Clinica Chimica Acta 455 (2016) 202208

Contents lists available at ScienceDirect

Clinica Chimica Acta

journal homepage: www.elsevier.com/locate/clinchim

A high-throughput mass spectrometry assay to simultaneously measure

intact insulin and C-peptide
Steven W. Taylor, Nigel J. Clarke , Zhaohui Chen 1, Michael J. McPhaul
Quest Diagnostics Nichols Institute, San Juan Capistrano, CA, United States

a r t i c l e

i n f o

Article history:
Received 6 October 2015
Received in revised form 14 January 2016
Accepted 21 January 2016
Available online 25 January 2016
Keywords:
C-peptide
Diabetes
Immunocapture
Insulin
LCMS/MS

a b s t r a c t
Background: Measurements of fasting levels of insulin and C-peptide are useful in documenting insulin resistance
and may help predict development of diabetes mellitus. However, the specic insulin and C-peptide levels associated with specic degrees of insulin resistance have not been dened, owing to marked variability among immunoassays and lack of standardization. Herein, we describe a multiplexed liquid chromatographytandem
mass spectrometry (LCMS/MS) assay for intact insulin and C-peptide.
Methods: Insulin and C-peptide were enriched from patient sera using monoclonal antibodies immobilized on
magnetic beads and processed on a robotic liquid handler. Eluted peptides were analyzed by LCMS/MS. Bovine
insulin and a stable isotopically-labeled (13C/15N) C-peptide were utilized as internal standards.
Results: The assay had an analytical measurement range of 3 to 320 IU/ml (18 to 1920 pmol/l) for insulin and
0.11 to 27.2 ng/ml (36 to 9006 pmol/l) for C-peptide. Intra- and inter-day assay variation was less than 11% for
both peptides. Of the 5 insulin analogs commonly prescribed to treat diabetes, only the recombinant drug insulin
lispro caused signicant interference for the determination of endogenous insulin. There were no observed interferences for C-peptide.
Conclusion: We developed and validated a high-throughput, quantitative, multiplexed LCMS/MS assay for intact
insulin and C-peptide.
2016 Elsevier B.V. All rights reserved.

1. Introduction
A number of studies have demonstrated the utility of fasting insulin
levels for predicting future diabetes mellitus. However, these studies
have often been retrospective and have employed a variety of immunoassay platforms for the measurement of insulin [15]. Specicity of assays, calibration procedures, specimen type, assay performance, and
conversion factors may all contribute to inter-assay platform variation
[6]. The lack of standardization of insulin assays has been clearly identied as restricting the capacity to compare measurements between different studies [7,8].
In contrast to immunoassays, mass spectrometry-based assays are
specic because they differentiate possible cross-reactive components
by molecular weight. Recently, we reported a mass spectrometrybased method for determining insulin B chain, liberated on chemical reduction of intact insulin, as a surrogate for intact insulin concentration
in serum [9]. Others have focused on the direct determination of intact
insulin by mass spectrometry in panels with exogenous insulin analog

Corresponding author at: Quest Diagnostics Nichols Institute, 33068 Ortega Hwy, San
Juan Capistrano, CA 92675, United States.
E-mail address: Nigel.J.Clarke@questdiagnostics.com (N.J. Clarke).
1
Current address: Cedars-Sinai Medical Center, 8700 Beverly Blvd, Los Angeles, CA
90048, United States.

http://dx.doi.org/10.1016/j.cca.2016.01.019
0009-8981/ 2016 Elsevier B.V. All rights reserved.

drugs, either by antibody-independent solid phase extraction (SPE)2

techniques [10] or with immunoenrichment [11,12]. Similarly,
proinsulin-derived C-peptide determined by immunoassay lacks standardization between assay platforms, and mass spectrometry-based approaches have been developed [13,14]. In addition to its use in diabetes,
an assay that simultaneously measures insulin and C-peptide may have
clinical utility in differential diagnosis of the cause of hypoglycemia: an
endogenous source or exogenous insulin administration [11].
2. Materials and methods
2.1. Chemicals, standards and quality controls
Human insulin World Health Organization (WHO) standard code
83/500 (National Institute for Biological Standards and Control
[NIBSC]) and C-peptide (Anaspec) were used as calibrators. Human insulin from US Pharmacopeia and C-peptide from Bachem were used as
controls. Bovine insulin (Sigma-Aldrich) and C-peptide labeled with
2
LC, liquid chromatography; AAA, amino acid analysis; CSH, charged surface hybrid;
ICMA, immunochemiluminometric assays; IS, internal standard; SRM, selected reaction
monitoring; NIBSC, National Institute for Biological Standards and Control; SPE, solid
phase extraction; SST, serum separator tubes; TEa, allowable total error; WHO, World
Health Organization.

