Insulin analogues as a new example of interference in insulin assays

In a recent article, Dayaldasani et al. ¹ carefully assessed cross-reactivity of five insulin analogues in three immunometric insulin assays. An excellent editorial ² also highlighted this complex phenomenon as a new source of error in immunoassay. Cross-reactivity was assessed by adding appropriate amounts of analogues to serum with known/constant insulin concentrations. Interferences of varying degrees of magnitude were found in each of the three assays.

In the 1950s, insulin (molecular weight 5808 Da) was considered to be non-immunogenic. Porcine and bovine insulins, which differ from human insulin by a single amino acid (porcine) and three amino acids (bovine), respectively, were widely used. Subsequently, such small differences from human insulin were shown to provoke production of endogenous insulin antibodies in some patients. The advent of human recombinant insulin reduced but did not eliminate the immunogenicity problem.

The five insulin analogues listed in the study of Dayaldasani et al. ¹ are also structurally different from naturally secreted human insulin. They are engineered recombinant forms in which amino acids in specific locations have been altered. For example, among the listed short-acting analogues, proline is substituted with aspartic acid at position 28 in the B-chain of insulin aspart; in insulin lispro, the penultimate C-terminal lysine and proline residues in the B-chain are reversed; in insulin glulisine there are two substitutions – the asparagine at position 3 in the B-chain is replaced by lysine, and the lysine at position 29 by glutamic acid. Further structural changes are introduced in long-acting analogues. In insulin glargine, the glycine at position 21 on the A-chain is substituted with asparagine, and two arginines are added to the C-terminus of the B-chain, while in insulin detemir, a fatty acid (myristic acid) is linked to the lysine at position 29 on the B-chain.

Interestingly, insulin aspart and porcine insulin, which both differ from human insulin by a single amino acid, have similar immunogenicity.^3,4 Engineering recombinant forms of insulin analogues was primarily aimed at favourably enhancing their pharmacokinetics (insulin concentration over time) and pharmacodynamics (resulting effects on glucose over time). Immunogenicity was a secondary effect, not necessarily detrimental to insulin action (for example, antibodies binding insulin might prolong its duration of action). In addition to the immunogenicity of insulin analogues, they have the potential to interfere with immunoassays and affect their accuracy.^3–8

Thus, biochemical investigation of patients treated with insulin analogues must take into account not only their known cross-reactivity in immunoassays as reported^1,2 but also their ‘known-unknown’ adverse potential, i.e. possible presence of endogenous antibodies and its impact on accuracy. In general, the immunogenicity of a substance is influenced by numerous factors such as its molecular size, chemical complexity/heterogeneity and individual’s genotype. Insulin analogues, whether administered subcutaneously, by infusion pump, or by inhalation, are not structurally identical to insulin secreted by the pancreas. These modes of administration can cause further structural and metabolic changes in situ, releasing dimeric or even hexameric forms, further increasing the repertoire of endogenous antibodies and their heterogeneity (class, subclass, affinity/avidity, titre, specificity, etc.). Cross-reactivity is predictable, but the presence of endogenous insulin antibodies is unpredictable and a complex source of analytical interference and inaccuracy. ⁹