Primary Endpoints and Pivotal Trial Designs Behind FDA and EMA Approvals in gMG, CIDP and ITP

Primary endpoints and pivotal trial designs behind FDA and EMA approvals in gMG, CIDP and ITP

Approvals in these three autoimmune indications follow three different endpoint patterns. Generalized myasthenia gravis (gMG) programs used the patient-reported MG-ADL scale in short placebo-controlled trials. CIDP programs used relapse or deterioration on the adjusted INCAT (aINCAT) disability scale, tested in withdrawal designs. ITP programs used sustained platelet counts of at least 50 x 10^9/L. FDA and EMA accepted the same core instruments. They differed on how to read the size of the treatment effect, how long trials needed to run, and how much enrichment or pre-selection was acceptable.

Key takeaways

• gMG: Every pivotal trial reviewed used MG-ADL as the primary endpoint, either as mean change from baseline or as a responder rate ^{5129546425391241}. FDA repeatedly called MG-ADL an "acceptable primary outcome" and pointed to precedent across earlier approvals ^302306325162. CHMP was stricter about effect size. It asked whether a placebo-adjusted mean change below 2 points is clinically relevant, and it said the literature MCID applies to within-patient change, not to between-group differences ^335340348.
• CIDP: Recent approvals used relapse or time-to-deterioration on aINCAT, with IVIg withdrawal or responder-enriched randomized-withdrawal designs ^934951322. Assessors described INCAT as validated and treated a 1-point change as clinically meaningful ²³²⁶⁵⁸. CHMP raised concerns about selection bias in the efgartigimod enrichment design and about how missing data were imputed in the HyQvia relapse analysis ⁶⁸⁰⁵⁵.
• ITP: Platelet response at or above 50 x 10^9/L, sustained over time and without rescue therapy, was the accepted surrogate at both agencies ^282273367488. Reviewers pressed on durability (for example, at least 6 of the last 8 weeks, or at least 4 of the last 6 visits), on rescue-therapy handling, and on missing-data imputation. They were consistently skeptical of bleeding scales as endpoints ^287291568240.

Generalized myasthenia gravis

Pivotal designs and primary endpoints

Rystiggo's EU entry criteria added a QMG floor of at least 11 ¹⁰¹. For Zilbrysq, the EU record notes entry criteria of MG-ADL of at least 6 and QMG of at least 12 ¹⁸⁰. For Soliris in the EU, a small supportive crossover trial (C08-001, N=14) used a QMG improvement of at least 3 points as its primary endpoint ⁴⁷³⁴⁷².

What FDA reviewers said

• MG-ADL is an established, agreed endpoint with precedent. For rozanolixizumab, FDA recorded prior Division agreement to MG-ADL and cited its use in the eculizumab, ravulizumab and efgartigimod programs ³⁰². The nipocalimab and zilucoplan reviews repeat the precedent argument and describe MG-ADL as an "acceptable primary patient-reported outcome" showing "important patient-reported impact and clinical benefit on daily function" ^162163325326.
• Responder vs continuous framing. For efgartigimod, FDA noted that mean change is more sensitive and that responder analyses can lose information. It still accepted the proposed responder definition, which builds in a durability requirement of at least 4 consecutive weeks ¹¹⁸¹³.
• QMG as corroboration. Reviewers described QMG as a validated physician-assessed motor measure. Concordant MG-ADL and QMG effects lowered concern that results were due to chance. Reviewers also cautioned that QMG examination findings on their own may have uncertain clinical meaningfulness ^301306161163.
• Population scope. For efgartigimod, the primary endpoint tested only AChR-Ab+ patients, and the overall-population result was driven by that subgroup. The approval was therefore limited to AChR-Ab+ gMG ¹²⁰¹²¹ (the qualifier was removed in May 2026 ⁶⁸²⁶⁸³).
• Analysis handling. For eculizumab, FDA flagged that patients who discontinued early might have met rescue criteria without being assessed. The sponsor added a worst-rank sensitivity analysis in response ¹¹³.
• Single-trial reliance. For zilucoplan, FDA found a single adequate and well-controlled trial persuasive enough in this rare disease, with pharmacodynamic confirmation ³²⁶.
• Novel PROs got more scrutiny than MG-ADL. In the rozanolixizumab review, FDA's validation concerns were about the new MG Symptoms PRO, not MG-ADL. The Division considered that instrument's validity and clinical meaningfulness uncertain ³⁰⁸.

