Effect of Long-Term Exposure to Insulin Lispro on the Induction of Antibody Response in Patients With Type 1 or Type 2 Diabetes

The studies were conducted according to the applicable laws of each participating country, the ethical principles of the Declaration of Helsinki, and guidelines for Good Clinical Practice.

The primary objective of the combined studies was to assess the safety of insulin lispro with respect to the long-term immunologic response. One-year immunologic effects were previously reported (29).

This composite multinational, multicenter, controlled and noncontrolled, open-label study evaluated 1,221 men and women with type 1 or type 2 diabetes (30) from parent studies (demographics detailed previously) (29,31,32) who continued in extension studies with 988 patients completing ∼4 additional years (Table 1). Exclusion criteria included significant renal, hepatic, or cardiac disease, cancer, drug or alcohol abuse, significant insulin allergy, recurrent severe hypoglycemia, anemia or hemoglobinopathy, BMI >35 kg/m², pregnancy, lactation, and treatment with oral hypoglycemic agents, systemic glucocorticoids, insulin therapy via pump, or insulin doses >2.0 units · kg⁻¹ · day⁻¹.

Human insulin analog insulin lispro ([LYS (B28), PRO (B29)], rDNA) (Humalog; Eli Lilly and Company, Indianapolis, IN) was provided in 10-ml vials or 1.5-ml cartridges (100 units/ml [3.5 mg/ml]). Basal insulin study drugs were Humulin N provided in 10-ml vials or 1.5-ml cartridges (100 units/ml [3.5 mg/ml]) and Humulin U provided in 10-ml vials (100 units/ml [3.5 mg/ml]). All insulins were administered subcutaneously. For the extension studies, patients used commercially available intermediate or long-acting formulations of human insulin as their basal insulin, and the investigator adjusted the dosage and time of injection of insulin lispro and basal insulin in accordance with the metabolic needs of the patient.

Data from these trials were segregated into three parts as described below.

Patients with type 1 or type 2 diabetes previously treated with human insulin for at least the last 2 years who were randomized to insulin lispro and successfully completed Studies A, B, C, or D (12-month parallel studies) continued treatment with insulin lispro in Extension Study K. Of 271 patients (145 men, 126 women) who were enrolled, 57 discontinued the study. Four patients discontinued the study due to unintended pregnancies and four patients died due to causes unrelated to the study. In addition, 7 patients changed their place of residence, 25 patients had personal conflicts or other personal reasons for discontinuation, 8 patients discontinued because of a protocol violation, 5 patients discontinued because of lack of efficacy (either patient’s or physician’s perception or both), and 4 patients were lost to follow-up. A total of 214 patients continued throughout the extension study.

Patients who were new to insulin therapy who were randomized to insulin lispro and successfully completed Studies E or F (12-month parallel studies) continued treatment with insulin lispro in Extension Study O. Of the 179 patients who entered from the parent studies E or F, 167 had received insulin lispro.

Of the 167 patients who received insulin lispro (39 patients with type 1 diabetes and 128 patients with type 2 diabetes), 46 discontinued the study. One patient developed squamous cell carcinoma, 3 patients died due to causes unrelated to the study, 7 patients were lost to follow-up, 5 patients changed their place of residence, 17 patients discontinued because of a personal conflict or other personal decision, 6 patients discontinued because of a protocol violation, and 7 patients discontinued because of their physician’s decision. A total of 121 patients completed the extension study.

Patients previously treated with insulin were randomized to insulin lispro or human insulin in a 6-month cross-over design and continued treatment with insulin lispro in the Extension Studies M and N. These studies allowed determination of whether an anamnestic response would occur after intermittent exposure to insulin lispro in patients who were randomized to the insulin lispro first and regular human insulin second followed by lispro in the extension.

A total of 445 men and 326 women with type 1 (n = 509) or type 2 (n = 262) diabetes continued on insulin lispro during the extension studies. A total of 653 patients completed the studies. Reasons for discontinuation of the study included adverse events (19 patients), death (9 patients), change of place of residence (9 patients), loss to follow-up (12 patients), withdrawal from the study (28 patients), violation of protocol (12 patients), and perceived lack of efficacy (29 patients).

Blood samples for the measurement of ISA, LSA, and cross-reactive antibodies were collected at various times during each study. Although these measurements were made in two separate laboratories, the assay protocols were identical and extensive cross-referencing of sera was performed. In brief, the antibody assay used was a self-blank assay in which prebound insulin is removed using dextran charcoal, samples are studied in duplicate, and specificity is established by differential displacement of bound, labeled insulin by addition of excess antigens (33). The interassay coefficient of variation of diluted antibody run at dilutions of 1:20,000 and 1:40,000 were 10% (mean binding 48.2%) and 14.9% (mean binding 23.2%), respectively, for cross-reactive antibodies. The reference ranges used (based on healthy subjects not previously exposed to exogenous insulin) for each antibody assayed were 0–2.5% for ISA, 0–3.2% for LSA, and 0–2.5% for cross-reactive antibodies in laboratory 1 and 0–0.80% for ISA, 0–0.85% for LSA, 0–1.57% for cross-reactive antibodies in laboratory 2. In one laboratory, the assays were performed during the parent phase of the trials, and in the other laboratory, the assays were performed during the extension phase. For assessing parent and extension phase cross-reactive antibody data together, the values from both laboratories were transformed using an empirical cumulative distribution function (CDF) transformation (34).

For evaluating antibody data over time, data from the parent and extension studies were combined. Data were analyzed by type 1 and type 2 diabetes and presented by visit. The baseline measurement from the parent studies was considered Month 0. Visit 1 of the extension studies was conducted at 12 months of therapy (except for Part 3, which was conducted at 6 months of therapy).

The statistical analyses were performed using SAS statistical software (SAS Institute, Cary, NC). Mean and 95% CIs are presented for each visit (Figs. 1–5). End point was compared with baseline using Wilcoxon’s signed-rank test. Repeated cross-reactive antibody measurements over time for patients were analyzed using a quadratic model (PROC MIXED in SAS) in which the exposure time on insulin lispro was the independent variable and the cross-reactive antibody binding was the dependent variable. The intercept and the linear and quadratic coefficients were defined as random effects, because these parameters differ for each patient, and an unstructured covariance term was used.

In patients who were previously treated with human insulin, data obtained were analyzed with respect to change from baseline to end point. For ISA, the mean values and upper 95% CIs for all visits were within or near the reference range. The ISA response (Fig. 1) in type 1 patients was consistently higher than in type 2 patients until after 40 months of the study. The end points were not statistically different from baseline. For LSA, the mean values and upper 95% CIs for most of the visits were within or near the reference range. There was a virtually complete overlap (Fig. 2) in LSA response between patients with type 1 and type 2 diabetes for a period of more than 25 months. Inconsistent differences between different visits and baseline were detected for both types of patients, but they remained within or near the reference range. The mean end point was statistically higher than baseline in patients with type 1 diabetes (P = 0.034), but the median of baseline and end point values for type 1 diabetes were still within the reference range in nonexposed subjects.

Cross-reactive antibody responses for patients with type 1 and type 2 diabetes are shown in Fig. 3. The mean values and 95% CIs for almost all visits were higher than the reference range (for non-insulin-exposed individuals). During most visits, cross-reactive antibody levels were statistically higher than baseline in patients with type 1 diabetes. The trend of cross-reactive antibody levels was downward after the peak near 18 months of therapy for patients with type 1 diabetes and 17 months of therapy for patients with type 2 diabetes. In both patient groups, cross-reactive antibody levels then decreased to within reference range values by the end of the study.

Individual cross-reactive antibody values (Fig. 3) show good fit of the data (r² = 0.741 for patients with type 1 diabetes and r² = 0.760 for patients with type 2 diabetes, between predicted and original values), indicating that the observed antibody response was consistent with the visit versus time analysis.

The antibody data were assessed in patients who were not previously treated with insulin with respect to change from baseline to end point. For ISA, the mean values and upper 95% CIs for all visits were within or near the reference range. The ISA response (data similar to Fig. 1 are not shown) in type 1 diabetes patients was slightly higher than in type 2 patients. The end points were statistically not different from baseline.

There was a virtually complete overlap in LSA response (data similar to Fig. 2 are not shown) between patients with type 1 and type 2 diabetes for a period of 20 months, followed by a sustained divergence of patients with type 1 diabetes, with a slightly higher response. Inconsistent differences at some individual visits and at baseline were detected for patients with both types of diabetes. The end point was not statistically different from baseline. Overall, the levels of both ISA and LSA remained within the reference range.

Baseline cross-reactive antibody levels were relatively low in these insulin-naïve patients. After initiation of insulin therapy, these levels increased during the first 12 months, followed by a gradual decrease that continued for the remainder of the study. The mean cross-reactive antibody levels in patients with type 2 diabetes were higher than the upper limit of the range for patients who were never exposed to insulin (Fig. 4). However, these levels were within the range for patients who were previously exposed to insulin (0–43.0%). Most patients had very low levels of antibodies with an occasional high value (50 and 75 percentile values were only 2 and 7%, respectively). The end point was statistically different from baseline in patients with type 1 diabetes (P = 0.001) and those with type 2 diabetes (P = 0.001).

In addition, as previously observed with other insulins, patients with type 1 diabetes and insulin-naïve patients had higher antibody binding levels than the patients with type 2 diabetes. Results of cross-reactive antibody binding (Fig. 4) show a good fit of the data (r² = 0.736 for type 1 diabetes, and r² = 0.776 for type 2 diabetes), indicating that the observed antibody response was consistent with the visit versus time analysis.

The antibody response in patients who were exposed to insulin intermittently was analyzed with respect to change from baseline to end point. For ISA, the mean values and upper 95% CIs for all visits were within or near the reference range (data very similar to Fig. 1 are not shown). Occasional differences between visits and baseline were also observed. The end points were statistically different from the baseline in patients with type 1 diabetes (P = 0.04) and those with type 2 diabetes (P = 0.002).

For LSA, the values for most of the visits were within or near the reference range (data similar to Fig. 2 are not shown). Differences between various visits and baseline were detected occasionally. The end points were not different from baseline.

For cross-reactive antibodies, a comparative analysis between two cross-over sequence groups for each diabetes type was performed. This analysis showed that there was no difference in cross-reactive antibody response between patients treated intermittently and patients treated consecutively with insulin lispro. Data were then analyzed by type of diabetes.

End point was not significantly different from baseline in patients with type 2 diabetes (P = 0.940), but for cross-reactive antibodies, the end point was higher than baseline in patients with type 1 diabetes (P < 0.001) (Fig. 5). Patients with type 1 diabetes demonstrated a sustained divergence; at most visits, cross-reactive antibody levels were statistically higher than baseline. Cross-reactive antibody binding (Fig. 5) results show a good fit (r² = 0.849 for type 1 diabetes, and r² = 0.886 for type 2 diabetes) of the data, indicating that the observed antibody response was consistent with the visit versus time analysis.

In all studies, these analyses did not show any statistically significant relationship between the change in the total daily insulin dose and the change in percent antibody binding for any of the antibody types (ISA, LSA, cross-reactive antibody).

The incidence of treatment-emergent adverse events potentially related to allergy was compiled from the database of all patients treated with insulin lispro, human insulin, or both in Studies A through H (Table 2). With respect to the general terms of allergic reaction, rash, and pruritus, the reported incidence was <3% and comparable between the treatment groups. When events of rash and pruritus were examined individually, 78 rash events and 17 pruritus events could be reasonably associated with an insulin-unrelated event (such as insect bite, sunburn, etc.). In the remaining 83 rash events (40 for human regular insulin and 43 for insulin lispro) and 26 pruritus events (11 for regular human insulin and 15 for insulin lispro), there were no statistical differences between treatment groups. For all other event categories, the incidence was <0.5% and there were no statistical differences between treatment groups. The incidence of allergic reactions for patients with the highest 1% of cross-reactive antibody or LSA binding was not different from the group as a whole.

Even with the use of human insulin, low-level antibody responses and allergic responses still occur, albeit with much less intensity and frequency. One can speculate as to whether these events are triggered by excipients in commercial insulin, such as protamine or preservatives, small amounts of insulin fragments, or dimers (35,36), or by the fact that therapeutic insulin is usually injected into the subcutaneous tissue. Several structural and pharmacological features of insulin lispro suggested that the immunogenic potential would be low. First, the reversal of the amino acids at B28 and B29 is in a relatively nonimmunogenic area of the molecule, and no additional or foreign amino acids were added to the structure. Secondly, the Lys B28-Pro B29 sequence duplicates the corresponding sequence in the structurally similar IGF-1 molecule, so the body normally “sees” this amino acid sequence. Thirdly, insulin lispro is very rapidly absorbed from the subcutaneous tissue, and prior insulin studies have suggested that the immunogenic potential of an insulin formulation is correlated with the duration of residence in the subcutaneous tissue. Insulins in suspension form have been shown to be more immunogenic than soluble preparations (25). This view is further supported by the findings of increased concentration of anti-insulin antibodies during continuous subcutaneous insulin infusion and multiple-injection therapy (37).

