRM 967 has a frozen serum matrix. Frozen serum
more closely matches the native state of clinical samples
than a lyophilized matrix. Reference materials supplied
by other metrology institutes are either lyophilized (9) or
do not have the target creatinine concentrations described
above (10 ).
Materials and Methods
We used 2 methods, based on GC-MS and LC-MS, to
confirm the accuracy of the quantification; both used
isotope dilution. The methods were independent; i.e.,
they used different procedures to measure the same
analyte. The sample preparation (ion exchange chroma-
tography vs protein precipitation), chromatography (gas
chromatography vs liquid chromatography), ionization
(electron impact vs electrospray), internal standards (cre-
atinine-
13
C
2
vs creatinine-d
3
), and quantification protocol
(bracketing vs linear regression) (11, 12) differed between
the methods.
preparation of srm 967
The pools of human serum used for SRM 967 were
prepared by Solomon Park Research Laboratories.
4
Blood
was collected from healthy, postmenopausal, adult
women following CLSI guidelines (13 ). The resulting
serum master pool of approximately 3 L was split into 2
pools. Pool 1 was not enriched with additional creatinine.
Pool 2 was enriched with an appropriate amount of
reagent-grade creatinine to bring the concentration up to
approximately 354
mol/L. Both pools were passed
through filters with a 0.2-
m pore size. No preservatives
were added. One-milliliter aliquots of the pools were
placed in 3-mL amber glass vials and capped with Teflon
stoppers and aluminum seals. The vials were frozen at
⫺80 °C until analysis.
preparation of the calibration standards
Certification of the creatinine concentrations in SRM 967
was performed at NIST. We made calibration solutions
that contained known unlabeled:labeled creatinine mass
ratios and internal standard solutions that contained
known masses of labeled creatinine. The internal standard
solutions were added to the samples at the beginning of
the sample preparation; the mass of the added labeled
creatinine was approximately equal to the mass of unla-
beled creatinine in the sample. To achieve this 1:1 ratio,
we performed a preliminary quantification in which a
wider range of mass ratios was used. Once the approxi-
mate creatinine concentration was measured, the quantity
of internal standard necessary for a 1:1 ratio was calcu-
lated. After sample processing, we ran the calibration
standards and samples in the same set on the mass
spectrometer. The unlabeled:labeled creatinine peak area
ratios in the samples were converted to mass ratios using
data from the calibration standard runs and either a
bracketing method or linear regression method, as de-
scribed below. The mass ratios were then solved for the
mass of the unlabeled creatinine, and the concentration of
unlabeled creatinine in each sample was calculated.
For each analytical method, we gravimetrically pre-
pared an independent stock solution of labeled creatinine
to make the calibration standards and sample internal
standard solutions. We weighed approximately 1.25 mg
labeled creatinine (either creatinine-
13
C
2
or creatinine-d
3
)
into a 50-mL volumetric flask, added 50 mL water, and
calculated the concentration of the solution (approxi-
mately 0.025 mg/g). This solution was split into 3 sets,
each set containing a different mass of labeled creatinine.
One set was used as internal standards for the samples
from serum pool 1, the 2nd set as internal standards for
the samples from serum pool 2, and the 3rd set to prepare
the calibration standards. The internal standard aliquots
were stored at ⫺20 °C.
The accuracy of the quantification was limited by the
accuracy of the mass of unlabeled creatinine in the cali-
bration standards. To test for bias in the GC-MS calibra-
tion standards, 2 independent solutions of unlabeled
creatinine were gravimetrically prepared as follows. We
weighed approximately 20 mg solid creatinine SRM 914a
(14) into a volumetric flask and added 20 mL water.
We transferred 2.5 mL of the solution to a 100-mL
volumetric flask and added water to a final volume of
100 mL. We then calculated the concentration (approxi-
mately 0.025 mg/g) of the stock solution.
