A randomized pharmacokinetic and pharmacodynamic trial of two regular human insulins demonstrates bioequivalence in type 1 diabetes and availability of biosimilar insulin may improve access to this medication

Abstract

Aims:

To compare the pharmacokinetic (PK) and pharmacodynamic (PD) effects and safety of therapeutic dosages of a regular insulin (experimental drug) produced by Bioton S.A. (Warsaw, Poland) versus Humulin^® R, a regular insulin (reference drug) produced by Eli Lilly (Indianapolis, Indiana).

Materials and Methods:

In a single-centre, randomized, double-blinded phase 1 crossover study, we used the manual euglycaemic clamp technique to compare PK and PD profiles between single subcutaneous doses (0.3 units/kg) of the two regular insulins in participants with type 1 diabetes (T1DM) with a washout period of 14 (± 7) days between tests.

Results:

We evaluated 56 participants. The mean participant age and body mass index were 32.9 years and 22.9 kg/m², respectively. The ratios (experimental/reference) of the geometric means of maximum plasma insulin concentration and for plasma insulin area under the curve (AUC) were 0.909 (90% confidence interval [CI] 0.822-1.01) and 0.993 (90% CI 0.944-1.04), respectively. The ratios of the geometric means of maximum glucose infusion rate (GIR) and for GIR AUC were 0.999 (95% CI 0.912-1.09) and 1.04 (95% CI 0.962-1.12), respectively.

Conclusions:

The experimental product regular human insulin and comparator Humulin^® R are bioequivalent in patients with T1DM. Wider entry to the pharmaceutical market of affordable, biosimilar regular insulins may substantially improve access to insulin for many socioeconomically disadvantaged patients with diabetes.

1 ∣. INTRODUCTION

Insulin is an indispensable, life-saving medication for people with type 1 diabetes (T1DM) and for many individuals with type 2 diabetes (T2DM). As the incidence of both conditions continues to rise worldwide, access to this vital therapy has increasingly challenged healthcare systems globally.¹

Rapidly rising costs have hindered patients' ability to afford the insulin they require for survival, forcing many to ration their insulin or forgo other important needs.^2,3 Moreover, in low- and middle-income countries, where 79% of people with diabetes reside, just 55% to 80% of medication-dispensing facilities even have insulin available. Clearly, economically viable solutions are sorely needed to improve insulin access and affordability.^4,5

Some experts have favoured wider use of human insulin as a more cost-effective option to manage diabetes rather than the more-expensive insulin analogues.^6,7 Despite greater consumer demand for the convenience associated with insulin analogues, systematic reviews and meta-analyses have found only modest differences in glycaemia between rapid-acting insulin analogues and regular human insulin.^8-12

These factors have likely contributed to a high percentage of patients remaining on human insulin in many countries, particularly in low- and middle-income countries. Thus, the demand for regular human insulin remains robust across the international market, suggesting that increasing the supply of this medication may reduce its cost.¹³ Moreover, other experts have suggested biosimilar insulins may increase competition in the insulin manufacturing sector, leading to increased availability and affordability.^1,3 At present, however, only three manufacturers control 99% of the global insulin market in terms of value and 96% in terms of volume.¹³

In March 2020, the US Food and Drug Administration (FDA) implemented a new, more efficient pathway for approving biosimilar insulins with the intent of facilitating greater market competition. Currently, new pharmaceutical companies are just beginning to market biosimilar insulins in the European and North American markets and whether greater competition will lower the consumer cost for these medications remains to be seen.¹⁴

Guidance from the FDA and the European Medicines Agency define “biosimilarity” to mean that the biological product is highly similar to the reference product with no clinical differences in terms of safety, purity and efficacy. To demonstrate bioequivalence, pharmacokinetic (PK) and pharmacodynamic (PD) studies are needed.^15,16

Although recombinant human insulins produced by Bioton S.A. (Warsaw, Poland) are used under the brand name Gensulin^® or SciLin^® in several countries, studies to determine whether this medication satisfies current regulatory biosimilarity standards with human insulins approved in other markets are as-yet unpublished. Thus, in the present study we aimed to compare the PK and PD profiles of subcutaneously administered doses of a regular insulin (experimental drug) produced by Bioton S.A. versus Humulin^® R regular insulin (reference drug), produced by Eli Lilly (Indianapolis, Indiana), in participants with T1DM.