What EMA assessors said

• Validated scales, but effect size scrutinized. CHMP described MG-ADL and QMG as validated standard instruments ¹⁹²³⁴⁸. For eculizumab, the prespecified worst-rank primary analysis at Week 26 was not statistically significant (p=0.0698), and the effect fell below the pre-defined 2-point clinically relevant between-group reduction ⁴⁷⁰³⁴⁸. CHMP called the applicant's MCID argument, which relied on individual responder rates rather than placebo-adjusted differences, "not optimal" but acceptable as positioning. Exploratory responder analyses were viewed as supportive ³⁴⁹³⁵⁰.
• The MCID applies within patients, not between groups. For ravulizumab, CHMP judged the 1.6-point placebo-adjusted difference statistically significant but modest. It clarified that the 2-point literature MCID reflects within-patient change and should mainly inform responder analyses ³³⁵³⁴⁰. CHMP found a 10.9% difference in 2-point responders (63.9% vs 53.0%) hard to interpret. It also flagged possible confounding because the analyses did not account for rescue therapy ³⁴⁰³³⁵.
• Responder endpoints should be read with mean change. For efgartigimod, assessors accepted the at-least-2-point, early-onset, sustained responder definition as suited to a fluctuating disease. They advised reading it alongside continuous mean changes, because dichotomizing loses information ¹⁹². Their major concern was that the primary analysis was restricted to AChR-Ab+ patients while a broad indication was being sought ¹⁹².
• Duration and durability. For zilucoplan, CHMP accepted the Week 12 MG-ADL result (about 2.09 points) as clinically relevant. It noted that it had advised in scientific advice a 24-week double-blind period to show maintenance of effect, and it requested more long-term data ^623617182. For rozanolixizumab, assessors called the lack of blinded data beyond the first 6-week cycle "concerning," with open-label MG0007 data only partly addressing durability ¹⁰⁸.
• Subgroups. For rozanolixizumab, MG-ADL showed little or no mean benefit in patients under 50 kg, and MuSK-positive and Asian subgroups were too small for firm conclusions ¹¹⁰.

Chronic inflammatory demyelinating polyneuropathy (CIDP)

Pivotal designs and primary endpoints

What reviewers said

• INCAT is validated, and a 1-point change is meaningful. For Hizentra, EMA assessors described the adjusted 10-point INCAT as "globally accepted, validated, and reliable" and consistent with prior IVIg trials. The adjustment excludes isolated arm-score changes from 0 to 1 that are not clinically meaningful ^2325350. EMA had endorsed the INCAT-based primary endpoint in scientific advice ²³⁸. For HyQvia, CHMP made the same point and noted that requiring two consecutive assessments was stricter than earlier CIDP trials, which could lower observed relapse rates ⁶⁵⁸.
• Withdrawal designs and carry-over. For Hizentra, assessors could not rule out a carry-over effect from prior IVIg (about 10 weeks) within the 24-week randomized period. They restricted the indication to maintenance after IVIg stabilization and asked for extension-study data ²³⁵¹¹⁶. For HyQvia, the remaining uncertainty included carry-over and possible enrolment of patients already in remission ⁶⁵⁹.
• Missing-data handling in relapse endpoints. For HyQvia, the original analysis imputed eight missing relapse determinations as "no relapse," even though some early discontinuations appeared to be due to CIDP worsening. CHMP called this anti-conservative, required revised estimands and tipping-point analyses, and accepted a composite that counted discontinuation for worsening as relapse ⁵⁵⁶⁵⁶.
• Enrichment and selection bias (efgartigimod). CHMP agreed that CIDP's rarity and diagnostic difficulty make conventional placebo-controlled trials hard to run. It accepted that a single compelling pivotal trial could suffice ⁶⁸⁰. It also found that enrolling only Stage A responders who tolerated treatment into Stage B introduced selection bias likely to overestimate efficacy. It considered it plausible that rebound after withdrawal of prior therapy, followed by spontaneous stabilization, inflated Stage A response ⁶⁸⁰. Historical placebo comparisons could not fix the internal-validity problem of the open-label Stage A ³²¹. Stage B maintenance benefit did not depend on whether patients had improved by at least 1 aINCAT point in Stage A ³¹⁹.
• FDA. For Vyvgart Hytrulo, the FDA source cited here is the label. It gives the design and endpoint definition and notes that aINCAT excludes the upper-limb 0-to-1 transition ⁹⁵. A label carries no reviewer commentary on the endpoint choice.
• EU extrapolation route for immunoglobulins. EMA's IVIg guidance allows extrapolation to CIDP without dedicated trials when efficacy in primary immunodeficiency and ITP is established and justified. Non-core dosing regimens need clinical data ⁶⁶⁶. SCIg maintenance after IVIg stabilization may also be extrapolated if supported by PK, clinical and literature data ⁶⁷¹.

