Insulin: Dosing Protocols

Research Dosing Disclaimer#

The dosing information below is derived from research studies and is provided for educational purposes only. Insulin is not approved for human use, and no official dosing guidelines exist.

Dose-Response Data#

We synthesized dose–response data for insulin across common animal models, focusing on weight-adjusted dosing and measurable outcomes in insulin tolerance tests (ITTs) and hyperinsulinemic–euglycemic clamps.

Key findings by species and paradigm

• Mouse ITT (C57BL/6J): A widely used protocol administers human regular insulin intraperitoneally at 0.75 U/kg after a 6-hour fast. This dose produces a characteristic fall in glucose over 15–90 minutes; high-fat diet mice show a blunted glucose decline versus lean controls, demonstrating reduced insulin sensitivity. Safety procedures include dextrose rescue if severe hypoglycemia occurs (e.g., <20 mg/dL).

• Rat clamp dose–response: In awake rats, graded IV insulin infusions at 2 and 18 mU/kg/min show a clear dose–response in pathway partitioning. At 2 mU/kg/min, whole-body glycolysis accounts for about 81% of glucose disposal while muscle glycogen synthesis contributes ~13%; at 18 mU/kg/min, glycogen synthesis rises to ~38% with glycolysis ~51%. Diabetic rats exhibit an overall reduction of insulin-mediated glucose metabolism to roughly 20–30% of control values across insulin levels, indicating decreased sensitivity.

• Dog clamp (conscious): With IV insulin infused at 0.6 mU/kg/min during euglycemic clamp, euglycemia was maintained with a glucose infusion rate (GIR) of approximately 6.8 mg/kg/min. This quantifies whole-body insulin action at this infusion rate and supports capillary transport as a rate-limiting step for insulin action.

• Pig ITT and clamp context: Female pigs (~57 kg) received an IV insulin bolus of 0.1 U/kg for an ITT. Under heat stress, insulin-stimulated glucose uptake increased relative to pair-fed thermoneutral controls, reflected by higher clamp glucose infusion requirements and improved insulin tolerance, indicating a context-dependent enhancement of insulin sensitivity at this fixed dose.

• Rat acute dose-ranging with insulin glargine (Wistar): Subdermal bolus dosing at 0, 4, 6, and 8 U/kg produced a dose-dependent rise in serum insulin and fall in glucose AUC. Glucose AUC fell by about 26% at 6 U/kg and ~42% at 8 U/kg; mild hypoglycemia occurred at 6 U/kg and severe hypoglycemia at 8 U/kg. A chronic model therefore used 4 U/kg/day to induce hyperinsulinemia without hypoglycemia.

• Additional clamp infusion rates reported in research contexts: Mouse and nonhuman primate clamp mixtures sometimes use higher insulin infusion rates (e.g., ~10 mU/kg/min) to achieve robust hyperinsulinemia, with GIR recorded to quantify insulin action in those models.

Embedded summary table

Interpretation and practical notes

• ITT dosing in mice at 0.75 U/kg IP after a 6-hour fast is a field standard for detecting between-group differences in insulin sensitivity; results are typically presented as time courses, percent decrease from baseline, nadir and recovery, and/or inverse AUC. Rescue glucose is recommended to mitigate severe hypoglycemia.

• In clamp studies, increasing insulin infusion rate (mU/kg/min) produces dose-dependent shifts in glucose disposal pathways and higher GIRs; diabetic or insulin-resistant states shift the dose–response downward, reducing GIR at any given infusion rate.

• Large-animal ITTs (e.g., pigs at 0.1 U/kg IV) and environmental modifiers (heat stress) demonstrate that at a fixed weight-adjusted dose, physiologic state substantially alters the glycemic response, reinforcing the need to report ambient conditions and fasting.

• Acute dose-ranging with basal insulin analogs in rats shows a steep dose–response for glucose lowering, with hypoglycemia risk rising sharply beyond ~6 U/kg SC in Wistar rats, informing selection of non-hypoglycemic doses for chronic modeling.

• For some entries (e.g., mouse/primate high-rate clamps), our excerpts identified infusion rates but did not include full quantitative GIR values; where precise numeric outcomes were unavailable in the extracted passages, we report qualitative dose–response and cite accordingly.

Administration Routes#

We compared insulin pharmacokinetics and bioavailability across subcutaneous (SC), intramuscular (IM), oral, and transdermal/microneedle (TD) routes using human studies and high‑level reviews, emphasizing route‑specific quantitative metrics, variability, and first‑pass effects.

