📚 Part of the TILA-278 Regulatory Dossier — Reader's Guide. This article shows the live document; edits to the source appear here automatically.

---

Statistical Analysis Plan (TILA-278)

Document ID: SAP-201 Version: 1.0 Change History: 1.0 — Initial issue. Standard(s): ICH E9(R1)

Statistical Analysis Plan (Synopsis)

Study: TILA278-201 · Compound: TILA-278 · Indication: Ulcerative Colitis (moderate-to-severe)

This Statistical Analysis Plan (SAP) pre-specifies, in advance of database lock and unblinding, the populations, estimands, hypotheses, statistical models, multiplicity control, missing-data strategy, and sensitivity/supplementary analyses for Study TILA278-201, sponsored by Virtual Biopharma Inc. TILA-278 is a Chinese Hamster Ovary (CHO)–derived humanized IgG1 bispecific monoclonal antibody administered subcutaneously; one arm antagonizes TL1A (delivering anti-inflammatory and anti-fibrotic effects) while the second arm agonizes the IL-22 receptor (promoting intestinal epithelial and mucosal healing). TILA278-201 is a Phase 2b, randomized, double-blind, placebo-controlled induction study in which subjects with moderate-to-severe Ulcerative Colitis (UC) were allocated 1:1:1 to a TILA-278 High-dose subcutaneous regimen, a TILA-278 Low-dose subcutaneous regimen, or matching Placebo, and treated over a 12-week induction period. The plan is written in accordance with ICH E9 and the ICH E9(R1) addendum on estimands and sensitivity analyses, and it is consistent with contemporary regulatory expectations for UC induction trials. All analyses are performed with validated statistical software (SAS version 9.4 or higher and/or R), and, unless otherwise stated, hypothesis tests are two-sided at a family-wise significance level of α = 0.05. This SAP governs the clinical-efficacy analyses only; product-quality, nonclinical, and clinical-pharmacology conventions relevant to a bispecific antibody are referenced where they bear on the statistical treatment of pharmacokinetic (PK), immunogenicity, and biomarker data.

Analysis populations

Efficacy and safety inferences are drawn from clearly separated, pre-defined analysis sets so that the randomization-based comparison is protected and the safety characterization reflects treatment actually received.

• Full Analysis Set (FAS): all randomized subjects with a baseline and at least one post-baseline assessment of the primary efficacy variable. Subjects are analyzed according to the treatment to which they were randomized (intention-to-treat principle). The FAS is the primary population for all efficacy analyses. Of the 900 subjects randomized (299 High-dose, 300 Low-dose, 301 Placebo), 840 subjects (284 High-dose, 283 Low-dose, 273 Placebo) meet the FAS criteria and constitute the primary efficacy population.
• Safety Set: all randomized subjects who received at least one dose (or partial dose) of study drug, summarized according to the treatment actually received; in this fully dosed study the Safety Set corresponds to the randomized set. Injection-site tolerability, infections, hepatic events, hypersensitivity/systemic administration reactions, and all other safety endpoints are summarized on this set.
• Per-Protocol Set (PPS): the subset of the FAS without important protocol deviations judged to affect the primary endpoint (e.g., major eligibility violations, prohibited concomitant therapy, or insufficient exposure). The PPS supports the primary endpoint as a supplementary robustness analysis; it is not used for primary inference.
• Pharmacokinetic (PK) Analysis Set: all subjects who received TILA-278 and have at least one evaluable post-dose concentration, used for population-PK characterization of the antibody's target-mediated disposition and for exposure–response analyses.
• Immunogenicity (ADA) Analysis Set: all treated subjects with at least one evaluable post-baseline anti-drug antibody (ADA) result, used to summarize ADA incidence, titer, persistence, neutralizing status, and the relationship of ADA to PK, efficacy, and safety.
• Biomarker/Pharmacodynamic (PD) Set: all subjects with evaluable baseline and post-baseline mechanistic biomarker data (e.g., fecal calprotectin, high-sensitivity C-reactive protein), used to characterize the dual TL1A-antagonist / IL-22R-agonist pharmacology.

Primary endpoint & estimand

Primary endpoint. Clinical remission per modified Mayo score at Week 12, defined as a modified Mayo score ≤ 2 with no individual subscore > 1. The modified Mayo score is the sum of three subscores — stool frequency (0–3), rectal bleeding (0–3), and centrally read endoscopy (0–3) — yielding a total range of 0–9. Endoscopy is read by a central, blinded reader to remove local-read variability, and remission requires an endoscopy subscore ≤ 1 through the "no subscore > 1" condition.

Estimand. Consistent with ICH E9(R1), the primary estimand is a treatment-policy estimand defined by the following five attributes:

1. Treatment condition: the TILA-278 subcutaneous induction regimen (High-dose or Low-dose) versus matching Placebo over the 12-week induction period.
2. Population: adults with moderate-to-severe active UC meeting the protocol entry criteria (the randomized population, operationalized as the FAS).
3. Variable (endpoint): the binary indicator of Week-12 clinical remission as defined above.
4. Intercurrent events (ICEs) and handling strategies: the treatment-policy strategy applies to the comparison of interest — the effect of treatment as assigned, irrespective of adherence. Anticipated ICEs and their pre-specified strategies are: discontinuation of study treatment (treatment-policy; values sought regardless); initiation of prohibited or rescue therapy for UC, including corticosteroid escalation, other advanced therapy, or colectomy (composite strategy — such subjects are classified as non-remitters at Week 12); and death (composite — non-remitter). Because Week-12 remission requires an on-schedule clinical and central-endoscopy assessment, subjects without an evaluable Week-12 status are treated as non-remitters (non-responder imputation), which operationalizes the treatment-policy target for an outcome that is unobservable once the subject is off study.
5. Population-level summary: the difference in clinical-remission proportions between each active dose and Placebo (risk difference), with an associated 95% confidence interval (CI).