2 ∣. MATERIALS AND METHODS

2.1 ∣. Trial design

In this Phase 1, single-centre, randomized, double-blinded, crossover study, we used the manual euglycaemic clamp technique to compare PK and PD characteristics between two recombinant regular human insulins in 56 adult participants with T1DM. The reference human insulin tested was manufactured by Eli Lilly (Humulin^® R) and the experimental insulin was produced by Bioton S.A. Each drug was tested as a single subcutaneous dose of 0.3 units/kg in random order, with a washout period of 14 days (± 7 days). This study was developed and conducted by the Institute of Pharmaceutical Sciences of Studies and Research (ICF) in Aparecida de Goiânia, Brazil.

2.2 ∣. Participants

Participants were recruited from an ICF clinical unit database, through referral from local physicians, and from a local diabetes association. There were no restrictions on gender and ethnicity. Supplementary Table 1 details the inclusion and exclusion criteria for study participants. In brief, volunteers were aged between 18 and 65 years, had a T1DM duration of at least 1 year, a body mass index (BMI) between 18 and 30 kg/m², and either a positive anti-glutamic acid decarboxylase (GAD) antibody test result, a positive anti-islet antigen-2 (IA-2) antibody test result, or a fasting plasma C-peptide level less than 0.5 ng/mL (0.2 mmol/L). Key exclusion criteria were T2DM, current treatment with oral hypoglycaemic agents, pregnancy or breastfeeding.

2.3 ∣. Interventions

After selection, all participants underwent a run-in period aiming for the best possible glycaemic control, with three visits, one per week. Basal insulin was suspended 36 hours before the clamp study. The day before admission, the study team instructed participants to control blood glucose with ultra-rapid insulin according to capillary blood glucose. Participants were hospitalized in the clinical unit on the evening prior to study and consumed a standardized meal at approximately 8:00 pm. Thereafter, participants continued fasting until the end of the study. An intravenous catheter was placed in each forearm, one for insulin and glucose infusion and the other for blood sampling. After the meal, all participants received a continuous intravenous infusion of rapid-acting insulin (insulin aspart; Novorapid^® FlexPen^®, Novo Nordisk, Bagsværd, Denmark) that was titrated based on our previously published algorithm¹⁷ to maintain plasma glucose within a target between 90 and 1105 and 6.1 mmol/L mg/dL until the beginning of the study the next morning at 7:00 am. Before beginning the study, the participants’ glucose levels had to fall within the target range for at least 1 hour without requiring any glucose infusion.

Based on the randomized treatment order, participants then received a 0.3-U/kg subcutaneous injection of either of two insulins in the abdomen. The onset of insulin action was defined as the time when glucose infusion was started in response to a decrease in plasma glucose concentration of 0.27 mmol/L below the baseline. Thereafter, the glucose infusion rate (GIR) was titrated to maintain plasma glucose at a target of 5.5 mmol/L (5-6.1 mmol/L). Blood glucose was measured every 5 minutes during the first 6 hours of study and every 10 minutes thereafter. Glucose concentrations were measured in duplicate in real-time during the clamp using a point-of-care meter (Accu-Chek; Roche Diabetes Care, Indianapolis, Indiana). Plasma glucose concentrations were later confirmed using a COBAS analyser (Roche Diagnostics, Mannheim, Germany). Study personnel collected blood samples to measure PK variables and performed safety evaluations at prespecified time intervals. End of insulin action was defined as the time when plasma glucose concentration consistently exceeded 7.2 mmol/L in the absence of glucose infusion. The duration of action was calculated as the difference between end and start of action.

Blood glucose concentration (mmol/L) and GIR were duly recorded in each participant’s source document, medical record and case report form. There were no important changes to study procedures after trial commencement.