We made calibration standards from both unlabeled
stock solutions such that the unlabeled:labeled creatinine
mass ratios of the 2 sets of calibration standards were
offset from each other by 0.1 units and ranged from 0.8 to
1.2. To test for bias, we ran the calibration standards as a
single set. The linearity of the resulting calibration curve
4
Certain commercial instruments and materials are identified in this
report to adequately specify the experimental procedures. Such identification
does not imply endorsement by the National Institute of Standards and
Technology, nor does it imply that the instruments and materials identified are
the best available for the purpose.
Clinical Chemistry 53, No. 9, 2007 1695
confirmed that the 2 independently prepared calibration
standard sets were not biased. The calibration standards
for the GC-MS method were derivatized as described
below and reconstituted in hexane to the same concentra-
tion as the GC-MS samples, approximately 10 mg/L, and
stored at ⫺20 °C until analysis. The calibration standards
for the LC-MS method were made in a similar manner,
except we prepared an independent solution of unlabeled
creatinine for each of the 3 sample sets. We diluted the
LC-MS calibration standards with 10 mmol/L ammonium
acetate to the same concentration as the LC-MS samples,
approximately 1.6 mg/L, and stored them at ⫺20 °C until
analysis.
quality control samples
SRM 909b Human Serum, which had been previously
certified for creatinine concentrations (15, 16), was ana-
lyzed along with SRM 967 to confirm the accuracy of the
analysis. SRM 909b consists of lyophilized serum (2
pools), and each vial was reconstituted with 10.00 mL
water and allowed to equilibrate for 1.5 h before being
prepared along with SRM 967.
quantification by gc-ms
Two aliquots from each SRM 967 vial and 1 aliquot from
each SRM 909b vial were measured. The aliquots were
added gravimetrically to the creatinine-
13
C
2
internal stan-
dard solutions and equilibrated overnight at 5 °C. Ion-
exchange chromatography was necessary to separate cre-
atine from creatinine before derivatization, because these
2 compounds will form the same derivative. Amberlite
IRC-50 ion-exchange resin (Chemical Dynamics) was
washed in water and soaked in 1.0 mol/L HCl for 3.5 h
with occasional agitation. The resin was then rinsed with
water and stored in excess 0.1 mol/L HCl until use. The
resin was slurry packed into 20 cm by 10-mm columns
using water. The volume of resin in each column was 5
mL. The resin was washed with 150 mL water, then the
samples were added to the columns. We eluted the
creatine with 75 mL water; this fraction was discarded.
We eluted the creatinine with 75 mL of 1.0 mol/L ammo-
nium hydroxide. We used a separate set of samples to
measure the densities using the Lang–Levy pipette
method (17 ).
The derivatization reaction required the samples to be
completely free of water. To remove the water and
ammonium hydroxide, the samples were freeze dried and
then reconstituted in 100% ethanol. We removed the solid
residue by passing the samples through polyvinylidene
fluoride syringe filters (25-mm diameter, 0.45-
m pore
size). The samples were evaporated to near dryness under
vacuum, and the creatinine was converted into a deriva-
tive according to the described procedure (5). The sam-
ples were then solvent exchanged into hexane. This
brought the final concentration of the samples to approx-
imately 10 mg/L. The samples were stored at ⫺20 °C until
analysis.
We performed the measurements by use of an Agilent
5972 GC-MS. The injection volume was 1
L. The GC
column was 30-m long, with a 0.25-mm internal diameter,
and a 0.25-
m thick DB-5ms stationary phase (Agilent).
The GC oven temperature program was 130 °C for 2 min,
12 °C/min to 250 °C, 250 °C for 0.5 min. The mass spec-
trometer was operated in the electron impact ionization
mode with selected ion monitoring of the [M-73]
⫹
ions at
m/z 150 and 152 for the unlabeled and labeled forms,
respectively.
We ran 3 separately prepared sets of samples on the
GC-MS. Each set consisted of 10 samples: 4 samples of
SRM 967 pool 1, 4 samples of SRM 967 pool 2, 1 sample of
SRM 909b pool 1, and 1 sample of SRM 909b pool 2.
