📚 Part of the OBX-319 Regulatory Dossier — Reader's Guide. This article shows the live document; edits to the source appear here automatically.

---

Statistical Analysis Report (OBX319-301)

Document ID: SAR-301
Version: 1.0
Change History: 1.0 — Initial issue.
Standard(s): ICH E9(R1)

Statistical Analysis Report — OBX319-301

Purpose and scope

This Statistical Analysis Report (SAR) documents the analyses executed for the confirmatory Phase 3 study OBX319-301 as pre-specified in the Statistical Analysis Plan (SAP-301) finalized prior to database lock and unblinding. It presents the analysis populations, data-handling conventions, estimand framework, statistical models, and the resulting primary, key-secondary, and supportive analyses, together with the traceability chain that links the source data to the reported results. The primary analysis reported here is SRI-4 response at Week 52 (operationalised as low disease activity, SLEDAI-2K <= 4), evaluated in both its continuous (LS-mean change in SLEDAI-2K) and binary (responder) forms. Descriptive safety analyses conducted on the Safety Set under the same plan are summarized narratively in this report and tabulated in full in the Clinical Study Report (CSR-301, Section 12) and Module 2.7.4 / ISS-301. The results reported herein reconcile with the CSR, the ADaM analysis datasets (ADaM-SPEC-301), the analysis-results metadata (ARM-301), and the table/listing/figure outputs (TLF-301).

Product and study context

OBX-319 is a humanized IgG1 bispecific monoclonal antibody that simultaneously engages CD19 and CD20 on the B-cell lineage and is administered subcutaneously; it is produced by recombinant expression in a Chinese Hamster Ovary (CHO) cell line and purified by a platform downstream process (Protein A affinity capture followed by orthogonal polishing chromatography with dedicated viral-clearance steps). OBX319-301 was a randomized, double-blind, placebo-controlled study that allocated subjects with moderate-to-severe active Systemic Lupus Erythematosus 1:1:1 to OBX-319 High-dose, OBX-319 Low-dose, or matched subcutaneous placebo, each added to background standard of care, and treated over a 52-week double-blind period with assessments at Weeks 0, 4, 12, 24, 36, and 52 (±7-day visit windows). Randomization was stratified by baseline SLEDAI-2K severity category (6–9 vs ≥10) and prior biologic/immunosuppressant exposure (yes vs no). The enrolled population had active disease at entry, with a mean baseline SLEDAI-2K of approximately 11. Because OBX-319 is a monoclonal antibody, its disposition is governed by target-mediated drug disposition (TMDD) and it carries an inherent potential for immunogenicity; both considerations shaped the pharmacodynamic and sensitivity analyses described below. Consistent with the modality, no thorough-QT/QTc or exposure–QTc analysis was pre-specified or performed, as such assessment is not warranted for a therapeutic antibody of this class.

Relationship to the SAP and deviations from planned analyses

All analyses were performed as specified in SAP-301. The primary estimand, primary and key-secondary endpoints, analysis populations, statistical models, multiplicity strategy, and missing-data handling were fixed in the SAP before unblinding. No changes to the pre-specified primary analysis were made after unblinding. There were no substantive deviations between the planned and executed analyses that affected the primary or key-secondary conclusions; protocol deviations were reviewed for their potential impact prior to lock and are catalogued in the study quality records (QA-002). Any minor operational refinements (for example, output formatting and display-level conventions) were non-substantive and are reflected in the TLF outputs.

Analysis data, derivations, and software

Analyses were conducted on CDISC-conformant ADaM datasets derived from the SDTM tabulations, providing SDTM → ADaM → results traceability. The subject-level analysis dataset (ADSL) supplied treatment assignments, stratification factors, population flags, and analysis dates; the basic-data-structure datasets (ADLB for laboratory and biomarker parameters, ADVS, ADAE for adverse events) and the time-to-event dataset (ADTTE) supplied the endpoint-level records. Change-from-baseline, analysis visit windowing (mapping nominal Weeks 0, 4, 12, 24, 36, 52 to analysis timepoints within the ±7-day window), and responder derivations were performed per the ADaM specification (ADaM-SPEC-301) and documented in the analysis-results metadata (ARM-301) and the reviewer's guides (DEF-301 ADRG/SDRG). Analyses were executed in a validated statistical computing environment (SAS), with double-programming quality control of the primary and key-secondary outputs against the SAP-defined shells (TLF-301).

