Introduction
Diabetes is estimated to affect 425 million people and to cost US$727 billion in healthcare worldwide [
1]. The growing pandemic of type 2 diabetes accounts for approximately 90% of this burden [
1]. Type 2 diabetes is often associated with an increased risk for cardiovascular disease, which is the leading cause of morbidity and mortality in individuals with diabetes [
2‐
5]. Other common coexisting conditions, such as hypertension, dyslipidaemia and obesity, further exacerbate the risk for cardiovascular disease [
5]. Hence, guidelines for type 2 diabetes recommend that, in addition to controlling glucose levels, any risk factors for cardiovascular disease should be treated [
5].
Blood glucose-lowering agents, in combination with lifestyle intervention, lipid-lowering drugs and/or antihypertensives, are often used to manage individuals with type 2 diabetes who are at risk for cardiovascular disease [
5,
6]. More than 40% of patients receiving statin therapy for dyslipidaemia in type 2 diabetes fail to achieve a reduction in LDL-cholesterol levels by a target decrease of 30–50% [
5,
7,
8]. Similarly, more than 40% of patients with type 2 diabetes treated with blood glucose-lowering agents do not reach their glycaemic goals [
9,
10]. Therefore, many individuals with type 2 diabetes have a residual risk for cardiovascular disease, despite receiving the recommended therapies for dyslipidaemia and hyperglycaemia.
Proprotein convertase subtilisin/kexin type 9 (PCSK9) is an endogenous enzyme that promotes the degradation of LDL receptors on hepatocytes [
11]. Inhibiting PCSK9 activity has been shown to maintain the expression of LDL receptors, which, in turn, decreases serum levels of LDL-cholesterol [
11]. Currently, alirocumab and evolocumab are the only PCSK9 inhibitor antibodies approved for clinical use [
12,
13]. In type 2 diabetes, PCSK9 inhibitors are recommended as adjuncts to maximally tolerated statin therapy in patients with atherosclerotic cardiovascular disease who require additional lowering of LDL-cholesterol levels [
5]. Moreover, inhibition of PCSK9 has been shown to significantly decrease cardiovascular outcomes in patients with type 2 diabetes and atherosclerotic disease [
14]. Glucagon-like peptide-1 (GLP-1) is an endogenous hormone that stimulates insulin secretion in response to postprandial increases in glucose levels (i.e. the incretin effect) and slows gastric emptying [
15]. Approved GLP-1 receptor agonists or GLP-1 analogues, such as liraglutide and exenatide, are resistant to degradation by dipeptidyl peptidase-4 (DPP-4) and thus provide robust and longer-acting efficacy [
16]. GLP-1 receptor agonists or GLP-1 analogues are an attractive and recommended therapeutic option for type 2 diabetes because of their demonstrated effects on glycaemic control and weight loss, and positive impact on cardiovascular outcomes [
6,
17‐
19].
MEDI4166 is an antibody–peptide fusion molecule comprised of a PCSK9 antibody and a GLP-1 analogue linked to the N-terminus of the antibody light chain using a peptide linker (M. Chodorge, MedImmune, Cambridge, UK, personal communication). MEDI4166 was designed to combine the mechanisms of action of PCSK9 inhibitors and GLP-1 agonists/analogues, described above, to treat patients with type 2 diabetes who require additional control of blood glucose and LDL-cholesterol levels, with the goal of reducing overall cardiovascular risk. Here, we report results from a phase 1, first-in-human, combined single ascending dose and multiple ascending dose study that evaluated the safety, efficacy, pharmacokinetics and immunogenicity of MEDI4166 in overweight or obese participants with type 2 diabetes.
Methods
Study assessments
The primary endpoint of Part A was safety. The incidence of treatment-emergent adverse events (TEAEs) and serious adverse events was recorded. Analyses of TEAEs included the type, severity and relationship to the study drug as summarised by the Medical Dictionary for Regulatory Activities, version 19.1 (
www.meddra.org). The co-primary endpoints of Part B were change from baseline (day 1 pre-dose) to day 36 in LDL-cholesterol levels and glucose area under the plasma concentration–time curve from 0 h to 4 h (AUC
0–4h) post-mixed-meal tolerance test (MMTT).
