📚 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 Plan (OBX-319)

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

Statistical Analysis Plan (Synopsis)

Study: OBX319-301 · Compound: OBX-319 · Indication: Systemic Lupus Erythematosus (moderate-to-severe active)

OBX-319 is an anti-CD19 × anti-CD20 B-cell-depleting bispecific humanized IgG1 monoclonal antibody (Chinese hamster ovary [CHO] cell culture, subcutaneous administration) developed for moderate-to-severe active Systemic Lupus Erythematosus (SLE) by Virtual Biopharma Inc. Study OBX319-301 is a Phase 3, randomized, double-blind, placebo-controlled, three-arm trial evaluating two dose levels of OBX-319 (High and Low) versus Placebo administered subcutaneously on a background of standard-of-care immunosuppression over a 52-week double-blind treatment period. Allocation is 1:1:1. This Statistical Analysis Plan (SAP) pre-specifies all populations, estimands, endpoints, statistical models, multiplicity control, and missing-data handling in accordance with ICH E9(R1); it is finalised prior to database lock and unblinding and supports the efficacy and safety analyses to be included in the Biologics License Application filed under 21 CFR 601. Unless otherwise stated, all hypothesis tests are two-sided and confidence intervals are two-sided at the 95% level.

Analysis populations

Full Analysis Set (FAS): all randomized subjects with a baseline and ≥1 post-baseline assessment, analysed according to randomized treatment assignment (intention-to-treat principle). The FAS is the primary population for all efficacy analyses. Safety Set: all randomized subjects who received ≥1 dose of study drug, analysed according to treatment actually received. Additional analysis populations are pre-specified to support the modality-specific endpoints of a therapeutic monoclonal antibody:

• Per-Protocol Set (PPS): FAS subjects without major protocol deviations affecting the primary endpoint (used for supportive robustness analyses only).
• Pharmacokinetic (PK) Analysis Set: subjects with ≥1 evaluable post-dose serum OBX-319 concentration.
• Pharmacodynamic (PD)/Biomarker Analysis Set: subjects with baseline and ≥1 post-baseline peripheral CD19+ B-cell count and/or serologic (anti-dsDNA, complement C3/C4) assessment.
• Immunogenicity (ADA-evaluable) Set: subjects with a baseline and ≥1 post-baseline anti-drug antibody (ADA) result.

A total of 480 subjects were randomized (162 High / 158 Low / 160 Placebo). Randomization was stratified by baseline SLEDAI-2K disease activity, baseline oral corticosteroid dose (prednisone-equivalent), and geographic region to protect the balance of prognostic factors across arms. Approximately 145 High, 145 Low, and 150 Placebo subjects are expected to be evaluable for the Week 52 primary endpoint after accounting for discontinuation; disposition and reasons for withdrawal are summarised by arm.

Primary endpoint & estimand

SRI-4 response at Week 52 (operationalised as low disease activity, SLEDAI-2K <= 4). The SLE Responder Index-4 (SRI-4) is a composite requiring, relative to baseline: a ≥4-point reduction in SLEDAI-2K; no new BILAG-2004 A organ-domain score and no more than one new BILAG-2004 B organ-domain score; no clinically meaningful worsening (increase of <0.30 points) in the Physician's Global Assessment (PGA); and no discontinuation of study treatment or use of prohibited/rescue medication beyond protocol-permitted thresholds. In this study the SLEDAI-2K component is operationalised as attainment of low disease activity, SLEDAI-2K ≤ 4, at Week 52.

The primary estimand is specified using the five ICH E9(R1) attributes:

• Treatment condition: OBX-319 High (subcutaneous) versus Placebo, and OBX-319 Low versus Placebo, each added to background standard-of-care.
• Population: randomized subjects with moderate-to-severe active SLE (FAS), baseline SLEDAI-2K approximately 11.
• Variable (endpoint): binary SRI-4 responder status at Week 52 as defined above.
• Intercurrent events and strategies: a treatment-policy strategy is applied for the primary estimand, so that the SRI-4 outcome is attributed regardless of premature discontinuation of study drug or changes in background therapy; permitted intercurrent events include treatment discontinuation, death, and initiation of prohibited immunosuppressants or additional biologics. Rescue/corticosteroid escalation above the protocol cap and treatment discontinuation are additionally handled through the responder definition (counted as non-response) in the supplementary composite estimand described under sensitivity analyses.
• Population-level summary: the risk difference in SRI-4/low-disease-activity response proportion between each active arm and Placebo.

