Introduction

Closed loop insulin delivery systems, often referred to as artificial pancreas systems, combine continuous glucose monitors (CGMs), insulin pumps, and control algorithms to automate insulin delivery in real time. These systems have demonstrated significant improvements in glycemic control, reducing time in hyperglycemia and hypoglycemia while increasing time in range for people with type 1 diabetes. However, their performance depends heavily on how quickly and predictably insulin acts after it is delivered. This is where pharmacokinetics—the study of drug absorption, distribution, metabolism, and excretion—becomes central. Understanding the pharmacokinetic profile of insulin allows engineers to build more accurate algorithms and enables clinicians to optimize therapy settings. This article explores the key pharmacokinetic principles that underpin closed loop systems and highlights how variations in insulin action affect system performance.

What Is Pharmacokinetics?

Pharmacokinetics (PK) describes what the body does to a drug over time. For insulin, the four main processes are:

  • Absorption – The rate at which insulin enters the bloodstream from the subcutaneous tissue.
  • Distribution – The spread of insulin throughout the body, binding to insulin receptors on target tissues (muscle, fat, liver).
  • Metabolism – The breakdown of insulin, primarily by insulin-degrading enzymes in the liver, kidneys, and peripheral tissues.
  • Excretion – Elimination of insulin and its metabolites, mainly via the kidneys.

In the context of diabetes management, the most clinically relevant aspect of PK is the time-action profile—how quickly insulin onset occurs, when it peaks, and how long it lasts. For closed loop systems, a rapid and reproducible time-action profile is essential because the algorithm continuously computes insulin doses based on recent glucose readings and predicts future needs. Any lag between subcutaneous insulin delivery and measurable glucose decline can lead to overshoot or undershoot of glucose targets.

Insulin Pharmacokinetics in Closed Loop Systems

Absorption

Subcutaneous injection or infusion remains the standard route for insulin delivery in closed loop systems. Absorption kinetics are influenced by several variables: insulin formulation, injection site, local blood flow, tissue composition, and the presence of lipohypertrophy (lumpy areas of fat accumulation from repeated injections). Rapid-acting insulin analogs such as lispro, aspart, and glulisine are designed to absorb quickly, with onset times of approximately 10–20 minutes and peak action at 1–2 hours. Newer ultra-rapid formulations (e.g., Fiasp, Lyumjev) use excipients like nicotinamide or treprostinil to accelerate absorption further, achieving onset as early as 4–8 minutes. In closed loop systems, faster absorption reduces the delay between algorithm-commanded delivery and the resulting glucose-lowering effect, allowing tighter control and less post-meal hyperglycemia.

Absorption rate also depends on the volume of insulin delivered. Large boluses for meals may be absorbed more slowly than small correction doses. Continuous subcutaneous insulin infusion (CSII) via pump uses a constant basal rate, but the total daily volume is moderate. Advances in microdosing and concentrated insulins (U-200, U-300) are being explored to minimize volume and improve absorption consistency.

Distribution and Action

Once absorbed, insulin enters the portal and systemic circulations. Unlike endogenous insulin secreted directly into the portal vein, subcutaneously administered insulin first enters the systemic circulation, leading to a delayed and less physiological distribution to the liver. This "peripheral hyperinsulinemia" is a known limitation, but closed loop algorithms can partially compensate by adjusting the timing and amount of insulin delivery. The distribution half-life of insulin is short—around 4–6 minutes—because insulin quickly binds to receptors and is cleared by hepatic and renal routes. The pharmacodynamic (PD) effect, however, lasts longer because receptor-bound insulin continues to promote glucose uptake even after free insulin has been metabolized. For closed loop systems, the combined PK/PD profile must be modeled accurately to predict glucose trajectory.

Metabolism and Excretion

Insulin is primarily metabolized in the liver (about 50–60%) and kidneys (30–40%), with the remainder broken down in peripheral tissues. The metabolic clearance rate is influenced by hepatic blood flow, renal function, and the presence of insulin antibodies. In patients with chronic kidney disease, insulin clearance is reduced, leading to prolonged action and increased risk of hypoglycemia. Closed loop systems need to account for such variations; some algorithms incorporate patient-specific parameters (e.g., total daily dose, body weight, renal function) to adjust model predictions. The excretion of intact insulin via urine is minimal, but when kidney function declines, the half-life can double, requiring careful dose reduction.

