Six Biologics, One Disease, Zero Head-to-Head Trials: How Do You Actually Choose in Myasthenia Gravis?

Bottom Line: Six biologics are now FDA-approved for generalized myasthenia gravis. Three target complement at C5 (eculizumab/Soliris, ravulizumab/Ultomiris, zilucoplan/Zilbrysq); three antagonize the neonatal Fc receptor (efgartigimod/Vyvgart, rozanolixizumab/Rystiggo, nipocalimab/Imaavy). Mean treatment-versus-placebo MG-ADL differences range from approximately −1.0 to −2.1 points — meaningful, but not curative. There are no head-to-head trials. The 2017 "refractory only" framing has crumbled, but choosing among the six is still a structural exercise rather than an evidence-based ranking. Antibody status, age, route preference, infection risk, and payer reality drive the decision more than trial efficacy.

The Controversy

For six decades, generalized myasthenia gravis was a problem of titration: pyridostigmine, prednisone, azathioprine or mycophenolate, IVIG or plasma exchange for crisis, and thymectomy for the right anatomic candidate. The MGTX trial (Wolfe, NEJM 2016) finally put numbers on thymectomy: in 126 non-thymomatous AChR-antibody-positive patients randomized to extended transsternal thymectomy plus prednisone versus prednisone alone, the surgical arm delivered a 2.85-point QMG advantage over three years, 41% lower prednisone exposure, and hospitalization for exacerbation in 9% versus 37% of patients (all P<0.001). That was the field's anchor — the first randomized evidence that an intervention could modify the trajectory of generalized MG, not merely its symptoms.

Eight years later, the anchor has shifted. Six biologics are now FDA-approved for gMG: eculizumab (Soliris), ravulizumab (Ultomiris), and zilucoplan (Zilbrysq) target complement; efgartigimod (Vyvgart IV and Vyvgart Hytrulo SC), rozanolixizumab (Rystiggo), and nipocalimab (Imaavy) block the neonatal Fc receptor. A seventh — inebilizumab — is at the FDA's door. An eighth — Descartes-08, a BCMA-targeted mRNA CAR-T — is in late-phase development without lymphodepletion or hypogammaglobulinemia. Batoclimab's Chinese phase 3 is positive. Satralizumab's sponsor walked away from a statistically positive trial because the effect was too small to matter.

The question is no longer whether to use a biologic in moderate-to-severe gMG. The question is which one, when, and for whom — and we are being asked to answer it without a single head-to-head trial. What the data show and what we conclude from them are two different things.

The Case For First-Line Biologics

The argument has matured. When eculizumab was approved in 2017 on the basis of REGAIN — a trial whose primary endpoint (MG-ADL worst-rank ANCOVA) technically missed at P=0.0698 and was rescued by a prespecified sensitivity analysis at P=0.0058 — the label was tightly restricted to refractory AChR-antibody-positive disease, defined as failure of ≥2 immunosuppressants or one plus chronic IVIG/PLEX. That framing has crumbled.

CHAMPION-MG (Vu, NEJM Evid 2022) enrolled MGFA class II–IV patients without requiring documented immunosuppressant failure, and ravulizumab delivered a clean primary endpoint hit: MG-ADL change at week 26 of −3.1 versus −1.4 (LSM Δ −1.6; 95% CI −2.6 to −0.7; P<0.001), with onset within one week. ADAPT (Howard, Lancet Neurol 2021) showed 67.7% of AChR+ efgartigimod-treated patients achieved a sustained ≥2-point MG-ADL improvement versus 29.7% on placebo (OR 4.95; P<0.0001). RINOMAX (Piehl, JAMA Neurol 2022) showed that a single 500 mg infusion of rituximab in new-onset gMG — though not FDA-approved for this indication — produced minimal disease manifestations in 71% versus 29% (P=0.007), with no MG hospitalizations in the rituximab arm.