Calibration was by bracketing, i.e., each sample was
measured in duplicate, in between duplicate measure-
ments of the 2 calibration standards with unlabeled:
labeled peak area ratios just below and above that of the
sample. We calculated the mass ratios by linear interpo-
lation between the bracketing standards for each sample.
We then repeated the measurements on a 2nd day, with
the order of the standards reversed. The results of the
2-day measurements were averaged to arrive at the mass
ratios from which the creatinine concentrations in the
samples were calculated.
quantification by lc-ms
Two aliquots from each SRM 967 vial and 1 aliquot from
each SRM 909b vial were measured. Aliquots of SRM 967
and SRM 909b were added gravimetrically to the creati-
nine-d
3
internal standard solutions and equilibrated over-
night at 5 °C. The proteins were precipitated by adding 3
volumes of ice-cold ethanol to each cold tube and vortex-
mixing. After standing for 5 min, the samples were
centrifuged at 900g for 20 min. The supernatant, contain-
ing the creatinine, was removed and concentrated to
dryness using a N
2
stream. Each sample was reconstituted
in 1 mL water and filtered through a polyvinylidene
fluoride syringe filter (13-mm diameter, 0.2-
m pore size).
We diluted the samples to a concentration of approxi-
mately 1.6 mg/L with 10 mmol/L ammonium acetate and
stored them at ⫺20 °C until analysis.
We performed the measurements by use of an Agilent
1100 series LC-MS. The injection volume was 4
L,
corresponding to approximately 6 ng creatinine. The
liquid chromatography column was a 15-cm-long, 2.0 mm
internal diameter, 3
m particle diameter, LUNA C18 (2)
(Phenomenex). The gradient mobile phase program was
10 mmol/L ammonium acetate for 7 min, ramped to 20%
10 mmol/L ammonium acetate and 80% acetonitrile by
7.1 min, and held for 13 min. The flow rate was 0.2
mL/min. The column temperature was 23 °C. The mass
spectrometer was operated using positive mode electros-
pray ionization and selective ion monitoring of the
(M⫹H)
⫹
ions at m/z 114 and 117 for creatinine and
creatinine-d
3
, respectively. The ionization source param-
eters were drying gas temperature 350 °C, N
2
gas flow 12
1696 Dodder et al.: Certification of Creatinine in Reference Material
L/min, nebulizer pressure 170 kPa (25 psi), capillary 1500
V, and fragmentor 120 V.
We ran 3 separately prepared sets of samples on the
LC-MS. Each set consisted of 14 samples: 6 samples of
SRM 967 pool 1, 6 samples of SRM 967 pool 2, 1 sample of
SRM 909b pool 1, and 1 sample of SRM 909b pool 2. Each
set was run as follows: the 5 calibration standards were
run 1st; followed by the samples; then the samples were
measured again in the reverse order; and last, the 5
calibration standards were run in reverse order. We
calculated a composite linear regression, using a slope-
intercept model, from the peak areas of the calibration
standards. We used the linear calibration to calculate the
mass ratios using the average of the duplicate sample
peak area measurements, which we then used to calculate
the creatinine concentration in the samples.
Results
Table 1 lists the measured concentrations of creatinine in
SRM 967. Examples of selected ion chromatograms from
each serum pool are shown in Fig. 1. The concentration
data were corrected for the purity (estimated uncertainty)
of the reference standard SRM 914a: 99.7% (0.3%). The
concentration unit conversion from mg/g to
mol/L was
performed using the densities of SRM 967. The pool 1
material had a density of 1.0231 g/mL; the pool 2 material
had a density of 1.0226 g/mL.
Among 12 GC-MS measurements, the largest was 69.2
mol/L. This point was identified as an outlier by Grubb
and Dixon tests and excluded from the statistical analysis
and calculation of the certified concentrations. The
GC-MS mean concentrations (SD) for SRM 967 were 67.0
(0.6)
mol/L for pool 1 and 346.1 (1.6)
mol/L for pool 2.