Immune thrombocytopenia (ITP)

Pivotal designs and primary endpoints

What reviewers said

• Platelet count as an accepted surrogate. For romiplostim, FDA agreed prospectively under a Special Protocol Assessment that maintained increases in platelet count were an appropriate primary outcome even without a demonstrated reduction in hemorrhage. It gave ITP pathophysiology and the impracticality of bleeding endpoints as reasons ²⁸². FDA accepted 50 x 10^9/L as clinically meaningful because it is at least 20 x 10^9/L above the highest possible baseline mean ²⁷³. For eltrombopag, FDA reviewers called platelet response a surrogate for bleeding risk. They explained that counts below 30 x 10^9/L are associated with bleeding, and 50 x 10^9/L is a safe range for hemostatic challenges ^566569574. EMA took the same line: platelet count is "generally used as a valid surrogate," and the goal is a haemostatically safe count, not normalization ³⁶⁷⁴⁹⁴.
• Durability over single time points. FDA told the fostamatinib sponsor that a Week 12 endpoint might not show durability. It recommended a 24-week endpoint (6 of the last 8 weeks) and then found the sponsor's "4 of the last 6 visits" proposal reasonable ²⁸⁷²⁹¹. FDA regarded single-time-point platelet endpoints as unacceptable because platelet counts vary ²²². In paediatric eltrombopag, FDA advised that durability was more clinically meaningful, and the PETIT2 primary endpoint was amended to sustained response (6 of 8 weeks) ⁵⁷⁷⁵⁸³. EMA guidance recommends platelet response at prespecified time points, without rescue, with durability characterized where possible ⁶⁰⁰.
• Rescue therapy and missing data. For romiplostim, FDA required that any patient who received rescue medication be classed as a non-responder ²⁶⁶²⁶⁵. For fostamatinib, FDA rejected LOCF and asked that missed weekly counts be treated as non-response ²⁹⁰²⁹². EMA did not endorse LOCF for fostamatinib and recalculated responder status after correcting Day 1, which left sensitivity results with very wide confidence intervals ⁴⁷⁹⁴⁸³.
• Bleeding scales are not trusted as primary measures. FDA found the WHO bleeding scale subjective and of unclear clinical meaning in both the eltrombopag and fostamatinib reviews ^568572287294. For rilzabrutinib, FDA asked for content validity, measurement properties and a meaningful within-patient change threshold for the proposed ITP Bleeding Scale ²⁴⁰. EMA noted that fostamatinib's bleeding outcomes were too sparse to support conclusions on bleeding risk ⁴⁸⁸.
• Enrichment within the blinded period (rilzabrutinib). FDA initially did not agree that the proposed phase 3 design was adequate and asked for justification of moving Week 12 non-responders to open-label treatment ³⁰. At review, FDA judged LUNA 3 adequate and well controlled. It found the missing data, driven by high dropout, did not change conclusions ^634628629. CHMP was more critical. Because most patients in the Weeks 13 to 24 endpoint window had been pre-selected by early response, the rate of durable response regardless of Week 12 status could not be determined. CHMP said this "severely challenged" interpretation of the results ¹⁶⁵¹⁶⁷.
• Durability after approval. Both agencies treated extension data as supportive only. FDA noted that romiplostim efficacy might decline with prolonged use ²⁶⁹. EMA noted that 4 of 17 fostamatinib responders lacked long-term response ⁴⁸⁷.

Cross-indication lessons for endpoint strategy

1. Use the agreed instrument, then defend the effect size. MG-ADL, aINCAT and a platelet threshold of 50 x 10^9/L were not disputed as instruments. Disputes were about whether a group-level difference is clinically relevant. CHMP explicitly separated within-patient MCIDs from between-group effects ³³⁵³⁴⁰.
2. Plan responder and continuous analyses together. Both agencies accepted responder definitions with durability requirements, but both noted that dichotomizing loses information ¹¹⁸¹⁹².
3. Expect EMA to ask for longer blinded exposure in chronic, fluctuating diseases. EMA's advice on 24 weeks for zilucoplan and its concerns about data beyond cycle 1 for rozanolixizumab are examples ¹⁸²¹⁰⁸.
4. Enrichment designs are accepted but examined closely. CHMP accepted responder-enriched randomized withdrawal in CIDP given the disease constraints, but flagged selection bias ⁶⁸⁰. Pre-selection inside the blinded period of an ITP trial drew major objections from CHMP ¹⁶⁵.
5. Pre-specify conservative missing-data and rescue rules. Imputing "no relapse" or using LOCF was challenged in CIDP and ITP. Treating rescue or discontinuation for worsening as failure was the accepted approach ^55266290479.

Evidence gaps worth a follow-up

For ravulizumab in gMG, efgartigimod SC in CIDP and avatrombopag in ITP, the FDA sources cited here are labels. They give designs and results but no reviewer commentary on the endpoint choice ^54695146. This article does not cover the EU record for Imaavy, US records for subcutaneous immunoglobulins in CIDP, MuSK-positive gMG subgroups or paediatric ITP endpoint amendments.