Educational simulator for insulin pump (CSII) therapy in type 1, type 2 and LADA diabetes. Enter body weight and insulin type to auto-compute TDD, ICR and ISF; enter meal carbs, pre-meal BG and active insulin to see the basal rate, carb bolus, correction bolus and predicted 4-hour blood glucose in real time.
Rapid-acting analogs use the 1800 Rule, regular human insulin uses the 1500 Rule
Meal carbohydrate
g
Pre-meal BG
mg/dL
Target BG
mg/dL
Active insulin (IOB)
U
Insulin still on board from a previous bolus; subtracted from the combined bolus to prevent stacking
Basal rate (% of TDD)
%
Fraction of TDD delivered as basal. 40–50% is typical in adults
Results
—
TDD (U/day)
—
Hourly basal (U/hr)
—
ICR (g/U)
—
ISF (mg/dL/U)
—
Meal bolus (U)
—
Predicted 4 h BG (mg/dL)
—
Pump diagram — basal + bolus delivery
The pump delivers a steady basal infusion via the subcutaneous cannula, with larger bolus pulses for meals. Yellow trace shows blood glucose rising after the meal and falling as insulin acts.
TDD: total daily dose (U/day, scaled with weight w); ICR: grams of carbohydrate metabolised per unit (g/U); ISF: mg/dL fall per unit (mg/dL/U); Bolus: combined meal + correction dose. Active insulin (IOB) is subtracted to give the final delivered dose.
Diabetic Insulin Pump Basal-Bolus — TDD, ICR and ISF
🙋
How is an insulin pump different from injections? I imagine it just delivers automatically when you press a button — why split things into a "basal" and a "bolus"?
🎓
Good question. The pump (formally CSII, Continuous Subcutaneous Insulin Infusion) trickles insulin nonstop through a cannula in the abdomen. A healthy pancreas does two things: a low constant background secretion plus a big burst at every meal. The pump mimics that — a 24-hour continuous basal, and on-demand boluses for meals and corrections. Splitting them lets you tune each independently.
🙋
So TDD is the total daily insulin. Is it really decided just by body weight?
🎓
As a starting point, yes: 0.5–0.7 U/kg/day. A 65 kg adult with type 1 starts around 65×0.7 ≈ 46 U/day — Holman's empirical estimate. But very active or lean people can need just 0.3 U/kg, while teenagers and pregnant women can need up to 1.2 U/kg. The final TDD is dialled in over weeks by watching CGM Time-in-Range and HbA1c, adjusting ±10–20% at a time.
🙋
Once TDD is set, the panel automatically gives ICR and ISF. Why are the constants 500 and 1800 — magic numbers?
🎓
Not magic — they're empirical constants back-calculated from thousands of type 1 patients. ICR = 500 / TDD gives the grams of carb 1 U can cover. With TDD = 45 that's 500/45 ≈ 11 g per unit. ISF = 1800 / TDD tells you how much 1 U drops your glucose in mg/dL. Old-style regular human insulin uses the 1500 Rule instead. These assume rapid-acting analogs (Lispro, Aspart) and routinely vary ±20% with exercise, body temperature and stress — so you verify and fine-tune with the CGM.
🙋
There's an "active insulin (IOB)" slider — when I increase it the final bolus drops. What is that?
🎓
That's critical. IOB is the leftover insulin from your last bolus that hasn't finished acting. Even rapid-acting insulin lasts 4–5 hours, peaking at about 90 minutes. If you took 5 U two hours ago and 3 U are still onboard, then "correcting" with another 4 U stacks the peaks and crashes your sugar 30 minutes later. That's insulin stacking, the #1 cause of pump newbie accidents. Smart pumps therefore auto-compute IOB and subtract it from the correction bolus — exactly what this tool does.
🙋
I keep hearing "artificial pancreas" — same hardware?
🎓
Same hardware plus an automatic control layer — a hybrid closed loop (HCL). Medtronic 780G, Tandem Control-IQ, Insulet Omnipod 5 and Beta Bionics iLet are the main examples. A CGM (Dexcom G7, Libre 3) feeds subcutaneous glucose every 5 minutes and a PID or MPC controller moves the basal up and down on its own. Meal boluses are still announced by the user, but overnight hypo prevention is fully automatic. Time-in-Range above 80% is common, and HbA1c in the low 6s is achievable. In Japan it's mainly Medtronic 770G/780G and Terumo Medisafe Insulink so far — Omnipod 5 and iLet aren't approved yet.
Frequently Asked Questions
This simulator uses the weight-based empirical formula of Holman et al.: TDD = body weight × U/kg/day. The defaults are 0.7 for type 1, 0.4 for insulin-dependent type 2 and 0.6 for LADA. In real practice the coefficient is individualised between 0.3 and 1.0 U/kg according to age, activity and insulin resistance, then fine-tuned every two weeks based on CGM Time-in-Range and HbA1c. New pump users typically start on the low side with about 50% as basal and verify that fasting glucose stays flat overnight.
The 500 Rule gives the Insulin-to-Carbohydrate Ratio (ICR) = 500 / TDD, i.e. how many grams of carbohydrate one unit of insulin can metabolise. The 1800 Rule gives the Insulin Sensitivity Factor (ISF) = 1800 / TDD, i.e. how many mg/dL one unit lowers blood glucose. For regular (non rapid-acting) human insulin the standard is the 1500 Rule. These approximations assume rapid-acting analogs and can vary by ±20% with exercise, stress or the menstrual cycle.
