In this chapter, we use simple models that run in the browser to check the content of the previous chapters by trying things yourself. When you move the sliders or buttons, the graphs update on the spot (all computation happens inside your browser).
WarningThese are “conceptual models”
The models used here are greatly simplified for educational purposes. They are meant for understanding the shapes and trends of the graphs and cannot be used to predict actual drug concentrations or doses. Pharmacokinetic parameters are approximate values based on representative figures from package inserts and published data.
7.1 Experiment 1: How fast do the native incretins disappear?
In Chapter 2, we learned that the half-life of active GLP-1 is 1–2 minutes and that of GIP is 5–7 minutes. Let’s look at the concentration profile after a single intravenous bolus.
Set the GLP-1 half-life to 1.5 minutes and confirm that almost nothing is left after 10 minutes
Try to explain why “injecting this hormone as it is and using it as a drug” is unrealistic
7.2 Experiment 2: Blood concentrations of a once-weekly drug
The GLP-1 receptor agonists seen in Chapter 6 extend the half-life to several hours to one week. Let’s simulate the blood concentration with repeated dosing.
The model is a one-compartment model with “first-order absorption from the subcutaneous tissue and first-order elimination”, and the concentrations from each dose are added together (the superposition principle). The vertical axis is a relative value with the peak concentration after the first dose set to 1.
pk = {const ke =Math.LN2/ t12;// Back-calculate the absorption rate constant ka from tmax (bisection in the range ka > ke)const tmax =Math.min(drug.tmax,0.9/ ke);const tmaxOf = ka =>Math.log(ka / ke) / (ka - ke);let lo = ke *1.0001, hi =50;for (let i =0; i <100; i++) {const mid = (lo + hi) /2;if (tmaxOf(mid) > tmax) lo = mid;else hi = mid; }const ka = (lo + hi) /2;const single = t => t <0?0: ka / (ka - ke) * (Math.exp(-ke * t) -Math.exp(-ka * t));const cmax1 =single(tmax);const total = weeks *168;const doses = d3.range(0, total, tau).filter((_, i) =>!(scenario.length>0&& i ===2));const dt =Math.max(0.25, total /1500);const curve = d3.range(0, total + dt, dt).map(t => ({t: t /24,c: doses.reduce((s, d) => s +single(t - d),0) / cmax1 }));const R =1/ (1-Math.exp(-ke * tau));// accumulation ratioconst t90 =Math.log(10) / ke;// time to reach 90% of steady statereturn { ke, ka, curve, doses, R, t90 };}
md`- **Accumulation ratio** R = 1 / (1 − e<sup>−k·τ</sup>) = **${pk.R.toFixed(2)}** — at steady state the concentration is about ${pk.R.toFixed(1)} times that after the first dose- **Time to reach 90% of steady state**: about **${(pk.t90/24).toFixed(1)} days** (about 3.3 times the half-life)- Half-life ${t12} h ÷ dosing interval ${tau} h = **${(t12 / tau).toFixed(2)}** (the more it exceeds 1, the smaller the fluctuation in concentration between doses)`
TipTry it
Select semaglutide and read off how many weeks it takes to reach steady state. Think about how this relates to the reason for dose escalation every 4 weeks in clinical practice
Keep the dosing interval at “once weekly” and lower the half-life to 13 hours (equivalent to liraglutide). What becomes a problem?
Check “Miss the 3rd dose”. How does the impact of a missed dose differ between drugs with long and short half-lives?
Compare tirzepatide and semaglutide and see how the ratio of peak to trough differs
NoteExplanation (click to expand)
Semaglutide’s half-life is about 1 week, so it takes about 3.3 weeks to reach 90% of steady state. Because it takes about a month for the effects and side effects to stabilize after a dose increase, escalation is done every 4 weeks. Starting at a low dose and increasing gradually is also meant to let the body get used to the gastrointestinal side effects (nausea, etc.).
With once-weekly dosing and a 13-hour half-life, the concentration falls to nearly zero before the next dose, creating periods when the drug is working and periods when it is not. The balance between half-life and dosing interval is important.
With a long half-life, missing one dose only lowers the concentration gradually, so the impact is relatively small. The flip side is that when side effects occur, the drug does not wash out quickly even after stopping it.
Tirzepatide, with its slightly shorter half-life (about 5 days), has a somewhat larger difference between peak and trough.
7.3 Experiment 3: Why do incretins rarely cause hypoglycemia?
Let’s check the idea of the triggering and amplifying pathways from Chapter 3 on a graph. The horizontal axis is blood glucose and the vertical axis is insulin secretion (relative value).
Glucose only: secretion by glucose itself (triggering pathway). An S-shaped curve that rises once blood glucose exceeds a certain threshold
+ Incretin: amplifies the secretion from the triggering pathway several-fold
+ SU drug: closes K_ATP channels regardless of blood glucose, so it shifts the curve to the left (toward lower glucose) and causes secretion even during hypoglycemia
viewof incStrength = Inputs.range([0,3], { value:1.5,step:0.1,label:"Strength of incretin action" })viewof hormone = Inputs.radio(["GLP-1","GIP"], { value:"GLP-1",label:"Type of incretin" })viewof suStrength = Inputs.range([0,1], { value:0,step:0.05,label:"Strength of SU drug" })viewof t2d = Inputs.toggle({ label:"Type 2 diabetes (reduced β-cell function)",value:false })
islet = {const n =5;const smax = t2d ?0.5:1;// β-cell functionconst K = t2d ?180:140;// blood glucose at half-maximal secretion// In type 2 diabetes the response to GIP is greatly reduced (Chapter 1)const responsiveness = (t2d && hormone ==="GIP") ?0.15:1;const amp =1+ incStrength * responsiveness;const hill = (g, k) =>Math.pow(g, n) / (Math.pow(g, n) +Math.pow(k, n));const Ksu = K * (1-0.55* suStrength);const f = {base: g => smax *hill(g, K),inc: g => smax *hill(g, K) * amp,su: g => smax * (hill(g, Ksu) * amp +0.25* suStrength) };const G = d3.range(40,301,2);const rows = G.flatMap(g => [ { g,y: f.base(g),s:"Glucose only" }, { g,y: f.inc(g),s:"+ Incretin" },...(suStrength >0? [{ g,y: f.su(g),s:"+ Incretin + SU drug" }] : []) ]);return { rows, f, amp };}
Confirm that even if you strengthen the incretin action, secretion at a blood glucose of 60 mg/dL barely increases
Increase the strength of the SU drug. What happens in the hypoglycemic range?
Turn on “Type 2 diabetes” and switch between GLP-1 and GIP. How does Chapter 1’s “GIP works poorly but GLP-1 works” show up?
NoteExplanation (click to expand)
Incretins only multiplicatively amplify the secretion from the triggering pathway, so in the hypoglycemic range where the trigger is not pulled, multiplying a value near 0 by anything still leaves it near 0. This is what “glucose-dependent” means.
SU drugs pull the trigger regardless of blood glucose, so insulin is released even in the hypoglycemic range. When incretin-based drugs and SU drugs are combined, the incretin amplifies even the secretion caused by the SU drug, raising the risk of hypoglycemia. This is why the SU dose is reduced when they are combined.
In this model, responsiveness to GIP in type 2 diabetes is lowered to 15% of the healthy level. Amplification remains with GLP-1, whereas GIP produces almost no amplification. In reality, however, it is known that the response to GIP partly recovers when blood glucose is improved (Chapter 6).