2. ligand. Unbound conjugate is washed off, and luminogenic
substrate is added. The resulting chemiluminescent signal is
proportional to the gastrin concentration. The assay is designed
to detect a range of bioactive gastrins.
Evaluation of specimen type suitability and sample stability
We compared gastrin measurements in 5 samples (20–279
ng/L; 10–133 pmol/L), each drawn into 4 different tube types:
serum, serum separator (SST), EDTA, and heparin plastic tubes
(Becton Dickinson, Franlin Lakes, NJ, USA). Serum or plasma
was separated and aliquoted within 1 h and assayed immediately
and after 1-day, 3-day, and 7-day storage at room temperature
and at 4°C, respectively, as well as after 1, 2, and 3 freeze–thaws.
Additional room temperature studies with hourly measurements
over 8 h were performed on a single serum specimen.
For specimen type comparisons, serum tubes served as the
reference. To be considered as a suitable specimen type, all
samples drawn into SST, EDTA, or heparin tubes had to yield
gastrin concentrations within ±20% of the concentrations
measured in the respective corresponding serum tubes.
For the stability studies, the gastrin concentrations measured
during the immediate analysis of each sample served as the
reference. For each sample type, storage condition, and storage
time point, satisfactory sample stability was defined as a
deviation of gastrin concentrations on subsequent measure-
ments of b20% compared to the corresponding original
immediate measurements.
Intra-assay and inter-assay imprecision
We assessed intra-assay and inter-assay imprecision at 5
levels between 20 ng/L (10 pmol/L) and 824 ng/L (393 pmol/
L), using 20–60 samples per level.
Spike and sample-mixing recovery
We spiked 3 patient samples, with previously determined
gastrin concentrations, with gastrin 1–17 (Peninsula Laborato-
ries, Belmont, CA, USA) at concentrations of 10 ng/L (5 pmol/
L), 100 ng/L (48 pmol/L), and 1 μg/L (478 pmol/L).
We also assessed recovery across 129–554 ng/L (62–264
pmol/L) by mixing of high (N = 3) and low patient samples
(N = 3).
Recovery on dilution
We diluted three patient samples (301 ng/L [144 pmol/L],
953 ng/L [455 pmol/L], 1.4 μg/L [674 pmol/L]) 1:2, 1:4, 1:8,
and 1:16 in multidiluent 2 (MD2) and gastrin diluent (LGAZ).
The measured values were expressed as percentages of the
calculated expected values.
Cross-reactivity
We spiked samples across four orders of magnitude of
concentration (all peptides N90% purity) with gastrin 1–34
(Peninsula Laboratories), sulfated gastrin 1–17 (Bachem
Laborotories, Torrance, CA, USA) (both desirable crossreactiv-
ities), gastrin 1–14, pentagastrin (both Bachem), Cholecysto-
kinin (CCK) 1–33, and CCK 26–33 (both Peninsula) (all
undesirable crossreactivities). Cross-reactivity was expressed as
the measured apparent increase in serum gastrin level as a
percentage of the spiked cross-reactant concentration.
Linearity of assay signal response and hook assessment
We spiked gastrin 1–17 (Peninsula) into patient samples at
levels of 14.3 ng/L (7 pmol/L) to 143 μg/L (68 nmol/L) in 10-
fold steps. Raw relative light units (RLU; background-
normalized light output) readings were used as the instrument
does not report concentrations N1.0 μg/L (N478 pmol/L).
Carryover
We spiked 143 μg/L (68 nmol/L) gastrin 1–17 into a patient
sample and placed it in front of 2 consecutive blank samples.
Susceptibility to heterophile antibody interference
We tested susceptibility to heterophile antibody interference
as described previously by running 203 unselected clinical
samples before and after treatment in heterophile blocking tubes
(HBT; Scantibodies Laboratory, Santee, CA, USA) [4].
Fasting reference intervals
We generated fasting reference intervals using 128 healthy
volunteers (68 female, 60 male), 22–81 years old: 22 from
Täby, Sweden and 106 from Rochester, MN, USA. Subjects on
gastrin-elevating drugs were excluded. The 97.5th percentile
was used as the upper boundary of the reference interval.
Method comparison
We compared the Immulite 2000 gastrin assay with an in-house
Mayo RIA, using 46 consecutive patient samples with gastrin
levels of 23–790 ng/L (11–377 pmol/L), and a commercial RIA
(Euria-Gastrin, Euro-Diagnostica, Malmö, Sweden), using 88
consecutive patient samples with serum gastrin levels of 23–1048
ng/L (11–500 pmol/L). Data were analyzed by Passing–Bablock
regression and Bland–Altman plots.