S.W. Taylor et al. / Clinica Chimica Acta 455 (2016) 202208

leucine (U-13C6, 15N) at position 30 (custom synthesized by New

England Peptide) were used as internal standards (IS). IgG-conjugated
magnetic beads were custom made by GenWay Biotech Inc. using
600 mg of 5-m aldehyde magnetic beads (Bioclone Inc. and 10 mg of
either (a) anti-intact insulin B chain antibody (Fitzgerald, 10R-I134f)
or (b) anti-proinsulin C-peptide antibody (Millipore, 051109) for insulin and C-peptide, respectively. Cleanascite lipid removal reagent and
clarication [15] was from Biotech Support Group LLC. Human stripped
serum SP1040 debrinated charcoal treated (not delipidated) was obtained from Golden West Biologicals Inc. Sodium chloride 5 mol/l and
1.5 mol/l Tris (trizma) base solution were from Sigma. Phosphatebuffered saline (PBS) 10 solution was from Fisher Scientic.
Stock solutions of human insulin WHO standard and Anaspec Cpeptide were prepared at nominally 2 mg/ml in 0.1% formic acid in
water, and aliquots were submitted for quantitative amino acid analysis
(AAA) in triplicate. For insulin, powders were weighed within
0.02 mg. For C-peptide, powders were provided by both manufactures at nominally 0.5 mg/vial. The insulin and C-peptide stocks were
subsequently diluted 1 in 10 in SP1040 stripped serum and stored frozen at 80 C in polypropylene tubes until the amino acid results
were returned. Based on the amino acid results, enough of the insulin
and C-peptide stocks were combined and further diluted in SP1040
stripped serum to a nal concentration of 16,000 IU/ml (96 nM) insulin
and 1360 ng/ml (450 nM) C-peptide. This stock of the combined peptides, designated Tube A, was stored frozen at 80 C in 200 l aliquots in 1.5 ml polypropylene tubes until the day of the assay.
Calibrators were prepared from Tube A mixed peptide stocks. Tube
A was thawed and serially diluted with SP1040 stripped serum to
320 IU/ml (1920 pmol/l) insulin; 27.2 ng/ml (9006 pmol/l) C-peptide
Tube B and 20 IU/ml (120 pmol/l) insulin; 1.7 ng/ml (563 pmol/l)
C-peptide Tube C. Each Tube A aliquot was only used once after
thawing to avoid freezethaw cycles. Tubes B and C were transferred
to the deck of a Hamilton Microlab Star robotic liquid handler along
with SP1040 stripped serum where the assay calibrators were prepared by automated parallel dilution of the Tube B and C stocks
into stripped serum delivered by 8 independent 1000 l pipetting
channels.
Controls of USP insulin and Bachem C-peptide were similarly prepared at 2 mg/ml in 0.1% formic acid, diluted in SP1040 stripped
serum and stored at 80 C until quantitative amino acid analysis results were returned. The insulin and C-peptide control stocks were
thawed, combined and diluted into SP1040 stripped serum to produce
the assay quality controls: QCH 173 IU/ml (1038 pmol/l) insulin;
7.7 ng/ml (2550 pmol/l) C-peptide, QCM 43 IU/ml (258 pmol/l) insulin; 2.07 ng/ml (685 pmol/l) C-peptide, QCL 13 IU/ml (78 pmol/l) insulin; 0.48 ng/ml (160 pmol/l) C-peptide. Each QC aliquot was only used
once after thawing to avoid freezethaw cycles.
Bovine insulin and heavy labeled C-peptide internal standards
were initially dissolved in 0.1% formic acid, combined, and diluted
in SP1040 stripped serum to approximately 7 nmol/l bovine insulin
and 11 nmol/l heavy C-peptide IS. The nal concentrations of IS
after addition to calibrators, controls, and samples were approximately 635 pmol/l bovine insulin and 1000 pmol/l heavy C-peptide
IS. No attempt was made to correct for peptide content in the internal standards.
2.2. Human subject approval and clinical sample collection
Serum samples were obtained from normal subjects following informed consent (Western IRB approval #1085473) and stored at
80 C until use. The use of anonymized discarded samples in these
studies was reviewed by the Western IRB and deemed exempt.
Unless otherwise stated, blood was collected into barrier-free
serum preparation tubes (red top) and allowed to clot. The resulting
serum was immediately processed and then stored at 80 C until
analysis.