The LC-MS results were 66.1 (0.2)
mol/L for pool 1 and
346.3 (0.8)
mol/L for pool 2. For pool 1, there was a 1.4%
difference between the mean GC-MS and LC-MS mea-
surements, and a 0.10% difference for pool 2. There was
no evidence of inhomogeneity within or among the vials,
or of a concentration trend corresponding to the vial
filling order. The results of the control measurements are
listed in Table 2 and were within 1.0% of the certified
values for SRM 909b.
Discussion
The measurement of creatinine in the SRM 967 pool 1
samples by GC-MS was slightly more variable and biased
toward higher values than the LC-MS method. This effect
was obscured at higher creatinine concentrations and led
Table 1. Quantification of creatinine in SRM 967 by GC-MS and LC-MS.
a
GC-MS method LC-MS method
SRM vial Pool 1,
mol/L SRM vial Pool 2,
mol/L SRM vial Pool 1,
mol/L SRM vial Pool 2,
mol/L
Set 1 Set 1
1 69.2 3 345.8 1 66.2 4 346.0
1 66.6 3 344.9 1 66.2 4 346.9
2 67.7 4 344.9 2 66.0 5 345.9
2 66.9 4 343.4 2 66.1 5 344.6
3 65.9 6 345.6
Set 2 3 65.9 6 345.7
5 68.1 7 346.8
5 67.2 7 346.7 Set 2
6 67.5 8 348.4 7 66.1 10 345.5
6 67.1 8 348.0 7 66.3 10 346.4
8 66.2 11 346.1
Set 3 8 65.9 11 346.3
9 66.4 11 343.6 9 66.0 12 346.0
9 66.6 11 346.5 9 65.9 12 346.1
10 66.4 12 346.8
10 66.4 12 346.7 Set 3
13 66.3 16 347.3
13 66.3 16 347.3
Mean 67.0 346.1 14 65.9 17 347.4
SD 0.6 1.6 14 66.3 17 347.9
CV 0.9% 0.45% 15 66.0 18 346.0
15 65.9 18 347.0
Mean 66.1 346.3
SD 0.2 0.8
CV 0.2% 0.2%
a
Two aliquots from each SRM vial were measured. The serum pool 1 data point 69.2
mol/L from the GC-MS method was excluded from the statistical analysis.
Clinical Chemistry 53, No. 9, 2007 1697
to the difference between the average results obtained by
the 2 methods to be greater for the pool 1 samples (1.4%)
than the pool 2 samples (0.10%). The difference between
the pool 1 measurements was considered small enough
such that the results could be combined in the same way
as the pool 2 measurements, as described below.
calculation of certified concentrations and
expanded uncertainties
The results from the 2 methods were combined using a
Bayesian approach (18–20). This approach, intended for
certifying data from a small number of analytical meth-
ods, assumes that both the means and the variances of the
methods could be different. The means were combined by
Eq. 1, where cˆ is the combined mean and c
1
and c
2
are the
means from the 2 methods. The combined mean is the
certified concentration.
cˆ ⫽
c
1
⫹ c
2
2
(1)
The variances of the combined means, u
2
, were calculated
by Eq. 2, where u
1
2
and u
2
2
are the variances of the
measurements of the 2 methods.
u
2
⫽
u
1
2
⫹ u
2
2
2
⫹
共c
1
⫺ c
2
兲
2
4
(2)
The expanded uncertainties of the measurements were
calculated following NIST guidelines (21). A combined
standard uncertainty for each serum pool was calculated
from type A and type B uncertainties. Type A uncertain-
ties are calculated using statistical methods; type B uncer-
tainties are estimated based on judgment (nonstatistical
methods). The combined variances, u
2
, were the type A
uncertainties. The corresponding type B uncertainties
were assigned a value of 1% of the combined mean, cˆ, for
each serum pool, to account for undetected interferences
and the uncertainty in the purity of the reference standard
SRM 914a (other sources of type B uncertainty were
considered negligible). The type A and B uncertainties
were combined using Eq. 3 to give combined standard
uncertainties, u
c
, for each serum pool.
u
c
⫽
冑
u
2
⫹ 共cˆ 䡠 0.01兲
2
(3)
The expanded uncertainty, U, for each serum pool was
calculated by multiplying the combined standard uncer-
tainties, u
c
, by the coverage factor k ⫽ 2; i.e., U ⫽ ku
c
. The
expanded uncertainty is half the length of the interval
about the certified concentration (cˆ ⫾ U) that is expected
Table 2. Quality control measurements of SRM 909b from
the GC-MS and LC-MS sample sets.