2.4 ∣. Assays

Blood samples for the measurement of glucose, hormone and antibody levels were collected through venipuncture in the forearm, with a closed vacuum collection system. Glucose was collected in a vacuum tube with sodium fluoride anticoagulant and the measurement was performed by enzymatic methodology with hexokinase after plasma centrifugation in the COBAS immunoassay analyser. Plasma insulin and C-peptide concentrations were assayed with an enzyme-like immunosorbent assay (ELISA), by auto-DELPHIA automatic fluoroimmunoassay (Wallac Inc., Turku, Finland).

Analysis of GAD and IA-2 antibodies was performed using an ELISA (EUROIMMUN AG, Lübeck, Germany). Anti-insulin antibody testing was performed with a NEXGEN 5 automatic analyser ELISA (Vyttra Diagnostics, São Paulo, Brazil).

The PK analysis of insulin was performed using high-performance liquid chromatography coupled with mass spectrometry (HPLC-MS/MS). For this, we used the Agilent 1290 series chromatograph (Agilent Technologies, Santa Clara, California) and the spectrometer used was the Sciex API5500 (SCIEX, Redwood City, California, EUA).

We used ELISA methodology for plasma insulin measurement at the initial and final visits, as part of the protocol-determined laboratory routine of the study. HPLC-MS/MS was used for PK measurements, as it has a higher sensitivity and specificity, avoiding cross-reaction with other insulins. We developed an HPLC method to quantify regular insulin. We validated the method using a selectivity test, aiming to verify the method’s ability to quantify regular insulin in the presence of insulin aspart and NPH insulin. For this, serum samples were contaminated with insulin aspart or NPH insulin, submitted to the extraction process, and analysed using the regular insulin quantification method. The results showed no interference of any tested insulin when quantifying human regular insulin.

2.5 ∣. Outcomes

This study aimed to compare the PK and PD profiles of two recombinant regular human insulins, namely, a regular human insulin produced by Bioton (experimental insulin) and Humulin^® R (reference insulin), in patients with T1DM. The PK variables analysed included area under the insulin concentration-time curve from time zero to the end of the clamp (AUC_0–T) and maximum insulin concentration (C_max). For the primary PK endpoint, we determined whether the 90% confidence intervals (CIs) of the ratio (reference/experimental) of geometric means for C_max and AUC_0–t were completely within 0.80 to 1.25. The PD variables analysed included area under the GIR-time curve from time zero to the end of clamp (AUC_GIR0–T) and maximum rate of glucose infusion (GIR_max). For the primary PD endpoint, we determined whether the 95% CIs for the ratio (reference/experimental) of geometric means for GIR_max and AUC_GIR0–T were completely bounded by 0.80 to 1.25.

Secondary endpoints included safety and tolerability after a single dose of either recombinant regular human insulin.

2.6 ∣. Randomization and blinding

The primary investigator, nursing team, and research participants were blinded to which medication was given during the PK and PD studies. The experimental and reference insulins were only handled inside the pharmacy by non-blinded personnel immediately prior to study and were administered according to a randomization list. No need arose to break the blinding code during the study.

2.7 ∣. Statistical methods

The sample size was determined by calculating the power function, which is based, among other factors, on an estimate of the intra-individual variation coefficient obtained from data published in the literature or from previous experiences with the analyte. Based on the literature available, 50 participants were expected to be sufficient to reject the null hypothesis of bio inequivalence (H₀), with a minimum power of 80%.

To test for equivalence, we applied the two one-sided tests (TOST) procedure. We used ANOVA to compare the means of each group.

The CI values were calculated for the mean difference in primary variables between treatments. For the PD analysis, a 95% CI was considered and, for the PK analysis, a CI of 90%. We determined the arithmetic difference for each outcome between experimental and reference drugs. Then, we transformed those differences into log values. Next, we determined the arithmetic mean of the log-transformed values. Finally, we calculated the antilog of this mean to obtain the geometric mean. Ratios of geometric means are considered bioequivalent if these limits are greater than 80% and less than 125%. Data are summarized as means and standard deviations, unless otherwise specified.