The case for moving earlier rests on three pillars: (1) onset of effect within 1–2 weeks for both complement and FcRn agents, versus 6–12 months for conventional immunosuppressants; (2) cumulative steroid toxicity avoided; (3) MGTX-style data suggesting that early aggressive treatment changes trajectory, not just symptoms.

The Case Against

Before we celebrate, let's read the fine print. The trial says yes — but look at who was in the room.

REGAIN enrolled a refractory population. CHAMPION-MG required MG-ADL ≥6 and excluded patients with thymoma in the prior three years. ADAPT required MG-ADL ≥5 and a stable immunosuppressant regimen. MycarinG (rozanolixizumab) required MG-ADL ≥3 — but its phase 3 population (n=200) had a mean disease duration of roughly seven years. None of these trials enrolled the moderately symptomatic, recently diagnosed patient who has not yet failed mycophenolate. None enrolled large numbers of patients over 75. Pregnant patients were excluded. Patients with active or recent malignancy were excluded.

Then there are the effect sizes. The mean treatment-versus-placebo MG-ADL difference is approximately:

• Ravulizumab (CHAMPION-MG): Δ −1.6
• Zilucoplan (RAISE): Δ −2.09
• Nipocalimab (Vivacity-MG3 phase 3): Δ −1.45
• Satralizumab (LUMINESCE): Δ −1.02 — sponsor halted development

An MG-ADL improvement of 2 points sustained for 4 weeks is the consensus minimum clinically important difference. Several of these means sit at or below it. Responder analyses look better because they measure the tail of the distribution — but a 1.5-point average improvement at a six-figure annual cost is a different proposition than a curative therapy.

And LUMINESCE deserves its own paragraph. Satralizumab hit its primary endpoint (MG-ADL Δ −1.02; P=0.020), hit QMG (Δ −1.63; P=0.0062), and then failed MG-QoL15r (P=0.11), breaking the statistical hierarchy. The sponsor halted development. That is what "statistically positive but clinically insufficient" looks like in real time. It should make us read every other positive trial more carefully — not less.

What the Data Actually Shows

Complement Inhibitors

Three agents, one mechanism (terminal complement blockade at C5), three delivery profiles:

• Eculizumab (Soliris) — REGAIN, n=125: IV every two weeks after weekly induction. Primary endpoint technically missed by worst-rank ANCOVA (P=0.0698); MG-ADL responder rate 55.7% versus 41.2%; approved on totality of evidence. Headache 24%, nasopharyngitis 18%. One meningococcal infection — vaccinated, treated successfully.
• Ravulizumab (Ultomiris) — CHAMPION-MG, n=175: IV every eight weeks after loading. MG-ADL Δ −1.6 (P<0.001); QMG Δ −2.0 (P<0.001); ≥5-point QMG improvement in 30.0% versus 11.3%. Open-label extension MG-ADL responder rate 76.4% at week 60. No meningococcal infections (universal vaccination per protocol).
• Zilucoplan (Zilbrysq) — RAISE, n=174: SC daily, self-administered. MG-ADL Δ −2.09; QMG Δ −2.97 (both P<0.001). The smallest molecule, the most flexible administration, the same mechanism.

None of these works in MuSK-antibody-positive disease, because MuSK antibodies are IgG4 and do not fix complement. That is a mechanism-based exclusion, not a trial-design quirk.

FcRn Antagonists

Three agents that increase IgG catabolism by blocking the recycling receptor — including pathogenic autoantibody catabolism:

