Phase II

Phase II Endpoint and Model Review — NTX-101

Week 24 efficacy, responder analysis, and benchmarked predictive modeling for NTX-101 proof-of-concept evaluation.

Gate Decision: GO

Primary model metrics

MetricValue
Threshold0.494
ROC AUC0.867
PR AUC0.845
Brier0.145
Sensitivity0.806
Specificity0.75
PPV0.758
NPV0.8
F10.781
N (test)63
Responder diff (pp)19.073
Gate decisionGO
Primary modellogistic_regression
Primary endpoint p-value0.0019
Threshold rationaleTraining ROC threshold maximizing specificity with sensitivity ≥ 0.78

Responder rates by arm

TreatmentResponder rate (%)
ntx101_high57.534
ntx101_low50.725
placebo38.462

Hypothesis testing summary

ComparisonMean differencet-statisticp-valueApprox CI
ntx101_low vs placebo -0.188 -2.142 0.034 [-0.313, -0.064]
ntx101_high vs placebo -0.266 -3.526 0.0006 [-0.355, -0.178]

Benchmark models

ModelROC AUCPR AUCBrierF1
logistic_regression 0.868 0.845 0.145 0.781
elastic_net 0.867 0.845 0.146 0.794
random_forest 0.838 0.849 0.173 0.75

Figures