We calculated three measures to assess the quality of clamp studies, as described by Benesch et al.¹⁸ To determine “precision”, we calculated the coefficient of variation for the COBAS plasma glucose measurements during each clamp as 100% × [SD plasma glucose (mmol/L)/mean plasma glucose (mmol/L)]. To evaluate the “accuracy” of each clamp we computed the mean absolute relative difference (MARD) from the target glucose concentration of 5.5 mmol/L as 100% ×∣ plasma glucose (mmol/L) −5.5 mmol/L∣/5.5 mmol/L. To evaluate the “control deviation” in each study, we determined the mean difference between the plasma glucose concentrations and the target concentration of 5.5 mmol/L.

2.8 ∣. Study approval

The study was approved by local institutional review board and by the Brazilian National Agency, ANVISA, before the first consultation, and registered at Plataforma Brazil (CAAE: 79890917.1.0000.5572). The protocol was registered on the International Clinical Trials Registry Platform (ICTRP)/REBEC website http://www.ensaiosclinicos.gov.br/ with the UTN code: U1111-1208-3749. The study was performed in accordance with the International Conference on Harmonization of Technical Requirements for Registration of Pharmaceuticals for Human Use Guidelines for Good Clinical Practice and adhered to the principles of the Declaration of Helsinki. All subjects read, understood, and signed the informed consent form agreeing to participate before any procedure.

3 ∣. RESULTS

A total of 64 volunteers were recruited. After Visit 1, two participants withdrew informed consent, one participant broke his arm in an accident, and another was diagnosed with breast cancer. At the first clamp study, four participants tested positive for illicit drugs at admission and were excluded. Fifty-six participants were randomized, and analysis was conducted in 53 (Figure 1). Studies were conducted between September 26, 2018 and June 24, 2019.

FIGURE 1.

Study flow diagram

3.1 ∣. Demographics and baseline characteristics

Among the 56 participants, 29 were women (51.8%). The participants' mean age (SD) was 32.9 (9.5) years, mean height was 168 (11.2) cm, mean weight was 65.2 (13.8) kg, and mean BMI was 22.9 (1.4) kg/m². The mean age of participants at T1DM diagnosis was 17.6 (9) years. The mean diabetes duration at first visit was 14.9 (9) years. Table 1 summarizes participant demographics.

TABLE 1.

Patient demographic at baseline (N = 56)

Characteristic
Gender: female, %	51.8
Age, years	32.9 (9.5)
Height, cm	168 (11.2)
Weight, kg	65.2 (13.8)
BMI, kg/m2	22.9 (1.4)
T1DM diagnosis, years	17.6 (9)

Notes: Data are presented as mean (SD).

Abbreviations: BMI, body mass index; T1DM, type 1 diabetes.

One participant used a continuous insulin infusion pump, but all the others were on a basal-bolus therapeutic regimen. Insulin glargine was used by 35 participants (62.5%), NPH was used by 17 (30.4%), insulin degludec by two (3.6%) and insulin detemir by one (1.7%).

The mean (SD) C-peptide level was 0.13 (0.19) ng/mL. Anti-insulin antibody results were negative in 23% of the participants. Glycaemic control was poor, with a mean fasting glycaemia of 11.8 (6.5) mmol/L and a mean glycated haemoglobin (HbA1c) of 76 (−1) mmol/mol.

3.2 ∣. Pharmacokinetics

Similar PK results were found in the two human insulins tested, and the ratios of the geometric means for AUC_0–t and AUC_0–inf were well centred (Table 2). Insulin concentrations were similar throughout the two studies, as plotted in Figure 2A.

TABLE 2.

Pharmacokinetic characteristics

Statistical results: PK variables	Cmax, pg/mL	AUC0–t, pmol h/L	AUC0–inf, pmol h/La
Geometric means obtained by least squares method
Comparator (A)	3647.68	15 335.24	19 917.03
Test (B)	3317.25	15 221.06	20 969.25
CIs (shortest) for the between-treatments ratio (transformed data)
Contrast	CI_90%	CI_90%	CI_90%
B vs. A	92.20-100.61	94.43-104.33	90.79-122.09
Between-treatment ratio, %
B/A	90.94	99.26	105.28
Posteriori A test power (%): calculated via TOST method
B/A	67.42	100.00	52.96
Mean square error (error variance)
Between-subject	0.1897	0.2032	0.1585
Intra-subject	0.0908	0.0221	0.1916
Variance coefficient, %
Between-subject	45.71	47.47	41.45
Intra-subject	30.83	14.94	45.95
P values obtained for ANOVA fixed effects (sequential)
Sequence	0.4458	0.3051	0.953
Treatment	0.1217	0.8025	0.5562
Period	0.2208	0.1075	0.3779

^aCalculated by a linear extrapolation of the last concentrations.

