Roche: CELESTIMO PFS topline and the evidence behind mosunetuzumab–lenalidomide

Confirmed event and its limits

Genentech’s September 16 announcement reports that the randomized CELESTIMO study met its primary progression-free survival endpoint at interim analysis. Mosunetuzumab plus lenalidomide was compared with rituximab plus lenalidomide after at least one previous systemic treatment for follicular lymphoma. Overall survival remains immature. Roche issued its corresponding announcement on September 17; these are two announcements of one event, not two independent trials. [1] [2]

The event is therefore a completed topline disclosure, recorded in History on September 16. It is not an FDA approval, a confirmed filing acceptance or a new PDUFA date. The sponsor intends the study to support both confirmation of the earlier monotherapy approval and a combination indication in an earlier treatment line; each regulatory action still needs its own dated evidence.

Biology and target

Mosunetuzumab is a CD20×CD3 bispecific antibody that brings T cells into proximity with CD20-expressing B cells. Lenalidomide provides an immunomodulatory partner. The rationale is to combine T-cell recruitment with a second immune mechanism, rather than assume that activity after repeated prior therapy automatically establishes benefit earlier in the disease course. [3] [7]

Analytical judgment: biological plausibility does not establish synergy. Because both randomized arms contain lenalidomide, the trial principally evaluates replacing the antibody component within a treatment strategy. A favorable comparison would not quantify how much benefit comes from lenalidomide itself, or prove superiority over every other contemporary treatment. Those questions require different comparisons.

Molecule, exposure and regimen

The IV label reports a geometric-mean half-life of 16.1 days and clearance that changes with time. These are monotherapy population-PK estimates, not a measured exposure-response threshold for the CELESTIMO combination. Step-up dosing addresses early immune activation; it does not eliminate the need to monitor cytokine release, infection and cytopenias. [3] [7]

The table preserves the 2022 design poster’s regimen version. The retrieved topline announcement does not reproduce final delivered doses, reductions or relative dose intensity. Do not substitute the approved monotherapy 60 mg loading schedule for the combination’s 30 mg schedule, or infer equal drug exposure from an equal number of nominal cycles.

Randomized design and populations

NCT04712097 is open-label and randomized 1:1. The design poster planned approximately 400 adults with CD20-positive grade 1–3a disease, ECOG performance status 0–2 and at least one prior systemic regimen. It excluded grade 3b or transformed disease, CNS lymphoma and lenalidomide-refractory disease. Stratification included early progression within 24 months, prior treatment count and anti-CD20 refractoriness. [3] [4]

The indexed May 2026 registry lists 478 total participants, but that study also includes a nonrandomized US extension. This total is not an actual randomized-arm denominator. Dividing it by two, or subtracting a previously reported cohort size, would manufacture allocation data. The final randomized, treated and analyzed populations remain to be reconciled against disposition and protocol amendments.

Interim results and statistical interpretation

The planned primary assessment is independent-review-committee PFS. The release states statistical significance but omits the hazard ratio, confidence interval, exact p value, arm medians and analysis cutoff. OS is immature. No Kaplan–Meier curve is reconstructed from this qualitative statement, and no survival advantage is inferred from a positive PFS headline. [1] [2] [3]

Analytical judgment: independent review reduces some assessment bias in an open-label study, but discontinuation patterns, scan timing and informative censoring still matter. The full analysis plan is needed to assess interim alpha spending, the endpoint hierarchy and sensitivity analyses. The final readout should show absolute disease-control duration alongside the relative effect, with subgroup interaction tests distinguished from within-subgroup nominal p values.

Prior monotherapy evidence

The original FDA accelerated approval drew on GO29781, a single-arm study with 90 efficacy-evaluable patients after at least two prior regimens. Independent review found 72 responses: 54 complete and 18 partial. Median response duration was 22.8 months, with a 95% confidence interval of 10 months to not reached. This establishes an activity benchmark in a different setting, not a randomized estimate against the CELESTIMO comparator. [5]

Analytical judgment: response duration is calculated among responders, whereas PFS follows an assigned analysis population from its defined starting point. The two endpoints cannot be exchanged in a benefit claim. Earlier-line selection, prior resistance and treatment combinations can each alter prognosis; numerical response-rate differences across these studies do not isolate a treatment effect.

