| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Editorials |
Institut de Physique Biologique, 1 place de l’Hôpital, F-67091 Strasbourg Cedex, France, Fax 33-3-90-24-40-57, e-mail remy.sapin{at}chru-strasbourg.fr
Insulin (51 amino acids, 5808 Da) is synthesized from its precursors preproinsulin and proinsulin (hPI; 86 amino acids) in the ß cells of the pancreatic islets of Langerhans. A human insulin molecule is chemically homogeneous, consisting of 2 polypeptide chains, the A chain (21 amino acids) and B chain (30 amino acids), connected by 2 disulfide bonds (A7-B7 and A20-B19); a 3rd disulfide bond links the A6 and A11 residues. In serum, insulin circulates in a free form (not bound to carrier proteins) together with small quantities of its precursors, mainly intact hPI and des (31,32) split hPI (hPI cleaved at the junction between the B chain and C-peptide linking the A and B chains in the hPI molecule). Des (64,65) split hPI (hPI cleaved at the junction between the A chain and C-peptide) is a minor component of hPIs in serum (1).
Insulin is the only hypoglycemic hormone. Its measurement in serum plays a central role in the assessment of ß-cell secretion and insulin resistance. Ideally, an insulin assay should be sensitive, specific, and applicable to a large number of samples. It should also be standardized to be efficiently used in large multicenter clinical studies and considered in guidelines.
Among insulin assay methods, only immunoassays are applicable to large numbers of samples. The 1st RIA for human insulin, described in 1959 by Yalow and Berson (2), relied on competitive binding of human insulin in plasma samples or calibrators and 131I-labeled bovine insulin to guinea-pig anti–bovine insulin polyclonal antibodies. Separation of the bound and free fractions was performed by chromatoelectrophoresis. Later the availability of human insulin in larger quantities (3) allowed the production of guinea-pig anti–human insulin antibodies, and an antibody precipitation technique made the assay easier to use. Except for 1 assay, RIAs detect both insulin and hPIs (intact and conversion intermediates between hPI and insulin) to various degrees, making it impossible to correctly assess insulin secretion under conditions in which circulating concentrations of hPIs are increased (type 2 diabetes, impaired glucose tolerance, and insulin resistance) (4)(5)(6).
In the mid-1970s, Kohler and Milstein described the principles allowing the production of monoclonal antibodies in large amounts (7). The development of antiinsulin monoclonal antibodies in the early 1980s (8) led to the introduction of 2-site immunometric assays (IMAs) requiring antibodies in excess. Insulin IMAs were recognized as more precise, more sensitive (with a lower detection limit), and more specific (selective) than RIAs. The greater specificity was achieved by use of monoclonal antibodies directed against the epitope lying at the free N-terminal region of the insulin A chain. This epitope is obscured by C-peptide in the hPI and des (31,32) hPI molecules (9). Subsequently, a large number of automated assays involving enzymatic or luminescent labels were commercialized, thereby improving intralaboratory reproducibility and practicability.
The main analytical pitfalls in insulin measurement in serum are related to (a) hemolysis, (b) circulating antiinsulin autoantibodies, and, for the past few years, (c) the reactivity (or lack of reactivity) of rapid- or long-acting pharmacological insulin analogs in sera (10). Hemolyzed samples contain an insulin-degrading enzyme and should not be analyzed unless they can be handled at 4 °C within 
2–3 h or an insulinase inhibitor has been added in the blood collection tube to prevent insulin degradation. Antiinsulin antibodies interfere with RIAs and IMAs, yielding overestimated values for free insulin, the biologically active form of insulin, that is not bound to antiinsulin antibodies in serum. Measurement of free insulin requires the removal of antiinsulin antibodies, which can be achieved by polyethylene glycol precipitation.
Highly purified recombinant human insulin preparations with a potency of 26 IU/mg are available from Novo Nordisk and Eli Lilly and Company (1 IU = 6 nmol = 34.85 µg of insulin) (11). This conversion factor, however, is not valid with some commercial assays calibrated against the 66/304 human insulin International Reference Preparation (IRP) established in 1974, for which the manufacturers must provide an assay-specific conversion factor.
In light of the advances in analytical methods, users expect insulin assays to provide high sensitivity (with low limits of detection) and good specificity (selectivity) and between-assay harmonization. Indeed, circulating insulin is homogeneous, unlike glycoprotein hormones (such as human chorionic gonadotropin, luteinizing hormone, and follicle-stimulating hormone), for which immunoassay standardization is problematic (12). Standardization is particularly important when results are interpreted by comparison with published results, for example when insulin assays are used in the assessment of insulin resistance.