Key Insulin Formulations for Closed Loop Systems

Rapid-Acting Analogs

The mainstays of closed loop insulin therapy are rapid-acting analogs: insulin lispro (Humalog), insulin aspart (NovoLog), and insulin glulisine (Apidra). All three have similar PK/PD profiles: onset 10–20 min, peak 1–2 h, duration 3–5 h. These are suitable for most systems, but subtle differences exist. For example, glulisine has a slightly faster onset in some studies, but the variability is high. More recent formulations like Fiasp (faster aspart) and Lyumjev (ultra-rapid lispro) have been developed specifically to reduce the lag time. Fiasp contains L-arginine and nicotinamide to enhance absorption; Lyumjev uses treprostinil, a prostacyclin analog that increases local blood flow. Clinical trials have shown that using ultra-rapid insulins in hybrid closed loop systems reduces postprandial glucose excursions and improves time in range by 5–10% compared to standard rapid analogs.

Concentrated Insulins

Insulin U-200 (Humalog 200) and U-300 (Toujeo, Basaglar) are used primarily for basal therapy in injections, but in pump-based closed loop systems, U-100 remains standard because pumps are calibrated for that concentration. However, concentrated insulins may reduce infusion set occlusion frequency at high delivered volumes. Newer investigational formulations aim to combine ultra-rapid action with prolonged stability for extended wear pump reservoirs.

Comparison of Onset, Peak, and Duration

A standard rapid analog takes about 10–20 min to show glucose-lowering effect, peaks at 60–90 min, and returns to baseline by 4–5 h. Ultra-rapid formulations show glucose effect within 4–8 min, peak at 40–60 min, and have a slightly shorter duration (3.5–4 h). This shorter duration can be an advantage in closed loop systems because it reduces the risk of "stacking" when multiple correction boluses are given. However, the faster peak also requires the algorithm to be more responsive, as the glucose nadir occurs earlier. The table below summarizes key profiles (approximate values from clinical trials):

  • Regular insulin (U-100): onset 30–60 min, peak 2–4 h, duration 6–8 h
  • Lispro/Aspart/Glulisine: onset 10–20 min, peak 1–2 h, duration 3–5 h
  • Fiasp (Faster Aspart): onset 4–8 min, peak 45–75 min, duration 3–4 h
  • Lyumjev (Ultra-Rapid Lispro): onset 4–8 min, peak 40–60 min, duration 3–4 h

These values are averages; individual variability is substantial due to the factors described below.

Factors Influencing Insulin Pharmacokinetics

Injection Site

Absorption rate varies by injection site: abdomen provides the fastest and most consistent absorption, followed by arms, thighs, and buttocks. For pump users, the infusion set is typically placed in the abdomen or hip area. Rotation of sites is critical to prevent lipohypertrophy, which can delay and unpredictably alter absorption. Clinical guidelines recommend using the abdomen for boluses before meals to maximize speed; closed loop systems that allow user-selected site recording can improve algorithm performance.

Physical Activity

Exercise increases blood flow to the injected area, speeding absorption and enhancing insulin sensitivity. In closed loop systems, this can lead to a mismatch if the algorithm does not account for upcoming activity. Many systems now include a "exercise mode" that raises target glucose and reduces insulin delivery. Understanding the PK change during exercise helps refine these modes. The risk is that faster absorption combined with increased glucose utilization during exercise can cause rapid hypoglycemia if not anticipated.

Meal Composition

Meals high in fat and protein slow gastric emptying, delaying the glucose peak. However, the insulin absorption profile remains unchanged. This dissociation can cause early hypoglycemia if the bolus is given too fast or late hyperglycemia if meal absorption exceeds insulin action. Advanced closed loop algorithms use meal announcement and, in some cases, meal composition estimation to modify insulin delivery. Pharmacokinetic models that incorporate gastric emptying rates are being developed.

Insulin Formulation and Concentration

As noted, formulation differences matter. Beyond rapid vs. ultra-rapid, the presence of excipients (e.g., nicotinamide, treprostinil) directly alters absorption kinetics. Insulin concentration also affects PK: higher concentrations (U-200, U-300) have slower absorption per unit volume due to reduced surface area-to-volume ratio when deposited in tissue. This is why concentrated insulins are used for basal, not bolus, delivery. In closed loop pumps, U-100 remains standard, but research into U-200 for pumps is ongoing.

Local Blood Flow and Temperature

Factors that increase local blood flow—heat, massage, inflammation—accelerate absorption. Cold, vasoconstriction, or scar tissue slow it. A hot shower or sauna shortly after a bolus can cause rapid hypoglycemia. Some closed loop systems are exploring temperature sensors on the infusion site as input to algorithm adjustments.

