📚 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 Report (TILA278-201)

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

Statistical Analysis Report — TILA278-201

This Statistical Analysis Report (SAR) documents the executed analyses for Study TILA278-201, a Phase 2b, randomized, double-blind, placebo-controlled, parallel-group induction study of TILA-278 in adults with moderately-to-severely active Ulcerative Colitis (UC), conducted by Virtual Biopharma Inc. TILA-278 is a subcutaneously administered, CHO cell–derived, humanized bispecific IgG1 monoclonal antibody that simultaneously antagonizes TL1A (TNFSF15) and agonizes the IL-22 receptor pathway; the TL1A-antagonist arm attenuates TH1/TH17-driven mucosal inflammation and pro-fibrotic signaling, while the IL-22R-agonist arm promotes epithelial regeneration and mucosal-barrier repair. All analyses were pre-specified in the Statistical Analysis Plan (SAP-201) and executed against the locked, validated analysis datasets; the results reproduced here are identical to those presented in the Clinical Study Report (CSR-201) and the underlying analysis outputs (Tables, Listings, and Figures). The report is prepared in accordance with ICH E9 and the estimand and sensitivity-analysis framework of ICH E9(R1).

The two subcutaneous dose regimens (TILA-278 High and TILA-278 Low) were compared with placebo over a 12-week induction period, with efficacy visits at Weeks 0, 2, 4, 8, and 12. The primary objective was to demonstrate superiority of TILA-278 over placebo for the induction of clinical remission at Week 12. Because TILA-278 is a therapeutic monoclonal antibody, no dedicated thorough-QT or proarrhythmic statistical analysis was included in this program, as such an evaluation is not warranted for a monoclonal antibody; cardiovascular safety was addressed through routine electrocardiogram and vital-sign monitoring summarized descriptively.

Analysis Populations and Subject Disposition

Analysis populations were defined prospectively in SAP-201. The Full Analysis Set (FAS) comprised all randomized subjects with a baseline and at least one post-baseline efficacy assessment and served as the primary population for the responder-based efficacy analyses under the intention-to-treat principle. The Safety Set comprised all randomized subjects who received any study drug and was the basis for safety summaries. The randomized (all-randomized) population supported the model-based analysis of the continuous modified Mayo endpoint under the treatment-policy estimand. Randomization was stratified by baseline modified Mayo severity (moderate, 4–6; severe, 7–9) and by prior biologic exposure (biologic-naïve; biologic-experienced) to balance these prognostic factors across arms.

Table 1. Subject Disposition and Analysis Sets (Study TILA278-201)

Disposition	TILA-278 High	TILA-278 Low	Placebo	Total
Screened	—	—	—	1700
Randomized	299	300	301	900
Full Analysis Set (FAS)	284	283	273	840
Discontinued study	17	17	29	63

Of 1700 subjects screened, 900 were randomized 1:1:1 to TILA-278 High (299), TILA-278 Low (300), or placebo (301). The FAS comprised 840 subjects (284 High, 283 Low, 273 placebo). Study discontinuations were more frequent under placebo (29 subjects) than under either active regimen (17 each), driven predominantly by lack of efficacy in the placebo group; this differential discontinuation pattern was accounted for in the estimand definition and in the missing-data handling described below. Baseline modified Mayo severity and prior biologic exposure were comparably distributed across arms, confirming successful stratified randomization.

Estimands and Analysis Framework (ICH E9(R1))

The primary estimand was defined using the five attributes of ICH E9(R1). Treatment condition: TILA-278 High or TILA-278 Low administered subcutaneously over the 12-week induction period, each versus placebo. Population: adults aged 18–75 years with moderately-to-severely active UC (baseline modified Mayo score 4–9). Variable (endpoint): for the primary analysis, clinical remission at Week 12, defined as a modified Mayo score ≤ 2 with no individual subscore > 1; for the key secondary continuous analysis, change from baseline in the modified Mayo score at Week 12. Intercurrent events: premature discontinuation of study drug, initiation of prohibited or rescue therapy (including corticosteroid escalation, additional biologic or advanced therapy, or UC-related surgery). A treatment-policy strategy was applied for the model-based continuous endpoint (all randomized subjects, irrespective of intercurrent events), whereas for the dichotomous remission endpoint a composite strategy was used in which subjects experiencing a confounding intercurrent event or lacking a Week-12 assessment were classified as non-responders. Population-level summary: the difference between each active arm and placebo, expressed as a risk difference in remission proportions and as a difference in LS-mean change in the modified Mayo score.

The modified Mayo score was derived from the stool-frequency, rectal-bleeding, and centrally read endoscopy subscores (the Physician Global Assessment subscore is excluded from the modified instrument). All endoscopies contributing to the endoscopy subscore and to the endoscopic-improvement endpoint were evaluated by a blinded central reader to remove site-level assessment variability. Type-I error was controlled at a two-sided α = 0.05, and the study was sized to provide 0.9 power for the primary comparison under the assumed effect size.

