A population pharmacokinetic model of SC insulin therapy for hyperglycemia during acute ischemic stroke among patients with pre-existing diabetes

Background: Regardless of pre-stroke diabetes status, patients admitted to the hospital for acute ischemic stroke are frequently observed to have elevated levels of blood glucose or hyperglycemia[1]. Hyperglycemia is associated with poor stroke outcomes[2]. Insulin therapy to achieve normoglycemia during the acute phase is associated with better clinical outcomes in other groups of critically ill patients[3].In contrast, no conclusive evidence exists that normalizing serum glucose during the acute phase of ischemic stroke improves outcomes[1]. Not stratifying patients according to their pre-stroke glycemic status[4] and differential pathophysiology of insulin regulation in both pre-stroke diabetic and non-diabetic patients during the acute phase[5] are proposed mechanisms to explain the inability to demonstrate expected benefit. The two groups had similar degrees of hyperglycemia but very different clinical outcomes[6]. In general, hyperglycemia in diabetic patients is secondary to a relative deficiency of insulin, whereas, in non-diabetic patients, it was suggested that they are not in a hypo-insulinemic state. Still, instead, hyperglycemia is part of the systemic stress response[7]. Thus, these may affect insulin pharmacokinetics, requiring different approaches to treating hyperglycemia in those groups. By understanding the differences between the pharmacokinetic behaviors of administrated insulin during AIS hyperglycemia in both populations, an appropriate clinical decision could be made to improve the treatment plan and, therefore, improve stroke outcomes.

Aims: To develop a pharmacokinetic model of SC insulin therapy for hyperglycemia during acute ischemic stroke among stroke patients with pre-existing diabetes.

Methods: This prospective study was conducted in two-stroke centers in Malaysia. Data from acute ischemic stroke patients who developed hyperglycemia within 72 hours of admission and treated with s/c insulin were collected. Ischemic stroke patients with or without pre-existing diabetes were recruited. Pre-existing diabetes status was confirmed by A1C level ≥ 6.5%. At least three blood samples were collected sparsely at 0hr, 0.5hr, 1.5hr, 3hr, 4hr, 6hr, and 8hr of the s/c insulin therapy. Population PK model of the s/c insulin was developed in both populations. One and two-compartment models with additive, exponential, and additive exponential error models and lag absorption models were tested through baseline model development, and potential covariates were added to improve the maximum likelihood estimation, good-to-fit plots, and predictive virtual checks. The model was validated using the sample importance resampling (SIR) technique. NONMEM v. 7.4 and PsN v.5.3.1 were used. Only the PK model of insulin in pre-existing diabetes was reported here.

Results: A compartment model with first-order absorption and elimination and an additive with an exponential error model for 53 diabetic patients were selected as the baseline model. Population parameters for the baseline model were: Ka: 0.805 h-1, Cl/F: 29.2 L/h, V/F: 162 L, endogenous insulin concentration: 8.93 µIU/mL—additive residual error: 10.1 and proportional residual error: 0.214. Random interindividual variability was below 50% for Ka and V/F, while it reached 112.2% on Cl/F. The mean age of 60 increased the absorption rate by 88% when it raised to a power of 7.32. In patients with concomitant hypertension, there was an exponential decrease in the clearance with a magnitude of effect -1.47. Full model retained with Ka: 1.54 h-1, Cl/F 72.2 L/h, V/d 243 L, endogenous insulin concentration: 7.76 µIU/mL, additive residual error: 9.5, proportional residual error: 0.189. Random IIV effects were reduced to a negligible value of 0.003% for Ka, 99% for Cl/F, and 53.2% for V/F. Relative standard errors (RSEs) were precisely estimated below 30% for fixed effect parameters and below 50% for ETAs.

Conclusions: Age and concomitant hypertension significantly affect the PK of regular s/c insulin in acute ischemic stroke state and determine an optimal insulin dose for acute hyperglycemia in this population. We are still developing the non-diabetic patients’ model, and later, we will link both models with the integrated glucose-insulin (IGI) model for the associated pharmacodynamic properties on glucose clearance and volume of distribution. This model will be instrumental in predicting personalized insulin doses based on the patient’s diabetes status and glucose regulation.


  1. Johnston, K.C., et al., Intensive vs Standard Treatment of Hyperglycemia and Functional Outcome in Patients With Acute Ischemic Stroke: The SHINE Randomized Clinical Trial. JAMA, 2019. 322(4): p. 326-335.
  2. Tsivgoulis, G., et al., Association of Baseline Hyperglycemia With Outcomes of Patients With and Without Diabetes With Acute Ischemic Stroke Treated With Intravenous Thrombolysis: A Propensity Score-Matched Analysis From the SITS-ISTR Registry. Diabetes, 2019. 68(9): p. 1861-1869.
  3. Krinsley, J.S., et al., Continuous glucose monitoring in the ICU: clinical considerations and consensus. Crit Care, 2017. 21(1): p. 197.
  4. Long, M.T., D.B. Coursin, and J.S. Krinsley, Treatment of Hyperglycemia in Patients With Acute Stroke. Jama, 2019. 322(22): p. 2248.
  5. Reshi, R., et al., Hyperglycemia in acute ischemic stroke: Is it time to re-evaluate our understanding? Med Hypotheses, 2017. 107: p. 78-80.
  6. Zhu, B., et al., Stress Hyperglycemia and Outcome of Non-diabetic Patients After Acute Ischemic Stroke. Frontiers in Neurology, 2019. 10.
  7. Tziomalos, K., et al., Stress hyperglycemia and acute ischemic stroke in-hospital outcome. Metabolism, 2017. 67: p. 99-105.