Abbreviations: AUC_0–inf, area under the zero to infinity time curve; C_max, maximum insulin concentration; CI, confidence interval; PK, pharmacokinetic; TOST, two one-sided tests.

FIGURE 2.

Pharmacokinetic (PK)/pharmacodynamic (PD) results. (A) Insulin concentrations in the PK study. (B) Glucose concentrations in the PD study. (C) Glucose infusion rate (GIR) in the PD study. Data are summarized as medians and 95% confidence intervals. Black dots correspond to Eli Lilly and white dots to Bioton insulin data

The geometric averages for insulin C_max were 3317.25 pg/mL for the reference (Eli Lilly) insulin and 3647.68 pg/mL for the experimental (Bioton) insulin which corresponded to a between-treatment ratio of 0.90 (90% CI 0.92-1.01). Insulin AUC_0–t was 15 221.06 pmol h/L for the Eli Lilly insulin and 15 335.24 pmol h/L for the Bioton insulin, corresponding to a between-treatment ratio of 0.99 (0.94-1.04). Thus, although insulin C_max was slightly lower for the Eli Lilly than the Bioton insulin, the insulin AUC_0–t was very similar in the two regular insulins. Moreover, the 90% CI for the ratios of the geometric means for C_max, AUC_0–t and AUC_0–inf were each within 0.80 to 1.25, consistent with PK bioequivalence criteria.

3.3 ∣. Pharmacodynamics

Subcutaneous injections of 0.3 U/kg of Bioton and Eli Lilly human insulin formulations resulted in similar PD characteristics (Table 3 and Figure 2). The ratios of the geometric means for each of the PD variables evaluated (GIR_max, AUC_GIR0–T, AUC_GIR0–4 and AUC_GIR0–6) were well centred.

TABLE 3.

Pharmacodynamic characteristics

Statistical results: PD variables	GIRmax, mg/kg/min	AUC_GIR0–t, mg/kg	AUC_GIR0–4, mg/kg	AUC_GIR0–6, mg/kg
Geometric means obtained by the least squares method
Experimental (A)	66.84	251.76	137.08	211.36
Reference (B)	66.75	260.82	138.76	212.74
CI (95%) for the ratio between treatments (transformed data)
CI (95%)	91.23-109.32	96.25-111.50	89.87-114.01	92.36-109.70
Between-treatment ratio, %
B/A	99.86	103.60	101.22	100.66
Posteriori A test power (%): calculated via TOST method
B/A	99.88	99.97	95.75	99.94
Mean square error (error variance)
Between-subject	0.1066	0.1607	0.1652	0.1602
Intra-subject	0.0559	0.0369	0.0967	0.0479
Variance coefficient, %
Between-subject	33.53	41.75	42.39	41.69
Intra-subject	23.98	19.40	31.86	22.16
P values obtained for ANOVA fixed effects (sequential)
Sequence	0.6823	0.4798	0.7912	0.7742
Treatment	0.9545	0.3726	0.8608	0.9155
Period	0.1414	0.0007	0.1159	0.0182
P values obtained for ANOVA fixed effects (partial)
Sequence	0.6823	0.4798	0.7912	0.7742
Treatment	0.9760	0.3393	0.8381	0.8792
Period	0.1414	0.0007	0.1159	0.0182

Abbreviations: CI, confidence interval; GIR, glucose infusion rate; PD, pharmacodynamic; TOST, two one-sided tests.