The corresponding continuous estimand for the key secondary endpoint (change from baseline in modified Mayo score at Week 12) is likewise defined under a treatment-policy strategy, summarized as the least-squares (LS) mean treatment difference from Placebo.

Primary analysis

Primary hypothesis and test. For each active dose, the null hypothesis of no difference from Placebo in Week-12 clinical-remission proportion is tested against a two-sided alternative. The primary responder analysis compares each active arm with Placebo using risk differences vs Placebo with a normal-approximation (Wald) 95% CI, complemented by a Cochran–Mantel–Haenszel test stratified by the randomization stratification factors (e.g., prior advanced-therapy exposure, baseline disease severity, and baseline corticosteroid use). Non-responder imputation is the primary handling for missing or post-ICE remission status, as specified in the estimand.

Continuous endpoint model. Change from baseline in Modified Mayo Score at Week 12 is analyzed by ANCOVA with treatment and baseline as covariates; a mixed-model repeated-measures (MMRM) analysis including treatment, visit, treatment-by-visit interaction, baseline, and baseline-by-visit terms with an unstructured covariance matrix provides the longitudinal, MAR-based supportive estimate and the LS-mean change and LS-mean difference from Placebo.

Multiplicity control. Type-I error is controlled at α = 0.05 (power 0.9) across the two dose comparisons and the ordered key secondary endpoints using a pre-specified hierarchical (fixed-sequence, optionally graphical/recycling) testing procedure. Testing proceeds from the High-dose primary comparison, to the Low-dose primary comparison, and then through the ranked key secondary endpoints — including endoscopic improvement (Mayo endoscopic subscore ≤ 1), clinical response, symptomatic remission, endoscopic remission, and corticosteroid-free remission — with formal significance claimed only while the sequence remains unbroken. This structure preserves the family-wise error rate at 0.05 two-sided without inflating the sample-size assumptions.

Missing data and sensitivity analyses. Missing data are handled under a MAR assumption with pre-specified sensitivity analyses. For the binary primary endpoint, non-responder imputation is primary; sensitivity to the missing-data mechanism is examined via multiple imputation under MAR, a tipping-point analysis that stress-tests departures from MAR (MNAR) in the active arms, and an as-observed (completers) analysis. For the continuous endpoint, MMRM under MAR is supported by control-based/jump-to-reference multiple imputation reflecting a conservative MNAR scenario. Concordance across these analyses establishes robustness of the primary conclusion.

Subgroup and supplementary analyses. Consistency of the primary treatment effect is assessed across pre-specified subgroups (e.g., prior advanced-therapy exposure, baseline corticosteroid use, baseline disease severity, region, sex, age, and body weight) via forest plots of risk differences with treatment-by-subgroup interaction terms interpreted descriptively. The PPS supports the FAS analysis as a robustness check.

Clinical pharmacology, immunogenicity, and biomarker analyses. Reflecting the target-mediated drug disposition (TMDD) expected of this bispecific antibody, PK are characterized by population-PK modeling accommodating nonlinear, target-mediated clearance, and exposure–response analyses relate individual exposure metrics to Week-12 remission and to key safety events to inform dose selection. Immunogenicity is summarized as treatment-emergent ADA incidence, titer, time-course, persistence, and neutralizing status, with descriptive assessment of the impact of ADA on PK, efficacy, and safety. Pharmacodynamic biomarkers consistent with the dual mechanism — TL1A antagonism (anti-inflammatory/anti-fibrotic) and IL-22R agonism (epithelial/mucosal healing), e.g., fecal calprotectin and high-sensitivity CRP — are summarized as change from baseline by treatment.

Safety analyses. Adverse events are coded with MedDRA and summarized as treatment-emergent event incidence by system organ class and preferred term, with particular attention to injection-site reactions and systemic administration reactions inherent to subcutaneous antibody dosing, infections and serious infections, and hepatic parameters. Laboratory values, vital signs, and standard 12-lead electrocardiograms are summarized descriptively; consistent with the pharmacology of a monoclonal antibody, no dedicated exposure–QTc modeling or thorough cardiac-repolarization analysis is planned, as such assessment is not warranted for this modality. Safety summaries are descriptive and are not part of the confirmatory testing hierarchy.

Interim analyses and data monitoring. An independent Data Monitoring Committee conducts unblinded periodic safety reviews. No efficacy interim analysis with alpha spending is planned; the full two-sided α = 0.05 is therefore available at the single final analysis, preserving the stated power of 0.9.

Sample size

The study randomized 900 subjects in a 1:1:1 allocation (approximately 300 per arm) to the TILA-278 High-dose regimen, the TILA-278 Low-dose regimen, and Placebo, supporting the primary comparison under the assumed effect size (see study.yaml statistics). At this size, and after allowance for a limited proportion of subjects without evaluable post-baseline data, the study provides at least 90% power (two-sided α = 0.05) to detect the anticipated between-group difference in Week-12 clinical remission for each active dose versus Placebo. The 900 randomized subjects were drawn from 1700 screened, and were allocated as 299 High-dose, 300 Low-dose, and 301 Placebo; the resulting Full Analysis Set comprises 840 subjects (284 High-dose, 283 Low-dose, 273 Placebo), which drives the realized power of the primary responder comparison. The two active-dose comparisons are ordered within the fixed-sequence testing hierarchy so that the family-wise error rate remains controlled at 0.05 without a downward adjustment to the per-comparison sample size, and the common per-arm allocation ensures balanced precision for both the binary primary endpoint and the continuous key secondary endpoint.