Discussion
There are two important aspects to this case. First, we provide a rare description of insulin allergy associated with hypereosinophilia, anti-insulin IgG antibodies, and anti-insulin receptor antibodies in a patient with type 2 diabetes. Second, the marked efficacy of liraglutide suggests that it is useful in all patients with type 2 diabetes.
In our case, we believed that our patient’s insulin allergy was predominantly a type 1 immediate reaction based on the fact that insulin lispro induced wheals, itching, and redness immediately after injection. Insulin allergy is typically classified into the following: type 1, which is local and immediate; type 3, which is related to autoimmune disease; and type 4, which is a delayed-type reaction. Symptoms typically manifest within 15 minutes after insulin injection in type 1 reactions, approximately 6 hours after type 3 reactions, and 8 to 24 hours after type 4 reactions [
2]. In our case, we confirmed that insulin lispro induced immediate symptoms but could not detect whether insulin aspart, glulisine, or glargine induced those reactions.
We did not perform either a skin prick test or an intradermal reaction test because of his past history of anaphylaxis. Out of concern for his safety, we substituted this with a non-invasive drug-induced lymphocyte stimulation test, which is normally considered best suited for delayed (type 4) reactions. Although this test was only positive for insulin glulisine, false-negatives have been detected in some cases, and we considered that insulin glargine was also associated with delayed-type allergy. This was because insulin glargine induced occasional induration and because its discontinuation resulted in improvement of both his induration and hypereosinophilia.
Based on the combination of these results, we concluded that our patient probably had a delayed-type allergic reaction that was complicated with a predominant immediate-type allergic reaction. Unfortunately, we did not obtain the list of additives in either the lispro or glargine, which could have been important given his history of protamine anaphylaxis. Further investigation of the cause of the allergic reactions could not be performed (for example, to other insulin preparations, including human neutral protamine Hagedorn, or the additives in the insulin types).
Previous reports of type 1 and type 4 allergic reactions to insulin indicate that hypereosinophilia does not necessarily occur. Therefore, the increase of eosinophilia in both local pathological findings and systemic peripheral blood is characteristic of this case. In general, although numerous eosinophils tend to be detected in local inflammation, such as allergy and parasitic infection, the complete role of eosinophils has not been elucidated [
10]. For example, although hypereosinophilia has been observed in invasive parasitic infection, it has recently been reported to be partially defensive
in vivo during such an infection [
10]. In clinical cases of hypereosinophilia, clinicians must quickly and accurately diagnose the cause [
11]. Potential causes in the differential diagnosis of this case included adrenal crisis, parasitic infection, and blood disease, but there was no evidence in support of these diagnoses. In addition, because hypereosinophilia simultaneously occurred with the observed clinical course of insulin allergy, we considered insulin allergy the more likely cause. This was consistent with the report by Nagai
et al., who reported a case of immediate-type allergy against human insulin associated with marked eosinophilia in a patient with type 2 diabetes [
12]. In that case, there was also the possibility of hypereosinophilia-induced renal dysfunction. Although there was no evidence of organ damage in this case, clinicians should remain vigilant for hypereosinophilic syndrome [
11,
13] and be prepared to initiate appropriate therapy. Currently, treatment options for insulin allergy include switching to another insulin device [
2], steroid therapy [
2], continuous subcutaneous insulin infusion therapy [
14], hyposensitization therapy [
15], pancreas transplantation [
16], and anti-IgE therapy [
17]. However, reports on the treatment of insulin allergy with hypereosinophilia are limited, and new options are needed. Therefore, we consider that liraglutide might be a breakthrough agent in the treatment of insulin allergy with hypereosinophilia in patients with type 2 diabetes. Indeed, a Medline search indicated that at the time of writing, we are the first to report the beneficial effects of liraglutide in this setting.
Treatment with liraglutide requires the residual ability to secrete insulin, but we do not know the appropriate range of insulin secretion that is necessary. Previously, Kozawa
et al. reported that the cut-off value for predicting the efficacy of liraglutide was 1.1 for the C-peptide index, which was defined as (fasting C-peptide/glucose)×100, and 1.5 ng/mL for fasting C-peptide [
18]. This index was also shown to be a predictive marker for the beta-cell area of the human pancreas [
18,
19]. Using a glucagon stimulation test, Usui
et al. reported that a ΔC-peptide value of 1.34 ng/mL was a useful cut-off point for switching from insulin to liraglutide without developing hyperglycemia [
20]. In this case, although the fasting C-peptide level was high, the ΔC-peptide in the glucagon stimulation test had a low value. Therefore, liraglutide alone could not inhibit postprandial hyperglycemia, and the antidiabetic drugs glimepiride, metformin, and voglibose were combined to achieve complete control. Moreover, the additional benefits of liraglutide include the prevention of body weight gain due to its appetite-suppressing and gastrointestinal peristaltic movement-suppressing effects. Liraglutide and other GLP-1 analogs have been reported to decrease body weight [
21,
22]. In our case, our patient’s body weight decreased by only 1.7 kg, but we could not administer liraglutide at a dose of more than 0.9 mg/day. Worldwide, liraglutide is administered at a maximum dose of 1.8 mg/day; however, Japan has approved that liraglutide is administered at a maximum dose of only 0.9 mg/day.