Skin Thickness and Body Mass Index

Subcutaneous tissue depth varies. In lean individuals, insulin may be injected into intramuscular tissue, which absorbs faster and unpredictably. In obesity, thicker adipose tissue can slow absorption. Pediatric and adolescent populations have different skin thickness, affecting PK. Closed loop systems designed for children must account for faster absorption and higher sensitivity.

Integrating Pharmacokinetics into Closed Loop Algorithms

Model Predictive Control vs Proportional-Integral-Derivative

Two main control strategies are used in closed loop systems. Proportional-integral-derivative (PID) algorithms adjust insulin delivery based on the current glucose error, cumulative error, and rate of change. PID is simple but does not explicitly incorporate a PK model. Model predictive control (MPC) uses a dynamic model of glucose-insulin interaction—often a compartmental PK/PD model—to predict future glucose levels and optimize insulin dosing over a horizon. MPC is better at handling delays and constraints, and it can incorporate patient-specific parameters such as insulin sensitivity, carbohydrate ratio, and PK profile of the chosen insulin. Many commercial hybrid closed loop systems (Medtronic 780G, Tandem Control-IQ, Omnipod 5) use MPC variations.

Modeling Insulin Action Curves

To build an accurate PK model, the insulin action curve must be parameterized. Common approaches use a two-compartment model (subcutaneous depot and plasma) or a one-compartment model with an absorption rate constant (ka) and elimination rate constant (ke). The peak time and duration are estimated from clinical data. However, these parameters vary by individual and over time. To improve robustness, adaptive algorithms continuously estimate insulin sensitivity and PK parameters from historical glucose data. For example, if the algorithm detects that glucose drops faster than predicted after a correction, it can infer faster absorption and adjust future dosing.

Accounting for Intra-Individual Variability

Even within the same person, insulin PK can vary day-to-day due to injection site, activity, meals, and hormonal cycles (e.g., menstruation). Closed loop systems that run on 24/7 operation can slowly adapt, but sudden changes (e.g., starting a new infusion set on a different site) require the algorithm to re-learn. Some systems prompt the user to enter site changes or activity levels. Advanced research explores using CGM data to detect changes in insulin absorption rate and trigger recalibration.

Challenges and Future Directions

Subcutaneous Delay and Sensing Lag

Even with ultra-rapid insulins, there is still a ~10–15 min lag between subcutaneous insulin delivery and peak glucose lowering. Additionally, CGM sensors measure interstitial glucose, which lags blood glucose by 5–10 min. Combined, the lag can cause oscillations. Future developments include intraperitoneal insulin delivery, which mimics portal physiology and eliminates subcutaneous absorption variability. Early studies show faster onset and more physiological hepatic insulinization, but it requires implanted devices and is currently invasive.

Dual-Hormone Systems

Adding glucagon to a closed loop system (bi-hormonal) can counteract insulin overdose and protect against hypoglycemia. Glucagon has its own PK: rapid onset (1–2 min) and short duration (~15–30 min). Integrating both hormone PK profiles into a single MPC is complex but promising. Studies show that bi-hormonal systems achieve >90% time in range with few hypoglycemic events, but they require dual pumps and frequent glucagon reconstitution.

Personalized Pharmacokinetics

No two patients have identical PK. Age, sex, ethnicity, genetics, and comorbidities all affect insulin clearance and sensitivity. Machine learning is being applied to CGM and pump history to create personalized PK models that update in real time. For example, recurrent neural networks can predict glucose with high accuracy using only past insulin and glucose data, implicitly learning the individual's PK/PD. These "black-box" models can outperform traditional compartment models but are harder to interpret.

Ultra-Fast and Stable Formulations

Researchers are developing insulins with onset times of 1–2 min and duration of 1–2 hours, essentially mimicking natural prandial insulin secretion. Inhaled insulin (e.g., Afrezza) has an even faster onset (3–4 min) but variable absorption and potential pulmonary side effects. Combining inhaled insulin for boluses with subcutaneous basal could create a fully closed loop without meal announcement. However, regulatory and practical challenges remain.

Conclusion

Pharmacokinetics lies at the heart of closed loop insulin delivery. Every element—from the choice of insulin formulation to the design of control algorithms—depends on understanding how insulin is absorbed, distributed, metabolized, and cleared. As ultra-rapid insulins become more widely adopted and algorithms become more adaptive, closed loop systems will continue to improve glucose outcomes and reduce patient burden. Continued research into personalized PK models, dual-hormone approaches, and novel delivery routes promises to bring us closer to a fully automated artificial pancreas. For clinicians and engineers alike, a deep appreciation of insulin pharmacokinetics is essential to designing, optimizing, and safely implementing these life-changing technologies.

External Resources: