Enspryng in thyroid eye disease: approval across a discordant Phase 3 program

October 15 approval and the actual label

Genentech announced US FDA approval on October 15, 2026; Roche published its global announcement on October 16. This dossier uses the earlier US announcement date for the completed regulatory event, rather than moving the event to the later publication date. The October label specifies adults with thyroid eye disease regardless of activity or duration. It is an approval, not merely acceptance or priority review. It is also distinct from the separate NMOSD indication and the still-pending MOGAD application. [1] [2] [5]

The label is broader than the active-disease primary analysis. That distinction matters: a regulatory indication spanning active and inactive disease does not make every subgroup a separate positive randomized trial. The practical change is a subcutaneous option that can be administered at home after training. Convenience does not establish greater efficacy or safety than an intravenous competitor because these trials used placebo, not an active head-to-head comparator.

IL-6 receptor biology in orbital inflammation

Thyroid eye disease combines autoimmune inflammation with orbital tissue expansion and remodeling. Proptosis measures forward displacement of the eye; diplopia reflects misalignment and double vision. They overlap clinically but are not interchangeable outcomes. Blocking IL-6 receptor signaling addresses an inflammatory pathway rather than physically decompressing the orbit. Reduced inflammation may improve symptoms while established tissue remodeling persists, so activity, duration and baseline anatomy are important effect modifiers to inspect rather than reasons to assume uniform response. [1] [2]

Satralizumab binds soluble and membrane-bound IL-6 receptors. Its recycling design supports repeated target engagement and prolonged circulation. The mechanistic rationale explains why a study is reasonable, but neither target binding nor experience in NMOSD establishes benefit in TED. The disease-specific randomized evidence below is the bridge that matters. This report does not substitute an antibody engineering claim for a measured improvement in eye position or vision.

Dose, exposure and what the label supports

The TED label uses actual body weight: 120 mg for 40–100 kg and 180 mg above 100 kg, given subcutaneously at weeks 0, 2 and 4, then every four weeks from week 8. Safety and effectiveness below 40 kg have not been established for TED. The registry lists a protocol dose for below 40 kg, but a protocol option must not be presented as an approved regimen. A larger body-weight dose is an exposure-management decision, not evidence that higher dosing produces a greater clinical response. [2] [3] [4]

Anti-drug antibodies were observed in 44.8% and 39.1% in the two TED studies. The label reports lower exposure in antibody-positive patients without an apparent clinically relevant impact on efficacy, safety or pharmacodynamics in the available data. Neutralizing ability was unknown. These findings should be read with assay and sample-size limitations; they are neither proof that immunogenicity is irrelevant nor a justification to infer individual loss of response.

Two trials and three different denominators

SatraGO-1 and SatraGO-2 were separate randomized, double-masked, placebo-controlled Phase 3 studies. Enrollment was 131 and 127, respectively, with 1:1 assignment. Across the program 258 were enrolled, 255 received treatment and 238 completed through week 24. Enrollment, safety exposure and the active-TED primary analysis are different populations. The primary active-population denominators were 50 versus 51 in Study 3 and 48 versus 49 in Study 4; they must not be replaced by the whole trial totals. [2] [3] [4]

Active TED required CAS at least 3 and proptosis at least 3 mm above normal; stable inactive disease had CAS below 3 with the same proptosis criterion. Patients were euthyroid or had limited biochemical thyroid dysfunction, and prior TED surgery or biologic treatment was excluded. The primary responder definition required at least a 2 mm reduction in the study eye without at least a 2 mm deterioration in the fellow eye at week 24. This is not synonymous with complete disease resolution, restored normal vision or avoidance of surgery.

The registry full-analysis set follows randomized assignment. Published endpoint counts are those available for the analysis, and the reported response percentages and adjusted contrasts should not be forced into reconstructed integer numerators. Complete estimand, imputation sensitivity and the full multiplicity hierarchy were not independently extracted. The unusual 95.03% confidence level is retained as reported, rather than silently rounded into a conventional 95% interval.