The geometric mean for GIR_max was 66.75 mg/kg/min for the Eli Lilly and 66.84 mg/kg/min for the Bioton insulin, yielding a between-treatment ratio of 1.00 (95% CI 0.91-1.09). The geometric mean for AUC__GIR0–6 was 212.74 mg/kg and 211.36 mg/kg for the Eli Lilly and Bioton insulins, respectively, corresponding to a between-treatment ratio of 1.01 (95% CI 0.92-1.10). Thus, for each PD characteristic, the 95% CI of the ratios of the geometric means was bounded by 0.80 to 1.25, consistent with PD bioequivalence.

The estimates for intra-individual coefficients of variation varied between 19% and 32%, with the minimum value being observed for AUC_GIR0–T, and the maximum value observed for AUC_GIR0–4.

Consequently, the test power obtained via the TOST method for all variables was greater than 95%. Regarding the P values obtained in the ANOVA model, there were no statistical differences for the sequence and treatment factors, respectively, at the level of 10% and 5% significance, in any of the evaluated characteristics.

3.4 ∣. Clamp quality variables

The coefficients of variation of plasma glucose measurements (ie, “precision”) were 10.4 (2.3)% for the Eli Lilly insulin studies and 10.5 (2.6)% for the Bioton insulin studies. The MARDs for plasma glucose levels and the target glucose of 5.5 mmol/L (ie, “accuracy”) was 2.3 (1.8)% and 1.6 (2.0)% for the Eli Lilly and Bioton insulin studies, respectively. Similarly, the mean differences between plasma glucose concentrations and the target glucose were 0.12 mmol/L (0.1) and 0.08 mmol/L (0.11) for the Eli Lilly and Bioton studies, respectively.

3.5 ∣. Adverse events

The study team monitored participants for potential adverse events and none were considered related to the study medications. Adverse events are listed in Supplemental Table 1.

4 ∣. DISCUSSION

The present randomized study demonstrated PK and PD bioequivalence between regular insulin produced by Bioton (experimental drug) and Humulin^® R regular insulin produced by Eli Lilly (reference drug) in T1DM using the “gold standard” euglycaemic clamp method. We also demonstrated safety and tolerability after a single subcutaneous dose of each drug.

As insulin costs around the world continue to increase, wider entry to the market of biosimilar human insulins may present a cost-effective solution, improving access to affordable insulin, particularly where resources are scarcer. In an analysis of insulin prices in 13 low- and middle-income countries, Ewen et al found the lowest paid unskilled government worker must work 6.1 to 7.9 days to purchase a 10-mL vial of analogue insulin. Further, purchasing the same amount of less-expensive human insulin still required 3.5 to 3.9 days' wages.⁵ Even in high-income countries, high costs have contributed to insulin rationing among many patients. Herkert et al found that one-quarter of patients with T1DM and T2DM at the Yale Diabetes Center reported cost-related insulin underuse. Those reporting cost-related rationing were three times more likely to have poor glycaemic control than those who did not.¹⁹

In Brazil, a country with the fourth-largest number of patients with diabetes, insulin access, glycaemic control, and medication adherence are major public health concerns. Although healthcare in Brazil is delivered free of charge and universally by the public system, health quality performance indicators are below preferred levels, with a national median score of 5.4 on a scale of 0 to 10. It is estimated that a Brazilian worker with T1DM earning minimum wage can spend up to 75% of his or her salary on the treatment expenses.²⁰ The Brazilian Type 1 Diabetes Study Group found that fewer than 20% of patients were meeting glycaemic goals and that lower economic status was a major barrier to medication adherence.^21,22 Similarly, participants in our study had an average HbA1c of 76 mmol/mol at baseline, exceeding the recommended target of 7%. As with other low- and middle-income countries, the entry of more affordable biosimilar human insulin would likely alleviate a key financial barrier that limits medication access and adherence. In one analysis, the entry of Gensulin to the Polish market in 2001 brought about a public resource savings of 2.0 billion Polish złotys by 2013 (486 million Euro in 2013).²³ In Poland, where patients pay the difference between the third-party reimbursement limit and the price of the drug, human insulin copay costs in 2017 were 3.6-fold and 1.5-fold lower with Gensulin (Bioton S.A.) versus human insulins from Novo Nordisk and Eli Lilly. Thus, these experiences provide early evidence that increased market competition is associated with a reduction in the cost of insulin to individual patients, third party payers, or both.