The separate 54-patient US extension

Roche’s December 2025 ASH release describes a single-arm US CELESTIMO extension of 54 patients receiving the combination in second-line or later disease. Complete response was 87.0% (95% CI 75.1–94.6). CRS occurred in 27.8%, neutropenia in 40.7% and infection in 57.4%. These data preceded the randomized topline and must retain their own source version. [6] [9]

The grade distribution below is a descriptive safety chart within that cohort. It does not compare treatments and does not establish a low risk in the randomized population. The full publisher abstract could not be retrieved during this review; detailed total AE, SAE, death and withdrawal counts have therefore not been independently extracted from that primary document. They remain unresolved, not absent.

Safety: keep each denominator separate

The September randomized announcement describes no new safety signals without supplying an arm-level safety table. The IV monotherapy label separately reports serious adverse reactions in 47% and adverse-reaction discontinuation in 3% of its 90-patient FL cohort. Its broader HLH warning covers 7/1,536 patients, including six fatal outcomes; that pooled population is not the CELESTIMO randomized denominator. [1] [6] [7]

Analytical judgment: an absence of a new signal does not establish equal safety or zero deaths. Compare unique patients, event severity, exposure and follow-up for both arms. Infection and immune toxicity can coexist, and cytopenias can complicate attribution in a combination regimen. The missing randomized table materially limits a net-benefit assessment even when the primary efficacy test succeeds.

Capital structure and management

Genentech is wholly owned by Roche; the listed-parent exposure tracked here is Roche’s US OTC ADR, RHHBY, consistent with the existing Roche coverage. Roche reported H1 2026 sales of CHF30.364 billion and core operating profit of CHF11.856 billion. These group figures are not cash on hand, program funding or a calculated cash runway. [6] [8]

Analytical judgment: a diversified parent changes financing context but does not remove trial, manufacturing or regulatory execution risk. No standalone Genentech share count or CELESTIMO budget is disclosed in these sources. Voting shares, participation certificates and ADR units must be reconciled before any per-share calculation; none is attempted here. Management’s next observable commitment is delivery of the detailed dataset and subsequent regulatory submissions, not the favorable adjective in its announcement.

What the next evidence must resolve

The confirmed change is a positive primary-endpoint disclosure in a randomized active-controlled study. The magnitude, durability and tolerability of that benefit remain incompletely specified. A clinically useful readout should reconcile actual arm sizes, assessment timing, censoring, absolute PFS, OS maturity and treatment discontinuations before a stronger benefit–risk conclusion is drawn. [1] [2] [3]

Keep this event separate from SUNMO in aggressive large B-cell lymphoma and from first-line follicular-lymphoma programs. Sharing an antibody does not make their controls, disease biology or analyses interchangeable. This dossier records what was available on September 17, 2026; a future conference presentation or filing should be appended as a new dated version, preserving the qualitative topline and older cohorts.

Sources / References

1. Genentech · CELESTIMO interim topline · September 16, 2026
2. Roche · CELESTIMO interim topline · September 17, 2026
3. Nastoupil et al. · CELESTIMO trial design · EHA 2022 poster
4. ClinicalTrials.gov · NCT04712097 · indexed update May 5, 2026; history retrieval incomplete
5. FDA · Drug Trials Snapshot: LUNSUMIO · original December 22, 2022 approval
6. Roche · ASH 2025 CELESTIMO US extension · December 8, 2025
7. Genentech · LUNSUMIO IV prescribing information · retrieved September 17, 2026
8. Roche · Half-year 2026 results · July 23, 2026
9. Sano et al. · CELESTIMO US extension · Blood 146 Supplement 1, abstract 1800 · November 3, 2025; publisher full-text retrieval incomplete