Statistical Methods

Responder (primary) analysis. Clinical remission at Week 12 was analyzed on the FAS with non-responder imputation for missing or intercurrent-event–confounded data. The treatment effect was summarized as the risk difference versus placebo with a normal-approximation (Wald) 95% confidence interval, as pre-specified in SAP-201; a Cochran–Mantel–Haenszel test stratified by the randomization factors (baseline severity, prior biologic exposure) provided the supportive stratified test of association. Each active regimen was compared with placebo.

Continuous (key secondary) analysis. Change from baseline in the modified Mayo score at Week 12 was analyzed by analysis of covariance (ANCOVA) with treatment group and baseline modified Mayo score as covariates, performed on the randomized population consistent with the treatment-policy estimand; missing Week-12 values were addressed by multiple imputation under a missing-at-random (MAR) assumption. LS-means, LS-mean differences versus placebo, associated standard errors (approximately 0.05 for each LS-mean estimate), 95% confidence intervals, and p-values were derived from the fitted model.

Multiplicity. Family-wise Type-I error was preserved at 0.05 across the two doses and the ordered endpoint family by a pre-specified fixed-sequence (hierarchical) testing procedure: within each dose, testing proceeded from clinical remission to change in the modified Mayo score to endoscopic improvement, with formal inference contingent on success at the preceding step. This structure protected the reported p-values against inflation from multiple comparisons.

Missing data and sensitivity analyses. The primary MAR-based results were stress-tested by pre-specified sensitivity analyses, including non-responder imputation for the binary endpoint, control-based (jump-to-reference) multiple imputation, and a tipping-point analysis to characterize the degree of departure from MAR that would be required to overturn the primary conclusion. Immunogenicity was evaluated as a potential effect modifier: anti-drug antibody (ADA) status was examined as a covariate and in sensitivity analyses (cross-referenced to the immunogenicity report, IMMUNO-001), and exposure–response was explored using population-PK–derived exposure metrics appropriate to the target-mediated drug disposition (TMDD) characteristics of the molecule. All analyses were executed in validated SAS® software (Version 9.4).

Primary and Key Secondary Efficacy Results

The executed analyses for TILA278-201 per the SAP. Primary analysis (Clinical remission (modified Mayo) at Week 12):

Arm	N	LS-mean Δ Modified Mayo Score @ Wk 12 (points)	Diff vs placebo (95% CI)	p
TILA-278 High	299	-3.36	-2.36 (-2.49, -2.23)	0.0000
TILA-278 Low	300	-2.76	-1.77 (-1.90, -1.64)	0.0000
Placebo	301	-1.00	— (reference)	—

The model-based ANCOVA showed a monotonic, dose-ordered reduction in the modified Mayo score at Week 12. The LS-mean change from baseline was −3.36 points for TILA-278 High and −2.76 points for TILA-278 Low, versus −1.00 points for placebo, corresponding to LS-mean differences versus placebo of −2.36 (95% CI −2.49 to −2.23) and −1.77 (95% CI −1.90 to −1.64), respectively. Both differences were statistically significant, and the confidence intervals were narrow — reflecting the size and precision of the study and the consistency of the effect — with neither interval approaching the null. The rank ordering of the LS-mean reductions (High −3.36 > Low −2.76 > Placebo −1.00) provides continuous-endpoint confirmation of the treatment effect that is concordant with the dichotomous remission result below.

Responder analysis — Clinical remission (Mayo <= 2)

Arm	N	Responders, n/N	Rate	Risk diff vs placebo (95% CI, %)	p
TILA-278 High	284	106/284	37.3%	36.6% (30.9, 42.3)	0.0000
TILA-278 Low	283	46/283	16.2%	15.5% (11.1, 19.9)	0.0000
Placebo	273	2/273	0.7%	— (reference)	—

Clinical remission at Week 12 was achieved by 37.3% (106/284) of TILA-278 High subjects and 16.2% (46/283) of TILA-278 Low subjects, versus 0.7% (2/273) of placebo subjects. The risk difference versus placebo was +36.6 percentage points (95% CI 30.9 to 42.3) for the High dose and +15.5 percentage points (95% CI 11.1 to 19.9) for the Low dose; both comparisons were statistically significant and were confirmed by the stratified Cochran–Mantel–Haenszel test. The near-absent placebo remission rate (0.7%) is consistent with the stringent composite remission definition and the moderate-to-severe baseline population and underscores the magnitude of the treatment effect. Both the absolute benefit of the High dose (+36.6 percentage points) and the clear separation of the Low dose from placebo (+15.5 percentage points) exceed thresholds generally regarded as clinically meaningful for UC induction, and the dose ordering (High > Low > Placebo) is preserved on both the continuous and the dichotomous scale.

Key Secondary Endpoint — Endoscopic Improvement at Week 12

Endoscopic improvement at Week 12, defined as a centrally read Mayo endoscopic subscore ≤ 1, was the objective mucosal-healing endpoint. Results are summarized in Table 2.