• Efgartigimod IV (Vyvgart) and SC (Vyvgart Hytrulo) — ADAPT, n=167: cyclic dosing. AChR+ MG-ADL responder 67.7% versus 29.7% (OR 4.95; P<0.0001); QMG responder 63% versus 14%. ADAPT+ long-term: more than 90% of AChR+ patients achieved a clinically meaningful improvement across the first 10 cycles. Mean IgG reduction approximately 56–60%. Five deaths in the long-term cohort, none adjudicated as treatment-related. ADAPT-SC established non-inferiority of the subcutaneous formulation to IV.
• Rozanolixizumab (Rystiggo) — MycarinG, n=200: SC weekly × 6 per cycle, approved for both AChR+ and MuSK+ disease. MG-ADL Δ −3.37 (7 mg/kg) and −3.40 (10 mg/kg) versus −0.78 placebo at day 43 (P<0.001).
• Nipocalimab (Imaavy) — Vivacity-MG3 phase 3, n=153: IV every two weeks, continuous (no cycling). MG-ADL over weeks 22–24: −4.70 versus −3.25 (Δ −1.45; P=0.002). Broadest label of the six: AChR+, MuSK+, age ≥12. Approved April 2025.

A Mechanistic Map

The Conventional Anchors Still Matter

MGTX (n=126) remains the cleanest disease-modification dataset in MG: 41% lower prednisone, 9% versus 37% hospitalization for exacerbation, 67% versus 47% minimal manifestation status at 36 months. The Pasnoor methotrexate trial (n=50) was negative on its primary endpoint (prednisone AUDTC Δ −488 mg; P=0.26) — yet methotrexate remains widely prescribed. That is a discipline-wide failure to update on negative data. BeatMG (rituximab, n=52) was negative for steroid-sparing in chronic AChR+ disease, while RINOMAX (n=47) was positive in new-onset gMG — a reminder that disease stage, not just antibody status, governs response.

For crisis, IGIV-C 2 g/kg over two days remains the workhorse: QMG Δ −6.4 ± 5.15 at day 14 (P<0.001), responder rates 77–88% across MG-ADL, MGC, and QMG. No biologic has displaced that.

Choosing in 2026: A Working Logic

Without head-to-head data, the sequencing logic is necessarily structural:

What's Coming — And One Cautionary Tale

MINT (inebilizumab, n=208) hit its primary endpoint cleanly: MG-ADL Δ −1.7 (P<0.001), QMG Δ −2.3 (P<0.001), with greater benefits in AChR+ patients. Anti-CD19 B-cell depletion would add a third mechanism to the field. Descartes-08, a BCMA-targeted mRNA CAR-T delivered outpatient without lymphodepletion, produced an MGC ≥5-point responder rate of 66.7% versus 27.3% at month 3 (P=0.0472), with 33.3% achieving minimal symptom expression (MG-ADL ≤1) by month 6. No B-cell depletion, no hypogammaglobulinemia — a different safety architecture than any biologic currently approved.

Batoclimab's Chinese phase 3 (n=132) showed sustained MG-ADL improvement in 58.2% versus 31.3% (OR 3.45; P=0.001), with FDA filing pending.

And then there is satralizumab. LUMINESCE met its primary MG-ADL endpoint (Δ −1.02; P=0.020). The sponsor halted development anyway. The question is not settled. Here is why that is actually a good thing — because the field is finally distinguishing statistical positivity from clinical transformation. We should hold every approved agent to that same standard going forward.

The Unresolved Question

The six approved biologics deliver, on average, 1.5–3 points of MG-ADL improvement above placebo. That is meaningful. It is not curative. None has demonstrated superiority over any other, and no trial has run long enough to tell us what year 5 or year 10 of continuous complement blockade or FcRn antagonism looks like — for infection risk, for malignancy, for the eventual ceiling of antibody-driven disease.

We have moved from one drug class to two, from refractory-only labeling to moderate-to-severe disease, from monthly infusions to daily self-injection. What we have not done is generate the comparative evidence that would let us choose rationally rather than structurally. The decision logic above is a workaround, not an answer.

The question to leave with you: If a payer told you tomorrow that your patient with newly diagnosed, moderately severe AChR+ generalized myasthenia gravis could have any one of the six biologics — and only one, for the next five years — which would you choose, and what evidence would you cite to defend the choice?

I suspect the honest answer, for most of us, would reveal how much of our current practice rests on inference rather than data. That is not a criticism. It is a roadmap for the trials we still need.