One missed primary endpoint, one positive comparison

Study 3 missed its prespecified primary endpoint: 49.0% versus 31.2%, adjusted difference 17.9 percentage points, with an interval crossing zero. Study 4 showed 52.9% versus 23.4%, adjusted difference 28.9 points. The differences are model-reported and need not equal subtraction of rounded percentages. The two trials should remain separate: combining only favorable endpoints would obscure the failure in the first study. Regulatory approval based on the program does not retrospectively turn that failed primary test into a positive one. [2] [3] [4]

In Study 4, overall-population response was 30/64 versus 14/63, and improvement of at least one diplopia grade among active patients with baseline diplopia was 17/28 versus 8/31. These support benefit beyond the single primary summary while using different denominators. They do not represent independent replication in a second trial, and the diplopia subgroup must not be treated as the whole randomized population.

A visible label-table inconsistency

The October label Table 9 prints the active-population proptosis-change contrast with percent signs and a positive upper confidence limit despite a millimeter endpoint. Visual inspection confirms this is present in the PDF, not merely text-extraction damage. The registry instead reports adjusted means −1.97 and −0.96 mm, contrast −1.02 mm, 95.03% CI −1.67 to −0.36, p=0.0027. This dossier preserves the conflict explicitly and uses the registry units and interval only with that source attribution; it does not silently amend the label. [2] [4]

Safety: treated populations and incomplete aggregate AE data

Registry safety denominators are 67/62 in SatraGO-1 and 64/62 in SatraGO-2. Through week 24, serious adverse events affected 3/67 versus 2/62 and 2/64 versus 3/62; deaths were zero in all four groups. These are observed participant counts, not repeated-event totals or evidence that long-term risk is zero. The registry nonserious-event listings use a 5% reporting threshold; their summary counts cannot be called all TEAEs. An overall any-TEAE rate was not available in the extracted results. [2] [3] [4]

Study withdrawal attributed to an adverse event was 0/68 versus 3/63 in Study 3 and 1/64 versus 1/63 in Study 4. These randomized-flow denominators differ from safety denominators, and withdrawal from the study is not identical to discontinuing drug. Separately the label describes two injection-reaction cases leading to treatment discontinuation without a complete per-arm total for all reasons. No pooled drug-discontinuation rate is inferred from those statements.

The pooled label table reports upper respiratory infection at 5.3% versus 2.4% and ALT increase coded as an adverse reaction at 5.3% versus 0.8%, in 131 versus 124 treated patients. Laboratory surveillance is broader: decreased neutrophils were 28.2% versus 9.7%, and ALT elevations 25.2% versus 7.3%. Coding an adverse reaction and detecting a laboratory shift are different definitions, explaining why these percentages should not be interchanged. Screening for hepatitis B and tuberculosis, liver-enzyme monitoring and neutrophil monitoring remain material even when short-term serious-event rates appear similar.

Sponsor execution and remaining evidence

Genentech is part of Roche, and Chugai developed satralizumab within the Roche group. This is not a standalone small-biotech financing event: an approval headline does not provide a product-level cash balance, new equity transaction or fully diluted capitalization. No new funding or management-capacity conclusion is inferred from approval. The relevant execution questions are safe home administration, laboratory follow-up, persistence and availability of longer-term randomized follow-up. [1] [2] [3] [4]

Part II rerandomizes responders and gives active treatment to nonresponders, so later durability analyses must identify original assignment, response-based selection and the new treatment allocation. They cannot be appended as if the original placebo comparison continued unchanged. The report includes the public approval, label and posted Part I results; it does not claim access to the complete clinical study reports or the full FDA review memorandum. Those gaps limit explanation of trial discordance and long-term benefit, even though approval itself is verified.

Sources / References

1. Genentech: US FDA approval announcement — October 15, 2026
2. ENSPRYNG US Prescribing Information, revised October 2026; sections 1,2,6,12,14
3. ClinicalTrials.gov NCT05987423 (SatraGO-1), results update July 31, 2026
4. ClinicalTrials.gov NCT06106828 (SatraGO-2), results update August 12, 2026
5. Roche global announcement — October 16, 2026