Rapid-acting insulin lasts about 4–5 hours with peak action around 1–2 hours. If you stack a new bolus while a previous one is still active, the peaks overlap and can cause severe hypoglycaemia. Smart pumps therefore compute Insulin on Board (IOB) and automatically subtract it from the correction bolus. This simulator does the same: enter the active insulin and it is removed from the combined bolus to give the final dose.
Hybrid closed-loop systems such as Medtronic 780G, Tandem t:slim X2 Control-IQ, Insulet Omnipod 5 and Beta Bionics iLet read CGM values (Dexcom G7, FreeStyle Libre 3) every five minutes and adjust the basal rate automatically via PID or Model Predictive Control. The meal bolus is still announced by the user, but overnight hypo avoidance and post-meal corrections are automatic. Time-in-Range of 75–85% (70–180 mg/dL) is commonly achieved — better glucose stability than this manual-calculation simulator can match.
Real-world applications
Standard of care for type 1 diabetes: More than one million people worldwide and 15,000+ in Japan alone use CSII. Large randomised trials (DCCT, STAR-3) show pump therapy lowers HbA1c by 0.3–0.5 % and reduces severe hypoglycaemia by 30–70 % compared with multiple daily injections (MDI), especially in children and young adults with type 1. The automatic ICR/ISF computation in this tool can be used in the outpatient clinic as the first step of carbohydrate-counting education.
Carb-counting education: 150 g of cooked rice (one bowl) ≈ 55 g carb, one slice of bread ≈ 26 g carb, one banana ≈ 23 g carb. Patients learn typical carb contents and mentally divide by ICR to estimate the meal bolus. The carb slider in this simulator lets you visualise instantly how much insulin is needed for, say, 60 g of carbs — useful as a teaching aid in dietitian sessions.
Foundation for hybrid closed loops: The auto-correction algorithms in Medtronic 780G SmartGuard, Tandem Control-IQ and Omnipod 5 SmartAdjust internally use the very same TDD/ICR/ISF. The only difference is that they read CGM trends and re-tune the basal every five minutes. Understanding manual basal-bolus calculation is the prerequisite for understanding HCL.
Intensified therapy for type 2 diabetes and LADA: Type 2 patients who fail to reach HbA1c targets on oral agents and GLP-1, and latent autoimmune diabetes in adults (LADA) once endogenous insulin secretion has dropped, are also CSII candidates. Type 2 typically needs a lower 0.4–0.6 U/kg TDD, and because insulin resistance is high the ICR must be set more conservatively than the bare number suggests.
Common misconceptions and pitfalls
The biggest pitfall is "using textbook TDD/ICR/ISF values as-is". The 500 Rule and 1800 Rule are approximations derived from population averages and individual variation reaches ±50 %. At the same body weight and age, an active person may have an ICR of 20 g/U while a sedentary person needs 8 g/U. The numbers from this simulator are an initial estimate only — they must be verified with two weeks of CGM tracking and adjusted in 10–15 % steps under the supervision of a physician or diabetes educator. Never run a pump on textbook values without clinical oversight.
Next, the "set one constant basal and forget it" mistake. Healthy basal insulin secretion rises before dawn (the dawn phenomenon) and dips at night. Type 1 patients commonly need 30–50 % more basal between 04:00 and 08:00, so smart pumps let you program 24 hourly segments. This tool only shows the daily average, but in real use you run "basal tests" (skip a meal and watch glucose drift over 6 hours) to tune each segment, and then apply temporary reductions of 50–80 % for exercise and increases up to 150 % for sick-day fevers.
Finally, do not treat the 4-hour BG prediction as "the post-meal glucose will land exactly here". It uses a coarse model with constant carb absorption (4 mg/dL/g) and a linear insulin effect. Real life depends strongly on glycaemic index (white rice vs. onigiri absorb at very different speeds), fat and protein content (delayed glucose rise), infusion-site absorption (peak time differs by ~30 min between abdomen and thigh) and whether you exercise after the meal. This is an educational sensitivity tool, not a clinical dosing engine.
How to Use
Enter body weight in kilograms (typical range 50–150 kg for adults with type 1 or type 2 diabetes)
Input total carbohydrate content of planned meal in grams (e.g., 45 g for a sandwich and juice)
Enter preprandial blood glucose in mg/dL and target glucose goal (typically 120–150 mg/dL for type 1 CSII therapy)
Simulator calculates total daily dose (TDD), hourly basal infusion rate, insulin-to-carb ratio (ICR), insulin sensitivity factor (ISF), mealtime bolus requirement, and predicted 4-hour postprandial glucose
Worked Example
Patient weighs 75 kg with type 1 diabetes on insulin pump. Body weight formula: TDD = 75 × 0.5 = 37.5 U/day; hourly basal = 37.5 ÷ 24 = 1.56 U/hr. For 60 g carbohydrate meal with preprandial glucose 110 mg/dL and target 130 mg/dL: ICR = 500 ÷ 37.5 = 13.3 g/U (mealtime bolus = 60 ÷ 13.3 = 4.5 U); ISF = 1800 ÷ 37.5 = 48 mg/dL/U (correction = (110 − 130) ÷ 48 = −0.4 U, omitted). Total bolus = 4.5 U. Predicted 4-hour glucose ≈ 145 mg/dL post-meal absorption.
Practical Notes
LADA (latent autoimmune diabetes in adults) typically requires lower TDD than type 1; adjust body weight factor from 0.5 to 0.3–0.4 U/kg if C-peptide positive