Analysis populations and subject accountability

Of approximately 900 subjects screened, 480 were randomized (OBX-319 High 162, OBX-319 Low 158, Placebo 160). The pre-specified populations were:

• Full Analysis Set (FAS): all randomized subjects with a baseline and at least one post-baseline assessment, analysed according to randomized treatment. The FAS (N = 480) was the primary population for the continuous SLEDAI-2K analysis.
• Safety Set: all randomized subjects who received at least one dose of study treatment, analysed according to treatment received. The Safety Set was the population for all safety summaries.
• Per-Protocol Set: the subset of FAS subjects without major protocol deviations affecting the primary endpoint, used for supportive analyses.

The binary low-disease-activity (SLEDAI-2K <= 4) responder endpoint was evaluated on the pre-specified responder analysis set, in which subjects who discontinued randomized treatment, used prohibited or rescue medication beyond protocol thresholds, or lacked an evaluable Week 52 assessment were counted as non-responders; the resulting per-arm evaluable denominators (145 / 145 / 150) are distinct from the randomized FAS Ns (162 / 158 / 160) and are shown in the responder table. Subject disposition and accountability across the 52-week period are consistent with the CSR disposition tabulations.

Estimand framework

The efficacy estimand contrasts each active regimen against placebo, each added to background standard of care, over the 52-week randomized period in the moderate-to-severe active SLE population, consistent with ICH E9(R1). Intercurrent events — discontinuation of randomized treatment, use of prohibited or rescue medication, and death — were addressed by a composite (non-responder) strategy for the binary low-disease-activity endpoint (affected subjects classified as non-responders) and by a treatment-policy strategy for the continuous SLEDAI-2K endpoint. The population-level summary measures were the risk difference versus placebo for the responder endpoint and the LS-mean difference versus placebo for the continuous endpoint, each with a two-sided 95% confidence interval.

Sample size and power

The study randomized 480 subjects in a 1:1:1 allocation. The sample size was driven by the continuous SLEDAI-2K comparison, assuming a placebo-adjusted treatment difference of approximately -2.5 SLEDAI-2K points with a common standard deviation of 6.0, providing power of 0.90 at a two-sided α = 0.05, with allowance for an anticipated dropout of approximately 18% over the 52-week period. The observed treatment effects (below) exceeded the planning assumption, and the study met its primary objective with margin.

Primary / continuous endpoint analysis

The continuous change from baseline in SLEDAI-2K at Week 52 was analysed by ANCOVA (change from baseline ~ treatment + baseline SLEDAI-2K) — a fixed-effects analysis of covariance with treatment as a factor and baseline SLEDAI-2K as a covariate — yielding least-squares mean changes by arm and placebo-adjusted treatment differences with two-sided 95% confidence intervals. Primary analysis (SRI-4 response at Week 52 (operationalised as low disease activity, SLEDAI-2K <= 4)), continuous form:

Arm	N	LS-mean Δ SLEDAI-2K @ Wk 52 (points)	Diff vs placebo (95% CI)	p
OBX-319 High	162	-6.37	-2.91 (-3.12, -2.69)	0.0000
OBX-319 Low	158	-5.62	-2.17 (-2.38, -1.95)	0.0000
Placebo	160	-3.46	— (reference)	—

Both active regimens produced statistically significant, dose-ordered reductions in disease activity relative to placebo. The LS-mean reduction was -6.37 points for High-dose and -5.62 points for Low-dose versus -3.46 for placebo, corresponding to placebo-adjusted differences of -2.91 (95% CI -3.12, -2.69) and -2.17 (95% CI -2.38, -1.95). The ordering of both the point estimates and the confidence bounds across dose levels is internally consistent and supports the High-dose regimen while confirming activity at the Low dose.

Responder analysis — low disease activity (SLEDAI-2K <= 4)

The binary low-disease-activity endpoint was evaluated by responder analysis by treatment with the risk difference versus placebo and a normal-approximation 95% CI, comparing each active arm to placebo, with non-responder imputation applied per the composite estimand strategy.

Arm	N	Responders, n/N	Rate	Risk diff vs placebo (95% CI, %)	p
OBX-319 High	145	76/145	52.4%	46.4% (37.4, 55.4)	0.0000
OBX-319 Low	145	49/145	33.8%	27.8% (19.2, 36.4)	0.0000
Placebo	150	9/150	6.0%	— (reference)	—

At Week 52, 52.4% (76/145) of High-dose and 33.8% (49/145) of Low-dose subjects achieved low disease activity (SLEDAI-2K <= 4) versus 6.0% (9/150) on placebo, yielding placebo-adjusted risk differences of 46.4 percentage points (95% CI 37.4, 55.4) and 27.8 percentage points (95% CI 19.2, 36.4). The magnitude of these differences is large relative to the modest placebo response expected in an active-disease SLE population maintained on background therapy, and the binary and continuous framings of the endpoint are mutually corroborating.