203

2.3. Sample preparation

Patient serum and controls were thawed, vortexed and transferred
to the deck of the Hamilton Microlab Star robotic liquid handler. The
deck can accommodate polypropylene 12 75 mm or equivalent
tubes and has adapters to accommodate 2 ml skirted-bottom tubes.
Separate programs were written for each tube type. One hundred fty
microliters of each calibrator, control and patient sample were transferred to a 96 Deep Well Plate (Thermo Fisher Scientic) using the 8 independent 1000 l pipetting channels tted with 300 l conductive tips.
To ensure that the 150 l was accurately transferred, liquid level sensing
would require the operator to check any sample that had 200 l before
transferring the sample to the plate. After the wells of the plate were
lled, 15 l of internal standard were added to each well followed by
50 l of Cleanascite delipidation reagent previously mixed into a uniform suspension by a brief aspiration/dispense cycle within its reagent
reservoir. The plate was subsequently sealed with adhesive lm and
transferred to a shaker on the deck of the robot by the robotic arm (2
channels tted with core grip paddles) where the suspension was
vortexed for 30 min at 850 rpm.
While the mixing was in progress, the magnetic beads were prepared for immunocapture. Anti-intact insulin B chain and anti-Cpeptide magnetic beads were provided as a slurry of 60 mg/ml beads
(1 g/l IgG in 1 PBS 0.05% azide). It was critical the beads were prepared as a homogenous suspension before dilution. Aliquots of bead
stocks (150 l anti-intact insulin B chain and 100 l anti-C-peptide)
were transferred to separate 1.5 ml microfuge tubes on a magnetic separator, the liquid was discarded, replaced by 1 ml of 1 PBS and
vortexed to a uniform suspension. The diluted beads (0.15 g/l IgG
anti-intact insulin B chain and 0.1 g/l IgG anti-C-peptide) were combined in a tube on the magnetic separator, the liquid discarded and
the beads successively washed with 1 mol/l NaCl and 1 PBS with
vortexing. Finally, the beads were resuspended in 2 PBS in a 1.5 ml
microfuge tube containing a Stir Stix stirrer bar (V&P Scientic, Inc.)
and transferred to the deck of the Hamilton robot adjacent to a Rotary
Magnetic Tumble Stirrer (V&P Scientic, Inc.) which keeps the beads
in constant suspension. During the delipidation, the robot transferred
10 l of beads from the suspension to each well of a 96 deep well
immunocapture plate.
Upon completion of delipidation/vortex mixing, the sample plate
was manually removed and centrifuged at room temperature at
5000 rpm (4696 RCF) for 10 min using a Beckman Coulter Allegra 25R
centrifuge. The centrifuged plate was then transferred back to the
deck of the robot and the plate sealer removed. Using the CO-RE 96Probe Head, the robot transferred (a) 150 l 2 PBS, then (b) 160 l of
the supernatant from the centrifuged samples to the deep well plate
containing the magnetic beads. The plate was then sealed with adhesive
lm and transferred to the shaker where the suspension was mixed by
vortexing at 1150 rpm for 60 min and the peptides were
immunocaptured by the beads.
After the 1 h incubation, the seal was removed from the
immunocapture plate which was transferred to a Magnetic Bead Separation Block for 96 deep well microplates (V&P Scientic, Inc.). Serum
supernatant removal and bead washing was facilitated by the CO-RE
96-Probe head and alternatively transferring the plate between the
magnetic separation block and shaker using the robotic arm. Beads
were washed for 15 s each with 150 l 1 mol/l NaCl (twice), 150 l 1
PBS (twice) and 100 l water (once). Finally, 30% ACN, 0.1% formic
acid (75 l) was added to the beads in each well and the plate sealed
and beads extracted on the shaker for 10 min at 1150 rpm. During,
the extraction, 15 l of Trizma base was added to the wells of the elution
plate. Trizma base enhances the stability of the peptides in the elution
plate while they await injection into the LC system from an autosampler
refrigerated at 4 C.
After the peptide extraction was complete, the seal was removed
from the bead plate and the extracted peptides transferred from the

204

S.W. Taylor et al. / Clinica Chimica Acta 455 (2016) 202208

bead plate to the elution plate containing Trizma base. The elution plate
was sealed with a pre-slit silicone 96 well mat and placed back on the
deck for a 15 s mix on the shaker at 1150 rpm prior to transfer to the
autosampler of the LC.
2.4. LCMS/MS
Analytical separation of intact insulin and C-peptide from remaining
matrix components prior to MS was achieved with a TurboFlow Aria
TLX-4 (Thermo-Fisher), a fully automated on-line two dimensional liquid chromatography system. The 4 columns in this LC system are operated in staggered parallel, facilitating high throughput. The sample
clean-up and enrichment was performed by on-line SPE using an
Oasis HLB 2.1 20 mm, 25 m cartridge column (Waters). Chromatographic resolution was accomplished using a 2.1 50 mm, 2.5 m,
XSelect CSH 130 C18 (Waters) heated to 60 C. Both SPE and analytical
columns use the same solvent A (water, 1% formic acid) and B
(acetonitrile).
Immunocaptured peptides in acetonitrile/water/trizma base (90 l)
were loaded onto the HLB column with 97% A solvent, 3% B solvent at
3 ml/min for 1 min. Peptides were then back-ushed off the extraction
cartridge with a plug of 35% solvent B from a prelled sample loop and
refocused onto the analytical column by means of a T valve delivering
97% solvent A at 0.25 ml/min to both the loading and eluting columns,
respectively for 1.5 min. Peptides were subsequently resolved at
0.5 ml/min using linear gradients from 3% to 20% solvent B for 30 s,
20% to 25% solvent B for 160 s and 25 to 28% solvent B over 70 s.
Data were acquired on the 6490 Triple Quadrupole Mass Spectrometer using SRM in positive ion mode under the following conditions: gas
temperature, 250 C; gas ow, 12 l/min; nebulizer, 31 psi; sheath gas
temperature, 350 C; sheath gas ow, 12; capillary 3500 V, nozzle voltage 500 V. MS1 resolution was set at wide, MS2 resolution was set at
unit, fragmentor was set at 380 and cell accelerator voltage was set to
1. In this assay, we consider the retention time and the ratio of the
mass transitions to support identication of intact insulin (tolerance of
25%) or C-peptide (tolerance of 20%).
For quantitative analysis of intact human insulin, the MH+6
6 ion (m/z
968.7 + 0.2) was used as the precursor ion (Supplementary Fig. 1a).
Transitions to a tyrosine immonium ion (136.0 + 0.2), and fragments
near the C-terminus of the B-chain ProlineLysine or y3y1
(226.1 + 0.2) and y3 (345.2 + 0.2) were monitored. The same frag6
ion (m/z 956.8 + 0.2)
ments from intact bovine insulin the MH+
6
were monitored however the mass of y3 (315.2 + 0.2) differs because
of the substitution of threonine for alanine at the C-terminus. Collision
energies were set at 40 for both human and bovine insulin but dwell
times were 120 ms for human insulin transitions compared with
30 ms for bovine insulin transitions. For human insulin the m/z 226.
ion was set as the quantier. The relative responses of the qualiers
m/z 136 and m/z 345 were 2.30 and 0.50 times the m/z 226 peak intensity, respectively. For bovine insulin the m/z 136 ion was set as the quantier. The relative responses of the qualiers m/z 226 and m/z 315 were
0.34 and 0.3 times the m/z 136 peak intensity, respectively. The bovine
insulin internal standard elutes approximately 0.25 min before the
human insulin analyte (Supplementary Fig. 2a).
ion
For quantitative analysis of intact human C-peptide, the MH+3
3
(m/z 1007.7 + 0.2) was used as the precursor ion (Supplementary Fig.
1b). Transitions to y5 (533.3 + 0.2), y6 (646.4 + 0.2), and y9
(927.5 + 0.2) were monitored. The analogous fragments from heavy laion (m/z 1009.2 + 0.2) i.e. y5 (540.3 + 0.2), y6
beled C-peptide MH+3
3
(653.4 + 0.2), and y9 (934.5 + 0.2) were monitored. Collision energies
were set at 27 for both C-peptide and the internal standard but dwell
times were 100 ms for C-peptide transitions compared with 25 ms for
the internal standard. For C-peptide and its internal standard the sum
of the transitions was used to quantitate the peptides whereas the ratios
of the intensities of the transitions were used to identify them. For Cpeptide the relative responses of the m/z 533 and 646 ions were 0.35