Method Set
Pool 1
(
mol/L)
Pool 2
(
mol/L)
GC-MS 1 55.6 467.0
2 55.8 468.0
3 55.4 466.7
Mean 55.6 467.2
SD 0.2 0.7
LC-MS 1 56.1 465.6
2 55.5 465.9
3 56.1 467.5
Mean 55.9 466.3
SD 0.4 1.0
Certified SRM value 56.18 467.4
Expanded uncertainty 0.55 5.3
Difference, GC-MS, % ⫺1.0 ⫺0.04
Difference, LC-MS, % ⫺0.50 ⫺0.23
Table 3. Calculation of the expanded uncertainty for
SRM 967.
Variable
Pool 1,
mol/L
Pool 2,
mol/L
Combined mean,
ˆ
c 66.53 346.20
Combined variance,
ˆ
u
2
0.38 1.60
Combined standard uncertainty, u
c
0.91 3.69
Expanded uncertainty, U 1.82 7.37
Fig. 1. Example GC-MS and LC-MS se-
lected ion monitoring chromatograms of
unlabeled and labeled creatinine.
1698 Dodder et al.: Certification of Creatinine in Reference Material
to encompass a large fraction (approximately 95%) of the
measurements obtained by subsequent analyses. Table 3
lists the values at each step of the calculation. The
certified, SI traceable, concentrations (expanded uncer-
tainty) for SRM 967 are 66.5 (1.8)
mol/L for serum pool
1 and 346.2 (7.4)
mol/L for serum pool 2, or 0.753 (0.021)
mg/dL for serum pool 1 and 3.916 (0.083) mg/dL for
serum pool 2.
commutability validation
A commutability validation study was organized by the
NKDEP. Commutability refers to the ability of the SRM to
give similar results to real patient samples when analyzed
by different analytical methods. The experimental design
followed a protocol recommended by the CLSI (22 ).
Briefly, creatinine was measured in SRM 967 and individ-
ual patient serum samples using routine laboratory meth-
ods and the NIST LC-MS method described above. SRM
967 was found to be commutable with 15 methods from 7
in vitro diagnostic equipment manufacturers (23).
Grant/funding support: None declared.
Financial disclosures: None declared.
Acknowledgments: We thank Karen Phinney, Mary Satter-
field, Katherine Sharpless, and Stephen Wise, all from NIST,
for reviewing the manuscript.
References
1. US Renal Data System. 2004 Annual Data Report/Atlas. http://
www.usrds.org/atlas_2004.htm (accessed April 2007).
2. US Renal Data System. 2005 Annual Data Report/Atlas. http://
www.usrds.org/atlas_2005.htm (accessed April 2007).
3. Myers GL, Miller WG, Coresh J, Fleming J, Greenberg N, Greene T,
et al. Recommendations for improving serum creatinine measure-
ment: a report from the laboratory working group of the National
Kidney Disease Education Program. Clin Chem 2006;52:5–18.
4. NIST. Certificate of analysis, standard reference material 967,
creatinine in human serum, 2007. http://ts.nist.gov/
MeasurementServices/ReferenceMaterials/232.cfm (accessed
April 2007).
5. Welch MJ, Cohen A, Hertz HS, Ng KJ, Schaffer R, Vanderlijn P, et
al. Determination of serum creatinine by isotope-dilution mass-
spectrometry as a candidate definitive method. Anal Chem 1986;
58:1681–5.
6. NCCLS. Development of definitive methods for the National Ref-
erence System for the Clinical Laboratory. NCCLS Publication
NRSCL 1-A. Wayne, PA: Approved Guideline, 1991.