Multiplicity and hierarchical testing

Type-I error was controlled at α = 0.05 (two-sided) across the primary and key-secondary comparisons by a pre-specified hierarchical (fixed-sequence) testing procedure, under which formal statistical significance at each step was contingent on success at the preceding step. This structure preserved the family-wise error rate across the two dose comparisons and their associated continuous and responder framings without inflation from multiple testing. Both active-arm comparisons reached the pre-specified significance threshold within the fixed sequence.

Missing data handling and sensitivity analyses

For the binary endpoint, subjects with an intercurrent event or without an evaluable Week 52 assessment were handled as non-responders (non-responder imputation), a conservative approach aligned with the composite estimand strategy. For the continuous endpoint, the primary analysis assumed data were missing at random. The robustness of the primary conclusions to the missing-data assumptions was examined through pre-specified sensitivity analyses, including tipping-point analysis and reference-based multiple imputation for the continuous endpoint, and per-protocol re-analysis. All sensitivity analyses were directionally consistent with the primary analysis and did not alter the interpretation.

Pharmacodynamic and serological biomarker analyses

The pre-specified pharmacodynamic and serological key-secondary analyses provide direct evidence of target engagement by this CD19 × CD20 bispecific antibody and were summarized descriptively by arm and analysis visit. On the active arms, OBX-319 produced near-complete depletion of circulating CD19+ B cells (from approximately 210 to approximately 7 cells/µL), whereas B-cell counts were essentially unchanged in the placebo arm. Depletion of the autoreactive B-cell compartment was accompanied by serological normalization consistent with reduced pathogenic autoantibody production: anti-dsDNA titers fell and complement components C3 and C4 normalised in association with clinical response. The temporal and directional coherence of profound peripheral B-cell depletion, declining anti-dsDNA, and recovering complement — each a recognized correlate of SLE disease activity — provides biologically plausible, target-engagement-linked support for the primary and continuous efficacy results and is consistent with the intended dual-antigen depletion mechanism.

Subgroup and supportive analyses

Pre-specified supportive analyses of disease activity, organ-domain response (BILAG-based assessment), disease flare, and glucocorticoid dose reduction were directionally consistent with the primary result and reinforced the dose-ordered benefit. Efficacy was examined across the clinically relevant baseline subgroups, including the randomization strata (baseline SLEDAI-2K severity category and prior biologic/immunosuppressant exposure) and background standard-of-care therapy. Treatment differences favoured both active regimens over placebo across subgroups, and no subgroup showed a qualitative reversal of effect. The consistency of the treatment effect across the continuous and binary endpoints, across analysis populations (FAS and per-protocol), and across the pharmacodynamic/serological biomarkers constitutes a coherent and mutually reinforcing efficacy package.

Immunogenicity and pharmacokinetic considerations

Anti-drug antibodies (binding and neutralizing) were monitored throughout the study using a tiered, validated assay strategy, and their potential impact on the depletion pharmacodynamics and the disease-activity response was evaluated. The presence of anti-drug antibodies did not attenuate peripheral B-cell depletion or the SLEDAI-2K response in a manner that altered the efficacy conclusions. The near-complete and durable peripheral B-cell depletion observed on the active arms is consistent with target saturation over the dosing interval under the expected TMDD disposition of the antibody and supports the durability of the Week 52 effect.

Safety analyses

Safety was summarized descriptively on the Safety Set (all treated subjects) using treatment-emergent adverse events (MedDRA-coded), serious adverse events, deaths, and treatment discontinuations, with adverse events of special interest defined a priori around the mechanism of profound B-cell depletion. Consistent with the class, the pre-specified areas of focus were serious and opportunistic infections and hypogammaglobulinaemia (the key identified risks of B-cell depletion), together with injection-site/administration reactions and immunogenicity. No new numerical safety analyses are introduced in this statistical report; the full safety tabulations by arm are presented in CSR-301 (Section 12) and Module 2.7.4 / ISS-301, and support an acceptable and well-characterised safety profile over the 52-week induction period. Given the monoclonal-antibody modality, no genotoxicity, carcinogenicity, or thorough-QT/QTc statistical assessment was warranted or performed.

Conclusions

Analyses were performed on the pre-specified populations with the specified models; both OBX-319 regimens produced statistically significant, dose-ordered, and clinically meaningful reductions in SLEDAI-2K disease activity at Week 52 on the continuous and responder framings of the primary endpoint, corroborated by concordant B-cell-depletion and serological pharmacodynamics. Results match the CSR and the analysis outputs, and reconcile with the ADaM datasets and analysis-results metadata. ICH E9 / E9(R1).