and 0.31 times the m/z 927 peak intensity, respectively. For the Cpeptide internal standard the relative responses of the m/z 540 and
653 ions were 0.46 and 0.34 times the m/z 934 peak intensity, respectively. For C-peptide, the internal standard and analyte co-elute (Supplementary Fig. 2b).
For both insulin and C-peptide, the ratio of the peak area of the analyte to the internal standard was used to calculate the concentrations
from the standard curve. A weighted quadratic model (1/x) was used
for generation of the standard curves ignoring the origin. Results were
reported as the concentration of insulin in IU/ml or pmol/l and Cpeptide in ng/ml or pmol/l. Software tools: MassHunter Workstation
Software, Acquisition Version B.04.01 and Quantitative Analysis Version
B.05.00 SP01 (Agilent), Aria 1.6 Software (Thermo Fisher Scientic), and
Hamilton Run Control Version 4.3.7270.

3. Method validation
3.1. Assay performance
Analytical performance specications, such as analytical measurement range, sensitivity, carryover, precision and accuracy, stability, interference, matrix effects, and recovery, were determined as part of
the validation, which meets internal policy in compliance with the Clinical Laboratory Improvement Amendment of 1988 (CLIA '88) regulation
(part 493.1253). The allowable total error (TEa) was set at 25% (proportional) or 5 IU/ml (constant) for insulin and 30% (proportional) or
0.2 ng/ml (constant) for C-peptide. We utilized quality control procedures according to Westgard Rules within the assay to allow determination of passing versus failing runs as well as to identify biases within the
data due to analytical issues. These data were then used to set the 2 and
3 SD limits for QC acceptability. During sample analyses, all QC levels
were tested at the beginning of each assay, with multiple replicates of
the QCs interspersed throughout the assay, thereby bracketing patient
samples.

3.2. Sample type, stability and interference

Four types of collection tubes (plain red top serum, serum separator
tubes [SST], EDTA plasma, and sodium heparin plasma) were assessed
using samples drawn from 17 individuals and analyzed with the LC
MS/MS assay. Sample stability over time was evaluated by spiking insulin and C-peptide into 6 patient samples, to achieve nal concentrations
of 20 to 110 IU/ml insulin (120 to 660 pmol/l) and 2.2 to 10.4 ng/ml Cpeptide (728 to 3443 pmol/l). Storage conditions were monitored and
maintained in the following temperature ranges: ultralow frozen
(60.0 to 90.0 C), frozen (10.0 to 30.0 C), refrigerated (2.0 to
8.0 C), and room temperature (18.0 to 26.0 C). Freezethaw stability
was evaluated using the same 6 individual patient serum samples.
Five aliquots were subjected to repeated freeze thaw cycles, which involved freezing samples in the ultralow freezer, thawing to room temperature, and then re-freezing at ultralow temperatures. For all the
stability studies, the rst set of the aliquots corresponded to the baseline
for that patient sample.
Interference was assessed by spiking the controls with low, medium,
and high levels of Intralipid (40, 100 and 200 mg/dl, appearance lightly cloudy to milky), bilirubin (2, 4 and 8 mg/dl, light to bright orange), or
red blood cell hemolysate (150, 375, and 1500 mg/dl, pink [mild] to
cherry [gross]). In addition, we examined the effects of the following
common insulin analogs on insulin and C-peptide determination: insulin lispro, insulin aspart, insulin glulisine, insulin glargine, and insulin
detemir. A stripped serum blank and a stripped serum containing
20 IU/ml insulin and 1.70 ng/ml C-peptide were separately spiked either with 75 or 150 IU/ml of each of the 3 analogs, respectively. Insulin
and C-peptide levels were then measured in each of these mixtures.