In Part A, secondary efficacy endpoints included LDL-cholesterol levels and glucose AUC0–4h post-MMTT. In Part B, drug safety and the change in fructosamine levels from baseline were evaluated as secondary objectives. The pharmacokinetics and immunogenicity of MEDI4166 were assessed as secondary objectives in both parts of the study. The following pharmacokinetic variables were assessed: maximum observed plasma concentration (Cmax); time to Cmax (Tmax); terminal phase elimination half-life (T½); AUC from time 0 extrapolated to infinity (AUC0–inf); AUC from time 0 to the last measurable plasma concentration (Part A) or the end of the dosing interval (Part B) and apparent clearance (Part A) or apparent clearance at steady state (Part B). A sequential flow-through sandwich method on the Gyrolab System (Gyros Protein Technologies, Uppsala, Sweden), validated to current regulatory guidelines, was used to quantify serum levels of MEDI4166 in participants with type 2 diabetes. The method was specific and selective for MEDI4166, using anti-idiotype antibodies to MEDI4166 to capture and detect the analyte. Immunogenicity was evaluated by the incidence of treatment-emergent anti-drug antibodies to MEDI4166, which was defined as the sum of treatment-induced and treatment-boosted anti-drug antibody-positive responses. To test immunogenicity, blood samples were collected on days 1, 15, 29 and 43 in Part A and on days 1, 15, 29, 43 and 71 in Part B.
Safety was analysed in a subgroup of participants who had LDL-cholesterol levels <0.65 mmol/l at any point during either portion of the study. Exploratory endpoints assessed in Part B included free PCSK9 levels, fasting GLP-1 activity, insulin AUC post-MMTT, HbA1c levels, fasting blood glucose levels, body weight and other lipid variables (total cholesterol, HDL-cholesterol, triacylglycerols, lipoprotein[a] and apolipoproteins A1 and B).
Fasting GLP-1 activity was determined ex vivo against a human GLP-1 receptor-expressing Chinese hamster ovary (CHO) cell line (AstraZeneca, Cambridge, UK; mycoplasma free) in serum samples following an overnight fast of ≥8 h. Using a competitive homogenous time-resolved fluorescence (HTRF) assay (Cisbio, Codolet, France), changes in CHO intracellular cAMP levels resulting from MEDI4166-receptor interaction, were assessed following cellular lysis. Within the assay, unlabelled cAMP produced from the CHO cell line competes with a known concentration of cAMP–d2 to bind to anti-cAMP–cryptate. The level of HTRF between the d2 and cryptate molecules was indirectly proportional to the concentration of intracellular cAMP and, therefore, MEDI4166 levels. Unknown levels of MEDI4166 in samples were then quantified from a MEDI4166–agonist curve. This validated method [
20,
21] was selective and specific for MEDI4166 and presented acceptable accuracy (within 25%) and precision (with 30%) for cell-based bioassays (MedImmune, data not shown).
Experiments in cells were replicated more than 5 times. No randomisation was carried out and experimenters were not blinded.
Discussion
In this phase 1, randomised, placebo-controlled, double-blind study in overweight or obese participants with type 2 diabetes, the primary endpoint of Part A was met, but only one of the two co-primary endpoints in Part B was met. In Part A, the incidences of TEAEs were comparable between treatment arms and MEDI4166 was generally well tolerated at all tested doses. In Part B, a significant decrease from baseline to day 36 in LDL-cholesterol levels was observed at all tested doses of MEDI4166 vs placebo but no significant differences in glucose AUC0–4h post-MMTT were observed between treatment groups. Based on these results, further clinical development of MEDI4166 as a dual-targeted therapy for patients with type 2 diabetes who are at risk for cardiovascular disease was discontinued.