In our case, it is possible that both the insulin allergy and the presence of anti-insulin IgG antibodies themselves aggravated our patient’s glycemic control. Therefore, we tested his human leukocyte antigen-antigen D related (HLA-DR) type based on the report by Uchigata
et al. that insulin autoimmune syndromes are strongly associated with HLA-DR4 [
23]. We showed that his HLA-DR types were DR-9 and DR-15 and not DR-4; thus, we did not consider HLA-DR to be a potential causative factor for producing anti-insulin IgG antibodies. Instead, we concluded that the anti-insulin IgG antibodies resulted from the exogenous insulin injection; thus, we used liraglutide without insulin to decrease the amount of anti-insulin IgG antibodies and to avoid further production.
Moreover, Scatchard analysis of the anti-insulin IgG antibodies in our patient indicated that there was a high binding capacity and a low affinity constant for high affinity sites. At high affinity sites, Eguchi reported two important patterns of Scatchard analysis of anti-insulin IgG antibodies: (1) high (11.5 to 53.2) binding capacity (10
−8 M) and low (0.04 to 0.21) affinity constant (10
8 M
−1) in insulin autoimmune syndrome; and (2) low (0.12 to 1.1) binding capacity (10
−8 M) and high (1.45 to 7.11) affinity constant (10
8 M
−1) in patients with diabetes requiring insulin therapy [
24]. Therefore, the binding capacity and low affinity constant for high affinity sites are clinically important. Our patient had extremely high binding capacity and a low affinity constant compared with previous reports. In particular, his extremely high binding capacity was consistent with his high serum insulin levels.
After insulin therapy was discontinued and liraglutide was started, the characteristics of his anti-insulin IgG antibodies changed. Specifically, there was a notable decrease in the binding capacity and an increase in the affinity constant for high affinity sites (Fig.
3c). Hara
et al. have reported that antibody-mediated insulin resistance was treated by ceasing insulin administration [
25], whereas Tamura
et al. reported that liraglutide ameliorated the anti-insulin IgG antibodies in a woman with type 2 diabetes receiving hemodialysis [
26]. In our case, it remains unclear how liraglutide might have affected our patient’s anti-insulin IgG antibodies; however, a recent report has suggested that GLP-1 receptor agonists interact with the immune system. In fact, Hadjiyanni
et al. reported that GLP-1 receptor signaling selectively regulated murine lymphocyte proliferation and maintained peripheral regulatory T cells
in vivo [
27]. In humans, the effects of GLP-1 receptor agonists have been investigated in patients with psoriasis [
28], but the detailed mechanism of action is unclear. Further reports are needed to improve our understanding of these agents.
Anti-insulin IgG antibodies have often been reported to induce hypoglycemia [
8,
21,
23,
24]. Unfortunately, we could not measure serum free insulin levels directly and do not know whether the anti-insulin receptor antibodies really affected glucose metabolism. Although Kim
et al. reported that anti-insulin receptor antibodies often coexist with anti-insulin IgG antibodies in Korean patients with insulin autoimmune syndrome [
29], the clinical significance of this finding has been not determined. Acquired anti-insulin receptor antibodies induce severe insulin resistance, called “type B insulin resistance,” which was first reported in 1976 [
30]. However, we found no other symptoms or complications of type B insulin resistance, such as acanthosis nigricans or autoimmune disease (Table
1). In addition, the total insulin dose was smaller than that reported in previous studies on type B insulin resistance [
9,
30]. Therefore, we considered the possibility that the presence of anti-insulin receptor antibodies is a false positive in this case. It is also possible that the method of measurement led to a false positive result for anti-insulin receptor antibodies because we used the insulin binding inhibition method [
31]. Perhaps the high insulin levels and high anti-insulin IgG antibodies levels had some unknown influence on the measured values.