A treatment-policy estimand compares each active arm with Placebo.

Primary analysis

ANCOVA of change from baseline in SLEDAI-2K at Week 52 with treatment and baseline as covariates; the responder analysis uses risk differences vs placebo with a normal-approximation 95% CI. Type-I error is controlled at α = 0.05 (power 0.9). Missing data are handled under a MAR assumption with pre-specified sensitivity analyses.

The binary SRI-4/low-disease-activity endpoint is analysed as the difference in response proportions (active − Placebo). Subjects with missing Week 52 responder status are classified as non-responders (non-responder imputation, NRI), consistent with the composite handling of discontinuation and rescue-medication intercurrent events; the primary risk-difference estimate and its normal-approximation 95% CI are supported by a Cochran–Mantel–Haenszel analysis and a logistic-regression model including treatment, baseline SLEDAI-2K, and the randomization stratification factors. The pre-specified response proportions are 52.4% (76/145) High, 33.8% (49/145) Low, and 6.0% (9/150) Placebo, corresponding to risk differences versus Placebo that anchor the powering of the study.

The continuous key secondary, change from baseline in SLEDAI-2K, is analysed both by the Week 52 ANCOVA above and, as the primary continuous model, by a mixed model for repeated measures (MMRM) including treatment, visit, treatment-by-visit interaction, baseline SLEDAI-2K, and stratification factors, with an unstructured covariance matrix and Kenward–Roger degrees of freedom. Least-squares (LS) mean changes are −6.37 (High), −5.62 (Low), and −3.46 (Placebo), giving differences versus Placebo of −2.91 (High) and −2.17 (Low).

Multiplicity control. To preserve the family-wise Type-I error at two-sided α = 0.05 across the two active doses and the ordered secondary endpoints, a hierarchical fixed-sequence testing procedure is applied. Testing proceeds only while each preceding null hypothesis is rejected, in the order: (1) OBX-319 High versus Placebo on the primary SRI-4/low-disease-activity endpoint; (2) OBX-319 Low versus Placebo on the primary endpoint; (3) OBX-319 High versus Placebo on SLEDAI-2K LS-mean change; (4) OBX-319 Low versus Placebo on SLEDAI-2K LS-mean change; followed by the remaining ranked secondary endpoints. No alpha is recycled, and the full α = 0.05 is available at each step reached.

Sample size

480 randomized (allocation 1:1:1) supports the primary comparison under the assumed effect size (see study.yaml statistics). The sample size is driven by the primary SRI-4/low-disease-activity comparison of each active arm versus Placebo, assuming an active-arm response proportion of up to approximately 52.4% against a Placebo response proportion of approximately 6.0%, a two-sided significance level of α = 0.05, and 90% power (power 0.9). Nominal allocation is 160 subjects per arm; provision for approximately 10% non-evaluability yields roughly 145 evaluable subjects per active arm and 150 in Placebo at Week 52, preserving power for the fixed-sequence comparisons. The design retains adequate power for the second (Low-dose) comparison within the hierarchy given the observed separation from Placebo.

Key secondary and other endpoints

Ranked key secondary endpoints, tested within the fixed-sequence hierarchy after the primary comparisons, comprise change from baseline in SLEDAI-2K (LS-mean changes −6.37 High / −5.62 Low / −3.46 Placebo; differences versus Placebo −2.91 / −2.17), BILAG-2004 organ-domain improvement, PGA change, corticosteroid-sparing (proportion of subjects tapering background prednisone-equivalent to ≤10 mg/day without flare), and annualised flare rate (SELENA-SLEDAI Flare Index). Serologic normalisation (anti-dsDNA and complement C3/C4) and time to first flare are analysed as additional (non-multiplicity-controlled) endpoints and summarised descriptively with nominal 95% CIs. Continuous endpoints use MMRM; binary endpoints use risk differences with normal-approximation 95% CIs and supportive logistic regression; time-to-event endpoints use Kaplan–Meier estimates with Cox proportional-hazards models stratified by the randomization factors.

