Integrated Quantitative Systems Pharmacology model of TED, linking the systemic TSHR-TRAb autoimmune axis of Graves' disease to its orbital cross-reaction with fibroblast IGF-1R — the key mechanistic node that drives T/B-cell infiltration, cytokine-mediated orbital fibroblast activation, hyaluronan/GAG-driven edema, adipogenesis and late-phase fibrosis — culminating in proptosis, eyelid retraction, diplopia and (rarely) compressive dysthyroid optic neuropathy (DON). The model spans the modern pharmacology stack (teprotumumab, IV methylprednisolone/oral prednisone, rituximab, tocilizumab, selenium) together with smoking as a key modifiable risk factor and Rundle's-curve natural history.

TED (Graves' orbitopathy) develops in ~25-50% of patients with Graves' disease when TSHR-stimulating autoantibodies (TRAb/TSI), generated by loss of tolerance to the thyroid TSH receptor, cross-react with a TSHR-IGF-1R signaling complex expressed on orbital fibroblasts (Bahn 2010 NEJM). This crosstalk, amplified by infiltrating Th1/Th17 T cells, B cells and macrophages secreting IL-1β/IL-6/TNF-α/IFN-γ, activates orbital fibroblasts to differentiate into adipocytes (orbital fat expansion) or myofibroblasts (TGF-β-driven fibrosis) and to overproduce hyaluronan, whose osmotic water-binding causes extraocular muscle (EOM) edema and conjunctival chemosis. Because the bony orbit cannot expand, increased fat/muscle volume raises intraorbital pressure, producing proptosis (Hertel exophthalmometry), eyelid retraction (sympathetic Müller-muscle overactivity), restrictive diplopia (fibrotic EOM tethering) and — when apex crowding compresses the optic nerve — sight-threatening dysthyroid optic neuropathy (DON). Disease activity is graded by the Clinical Activity Score (CAS 0-7, Mourits 1997) and EUGOGO severity class; the natural history typically follows Rundle's curve — an active inflammatory phase (6-24 months) that gradually transitions to a quiescent fibrotic phase in which inflammation resolves but structural change (fibrosis, fixed proptosis/diplopia) may persist. Smoking is the single strongest modifiable risk factor, dose-dependently worsening severity and blunting response to glucocorticoids and radiotherapy. First-line therapy for active moderate-severe disease is IV methylprednisolone or, since 2020, the IGF-1R-blocking antibody teprotumumab (OPTIC trial, 83% proptosis responders vs 10% placebo); rituximab (anti-CD20) and tocilizumab (anti-IL-6R) are used in steroid-resistant disease with more heterogeneous trial evidence, and selenium is EUGOGO-endorsed for mild disease.

1. Systemic autoimmunity — TSHR-TRAb axis (Graves' disease, RAI/thyroidectomy, ATD)
2. Orbital immune cell infiltration (Th1/Th17, Treg deficiency, B cells, macrophages, cytokines)
3. TSHR-IGF-1R crosstalk on orbital fibroblasts (the key teprotumumab target node)
4. Orbital fibroblast fate decision (adipogenic PPAR-γ vs myofibroblast/TGF-β-CTGF)
5. Glycosaminoglycan/hyaluronan & ECM dynamics
6. Orbital tissue remodeling (fat volume, EOM volume/edema, fibrosis)
7. Mechanical/compressive consequences (proptosis, apex crowding, DON, corneal exposure)
8. Clinical activity, severity & PRO endpoints (CAS, EUGOGO class, GO-QoL, NOSPECS)
9. Drug PK/PD — teprotumumab (IGF-1R mAb)
10. Drug PK/PD — glucocorticoid, rituximab, tocilizumab, selenium
11. Non-pharmacologic/procedural therapy (smoking cessation, orbital RT/decompression, strabismus/lid surgery)
12. Adverse effects & safety monitoring (hearing loss, hyperglycemia, Cushingoid, hepatotoxicity)

• Drug PK (4 agents, 8 compartments) — teprotumumab (2-cmpt IgG), glucocorticoid genomic-signal compartment (IV methylprednisolone bolus or oral prednisone with absorption), rituximab (2-cmpt IgG), tocilizumab (2-cmpt, simplified from TMDD).
• Orbital autoimmune/fibroblast disease network (10 compartments) — B-cell pool, TRAb titer, T-cell activation, cytokine drive, fibroblast activation, hyaluronan/GAG index, orbital fat volume, EOM volume/edema, fibrosis index, Clinical Activity Score.
• Safety trackers (2 compartments) — cumulative teprotumumab AUC (hearing-impairment risk) and cumulative glucocorticoid dose (Cushingoid/hepatotoxicity risk).

1. Patient profile — adjustable age/weight/smoking/baseline CAS/severity/duration.
2. Drug PK — plasma concentrations of the three tracked mAbs + glucocorticoid signal.
3. Immune / Fibroblast PD — receptor occupancy, B-cell/TRAb, T-cell/cytokine, fibroblast/hyaluronan.
4. Orbital tissue remodeling — orbital fat volume, EOM volume/edema, fibrosis index.
5. Clinical endpoints — CAS, Hertel proptosis, GO-QoL, EUGOGO active-disease flag.
6. Scenario comparison — runs all 9 scenarios with the chosen profile.
7. Biomarkers & safety — teprotumumab AUC/hearing-AE risk, cumulative glucocorticoid/Cushingoid-AE risk.
8. References — key citations and link to the full bibliography.

# 1) Render the mechanistic map
dot -Tsvg ted_qsp_model.dot -o ted_qsp_model.svg
dot -Tpng -Gdpi=150 ted_qsp_model.dot -o ted_qsp_model.png

# 2) Simulate scenarios in R
install.packages(c("mrgsolve","dplyr","tidyr","ggplot2","shiny",
                   "shinydashboard","DT"))
library(mrgsolve)
mod <- mread_cache("ted_mrgsolve_model.R")
out <- mod %>% ev(amt=10*70, cmt="CEN_TEP", time=0) %>%
  ev_seq(ev(amt=20*70, cmt="CEN_TEP", ii=21*24, addl=6, time=21*24)) %>%
  mrgsim(end=17520, delta=24)
plot(out, c("Proptosis_mm","CAS_score","IGF1R_occupancy","GOQOL_score"))

# 3) Launch the dashboard
shiny::runApp("ted_shiny_app.R")

• Designed for research, education, and hypothesis generation — not clinical decision support.
• All PK/PD parameters are population means or literature-informed surrogates; inter-individual variability needs omega() blocks for population simulations.
• Thyroid hormone/TSH dynamics themselves are not modeled explicitly (TRAb is tracked as an orbital-disease driver, not a thyroid-function simulator) — pair with a dedicated Graves'-disease thyroid-axis model for combined thyrotoxicosis + orbitopathy simulation.
• The glucocorticoid compartment represents genomic anti-inflammatory persistence (not true plasma steroid pharmacokinetics); cumulative-dose bookkeeping (CUMGC) must be incremented via a matching dosing event in R alongside each GC_C/GUT_GC dose (see usage notes in the model file).
• Diplopia and dysthyroid optic neuropathy are represented qualitatively via the EOM-volume/fibrosis and apex-crowding nodes in the mechanistic map rather than as separate quantitative ODE endpoints.

Inherits the repository LICENSE.