Previous short-term studies have demonstrated that the immunogenicity of insulin lispro was not different from regular human insulin in patients with type 1 or type 2 diabetes treated for 12 months (29). The present analyses of these combined studies provide the first opportunity to define the immunologic profile of insulin lispro over an extended treatment of period of more than 4 years. For both patients with and without prior insulin therapy, the values of ISA and LSA antibodies generally remained within the nonexposed reference ranges for the duration of the study. Most antibodies generated in these clinical trials were cross-reactive rather than specific (29). In both patients with type 1 or type 2 diabetes previously treated with insulin, there were small increases in cross-reactive antibody levels from baseline to peaks at 14–18 months during insulin lispro therapy. Subsequent to these responses, the cross-reactive antibody levels gradually declined to the baseline levels by 36–48 months of treatment. Although it has been reported that in patients started on either highly purified porcine or recombinant human insulin antibodies develop by 6 months followed by a decline thereafter, earlier work has suggested that less immunogenic insulins are associated with a later peak of smaller magnitude (25,38,39). Patients not previously treated with insulin also exhibited predominantly cross-reactive antibody responses with peak responses by 12 months and a declining over the remaining years of the study. In both treatment groups, patients with type 1 diabetes developed insulin antibodies more readily than patients with type 2 diabetes (29).

In patients who received intermittent therapy with insulin lispro in the cross-over followed by extension studies, there was no anamnestic response in either the cross-reactive or the lispro-specific antibodies. In most studies, as previously reported (39), there were no relationships between changes in cross-reactive antibodies and changes in insulin dosage over time. The few statistically significant changes occurred only in patients with type 1 diabetes, and these changes were inconsistent with dosages moving in opposite directions with slight increases in antibody. The incidence of nonspecific allergic response was slightly >2% and occurred with similar frequency in treatment with insulin lispro and human insulin. Therefore, the allergenic profile of insulin lispro is low and similar to that of human insulin (26,29).

Insulin lispro has been successfully used to manage patients with cutaneous insulin allergy or immunologic resistance (9,24,40–46). Case reports indicate that insulin lispro seems to have lower immunogenic potency (9,10,24,40). In intradermal tests, the initial wheal-flare response generated by insulin lispro was 50% less intense than human insulin, which progressively declined and vanished by week 25 (9). This reduced immunogenic response may be attributable to the rapid monomeric state of insulin lispro after injection (10). Additionally, a reduction in insulin autoantibodies has been observed after treatment with insulin lispro (41).

Long-term treatment with insulin lispro, similar to recombinant human insulin, elicits a low and clinically inconsequential immunogenic response. As with other insulins, treatment with insulin lispro is associated with slightly greater antibody responses in patients with type 1 diabetes than in those with type 2 diabetes. Intermittent treatment does not exaggerate specific or cross-reactive antibody response. Patients treated with insulin lispro do not develop increased insulin dosage requirements, nor do they experience an increase in events related to insulin allergy.

This work was sponsored by Eli Lilly and Company.