Although rapid-acting insulin analogues were designed to more closely mimic the PK/PD characteristics of endogenous insulin,^24-26 some authors question whether the benefits of more-rapid insulins justify the substantially greater costs of these medications in patients with T1DM.⁶ In one Cochrane systematic review, the authors examined whether short-acting insulin analogues were more useful than regular human insulin for adults with T1DM.¹⁰ After reviewing nine randomized trials involving 2693 participants, the analysis suggested only a minor benefit of short-acting insulin analogues with regard to blood glucose control, with a mean difference in HbA1c of just −22 mmol/mol in favour of insulin analogues. Similarly, in a recent meta-analysis of 22 randomized controlled trials including 6235 patients with T1DM, Melo et al found a modest mean difference in HbA1c of −22 mmol/mol favouring insulin analogues.⁸

Along similar lines, another Cochrane review compared the glycaemia-lowering effects in 10 randomized trials of rapid-acting insulin analogues versus regular human insulin in nonpregnant adults with T2DM and found a mean difference in change in HbA1c of only −23 mmol/mol.²⁷ Likewise, Davidson⁶ compared human versus analogue insulins in an analysis of 60 randomized control trials involving 21 534 patients with T1DM and T2DM and found minimal differences in efficacy and only minor differences in hypoglycaemia. Particularly in the absence of glucose self-monitoring, the benefits of insulin analogues are believed to be minimal.²⁸

Interestingly, Melo et al found monthly rates of total and severe hypoglycaemic episodes were 7% and 32% lower, respectively, in patients with T1DM using short-acting insulin compared with patients using regular human insulin. This robust reduction in hypoglycaemia was not seen in previous systematic reviews,^10,12 possibly because advancements in home glucose monitoring have enhanced the ability of short-acting insulin analogues to decrease hypoglycaemic exposure. Additionally, many patients with T1DM prefer short-acting analogues over regular insulin because of increased mealtime flexibility.¹² Thus, healthcare leaders must consider these important differential costs and benefits in choosing between regular human insulin and short-acting insulin analogues.

Insulin PK/PD clamp studies have investigated a wide range of doses, with very low doses such as 0.05 U/kg to doses of at least 1.4 U/kg.²⁹ While high doses have posed challenges for investigators to maintain high-quality clamp studies (particularly with rapid-acting insulin), low doses could provoke minimal responses, substantially increasing the variability both between subjects and within subjects studied repeatedly.¹⁸ Thus, although there is no consensus regarding the dose of subcutaneous insulin to use in PK/PD clamp studies, we felt the 0.3-U/kg dose of human regular insulin optimized the need for a detectable response while remaining within a clinically relevant dosing range. Additionally, we note that current clinical study guidelines for determining biosimilarity between insulins recommend testing participants who are either healthy volunteers or patients with T1DM.³⁰ Thus, separate biosimilarity studies in patients with T2DM are not necessary in most cases.

Our study population demonstrated chronically suboptimal glycaemic control. According to Gardner et al, acute hyperglycaemia can potentially prolong insulin action time.³¹ To minimize this potential confounder, glycaemia was optimized in all patients during a run-in period. Additionally, participants were hospitalized overnight to receive a variable intravenous insulin infusion targeting a fasting plasma glucose level between 5 and 61. mmol/L 1 hour before beginning the clamp study.¹⁷

A strength of our study is the use of the euglycaemic clamp technique combined with a crossover design to precisely quantify PK and PD variables in patients with T1DM. This approach allowed us to show that there were no clinically meaningful differences between the two formulations of insulin. Further, because individuals with T1DM have little-to-no endogenous insulin secretion, the outcomes measured in the two insulins most likely reflected the effects of the exogeneous insulin per se, without the confounding effect of endogenous insulin.

In conclusion, this study demonstrates the PK and PD bioequivalence of recombinant regular human insulin produced by Bioton S.A. to the reference brand. This study may represent an important step towards improving insulin access and affordability for patients with diabetes worldwide.

DATA AVAILABILITY STATEMENT

Data available on request due to privacy/ethical restrictions