Most TEAEs observed with MEDI4166 were mild or moderate in severity. The incidence of gastrointestinal-related TEAEs with MEDI4166, such as vomiting and nausea, is consistent with the safety profile of other GLP-1 receptor agonists approved for type 2 diabetes [
23]. Furthermore, the TEAE of injection-site reactions aligns with results from previous clinical trials with approved GLP-1 receptor agonists [
24] and PCSK9-antibody inhibitors [
25]. No potential safety concerns with MEDI4166 were identified in the subgroup analysis of participants exhibiting very low levels of LDL-cholesterol following treatment. These results should, however, be interpreted with caution due to the relatively small sample sizes (typical of early-phase clinical trials).
The observed exposures to MEDI4166 at all tested doses were much lower compared with predictions based on preclinical studies in rats and cynomolgus monkeys. For example, in the preclinical studies, the observed vs predicted exposure following a single dose of MEDI4166 at 400 mg by AUC was 824.1 μmol/l × day vs 2228.1 μmol/l × day and by Cmax was 65.2 μmol/l vs 176.3 μmol/l (M. Chodorge, MedImmune, personal communication). The presence of anti-drug antibodies did not affect the safety, pharmacokinetics or pharmacodynamics of MEDI4166 compared with participants who did not have anti-drug antibody-positive responses. Overall, MEDI4166 was well tolerated and was associated with a pharmacokinetic profile potentially suitable for a once-weekly dosing regimen, despite being associated with a high degree of interindividual variability.
Robust and significant dose-dependent decreases in LDL-cholesterol levels were observed with single (33–63%) and multiple doses (44–68%) of MEDI4166. In addition to reduced LDL-cholesterol levels, at the higher dose levels of MEDI4166 there were significant decreases from baseline in total cholesterol, apolipoprotein B and lipoprotein(a) levels in Part B. The effect of MEDI4166 on LDL-cholesterol levels is comparable to effects seen in early-phase trials with other PCSK9 inhibitors. Specifically, dose-dependent decreases from baseline in LDL-cholesterol levels were observed with alirocumab (25–75%) in individuals with primary hypercholesterolaemia [
26] and with evolocumab (66%) in individuals with heterozygous familial hypercholesterolaemia [
27]. Moreover, both alirocumab and evolocumab significantly decreased other lipid variables, including total cholesterol, apolipoprotein B and lipoprotein(a) levels [
26,
27]. The efficacy of GLP-1 receptor agonists in improving glycaemic control and decreasing body weight has been clearly demonstrated in type 2 diabetes [
19,
28]. Treatment with MEDI4166, however, had no clinically relevant impact on postprandial glucose levels, fasting glucose levels, HbA
1c levels or body weight at any of the doses used in our study. The ability of MEDI4166 treatment to decrease LDL-cholesterol levels but not glucose levels may be explained by the decrease in free PCSK9 levels and lack of sufficient GLP-1 activity, respectively. Our results suggest that the effects of MEDI4166 on GLP-1 activity were inadequate for improving glycaemic control as measured by glucose AUC
0–4h post-MMTT. Conversely, the clear impact of MEDI4166 treatment on decreasing free PCSK9 levels parallels its robust effect of reducing LDL-cholesterol levels. It is unclear why MEDI4166 differentially affects LDL-cholesterol and glucose levels but this may be related to insufficient potency at the GLP-1 receptor combined with lower-than-expected exposure, as described below.
The tested doses of MEDI4166 in this study were based on a target-mediated drug disposition pharmacokinetic–pharmacodynamic model tailored for the study drug. Based on simulations using this model, MEDI4166 at the tested doses was expected to achieve ≥90% suppression of endogenous levels of PCSK9 levels, similar to that observed with alirocumab 150 mg (M. Chodorge, MedImmune, personal communication) [
29]. Moreover, these doses were chosen to obtain GLP-1 activity similar to that observed with daily liraglutide 1.8 mg and weekly dulaglutide 1.5 mg (M. Chodorge, MedImmune, personal communication) [
30]. Multiple doses of MEDI4166 at the highest dose tested (400 mg) suppressed PCSK9 almost completely and durably, and this translated to a sustained decrease in plasma LDL-cholesterol levels. However, the increase in GLP-1 activity observed with MEDI4166 did not lead to a clinically meaningful effect on glucose control. These results suggest an unbalanced dual pharmacology of MEDI4166 in humans that was not expected from early preclinical data. The pharmacokinetic–pharmacodynamic model was updated during the clinical study to determine feasibility for dose escalation. The lower-than-expected exposure, high interindividual variability, nonlinear pharmacokinetics and lack of glucose-lowering effects with MEDI4166 introduced a high level of uncertainty into simulated scenarios and did not support further dose escalation in the clinic (MedImmune, data not shown).