The authors thank Kenneth Robertson, PharmD, for his valuable contributions in the preparation of the manuscript. We also thank the following investigators and their site personnel who participated in these multicenter insulin lispro studies: A. de Leiva, Barcelona, Spain; A. Estour, Saint-Etienne, France; A. Jara, P. Sanchez, Madrid, Spain; A. Karasik, Tel-Hashomer, Israel; A. Scheen, Liege, Belgium; A. Tiengo, Padova, Italy; A.A. Motala, Durban, South Africa; A.H. Barnett, Green East, U.K.; A.J. Garber, Houston, Texas; A.J. Nankervis, Parkville, Australia; A.J. Spijker, Roermond, the Netherlands; A.P. Roberts, Adelaide, Australia; A.R. Daniels, Auckland, New Zealand; A.V. Greco, Roma, Italy; B. Charbonnel, Nantes, France; B. Goldberg, Miami, FL; B. Schulze-Schleppinghoff, Essen, Germany; B.O. Boehm, Ulm, Germany; B.R. Zimmerman, Rochester, Minnesota; C. Coscelli, Parma, Italy; C. Rosak, Frankfurt, Germany; C.D. Agardh, Lund, Switzerland; C.M. Kesson, Glasgow, U.K.; D. Andreani, Rome, Italy; D. Maingay, Blaricum, the Netherlands; D. Fedele, Padova, Italy; D.H. Clarke, Salt Lake City, Utah; D.K. Yue, Camperdown, Australia; D.P. Cameron, Woolloongabba, Australia; D.S. Schade, Albuquerque, New Mexico; D.W. Erkelens, Utrecht, the Netherlands; E.J. Bastyr, III, Galveston, Texas; E.L. Toth, Edmonton, Canada; EL Vilardell, Barcelona, Spain; E.N. Janssens, Dordecht, the Netherlands; F. Fery, Brussels, Belgium; F. Stockmann, Goettingen, Germany; F.A. Gries, Auf’m Hennekamp, Germany; F.B. Bonnici, Cape Town, South Africa; F.J. Seif, Tügbingen, Germany; G. Bompiani, Palermo, Italy; G. Cathelineau, Paris, France; G. Crepaldi, Padova, Italy; G. Delluc, Perigueux, France; G. Pagano, Torino, Italy; G. Pozza, Milano, Italy; G. Schernthaner, Vienna, Austria; G. Somers, Brussels, Belgium; G. Menzinger, Rome, Italy; G.A. Hitman, London, U.K.; G.C. Ellis, Somerset West, South Africa; G.E. Storms, Bilthoven, the Netherlands; H.J. Hazenberg, Deventer, the Netherlands; H. Kaplan, Cape Town, South Africa; H. Schatz, Bochum, Germany; H..J. Ruessmann, Dinslaken, Germany; H.P. Chase, Denver, Colorado; H.U. Jastram, Kaiserslautern, Germany; I. Folling, Trondheim, North Dakota; I. Koop, Marburg, Germany; I.H. De Leeuw, Edegem, Belgium; I.M. Fettes, Toronto, Canada; J. Beyer, Mainz, Germany; J. Cassar, Isleworth, U.K.; J. Huisman, Enschede, the Netherlands; J. Mahoudeau, Caen, France; J. Viikari, Turku, Finland; J.A. Davidson, Dallas, Texas; J.A. Vazquez, I. Goikolea, Barakaldo, Spain; J.J. Altman, Paris, France; J.K. Wales, Leeds, U.K.; J.L. Chiasson, Montreal, Canada; J.L. Rosenzweig, Boston, Massachusetts; J.L. Selam, Paris, France; J.P. Lauvaux, Brussels, Belgium; J.P. O’Hare, Bath, U.K.; J.P. Riou, Lyon, France; J.R. Attali, Bondy, France; J.R. Patsch, Innsbruck, Austria; J.S. Soeldner, Sacramento, California; K. Federlin, Giessen, Germany; K. Huth, Frankfurt, Germany; K.E. Schroder, Duisburg, Germany; K.G. Petersen, Freiburg, Germany; K.M. Shaw, Cosham, U.K.; L. Blonde, New Orleans, Louisiana; L. Sacca, Napoli, Italy; L. Van Gaal, Edegem, Belgium; L.G. van Doorn, Tilburg, the Netherlands; L.I. Robertson, Durban, South Africa; M.C. Mengel, Orlando, Florida; M. Cohen, Caulfield, Australia; M. Frank, Hamburg/S, Germany; M. Haslbeck, Munich, Germany; M. Mancini, Napoli, Italy; M. Muggeo, Verona, Italy; M. Dreyer, Hamburg, Germany; M.A. Omar, Durban, South Africa; M.H. Borkenstein, Graz, Austria; M.H. Tan, Halifax, Canada; M.J. Prince, Indianapolis, Indiana; M.L. Reeves, Chattanooga, Tennessee; M.L. Spencer, Minneapolis, Minnesota; O. Lnenicka, Orebro, Switzerland; P. Diem, Bern, Switzerland; P. Drouin, Dommartinlestoul, France; P. Fossati, Lille, France; P. Salmela, Oulu, Finland; P.A. Boyce, Indianapolis, Indiana; P.B. Vialettes, Marseilles, France; P.F. Bougneres, Paris, France; P.J. Guillausseau, Paris, France; R. Astorga, Seville, Spain; R. Bouillon, Leuven, Belgium; R. Giorgino, Via Ennio, Italy; R. Landgraf, Munich, Germany; R. Lauro, Rome, Italy; R. Marechaud, Poitiers, France; R. Moore, Durban Natal, South Africa; R. Moore, Durban, South Africa; R. Vigneri, Catania, Italy; R.A. Guthrie, Wichita, Kansas; R.K. Mayfield, Charleston, West Virginia; R.M. Bergenstall, Minneapolis, Minnesota; S. Vaaler, Oslo, Norway; S. Weitzman, Beersheba, Israel; S.A. Arslanian, Pittsburgh, Pennsylvania; S.R. Heller, Sheffield, U.K.; T. Halmos, Budapest, Hungary; T.A. Welborn, Perth, Australia; V.A. Koivisto, Helsinki, Finland; W. Lourens, Pretoria, South Africa; W. Sturmer, Wuerzburg, Germany; W.D. Alexander, Sidcup, U.K.; W.F. Mollentze, Bloemfontein, South Africa; W.K. Waldhausl, Lasarettgasse, Austria.