7. Joint Committee on Traceability in Laboratory Medicine. Database
of higher order reference materials and reference measurement
procedures for laboratory medicine and in vitro diagnostics. JCTLM
List I, 2006. http://www.bipm.org/utils/en/xls/jctlm_listI.xls (ac-
cessed April 2007).
8. Stokes P, O’Connor G. Development of a liquid chromatography-
mass spectrometry method for the high-accuracy determination of
creatinine in serum. J Chromatogr B Analyt Technol Biomed Life
Sci 2003;794:125–36.
9. Institute for Reference Materials and Measurements. Certificate
of analysis for BCR-573, BCR-574, and BCR-575, creatinine in
human serum, 2005. http://www.irmm.jrc.be/html/reference_
materials_catalogue/catalogue/index.htm (accessed April 2007).
10. Laboratory of the Government Chemist. Certificate of analysis,
ERM-DA252a and ERM-DA253a, creatinine in frozen human se-
rum, 2007. http://www.lgcpromochem.com/home/home_en.
aspx (accessed April 2007).
11. Cohen A, Hertz HS, Mandel J, Paule RC, Schaffer R, Sniegoski LT,
et al. Total serum cholesterol by isotope dilution/mass spectrom-
etry: a candidate definitive method. Clin Chem 1980;26:854 – 60.
12. Yap WT, Schaffer R, Hertz HS, White E, Welch MJ. On the
difference between using linear and non-linear models in bracket-
ing procedures in isotope-dilution mass-spectrometry. Biomed
Mass Spectrom 1983;10:262–4.
13. CLSI. Preparation and validation of commutable frozen human
serum pools as secondary reference materials for cholesterol
measurement procedures; approved guideline C37-A. CLSI, 1999.
http://www.clsi.org (accessed April 2007).
14. NIST. Certificate of analysis, standard reference material 914a,
creatinine, 1994. http://ts.nist.gov/MeasurementServices/
ReferenceMaterials/232.cfm (accessed April 2007).
15. NIST. Certificate of analysis, standard reference material 909b,
human serum, 2004. http://ts.nist.gov/MeasurementServices/
ReferenceMaterials/232.cfm (accessed April 2007).
16. Phinney CS, Murphy KE, Welch MJ, Ellerbe PM, Long SE, Pratt KW,
et al. Definitive method certification of clinical analytes in lyophi-
lized human serum: NIST standard reference material (SRM)
909 b. Fresenius J Anal Chem 1998;361:71– 80.
17. Sniegoski LT, Moody JR. Determination of serum and blood
densities. Anal Chem 1979;51:1577– 8.
18. Hornikova A, Zhang NF, Welch MJ, Tai S. An application of
combining results from multiple methods: statistical evaluation of
uncertainty for NIST SRM 1508a. Metrologia 2006;43:205–12.
19. Liu HK, Zhang NF. Bayesian approach to combining results from
multiple methods. Proceedings of the Section on Bayesian Sta-
tistical Sciences, American Statistical Association Meeting,
2001, Atlanta, GA.
20. Satterfield MB, Sniegoski LT, Sharpless KE, Welch MJ, Hornikova
A, Zhang NF, et al. Development of a new standard reference
material: SRM 1955 (homocysteine and folate in human serum).
Anal Bioanal Chem 2006;385:612–22.
21. Taylor BN, Kuyatt CE. NIST Technical Note 1297: guidelines for
evaluating and expressing the uncertainty of NIST measurement
results. http://physics.nist.gov/Pubs/guidelines/contents.html
(accessed April 2007).
22. CLSI. Evaluation of matrix effects. CLSI Document EP14-A2,
Approved Guideline, 2nd ed. 2005. http://www.clsi.org (ac-
cessed April 2007).
23. National Kidney Disease Education Program. Commutability study
results: NIST SRM 967 and CAP LN24: spring/summer 2006.
http://www.nkdep.nih.gov/labprofessionals/commutabilitystudy.
htm (accessed April 2007).
Clinical Chemistry 53, No. 9, 2007 1699