# Development and External Validation of a Multivariable Model to Predict Early Minimal Symptom Expression Response in Adult Generalized Myasthenia Gravis Patients Treated With Efgartigimod

## TL;DR

A new model predicts which patients with a muscle disease will respond well to a specific treatment within four weeks.

## Contribution

A three-factor nomogram was developed and validated to predict early response to efgartigimod in generalized myasthenia gravis patients.

## Key findings

- Lower bulbar MG-ADL, higher FVC%, and lower IgG independently predicted early MSE response.
- The nomogram showed strong discrimination with AUCs above 0.83 in both derivation and validation cohorts.
- The model demonstrated good calibration and meaningful net clinical benefit across decision thresholds.

## Abstract

Efgartigimod, a neonatal Fc receptor (FcRn) blocker, is approved for generalized Myasthenia Gravis (gMG), but predictors of early response are unclear. Using minimal symptom expression (MSE) as the therapeutic target, we developed and externally validated a clinical model to predict early MSE response after starting efgartigimod. This efgartigimod‐tailored model estimates the baseline probability of achieving early MSE and may assist in individualized treatment selection at therapy initiation.

We retrospectively analyzed 118 adults with gMG treated at five tertiary centers in China (efgartigimod 10 mg/kg IV weekly ×4). MSE was defined as Myasthenia Gravis–Activities of Daily Living (MG‐ADL) ≤ 1 sustained ≥ 4 weeks; early MSE response was achievement within 4 weeks of initiation. Patients were split into a derivation cohort (n = 64; three centers) and an external validation cohort (n = 54; two centers). Candidate predictors included demographics, baseline severity, and laboratory indices. Variables associated with early MSE in univariable analyses entered multivariable logistic regression to construct a nomogram. Discrimination (AUC‐ROC), calibration (curves and Spiegelhalter Z‐test), and clinical utility (decision curve analysis, DCA) were assessed, with bootstrap internal validation.

Early MSE response occurred in 26/64 derivation and 22/54 validation patients. Lower bulbar MG‐ADL (OR 0.633, p = 0.040), higher FVC% (OR 1.042, p = 0.048), and lower IgG (OR 0.795, p = 0.036) independently predicted early MSE response. The nomogram showed strong discrimination—derivation AUC 0.869 (95% CI 0.797–0.941), bootstrap AUC 0.880 (0.806–0.954), and external AUC 0.839 (0.760–0.919)—and good calibration (Spiegelhalter Z: 1.03, p = 0.303; 0.94, p = 0.347; 1.17, p = 0.242). DCA indicated net benefit across thresholds 0.05–0.82, with validation curves mirroring derivation.

A three‐factor nomogram (bulbar MG‐ADL, FVC%, and serum IgG) provides an efgartigimod‐specific baseline estimate of early sustained‐MSE response probability, which may help neurologists select appropriate candidates, counsel expected benefit, and tailor follow‐up intensity or alternative escalation strategies.

Chinese Clinical Trial Registry (ChiCTR2500101971)

A multicenter adult gMG cohort from five tertiary centers in China (N = 118) received efgartigimod 10 mg/kg IV weekly × 4. Early minimal symptom expression (MSE) response was defined as MG‐ADL ≤ 1 sustained ≥ 4 weeks and assessed within 4 weeks. A three‐factor nomogram (bulbar MG‐ADL subscore, FVC% predicted, serum IgG) was developed in the derivation cohort (n = 64), internally validated by bootstrap, and externally validated (n = 54). The model showed strong discrimination (ROC/AUC), good calibration, and meaningful net clinical benefit on decision‐curve analysis, enabling individualized prediction of early MSE response to efgartigimod.