The original pharmacokinetic–pharmacodynamic model that was used to predict study doses was based, in part, on potency assessments conducted in the GLP-1 receptor cAMP accumulation CHO cell assay (for a brief description, please see
ESM Methods) [
31]. As MEDI4166 is an antibody–peptide fusion molecule, the measurements and dose–response relationship were compared with a GLP-1-Fc(γ4) fusion protein, synthesised in-house. In this GLP-1 receptor overexpressing CHO cell line, MEDI4166 demonstrated a half maximal effective concentration of 2560 pmol/l compared with 10.4 pmol/l for the reference compound GLP-1-Fc(γ4) (based on the half-maximal effective concentration; ESM Table
2). This corresponds to a 246-fold difference in potency between the two compounds (ESM Fig.
2a). However, in the EndoC-βH1 (human insulinoma) cell line, which express endogenous GLP-1 receptors, the potency of MEDI4166 at the human GLP-1 receptor was 472-fold lower than the GLP-1-Fc(γ4) (ESM Table
2 and ESM Fig.
2b). This suggests that the transfected GLP-1 receptor cAMP accumulation CHO cell assay, in which the original potency assessments were made and on which the pharmacokinetic–pharmacodynamic modelling was based, overestimated the potency of MEDI4166 due to a high level of GLP-1 receptor expression. The relatively more physiological cell line, EndoC-βH1, may have provided a more accurate estimate for the dose predictions in relation to GLP-1 activity. Taken together, these results suggest that, at the doses tested in this study, MEDI4166 demonstrated insufficient GLP-1 receptor stimulation and that substantially higher doses of MEDI4166 would be required to achieve clinical efficacy in terms of glycaemic control. Such high doses, however, would not be feasible in humans and, hence, further clinical development of MEDI4166 was terminated.
Overall, the safety and pharmacokinetic profiles of MEDI4166 supported once-weekly dosing. Although treatment with MEDI4166 robustly and significantly decreased LDL-cholesterol levels, there were no significant or clinically relevant reductions in postprandial glucose levels or sufficient increases in GLP-1 activity. A particular strength of this study was the inclusion of individuals with type 2 diabetes in both parts of the study, allowing efficient conclusions on the potential clinical utility of MEDI4166 in the target population to be made. Based on the results from this phase 1 combined single and multiple ascending dose study, further clinical development of MEDI4166 as a dual-targeted therapy for patients with type 2 diabetes who are at risk for cardiovascular disease was discontinued.
Acknowledgements
The authors thank W. Pu (formerly Clinical Biostatistics & Data Management at MedImmune LLC, Gaithersburg, MD, USA) for supporting the study design and statistical analyses for Part A of the study. The authors thank S. DiCostanza (Clinical Operations at MedImmune LLC, Gaithersburg, MD, USA) for study management and T. Yang (Clinical Biostatistics & Data Management at MedImmune LLC, Gaithersburg, MD, USA) for programming support. The authors acknowledge D. Hornigold (Cardiovascular, Renal, and Metabolism Research, MedImmune Ltd., Cambridge, UK) who contributed to the in vitro experiments described in the
ESM. Some data from Part A of this study was presented at the 53rd Annual Meeting of the European Association for the Study of Diabetes, Lisbon, Portugal (11–15 September 2017). Some data from Part B of this study were presented at the 54th Annual Meeting of the European Association for the Study of Diabetes, Berlin, Germany (1–5 October 2018). Editorial assistance was provided by V. R. Jayasinghe, of Oxford PharmaGenesis Inc., Newtown, PA, USA, and was funded by MedImmune LLC, Gaithersburg, MD, USA.
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.