Subcutaneous (SC). SC injection remains the benchmark for systemic insulin delivery. For fast‑acting insulin aspart formulated for accelerated absorption, absolute bioavailability after SC dosing is about 80% overall, with region‑specific estimates of 83% (abdomen), 77% (upper arm), and 77% (thigh). Early exposure and Cmax are greater from abdomen/arm than thigh, though total exposure (AUC) is similar across sites. Typical appearance occurs within ~3 minutes, tEarly50%Cmax around ~20 minutes, and tmax ~50–57 minutes. Variability is influenced by subcutaneous blood flow (increased by heat/exercise; reduced by obesity/smoking), injection region (abdomen > arm > thigh > buttock for speed), depot volume and geometry, excipients, and tissue changes such as lipohypertrophy. Coadministration of recombinant human hyaluronidase (rHuPH20) approximately doubles first‑hour exposure, shortens tmax from roughly 68–86 to ~41–44 minutes, and shortens mean duration of action by ~40–50 minutes, producing an ultrarapid profile. SC delivery is not subject to hepatic first‑pass extraction.

Intramuscular (IM). IM injection accelerates early absorption compared with SC. With faster aspart given IM, onset of appearance is ~2.6 minutes, tEarly50%Cmax ~14 minutes, and tmax ~45 minutes, with earlier and higher early exposure than SC; however, total exposure can be similar or lower relative to SC in the same study. IM dosing carries greater risk of rapid, unpredictable insulin action and hypoglycaemia, so it is generally discouraged outside specific clinical circumstances. IM delivery is systemic and not subject to hepatic first‑pass.

Oral (enteral). Oral insulin faces enzymatic degradation and poor intestinal permeability, yielding very low systemic bioavailability. In a healthy‑volunteer clamp study of an investigational prandial oral insulin (N11005; 300 IU), relative serum bioavailability versus injected reference was ~0.71%, with oral Cmax 16.3 mU/mL vs 40.3 mU/mL for injection and median Tmax ~5 hours; despite low systemic exposure, early pharmacodynamic onset (tGIR10%max) occurred at ~11 minutes, reflecting formulation‑dependent gut absorption dynamics. For an investigational basal oral insulin 338 tablet, human PK showed Tmax ~40–60 minutes and a long half‑life ~55 hours, and exposure was not materially affected by meals given ≥30 minutes post‑dose; these reflect the molecule’s modified kinetics and enhancer‑assisted absorption. Reviews emphasize that most oral formulations achieve <1% bioavailability but may preferentially deliver insulin to the portal circulation, potentially mimicking physiological hepatic exposure and reducing peripheral hyperinsulinaemia. Oral delivery, uniquely among routes discussed, is subject to first‑pass hepatic extraction.

Transdermal/microneedle (TD). Human trials and a systematic review indicate that intradermal microneedle or needle‑free jet systems can produce faster onset and earlier peak compared with SC, often with similar total exposure. In a randomized crossover trial in patients with diabetes comparing SC injection to a microneedle patch, oral capsule, and inhaled insulin, the microneedle patch achieved Tmax ~1.8 hours vs ~2.5 hours for SC, with AUC ~94% of SC and duration ~6.0 vs ~6.5 hours. The systematic review of clinical trials found multiple studies where microneedle or jet delivery shortened time‑to‑peak by ~20–30 minutes and reduced early post‑prandial hyperglycaemia while showing comparable overall exposure. Classical non‑breaching topical delivery is limited by the stratum corneum, and transdermal enhancement requires physical methods (microneedles, jet, electroporation/iontophoresis) or chemical enhancers; these approaches deliver systemically and are not subject to hepatic first‑pass. Skin/device factors (application technique, site, microneedle characteristics) influence variability.

Comparative interpretation. SC provides reliable systemic exposure with absolute bioavailability near 80% for rapid analogs, moderate tmax (~50–57 min), and well‑characterized variability determinants; absorption can be further accelerated with hyaluronidase at the cost of shorter duration. IM accelerates onset and peak beyond SC but increases risk of hypoglycaemia due to rapid, less predictable uptake. Oral remains investigational with sub‑percent systemic bioavailability; however, it offers potential hepatic targeting and distinct PK profiles depending on formulation (e.g., insulin 338 vs prandial N11005). Transdermal microneedle/jet delivery can advance onset and peak without major loss of overall bioavailability relative to SC in several human studies, though device factors and skin tolerance are practical considerations.

Key caveats across routes. SC absorption is sensitive to site, temperature, and tissue pathology (lipohypertrophy), and abdomen is preferred for fastest uptake; thigh is slower in early exposure. IM is generally avoided for routine insulin because of hypoglycaemia risk from rapid uptake. Oral insulin studies vary in quality and often small sample sizes; systemic bioavailability is typically <1% and highly formulation‑ and timing‑dependent. Transdermal systems require specialized devices and training; most data to date are from small trials, and approaches without barrier‑breaching have very low permeability.

Human-Equivalent Dosing#

Objective: Provide a sourced overview of how animal study doses of insulin are translated to human-equivalent doses and summarize allometric scaling methods used in the literature.

Summary of methods and how they are applied

• BSA-based human-equivalent dose (HED) using Km factors. Many translational workflows compute HED via body surface area normalization using species-specific Km values: HED (mg/kg) = Animal dose (mg/kg) × (Km_animal / Km_human). Km is defined as weight/BSA, and tabulated values are often used; this approach underlies common calculators and is consistent with regulatory practice for conservative first-in-human (FIH) dose selection. However, it does not account for insulin’s PK/PD nuances and should be complemented by exposure–response information when possible.