S.W. Taylor et al. / Clinica Chimica Acta 455 (2016) 202208

3.3. Method comparison

The new method was compared to (a) an FDA-approved commercial
(Beckman Access ICMA) platform for measurement of insulin in patients; and (b) ADVIA Centaur C-peptide assay. These patient samples
were de-identied discards that were submitted previously for routine
clinical testing. Measured concentrations covered the expected normal
ranges and beyond, from low to very high insulin and C-peptide levels.
Correlation between methods was evaluated through the use of Deming
regression.
4. Results
4.1. Assay performance
The assay demonstrated an analytical measurement range of 3 IU/ml
to 320 IU/ml (18 pmol/l to 1920 pmol/l) for insulin and 0.11 ng/ml to
27.2 ng/ml (36 pmol/l to 9006 pmol/l) for C-peptide. The limit of quantitation (LOQ) was determined by assaying 5 different samples at concentrations close to the expected LOQ (1.25, 2.5, 5, 10, and 20 IU/ml for
insulin and 0.11, 0.22, 0.44, 0.85, and 0.17 ng/ml for C-peptide), and
then evaluating the intra-assay reproducibility in 7 runs and inter-assay
reproducibility in another 8 runs. LOQs of 2.5 IU/ml (rounded up to
3 IU/ml [18 pmol/l] because the assay will report in whole numbers)
for insulin and 0.11 ng/ml (36 pmol/l) for C-peptide were the lowest concentrations that yielded acceptable performance (ie, 95% condence interval for the coefcient of variation [CV] remained below 20%). A
stripped serum blank was measured 20 times, and the resulting area ratios were back-calculated to establish a limit of detection (4 SD from the
zero concentration) of 1.5 IU/ml for insulin and 0.10 ng/ml for Cpeptide. The limit of blank (2 SD from the zero concentration) was
0.9 IU/ml for insulin and 0.06 ng/ml for C-peptide.
Evidence of carryover was evaluated by running matrix blanks immediately after high standards on each of the 4 channels on the
TurboFlow Aria LC system. The readings of blanks run immediately
after the insulin (320 IU/ml) and C-peptide (27.2 ng/ml) high calibrators showed carryover ranging from 0.4% to 0.7% for insulin and 0.3% to
0.4% for C-peptide, corresponding to the LOQ or less. Consequently,
blanks should be run after the high calibrator to avoid carryover. Patient
samples should be rerun if they immediately follow an LC column
exposed to a patient sample with 320 IU/ml insulin or 27.2 ng/ml
C-peptide or greater. No signicant carry over was observed at
lower calibration levels.
Intra- and inter-assay precision and accuracy were established using
5 replicates of the controls at each level over 5 days. Intra-assay CVs
ranged from 4% to 9% for insulin and 4% to 8% for C-peptide. Interassay CVs ranged from 7% to 11% for insulin and 7% to 10% for Cpeptide (Table 1).
The accuracy of the assay was established by comparing the concentrations of insulin and C-peptide in the quality controls (QCs) determined using the calibration curve with the calculated concentrations
of insulin and C-peptide spiked into stripped serum. Concentrations
for both calibrators and QCs were calculated by weighing the amount
Table 1
Performance of the LCMS/MS assay for insulin and C-peptide (5 replicates over 5 days).
Insulin (IU/ml)

Target
Overall mean
Overall SD
Overall CV
Overall accuracy
Diff (target-mean)
TEa/4

C-peptide (ng/ml)