Intercurrent events, missing data & sensitivity analyses

Under the primary MAR assumption, missing continuous data are addressed by the likelihood-based MMRM and missing responder status by NRI. Pre-specified sensitivity and supplementary analyses assess robustness to departures from MAR and to alternative intercurrent-event handling: (a) a supplementary composite (while-on-treatment) estimand counting treatment discontinuation and prohibited rescue-medication use as non-response; (b) multiple imputation and jump-to-reference (control-based) imputation for the continuous endpoint under plausible missing-not-at-random scenarios; (c) tipping-point analysis to identify the imputation assumption that would overturn a statistically significant primary result; and (d) a per-protocol supportive analysis. Subgroup analyses of the primary endpoint are pre-specified for baseline SLEDAI-2K stratum, baseline corticosteroid dose, geographic region, prior biologic exposure, and baseline autoantibody status, presented as forest plots of risk differences with nominal 95% CIs and treatment-by-subgroup interaction tests for consistency.

Pharmacokinetic and pharmacodynamic analyses

Serum OBX-319 concentrations are summarised by nominal time and dose; population PK is characterised using a model that accommodates target-mediated drug disposition (TMDD), reflecting binding to CD19 and CD20 on B cells and the nonlinear clearance expected of a B-cell-directed bispecific antibody. Individual exposure metrics (e.g., steady-state trough and average concentration) are derived for exposure–response analyses linking exposure to the SRI-4/low-disease-activity response and to key safety events. Pharmacodynamic analyses summarise peripheral CD19+ B-cell counts over time by arm; near-complete depletion is expected on both active arms (approximately 210 → approximately 7 cells/µL) versus essentially unchanged counts on Placebo. Serologic PD markers (anti-dsDNA titre and complement C3/C4) are summarised as change from baseline and as shift-to-normalisation tables, with the expected pattern of falling anti-dsDNA and normalising C3/C4 accompanying clinical response. Exposure–response and PK/PD relationships (drug exposure → B-cell depletion → serologic normalisation → SRI-4 response) are explored graphically and, where appropriate, by regression modelling.

Immunogenicity analyses

Anti-drug antibodies are assessed using a validated tiered assay strategy (screening, confirmatory, and titre) with neutralising-antibody (NAb) evaluation of confirmed-positive samples. Immunogenicity is summarised on the ADA-evaluable set as the incidence of treatment-emergent and treatment-boosted ADA and NAb by arm and over time. The potential impact of ADA/NAb status on PK (OBX-319 exposure), PD (B-cell depletion and serologic markers), efficacy (SRI-4/low-disease-activity response), and safety (including injection reactions and hypersensitivity) is examined descriptively by ADA status.

Safety analyses

All safety analyses are performed on the Safety Set and are descriptive. Treatment-emergent adverse events (TEAEs) are summarised by MedDRA system organ class and preferred term, by severity, and by relationship to study drug, with exposure-adjusted incidence rates (EAIR per 100 subject-years) to account for differential exposure. Adverse events of special interest reflect the identified risks of B-cell depletion with this modality and are analysed with dedicated summaries: serious and opportunistic infections; hypogammaglobulinaemia, supported by longitudinal summaries and shift tables of serum immunoglobulin (notably IgG) concentrations; injection/infusion reactions; and immunogenicity-related hypersensitivity. Laboratory data, vital signs, and 12-lead ECGs are summarised descriptively with shift tables; consistent with a therapeutic monoclonal antibody, no formal thorough-QT/cardiodynamic analysis is planned, as centralised QT interval characterisation is not warranted for this modality. Deaths, serious adverse events, and events leading to discontinuation are tabulated by arm.

Interim analyses and data monitoring

An independent Data Safety Monitoring Board (DSMB) periodically reviews unblinded safety data throughout the double-blind period, with particular attention to the class risks of serious/opportunistic infection and hypogammaglobulinaemia. Any planned interim assessment is limited to non-binding futility and safety monitoring; no efficacy stopping boundary is applied and no Type-I error is spent at interim, so the full two-sided α = 0.05 is preserved for the final analysis. All DSMB analyses are conducted by an unblinded independent statistician external to the study team to protect the integrity of the double blind.