We also classified gastrin measurements from 28 Mayo
patients with known diagnoses as being either consistent
(N = 13) or not (N = 15) with elevated serum gastrin levels.
Diagnoses considered consistent were: gastrinoma (N = 5),
current treatment with proton pump inhibitors (N = 4), atrophic
gastritis or pernicious anemia (N = 3), and severe H. pylori
gastritis (N = 1). Inconsistent diagnosis included: irritable bowel
and non-specific abdominal pain (N = 3), non-endocrine
diarrhea or malabsorption (N = 3), resolving acute gastritis
(N = 2), orthostatic hypotension (N = 1), past pancreatitis
(N = 1), chronic nausea without definite pathology (N = 1),
pituitary stalk cyst (N = 1), suspected mast cell disease (N = 1),
388 M.L. Eastvold et al. / Clinical Biochemistry 39 (2006) 387–390
3. islet cell tumor (gastrin negative on histology; N = 1), and
hyperkalemia of unknown cause (N = 1). The clinical
performance of the Mayo RIA and the Immulite 2000 gastrin
assay was compared in this dataset by receiver operating curve
(ROC) analysis.
Results
Gastrin measurements from samples collected in SST,
EDTA, and heparin tubes deviated from corresponding serum
tube results by −15.4% to +11.5% (average −4.3%). At room
temperature, aliquots from all primary tube types showed an
average degradation of 27% (range 23–30%), 36% (range
34–40%), and 44% (range 40–46%) after 1 day, 3 days, and
7 days, respectively. The hourly measurements at room
temperature showed a linear 2.8%/h degradation of gastrin
(R = 0.93). Refrigerated serum and plasma aliquots were
stable for 3 days. Frozen aliquots were stable through 3
freeze–thaws.
Intra-assay and inter-assay coefficients of variation were
b8% across the tested range, mostly b5%. Mean recovery for
spiked gastrin 1–17 was 98% (range: 88.5%–106%), compa-
rable to the results of the mixing experiments (average 104%;
range: 98–108%). Mean recovery on dilution was 108% (range:
93–135%) with LGAZ and 87% (range: 62–131%) with MD2.
Cross-reactivity with gastrin 1–34 ranged from 11.1 to 13.6%,
and, for sulfated gastrin 1–17, cross-reactivity ranged from 153
to 213%. There was no detectable cross-reactivity with the other
peptides. The raw RLU response to spiked gastrin 1–17 was
linear between 10 ng/L (5 pmol/L) and 10 μg/L (4780 pmol/L),
as judged by b10% deviation from predicted results based on
extrapolation of the calibration curve. Between 10 μg/L (4780
pmol/L) and 100 μg/L (48 nmol/L), the response curve
flattened. A hook was observed with spiked concentrations of
Fig. 1. Method comparison of two gastrin RIAs with the Immulite gastrin assay. (A) Scatterplot (Passing–Bablock linear fit with 95% CI superimposed) of serum
gastrin concentrations in 46 patient samples measured by an in-house Mayo RIA (abscissa) plotted against the corresponding concentrations measured with the
Immulite 2000 assay (ordinate). (B) Bland–Altman difference plot of Mayo RIA and Immulite 2000 gastrin measurements (horizontal solid line: mean
difference = −15.17 ng/L; broken lines: 95% agreement CI). (C) Scatterplot (Passing–Bablock linear fit with 95% CI superimposed) of serum gastrin concentrations in
88 patient samples measured by Euria RIA (abscissa) plotted against the corresponding concentrations measured with the Immulite 2000 assay (ordinate). (D) Bland–
Altman difference plot of Euria RIA and Immulite 2000 gastrin measurements (horizontal solid line: mean difference = −26.02 ng/L; broken lines: 95% agreement CI).
389M.L. Eastvold et al. / Clinical Biochemistry 39 (2006) 387–390
4. 143 μg/L (68 nmol/L). No carryover was observed up to the
same concentration. Results before and after heterophile
blocking tube pretreatment correlated with a slope of 0.99,
intercept of 0.45, and an R of 0.99. No results were discrepant
by more than 3 SD difference percentages, which would have
been indicative of possible heterophile interference.
The fasting reference interval was determined to be b100
ng/L (b48 pmol/L). The distribution of gastrin concentrations
was skewed towards low values. Approximately 90% of
measurements were b60 ng/L (28 pmol/L). We found no age-
or gender-related differences. The values in the Swedish
population were comparable to those observed in the US
Minnesotan population.
The Immulite assay showed good agreement with the Mayo
RIA (slope = 0.90, CI: 0.75 to 1.01; intercept = 4.63, CI: −5.31
to 26.65; R = 0.93, and the Euria RIA (slope = 0.97, CI: 0.92 to
1.01; intercept = −17.48, CI: −28.81 to −14.39; R = 0.98) (Fig.
1). When results N200 ng/L (96 pmol/L; Mayo: N = 17, Euria:
N = 15) were analyzed separately, the slopes were 1.2 (CI: 0.83
to 1.63) and 0.97 (CI: 0.77 to 1.17), respectively.
The clinical performance of the Immulite assay in the
patients with known diagnoses was similar to that of the
Mayo RIA (Fig. 2). The areas under the ROC for the two
assays did not differ significantly (Immulite 2000: 0.892;
Mayo RIA: 0.946). The cut-off points for 100% sensitivity for
the Mayo and Immulite 2000 assays were 28 ng/L and 25 ng/
L, respectively, while 100% specificity was achieved at 200
ng/L (96 pmol/L) with the Mayo assay and at 319 ng/L (155
pmol/L) with the Immulite 2000 assay. Maximum combined
sensitivity and specificity were at 42 ng/L (20 pmol/L) for the
Mayo assay (sensitivity 92%, specificity 87%) and at 70 ng/L
(33 pmol/L) for the Immulite 2000 assay (sensitivity 92%,
specificity 80%). Using 100 ng/L (48 pmol/L) as a diagnostic
cut-off, which is the upper limit of the healthy population
reference range, the Immulite 2000 had a sensitivity of 73%
and a specificity of 87%. The corresponding figures for the
Mayo RIA at this level were 61% sensitivity and 93%
specificity.
Discussion
The Immulite 2000 automated gastrin assay is precise,
accurate, and fast. The analytical turn-around time for the
Immulite assay is just over 1 h, compared with N12 h for the
Mayo RIA (overnight incubation) and 3–4 h for the Euria-
Gastrin assay. There are no significant problems with
undesirable cross-reactivity, carryover, or analytical interfer-
ence by heterophile antibodies. The assay correlates well with
the two established RIAs and appears to offer comparable
diagnostic performance while improving dynamic range by
more than two-fold. The dynamic range could likely be
extended further by adding additional calibrators, given that
the raw signal response was linear well beyond the highest
calibrator and that hooking was only observed at extremely
high analyte levels. It might be useful for the manufacturer to
consider this option as dilution linearity was borderline with
LGAZ and sub-optimal with MD2.
The only significant bias, which was observed between the
Immulite 2000 assay and the two RIAs, was a small constant
bias (negative intercept) against the Euria assay, possibly due to
the use of different calibrators (Immulite 2000: UK-MRC
standard A NIBSC 66/138, porcine gastrin II; Euria RIA:
synthetic human gastrin 1–17; Mayo RIA: purified human
gastrin from Calbiochem). Since our study included only a
modest number of samples with gastrin concentrations N200 ng/
L (96 pmol/L), it is also possible that in the high range the
method agreement could be slightly worse. However, at these
elevated levels, this is not likely to result in a change in clinical
decision making [2].
Finally, given the demonstrated poor specimen stability of
gastrin, the faster analytical speed of the Immulite 2000 assay
may have a significant positive impact on analytical quality,
beyond the improved turn-around time. In particular, test
accuracy and result precision of serial monitoring of patients are
likely to improve.
References
[1] Dockray GJ, Varro A, Dimaline R, Wang T. The gastrins: their production
and biological activities. Annu Rev Physiol 2001;63:119–39.
[2] Berg CL, Wolfe MM. Zollinger–Ellison syndrome. Med Clin North Am
1991;75:903–21.
[3] Ward CJ. Modern approaches to the investigation of vitamin B12
deficiency. Clin Lab Med 2002;22:435–45.
[4] Preissner CM, Dodge LA, O'Kane DJ, Singh RJ, Grebe SKG. Prevalence of
heterophilic antibody interference in eight automated tumor marker
immunoassays. Clin Chem 2005;51:208–10.
Fig. 2. ROC plot of serum gastrin measurements obtained by Mayo RIA (solid
circles) and Immulite 2000 (open circles) in 13 patients with and 15 patients
without confirmed clinical conditions consistent with serum gastrin elevations.
The gray diagonal line denotes the line of no diagnostic value. Tests of
diagnostic value have ROCs above this line. Optimal ROC-based diagnostic cut-
offs (sensitivity and specificity N85%) in this patient population were between
40 and 50 ng/L for the Mayo RIA and between 70 and 80 ng/L for the Immulite
2000 assay.
390 M.L. Eastvold et al. / Clinical Biochemistry 39 (2006) 387–390