QC low

QC
medium

QC high

QC low

QC
medium

QC high

13.3
14.1
1.5
11%
106%
0.8
1.25

43.1
43.3
3.6
8%
100%
0.2
2.69

172.7
173.0
12.7
7%
100%
0.3
10.79

0.48
0.51
0.05
10%
106%
0.03
0.05

2.07
1.96
0.14
7%
95%
0.11
0.16

7.70
7.77
0.35
5%
101%
0.07
0.58

205

of peptide and correcting for peptide content by quantitative AAA. All

QCs passed with accuracy ranging from 100% to 106% for insulin (TEa/
4 = 6.3% or 1.25 IU/ml), and 95% to 106% for C-peptide (TEa/
4 = 7.5% or 0.05 ng/ml).
Recovery studies were performed by spiking known levels of insulin
(10, 20, or 40 IU/ml) and C-peptide (1.02, 1.70 or 3.40 ng/ml) into patient sera with known low levels of both. Analysis was corrected for
background levels of insulin and C-peptide. Mean recoveries ranged
from 96% to 106% for insulin and 91% to 104% for C-peptide. Recoveries
of both peptides from patient sera at the three levels were within 10%
(TEa/3) (Supplementary Tables 1a and b).
Patient sera with high endogenous levels of insulin and C-peptide
were diluted with stripped serum, and the recoveries were measured
as a percentage of the value, which was calculated by dividing the
neat levels by the dilution factor. Patient 1 serum had an insulin level
of 69.3 IU/ml and a C-peptide level of 6.37 ng/ml, whereas patient 2
serum had an insulin level of 753.5 IU/ml (by extrapolation) and a Cpeptide level of 20.84 ng/ml. Recoveries were within 10% (TEa/3)
for both insulin and C-peptide for Patients #1 and 2 (Supplementary
Tables 2a and b).
4.2. Sample type, interference, and stability
One type of serum collection tube and two types of plasma collection
tubes were compared with the plain red top serum tubes used for the
validation. Insulin and C-peptide levels determined in blood collected
in SST gave equivalent results to the red top serum tubes for both peptides. For insulin, neither EDTA nor sodium heparin plasma were acceptable tube types. For C-peptide, both plasma tube types were acceptable
if the samples were manually transferred from the sample tubes to the
sample plate. However, because of the viscosity of the plasma, an unacceptable number of samples were inefciently transferred by the robot.
Consequently, the assay is restricted to serum tubes.
Storage and freeze/thaw stability of C-peptide and insulin were evaluated in 6 patient serum pools. Both insulin and C-peptide were stable
for at least 28 days frozen, at least 39 weeks at ultralow temperatures,
and at least 5 freeze/thaw cycles. At room temperature, both peptides
degraded after 24 h of storage. However, insulin and C-peptide differed
in stability when refrigerated: insulin was stable for at least 28 days in
refrigerated serum, whereas C-peptide degraded at 7 days.
The determination of insulin and C-peptide levels was not affected
by lipemic or icteric interference at any of the concentrations tested
(Supplementary Tables 3a and b). However, for insulin, all levels of hemolysis were found to be unacceptable, with progressive loss of the insulin peak until it was undetectable. For C-peptide, mild to moderate
hemolysis was acceptable, but gross hemolysis was unacceptable.
Several common insulin analogs were tested for possible interference. Insulin glargine was not immunocaptured by the antibodies
used, and insulin detemir eluted almost 2 min after insulin under the
gradient conditions employed. This left the 3 analogs (insulin aspart, insulin glulisine, and insulin lispro) as possible sources of interference. All
3 were immunocaptured and eluted within the window used to detect
human and bovine insulin IS. Consequently, all 3 were tested for interference. Neither insulin aspart nor insulin glulisine caused interference
at any level. Insulin lispro caused signicant interference which became
worse with increasing concentration (Supplementary Table 3c). The
presence of insulin lispro was obvious by its prole, causing a characteristic qualier failure (Supplementary Fig. 2e). As expected, no insulin
analogs interfered with C-peptide determination (Supplementary
Table 3c).
4.3. Method comparison
Previously tested serum samples were collected from the FDAapproved Beckman Access ICMA platform for insulin and the ADVIA
Centaur C-peptide assay. All samples were tested with the current

206

S.W. Taylor et al. / Clinica Chimica Acta 455 (2016) 202208

examined. The discrepancy is consistent with the behavior of our calibrators in the immunoassay (Fig. 3).
4.4. Reference interval determination
A reference interval was established with 33 subjects (20 women, 13
men 19 to 58 years of age) who were apparently healthy, ambulatory,
community-dwelling, non-medicated individuals with no history of diabetes; each individual had normal fasting glucose (b 100 mg/dl),
HbA1c b 5.7, and BMI b 26.0. For both insulin and C-peptide, the distributions were non-Gaussian by any transformation means and had no
outliers. Therefore, the observed reference interval was 16 IU/ml
(96 pmol/l) for intact insulin and 0.68 to 2.16 ng/ml (225 to
715 pmol/ml) for C-peptide.
5. Discussion

Fig. 1. Method comparison (Deming regression) for patient samples: Current assay for
insulin versus FDA-approved ICMA platform for insulin (n = 117). Correlation rs = 0.97
(Spearman); tau = 0.87 (Kendall).

assay, and those with sufcient sample volume were also tested with
the other assays for determination of insulin and/or C-peptide levels.
Insulin test results from 117 samples obtained using Beckman ICMA
platform were compared with results from the current assay and
yielded a good correlation (Deming regression, 1.04x 0.64, Fig. 1).
In contrast, the C-peptide test results from 121 samples obtained the
ADVIA Centaur C-peptide assay exhibited a strong negative bias
(Deming regression of 0.70x + 0.18, Fig. 2) for the 121 samples

Fig. 2. Method comparison (Deming regression) for patient samples: Current assay for Cpeptide versus Centaur ICMA platform for C-peptide (n = 121). Correlation rs = 0.98
(Spearman); tau = 0.90 (Kendall).

We report the validation of a multiplexed method for the simultaneous determination of intact insulin and C-peptide by immunocapture
LCMS/MS techniques. The method is almost entirely automated and
high throughput, and can distinguish native insulin from all tested exogenous analogs with only insulin lispro causing interference. This contrasts with the Beckman Access ICMA assay, in which insulin aspart,
glargine, in addition to lispro, are indistinguishable from human insulin
having cross-reactivities of 90% to 110% [16]. Sensitivity is equivalent to
those of current immunoassays for both peptides. Run as a multiplex
test, our assay has some distinct advantages. For example, hormone
levels useful for the differential diagnosis of hyperinsulinemic hypoglycemia related to an insulinoma (high insulin and C-peptide) versus surreptitiously injected insulin (high insulin and normal/low C-peptide)
are efciently provided to the clinician in a single assay [17].
Key components of the methodology involve the use of a
delipidation reagent to enhance immunocapture. Early trials on
delipidated stripped serum showed much higher IS responses compared with patient sera indicating that the endogenous lipid had significant matrix effects. We subsequently employed Cleanascite reagent
[15] to remove lipid from patient samples and reverted to a less processed form of stripped serum, which still contained lipid, for a better
matrix match. The result was greatly enhanced recoveries and tighter

Fig. 3. Method comparison (Deming regression) for current assay C-peptide calibrators in
the Centaur ICMA platform.