• Weight-based power-law HED without explicit Km tables. When explicit Km values are unavailable, HED can be computed as HED = Animal dose × (W_animal/W_human)^(1−b), where b is an allometric exponent. Conventional choices are b = 0.67 (BSA-related) or b = 0.75 (metabolic-rate based), leading to exponents 1−b = 0.33 or 0.25, respectively. These choices influence how conservative the translation is, especially from small species, and have been discussed in translational method reviews.

• Allometric scaling of PK parameters (preferred for insulin translation). Rather than directly scaling dose, investigators commonly scale pharmacokinetic parameters by body size, then simulate human exposures to select doses. Typical assumptions are CL ∝ W^0.75, V ∝ W^1, and first-order rate constants ∝ W^(−0.25). This framework is used to predict human AUC and Cmax from animal PK, and then combined with PD targets to pick an initial clinical dose.

• Exposure matching and NOAEL/MABEL workflows. Translational guides recommend selecting clinical starting doses by matching predicted human exposure (e.g., AUC at NOAEL) or by MABEL when pharmacology is potent. Practically, one predicts human CL (often via allometry), computes the human dose that reproduces a target AUC or minimal biological effect, and applies safety factors (commonly ≥10) for FIH. This approach is applicable to insulin if exposure–effect relationships are available.

• Model-based insulin translation using integrated glucose–insulin (IGI) models. Interspecies scaling of mechanistic glucose–insulin models has been performed with exponents near theoretical allometry (e.g., fitted exponents ~0.87 for clearances, ~0.9–0.98 for volumes, and ~−0.25 for rate constants), while allowing a few species-specific adjustments. Such models incorporate baseline glucose/insulin and sensitivity parameters per species and have successfully described IVGTT datasets, facilitating translation of insulin PD across species (alskarUnknownyearinterspeciesscalingof pages 1-1).

• Clamp-based pharmacodynamic translation for insulin. For insulin and insulin analogs, the gold-standard PD assessment is the euglycemic glucose clamp. Translation often targets reproducing clamp-derived PD metrics—glucose infusion rate (GIR) vs. time profiles, GIR AUC (overall glucose-lowering effect), and duration/peak characteristics—rather than relying on a fixed U/kg conversion. Reviews of clamp methodology detail how GIR and insulin concentration define time–action profiles used to compare analogs and guide clinical dosing decisions.

Notes specific to insulin dosing units and examples

• Units and dosing. Animal and human insulin doses are frequently reported in U/kg. Simple proportional conversion by body weight (mg/kg or U/kg) is discouraged without considering PK/PD, since bioavailability, clearance, depot behavior, and insulin sensitivity differ across species.

• Example doses in interspecies datasets. IVGTT datasets used in interspecies IGI modeling include insulin bolus doses around 0.03 U/kg in dogs and humans, providing a cross-species anchor for PD model translation rather than a rule for routine HED conversion (alskarUnknownyearinterspeciesscalingof pages 1-1).

• Practical workflow seen in development programs. A common path is: (1) scale PK (CL, V) allometrically to predict human exposure for candidate insulin or analog; (2) use clamp-derived exposure–response relationships to choose a human dose that matches desired GIR AUC/shape; (3) cross-check with conservative BSA/Km HED if a safety-relevant animal NOAEL exists, applying safety factors for FIH.

Key equations and parameters

• BSA/Km HED: HED (mg/kg) = Dose_animal (mg/kg) × (Km_animal/Km_human); Km = weight/BSA; also equivalently dose_mg/m2 = dose_mg/kg × Km.
• Weight power law HED: HED = Dose_animal × (W_animal/W_human)^(1−b) with b ≈ 0.67 or 0.75.
• PK allometry: CL_h = CL_a × (W_h/W_a)^0.75; V_h = V_a × (W_h/W_a)^1; k (rate) ∝ W^(−0.25).
• Model-based IGI scaling: fitted exponents near theory (clearance ~0.87; volumes ~0.9–0.98) plus species-specific PD parameters (alskarUnknownyearinterspeciesscalingof pages 1-1).
• Clamp PD matching: select dose to match GIR AUC/shape at steady-state or single-dose; no single algebraic formula.

Caveats and recommendations

• For insulin, prioritize PK/PD-informed translation (allometric PK to predict exposure + clamp PD matching) over standalone mg/kg or U/kg conversions. Use BSA/Km-based HED mainly as a conservative starting point in safety contexts, or when required for MRSD frameworks.
• Species differences in insulin sensitivity and depot kinetics mean that model-based approaches (e.g., IGI) and clamp endpoints provide more reliable guidance for clinical dose selection than simple body-size corrections.

Evidence Gaps#

• No human dose-finding studies have been completed
• Allometric scaling from animal models has inherent limitations
• Route-specific bioavailability data in humans is absent
• Optimal treatment duration has not been established

Related Reading#

• Insulin overview and research guide
• Insulin side effects profile
• Insulin research evidence