| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Technical Briefs |
1 Department of Clinical Laboratory, Kobe University Hospital, Kobe, Hyogo, Japan; 2 Department of Clinical Pathology and Immunology, Kobe University Graduate School of Medicine, Kobe, Hyogo, Japan;
aaddress correspondence to this author at: Department of Clinical Pathology and Immunology, Kobe University Graduate School of Medicine, 7-5-1 Kusunoki-cho, Chuo-ku, Kobe, Hyogo 650-0017, Japan; fax 81-78-382-6209, e-mail kumagais{at}kobe-u.ac.jp
Abstract
Background: Insulin measurement is used for the diagnosis of hypoglycemia and for insulin pharmacokinetic evaluations. We assessed the analytical and clinical performance of the ARCHITECT® insulin assay, a chemiluminescent immunoassay recently introduced for the ARCHITECT i2000 fully automated immunoassay analyzer (Abbott Laboratories). We also tested whether major insulin analogs cross-reacted with the immunoassay reagents.
Methods: We used Clinical and Laboratory Standards Institute protocols to assess the analytical performance of the ARCHITECT insulin assay and compared its accuracy with that of the E-test TOSOH II (IRI) from TOSOH Corporation. We used 3 recombinant insulin analogs (lispro, aspart, and glargine) to evaluate the cross-reactivity of insulin analogs with the ARCHITECT immunoassay reagent.
Results: The total CV for the ARCHITECT assay was <5%. Correlation between the ARCHITECT insulin assay and the E-test TOSOH II (IRI) was satisfactory in the measured range, but we detected a slope deviation between the assays. The ARCHITECT insulin assay showed low cross-reactivity to the insulin analog aspart, whereas it detected the other insulin analogs, lispro and glargine, in concentrations as high as the theoretical concentrations.
Conclusions: The ARCHITECT insulin assay showed favorable basic performance, including reproducibility, dilution linearity, detection limit, and effects of interfering substances. When interpreting results, clinicians and laboratory pathologists should be aware of the cross-reactivity of the ARCHITECT and other immunoassays to specific insulin analogs prescribed to diabetes patients.
Insulin assays have been widely used to provide diagnostic information for diabetes mellitus and rare hypoglycemic syndromes (1)(2)(3) and to quantify insulin for pharmacokinetic evaluations.
The ARCHITECT® insulin assay (Abbott Laboratories) is a 1-step chemiluminescent immunoassay that uses paramagnetic microparticles coated with anti-insulin monoclonal antibody and acridinium-labeled anti-insulin monoclonal antibody conjugate. Samples, microparticles, and conjugate are added to a reaction vessel to form a particle–insulin–conjugate sandwich. After incubation, washing removes materials not bound in the 1st step. Addition of the pretrigger reagent, which includes hydrogen peroxide, and the trigger reagent, which includes sodium hydroxide, leads to acridinium-produced chemiluminescence, measured as relative light units (RLUs). The RLUs are proportional to the insulin concentration in the sample. The assay is traceable to the WHO Insulin 1st International Reference Preparation (WHO IRP 66/304; National Institute for Biological Standards and Control). The calibrators includes 6 concentrations (0, 3, 10, 30, 100, and 300 mIU/L). These are measured in duplicate to establish the calibration curve calculated by a 4-parameter logistic curve method.
We assessed the analytical and clinical performance of the ARCHITECT insulin assay. In addition, because a recent study revealed that the insulin analogs used for patient treatment cross-reacted with reagents used for insulin measurement (4), we tested for cross-reactions between major insulin analogs and the ARCHITECT immunoassay reagents. Following the protocols of the Clinical and Laboratory Standards Institute for total imprecision, we used the ARCHITECT insulin assay to measure Bio-Rad® Immunoassay Plus Controls (Bio-Rad), which include control sera containing low, moderate, and high concentrations of insulin, in duplicate in 2 runs each day for 20 days. The within-run CVs were 1.1% to 
1.8%, and the total CV did not exceed 5% in the measurement range.
We performed a dilution linearity study on 2 serum samples with increased insulin values and on the WHO standard. The samples were diluted with calibrator A (0 mIU/L). Each sample was measured twice in a 10-fold serial dilution. The ARCHITECT insulin assay was confirmed to be linear up to 273.5 mIU/L on the basis of a recovery acceptance range of 10% from the expected concentration.
Using serum pools containing low, middle, and high insulin concentrations to measure interference, we detected no interference from free and complex bilirubin (up to 194 mg/L), hemoglobin (up to 7650 mg/L), or chyle (up to 2800 formazine turbity units). In addition, a 1/10th volume of 300 mIU/L insulin solution was added to the sera with high concentrations of rheumatoid factor (223–2350 kIU/L), hepatitis B surface antigen (3798–4337 kIU/L), and hepatitis C virus RNA (2.0 x 106 to 3.5 x 106 kIU/L), respectively. These samples were measured in duplicate to obtain analytical recoveries. These potential interferents did not affect the measured insulin values. According to the manufacturer, human anti-mouse antibody (HAMA) blockers were included in the ARCHITECT insulin. Results from the analytical recovery test showed no significant interference of HAMAs (data not shown). In addition, no high-dose hook effect was detected in 1 very highly concentrated WHO reference preparation containing 30 000 mIU/L insulin.
We assayed for insulin concentrations 5 times in solutions of proinsulin [relative molecular mass (Mr) 9 390; adjusted to 1 x 106 ng/L (3.31 x 10–6 mol/L)] and C-peptide [Mr 3020; adjusted to 1 x 107 ng/L (0.106 x 10–6 mol/L)]. The cross-reactivities, defined as a percentage of the amount of insulin (Mr 5807) corresponding to the analysis results in each solution for the amount of the proinsulin or C-peptide in the sample (insulin: 1 mIU/L = 41.67 ng/L = 7.18 x 10–12 mol/L), were <0.005% for proinsulin and <0.00001% for C-peptide.
To determine the minimum detectable insulin concentration, we prepared several samples by diluting calibrator B (3 mIU/L) serially with calibrator A (0 mIU/L) and calculated the mean (SD) RLU values from 10 measurements of each of those samples. The detection limit was defined as the lowest measurable concentration of insulin for which the mean – 3 SD was higher than the mean + 3 SD of calibrator A, measured 20 times. In this study, the detection limit was determined to be 0.5 mIU/L.
The E-test TOSOH II (IRI) from TOSOH Corporation is an enzyme immunoassay designed for use on the AIA-21. We used this assay and the ARCHITECT assay to measure insulin concentrations in 100 hospital serum samples (Fig. 1
). The regression equation between the ARCHITECT insulin (y) and the E-test TOSOH II (IRI) (x) results was: y = 0.832x + 0.068 mIU/L [95% confidence interval of the slope, 0.815–0.849; for the y-intercept; 0.775–0.912 mIU/L; correlation coefficient (r) = 0.990; and Sy|x = 3.17 mIU/L]. Although the correlation of the ARCHITECT insulin assay with the E-test TOSOH II (IRI) was good across the clinical measurement range, we detected slope deviation. To investigate the cause of the slope deviation, we confirmed that both assays were exactly calibrated against the WHO reference preparation. We prepared 3 concentrations of insulin (50, 100, and 200 mIU/L) by diluting the WHO reference preparation with the specialized diluents provided for each assay and measured each diluted sample 5 times. The mean percentage of measured value to theoretical value was 104.5% (range, 101.5%–106.8%) for the ARCHITECT insulin and 105.4% (range, 104.9%–105.9%) for the E-test TOSOH II (IRI). These results indicated that the 2 methods were exactly calibrated against the WHO reference preparation. To examine the influence of serum effects, we added 2 different concentrations of WHO standard solution (100 and 200 mIU/L) to 8 different sera at a ratio of 1 to 9. These samples were measured 3 times to calculate recovery ratios. The mean recovery ratio was 98.7% (range, 95.5%–103.8%) for the ARCHITECT insulin assay and 111.9% (range, 105.6%–119.6%) for the E-test TOSOH II (IRI). The mean recovery ratio of the E-test TOSOH II (IRI) thus was higher than that of ARCHITECT insulin. A Bland–Altman analysis indicated that proportional error accounted for the differences between these 2 assays (data not shown). Although we could not identify the causative components in sera that contributed to the slope deviation, the difference in recovery ratios suggests that some kind of matrix component in the sera affected the correlation slope.
|
We obtained a reference interval for fasting insulin concentrations from results of the analysis of 52 serum samples from healthy volunteers with fasting glucose concentrations within the reference interval (glucose <1 g/L after fasting for at least 8 h) defined in the American Diabetes Association revised standard criteria (5). The reference interval (mean ± 2 SD) for fasting insulin obtained with a Box–Cox power transformation and a parametric method was 2.7–10.4 mIU/L.
We obtained 3 kinds of recombinant insulin analogs, lispro (Humalog®; Eli Lilly and Company), aspart (NovoRapid® 300 FlexPen®; Novo Nordisk Pharmaceuticals), and glargine (Lantus®; Aventis Pharmaceuticals) and investigated whether they were detectable by these assays. Each of these recombinant insulin analogs has a nominal concentration of 100 kIU/L and is suitable for injection. Each was diluted volumetrically with 10 g/L bovine serum albumin to final insulin concentrations of 10 and 100 mIU/L. We measured all dilutions of the insulin preparations in triplicate and calculated the percentage cross-reactivity from the ratio of the measured and nominal concentrations. All dilutions were assayed on the ARCHITECT insulin and the E-test TOSOH II (IRI). A summary of the cross-reactivity percentages is shown in Table 1
.
|
The ARCHITECT insulin and E-test TOSOH II (IRI) assays demonstrated similar recoveries, close to the theoretical concentration, for insulin lispro (insulin B28 Pro
Lys and B29 LysPro). The ARCHITECT insulin assay had a low percentage of reactivity (76%) to insulin aspart (insulin B28 ProAsp), whereas the E-test TOSOH II (IRI) measured an almost nominal concentration. Insulin glargine (insulin A21 GlyAsn, and addition of 2 arginine residues to the COOH terminus of the B chain) had 22% cross-reactivity with E-test TOSOH II (IRI), whereas it showed concentration-dependent cross-reactivity with the ARCHITECT insulin assay (105% at 10 mIU/L and 83% at 100 mIU/L). Human insulin has 2 antigenic determinants. One is the terminus of the B chain (6)(7)(8)(9), and the other is the immunodominant A-chain loop, which comprises residues A8–A10(10)(11)(12)(13)). Insulin analogs have different antigenic sites than does human insulin. These results indicated that the insulin analogs could show various degrees of cross-reactivity depending on the selection of the monoclonal antibody used in the commercial reagents.
The ARCHITECT insulin assay showed favorable results in the basic performance evaluation, including reproducibility, dilution linearity, detection limit, and effects of interfering substances. In the comparison test, however, we observed slope deviation. Standardization of immunoassays for insulin remains a problem despite the availability of large quantities of human insulin through recombinant DNA technology (2). A task force on standardization of insulin assays noted significant variability in insulin results and suggested a process for the assessment and certification of insulin assays(14). Although we could not completely explain the cause of the slope deviation, it does suggest the influence of a matrix component and the need for further investigation.
Several immunoassays are available for detecting human insulin and insulin analogs. Bowsher et al. (15) reported on a sensitive RIA specific for insulin lispro. For all such assays, users should be aware of the various cross-reactivities of insulin analogs with the immunoassays and choose the most suitable reagent.
In summary, the ARCHITECT insulin assay demonstrated good performance and is a suitable assay for the quantitative determination of insulin in human serum.
References
The following articles in journals at HighWire Press have cited this article:
|
|
 |
|
  T. Linn, B. Fischer, N. Soydan, M. Eckhard, J. Ehl, C. Kunz, and R. G. Bretzel Nocturnal Glucose Metabolism after Bedtime Injection of Insulin Glargine or Neutral Protamine Hagedorn Insulin in Patients with Type 2 Diabetes J. Clin. Endocrinol. Metab., October 1, 2008; 93(10): 3839 - 3846. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. E. Manley, I. M. Stratton, P. M. Clark, and S. D. Luzio Comparison of 11 Human Insulin Assays: Implications for Clinical Investigation and Research Clin. Chem., May 1, 2007; 53(5): 922 - 932. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