S.W. Taylor et al. / Clinica Chimica Acta 455 (2016) 202208

CVs for the IS throughout the plate. Chromatography was also critical,
the CSH media [18], heated columns and relatively at gradients resolved insulin and C-peptide from residual non-specically bound material. Chromatographically, we found C-peptide to be a better
behaved analyte than insulin, which has known issues because of its
physiochemical properties including nonspecic adsorptive loss and
broad peaks on certain chromatographic media [18]. In terms of trouble
shooting, a strong linear response for C-peptide rules out automation
failures when insulin exhibits a weak response in the same assay.
Reference intervals for commercially available kits are often
established for healthy populations that are insufciently described
[19]. We originally sampled 103 healthy volunteers to establish a reference interval but only 33 of the subjects met the criteria for normal
HbA1c and fasting glucose with a BMI b 26.0. Our reference interval
for fasting insulin (16 IU/ml) is consistent with that reported in the
literature using healthy volunteers with normal glucose tolerance and
similar BMIs (1.9 to 15 IU/ml) [19]. The C-peptide reference interval
was lower in our study (0.68 to 2.16 ng/ml) compared with immunoassay (0.78 ng/ml to 5.19 ng/ml) for a healthy, fasting, but otherwise undened, adult population [20]. In addition to differences in patient
characteristics, the disparity may reect the lack of standardization between methods, as has been reported in the literature for C-peptide [21]
and which is further discussed below.
An important feature of our work is the use of quantitative AAA to
establish the peptide content of both insulin and C-peptide. We found
excellent agreement between the values for the controls, which were
calculated by quantitative AAA and corrected for dilution into stripped
serum, and the values determined from using the calibration curves
(Table 1). This assessment of accuracy is powerful because quantitative
AAA is independent of the assay because materials used for the calibrators and controls are from different sources.
For insulin, we obtained excellent agreement between values we obtained for patient samples using our assay and those obtained in the
Beckman Access ICMA assay (Fig. 1), which is standardized to the original WHO International Reference Preparation 66/304. For C-peptide,
we observed excellent correlation but a signicant negative bias in the
values obtained for the patient samples compared with the ADVIA Centaur C-peptide assay, which is standardized against WHO C-peptide reference material 84/510 (Fig. 2). Consequently, in contrast to insulin, it is
unlikely that our C-peptide assay would standardize to the WHO reference material. Unfortunately, we cannot standardize our assay against
WHO insulin 66/304 or C-peptide 84/510, the materials are no longer
supplied by the NIBSC.
It is noteworthy that immunoassays consistently overestimate Cpeptide in patient samples compared with mass spectrometry-based
assays [13,14]. These data have been explained by assay specicity,
with the values from immunochemical assays being inated by the
cross-reactivity of partially processed or degraded forms of C-peptide
that share the detected epitope [14]. However, we observed similar negative bias in the values for our calibrators in the Centaur assay (Fig. 3),
suggesting that the discrepancy is from the calibrator responses rather
than assay specicity.
A possible explanation for calibrator bias is discordant peptide content. The WHO C-peptide reference reagent was provided by NIBSC as
10 g of human C-peptide lyophilized with bovine serum albumin
(BSA) and lactose in a glass ampoule. Although amino acid analysis
was performed by the NIBSC to establish peptide purity of the bulk material, a relationship with the amount of material weighed i.e. peptide
content was not [22]. The value of 10 g was assigned by immunoassay
results performed by independent laboratories using their own preparations of C-peptide as secondary calibrators [22]. The concentrations
of C-peptide in our calibrators have been veried by quantitative AAA.
Thus, the negative bias for our assay compared to the Centaur assay
may reect higher peptide content in our calibrators (Fig. 3) and
lower values for patient samples measured against our calibrators
(Fig. 2). It is also signicant that commercial immunoassays, all