In 1996 the American Diabetes Association (ADA) task force on standardization of insulin assays reported widely disparate results for research-grade insulin assays (13). In the April issue of Clinical Chemistry a new report of the ADA Workgroup (14) presents the performance characteristics (imprecision, specificity, and between-assay comparability) of 12 insulin assays commercialized between 2004 and 2006. Studies of within-run imprecision revealed CVs >10% in almost half the methods. Three of the methods with CVs >10% had detection limits 
12 pmol/L. As underlined by Manley et al. in the comparison study of 11 commercially available insulin assays published in this issue of Clinical Chemistry (15), the detection limits quoted by the manufacturers are generally estimated from the mean of zero signals plus 2 SD (a parameter also called, in newer nomenclature, the limit of the blank). The more important characteristic of assays is their lower limit of quantification, the lowest concentration that can be measured with a defined uncertainty. In endocrinology, the limit of quantification is often considered to be the lowest concentration that can be measured with a between-day CV of 20%. This value is sometimes referred to as "functional sensitivity," a term that unfortunately confounds sensitivity and lower limit of quantification. Patients with hyperinsulinism do not usually have dramatically high insulin concentrations (16); as a result, investigations of hypoglycemia require insulin assays that have low imprecision in the low concentration range. This characteristic requires high sensitivity at low concentrations (that is, a large change in signal per unit change of concentration) and a good precision of signal measurement.
The ADA Workgroup assessed the degree of cross-reactivity of hPI, split (32,33) hPI, and des (64,65) hPI in insulin assays. It considered assays to be adequately selective ("specific") if they showed cross-reactivities with hPI and split (32,33) hPI <3%. Among the 10 assays studied, 8 met this criterion; 1 reacted with both hPI and the 2 hPI-related peptides; and the remaining assay recognized split (32,33) hPI only. Many highly specific assays are now available commercially and are preferable whenever an insulin-specific measurement is required, as in samples from patients with increased hPIs (5)(6). Nevertheless, specificity is certainly not the only important source of intermethod variability.
Lacking a gold standard against which to evaluate insulin assays, the ADA Workgroup judged harmonization of results by comparing results yielded by each method with the overall mean of all the methods. In a trial involving 40 heparin plasma samples, only 6 of 10 assays were considered to have an acceptable agreement (total error allowance of 25.6% from the mean value). The use of a common recombinant insulin preparation, proposed reference material insulin (PRMI), instead of the manufacturers’ preparations, as calibrators did not improve the result. All assays were calibrated against the 66/304 IRP defined in terms of international units. The use of different assay-specific factors for converting concentrations expressed in international units (mIU/L) to the units used in the ADA study (pmol/L) may have contributed to the observed biases. Poor recovery (64%–79%) of PRMI added to a serum pool was observed in 3 of 10 assays, which suggests that some assays are not properly calibrated. Matrix effects may also, at least in part, account for the lack of harmonization of the results. Indeed, the recovery of PRMI added to a serum with a low insulin concentration differed significantly from the recovery of PRMI diluted in the assay manufacturer’s recommended diluent. The lack of harmonization of insulin assays and the matrix effect highlighted in the ADA report are supported by data presented by Manley et al. (15): the results obtained from 150 sera varied by a factor of 2 for samples with high and midrange insulin concentrations and by a factor of >20 for samples with low insulin concentrations. With 1 assay, a significant difference (–14%) was observed between the results of insulin measurements in serum and plasma samples.
All things considered, some assays need improvement in reproducibility and detection limits. The detection limit can be improved by decreasing noise (in this respect, better washing using magnetic particles may help) and by using efficient signals (such as luminescent signals) and antibodies with high affinities. Specificity remains an important issue: specific assays should be the method of choice unless an insulin assay is used to investigate hypoglycemia, which requires the detection of hPIs and of endogenous and exogenous insulin. Specificity against the widely used rapid- and long-acting insulin analogs should be systematically assessed. It is also clear that insulin assays still call for standardization. A better between-assay harmonization of results could have been expected for the IMAs of a homogeneous peptide, not bound to carrier proteins and available in a human recombinant form.
With the aim of achieving correct standardization of insulin assay results, the ADA report highlights the necessity of (a) preparing a new reference preparation defined in terms of mass, (b) validating the commutability of this reference preparation in assay diluents with native patient sera, and (c) developing a reference method to achieve the traceability of results over long periods. With the efforts of the ADA Workgroup, we can expect that improvements in assay standardization will be achieved within the next couple of years and will have a greater impact than those noted since the previous ADA report, published 10 years ago (13). The introduction in insulin immunoassays of monoclonal antibodies produced in serum-free cell culture medium may lead to reduction in undesirable variables and to improved standardization (17). Consistent measurements of insulin secretion will be achieved only with better-standardized insulin immunoassays, which will also play a key role in the development of guidelines for the investigation of insulin resistance and diabetic states. We will then approach the 2nd half-century of insulin assays in a good position to make use of all the services awaited since they were first described.
Acknowledgments
Grant/funding support: None declared.
Financial disclosures: None declared.
References
The following articles in journals at HighWire Press have cited this article:
|
|
 |
|
  A. N. Hoofnagle, J. O. Becker, M. H. Wener, and J. W. Heinecke Quantification of Thyroglobulin, a Low-Abundance Serum Protein, by Immunoaffinity Peptide Enrichment and Tandem Mass Spectrometry Clin. Chem., November 1, 2008; 54(11): 1796 - 1804. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. E. Manley, S. D. Luzio, I. M. Stratton, T. M. Wallace, and P. M.S. Clark Preanalytical, Analytical, and Computational Factors Affect Homeostasis Model Assessment Estimates Diabetes Care, September 1, 2008; 31(9): 1877 - 1883. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