Table 2. Endoscopic Improvement at Week 12 (centrally read endoscopic subscore ≤ 1; FAS)

Treatment	n/N	Rate	Difference vs placebo (percentage points)	p
TILA-278 High	139/284	48.9%	+42.7	0.0000
TILA-278 Low	79/283	27.9%	+21.7	0.0000
Placebo	17/273	6.2%	— (reference)	—

Endoscopic improvement was achieved by 48.9% of TILA-278 High subjects and 27.9% of TILA-278 Low subjects, versus 6.2% under placebo (High vs placebo +42.7 percentage points; Low vs placebo +21.7 percentage points; both statistically significant within the fixed-sequence hierarchy). The objective endoscopic benefit — evaluated by blinded central reading — is consistent with the IL-22R–agonist, mucosal-healing component of the TILA-278 mechanism and corroborates the symptomatic (remission) and continuous (Mayo-change) findings.

Time-Course of the Continuous Endpoint

Change from baseline in the modified Mayo score was assessed longitudinally to characterize the onset and durability of effect across the induction period (Table 3).

Table 3. LS-Mean Change from Baseline in Modified Mayo Score Over the Induction Period (FAS)

Visit	TILA-278 High	TILA-278 Low	Placebo	High − Placebo	Low − Placebo
Week 2	−1.10	−0.90	−0.45	−0.65	−0.45
Week 4	−1.95	−1.55	−0.68	−1.27	−0.87
Week 8	−2.80	−2.25	−0.88	−1.92	−1.37
Week 12	−3.36	−2.76	−1.00	−2.36	−1.77

Separation from placebo was evident at the first post-baseline assessment (Week 2) and deepened progressively through Week 12, with the dose-ordered ranking maintained at every visit and the treatment difference versus placebo increasing monotonically over time. The Week-12 estimates correspond to the primary ANCOVA analysis; minor differences from the arithmetic difference of the rounded LS-means reflect rounding of the model-based estimates. The early onset followed by continued deepening of response is consistent with the combined mechanism of TILA-278 — rapid dampening of mucosal inflammation via TL1A antagonism followed by progressive IL-22R–driven epithelial regeneration — and supports the adequacy of a 12-week induction period.

Subgroup Analyses

Pre-specified subgroup analyses of the primary endpoint were performed by the randomization stratification factors. The treatment effect was consistent in direction and clinically relevant in magnitude across all subgroups, with the dose ordering preserved throughout.

Table 4. Clinical Remission at Week 12 by Baseline Modified Mayo Severity (FAS)

Subgroup	TILA-278 High, n/N (%)	TILA-278 Low, n/N (%)	Placebo, n/N (%)
Moderate (baseline Mayo 4–6)	76/170 (44.7)	34/169 (20.1)	2/163 (1.2)
Severe (baseline Mayo 7–9)	30/114 (26.3)	12/114 (10.5)	0/110 (0.0)

Table 5. Clinical Remission at Week 12 by Prior Biologic Exposure (FAS)

Subgroup	TILA-278 High, n/N (%)	TILA-278 Low, n/N (%)	Placebo, n/N (%)
Biologic-naïve	68/156 (43.6)	30/155 (19.4)	2/150 (1.3)
Biologic-experienced	38/128 (29.7)	16/128 (12.5)	0/123 (0.0)

Remission rates were higher in subjects with moderate baseline disease than in those with severe disease, and higher in biologic-naïve than in biologic-experienced subjects, as expected; nonetheless, both TILA-278 doses produced a clinically meaningful benefit over placebo in every subgroup, and the dose ordering (High > Low) was preserved. Retained efficacy in the severe stratum (placebo remission 0.0%) and in biologic-experienced subjects (High 29.7% vs placebo 0.0%) is consistent with the distinct dual anti-TL1A/IL-22R mechanism, which does not rely on pathways targeted by conventional biologic therapies.

Sensitivity, Immunogenicity, and Exposure Analyses

The primary MAR-based conclusions were robust to the pre-specified sensitivity analyses. Non-responder imputation and control-based (jump-to-reference) multiple imputation yielded treatment effects consistent in direction and significance with the primary result, and the tipping-point analysis indicated that implausibly extreme departures from MAR would be required to overturn the primary remission finding. ADA status, examined as a covariate and in supplementary analyses, did not materially alter the estimated treatment effect; the immunogenicity findings are detailed in IMMUNO-001. Exploratory exposure–response analyses using population-PK–derived exposure metrics were consistent with the observed dose ordering, in keeping with the target-mediated disposition of the molecule. Supportive biomarkers (C-reactive protein and faecal calprotectin) declined in the active arms in proportion to modified Mayo improvement, with the greatest reductions in the High-dose group, providing biochemical concordance with the symptomatic and endoscopic results.

Conclusions

Both TILA-278 dose regimens were statistically significantly and clinically meaningfully superior to placebo for the induction of clinical remission at Week 12, with concordant, dose-ordered benefits on the continuous modified Mayo endpoint, on centrally read endoscopic improvement, and on supportive biomarkers, and with a treatment effect that was consistent across baseline severity and prior biologic exposure. The multiplicity-protected primary and key secondary analyses, together with robust sensitivity analyses, support a genuine disease-modifying induction effect rather than an isolated finding on a single measure.

Analyses were performed on the pre-specified populations with the specified models; results match the CSR and the analysis outputs. ICH E9 / E9(R1).