207

traceable to the WHO material, have exhibited variable bias when compared with each other and mass spectrometry, despite excellent correlation between methods [13].
6. Conclusion
We developed a multiplexed assay for insulin and C-peptide and validated it in accordance with CLIA '88 guidelines. Results are independently traceable to SI units to enable standardization with assays
developed using similarly traceable reference materials [23]. This standardization is critical to enable future comparison between studies
and establish clinically important ranges for insulin and C-peptide levels
in healthy, insulin resistant, prediabetic and diabetic individuals as well
as pediatric and elderly populations.
Acknowledgments
The authors thank Amanda Albi and Robert Barham for programming and assistance with the automation. We also thank Anh Nguyen
for technical support and Dr. Michael Cauleld and Dr. Andrew Hellman
for their thoughtful comments on the manuscript.
Appendix A. Supplementary data
Supplementary data to this article can be found online at http://dx.
doi.org/10.1016/j.cca.2016.01.019.
References
[1] H. Wang, N.M. Shara, D. Calhoun, J.G. Umans, E.T. Lee, B.V. Howard, Incidence rates
and predictors of diabetes in those with prediabetes: the Strong Heart Study, Diabetes Metab. Res. Rev. 26 (2010) 378385.
[2] J.L. Johnson, D.S. Duick, M.A. Chui, S.A. Aldasouqi, Identifying prediabetes using
fasting insulin levels, Endocr. Pract. 16 (2010) 4752.
[3] K.C. Sung, M.H. Seo, E.J. Rhee, A.M. Wilson, Elevated fasting insulin predicts the future incidence of metabolic syndrome: a 5-year follow-up study, Cardiovasc.
Diabetol. 10 (2011) 108.
[4] R. Dankner, A. Chetrit, M.H. Shanik, I. Raz, J. Roth, Basal-state hyperinsulinemia in
healthy normoglycemic adults is predictive of type 2 diabetes over a 24-year
follow-up: a preliminary report, Diabetes Care 32 (2009) 14641466.
[5] R. Dankner, A. Chetrit, M.H. Shanik, I. Raz, J. Roth, Basal state hyperinsulinemia in
healthy normoglycemic adults heralds dysglycemia after more than two decades
of follow up, Diabetes Metab. Res. Rev. 28 (2012) 618624.
[6] S.E. Manley, I.M. Stratton, P.M. Clark, S.D. Luzio, Comparison of 11 human insulin assays: implications for clinical investigation and research, Clin. Chem. 53 (2007)
922932.
[7] S. Marcovina, R.R. Bowsher, W.G. Miller, M. Staten, G. Myers, S.P. Caudill, et al., Standardization of insulin immunoassays: report of the American Diabetes Association
workgroup, Clin. Chem. 53 (2007) 711716.
[8] I.M. Stratton, A.I. Adler, H.A. Neil, D.R. Matthews, S.E. Manley, C.A. Cull, et al., Association of glycaemia with macrovascular and microvascular complications of type 2
diabetes (UKPDS 35): prospective observational study, BMJ 321 (2000) 405412.
[9] Z. Chen, M.P. Cauleld, M.J. McPhaul, R.E. Reitz, S.W. Taylor, N.J. Clarke, Quantitative
insulin analysis using liquid chromatographytandem mass spectrometry in a highthroughput clinical laboratory, Clin. Chem. 59 (2013) 13491356.
[10] E.E. Chambers, K.J. Fountain, N. Smith, L. Ashraf, J. Karalliedde, D. Cowan, et al., Multidimensional LCMS/MS enables simultaneous quantication of intact human insulin and ve recombinant analogs in human plasma, Anal. Chem. 86 (2014) 694702.
[11] S. Peterman, E.E. Niederkoer, D.A. Phillips, B. Krastins, U.A. Kiernan, K.A. Tubbs,
et al., An automated, high-throughput method for targeted quantication of intact
insulin and its therapeutic analogs in human serum or plasma coupling mass spectrometric immunoassay with high resolution and accurate mass detection (MSIAHR/AM), Proteomics 14 (2014) 14451456.
[12] A. Thomas, W. Schanzer, M. Thevis, Determination of human insulin and its analogues in human blood using liquid chromatography coupled to ion mobility mass
spectrometry (LC-IM-MS), Drug Test Anal. 6 (2014) 11251132.
[13] D.R. Cabaleiro, D. Stockl, J.M. Kaufman, T. Fiers, L.M. Thienpont, Feasibility of standardization of serum C-peptide immunoassays with isotope-dilution liquid chromatographytandem mass spectrometry, Clin. Chem. 52 (2006) 11931196.
[14] T. Kinumi, R. Mizuno, A. Takatsu, Quantication of serum C-peptide by isotopedilution liquid chromatographytandem mass spectrometry: enhanced detection
using chemical modication and immunoafnity purication, J. Chromatogr. B
Analyt. Technol. Biomed. Life Sci. 953954 (2014) 138142.
[15] A.R. Castro, W.E. Morrill, V. Pope, Lipid removal from human serum samples, Clin.
Diagn. Lab. Immunol. 7 (2000) 197199.

208

S.W. Taylor et al. / Clinica Chimica Acta 455 (2016) 202208

[16] B. Heurtault, N. Reix, N. Meyer, F. Gasser, M.J. Wendling, C. Ratomponirina, et al., Extensive study of human insulin immunoassays: promises and pitfalls for insulin analogue detection and quantication, Clin. Chem. Lab. Med. 52 (2014) 355362.
[17] P.E. Cryer, L. Axelrod, A.B. Grossman, S.R. Heller, V.M. Montori, E.R. Seaquist, et al.,
Evaluation and management of adult hypoglycemic disorders: an Endocrine Society
Clinical Practice Guideline, J. Clin. Endocrinol. Metab. 94 (2009) 709728.
[18] E.E. Chambers, C. Legido-Quigley, N. Smith, K.J. Fountain, Development of a fast
method for direct analysis of intact synthetic insulins in human plasma: the large
peptide challenge, Bioanalysis 5 (2013) 6581.
[19] R. Haeckel, D. Colic, W. Wosniok, Reference interval of serum insulin concentrations,
Clin. Chem. Lab. Med. 42 (2004) 10691070.

[20] J. Schultess, C. van Duren, M. Martens, M. Costa, T. Llop, T. Marti, et al., Diagnostic
performance of the ARCHITECT C-peptide immunoassay, Clin. Chem. Lab. Med. 47
(2009) 834841.
[21] A.G. Jones, A.T. Hattersley, The clinical utility of C-peptide measurement in the care
of patients with diabetes, Diabet. Med. 30 (2013) 803817.
[22] National Institute for Biological Standards and Control (NIBSC). C-peptide of human
insulin, International Reference Reagent NIBSC code: 84/510. Instructions for use.
Version 5.0. http://www.nibsc.org/documents/ifu/84-510.pdf. Updated December
2013. Accessed December 2015.
[23] K. Van Uytfanghe, L.M. Thienpont, Standardization of insulin and C-peptide a status report, Mdecine Nuclaire 34 (2010) 566570.