AIPH-TB: AI-Optimised Pyrazinamide-Hydroxychloroquine vs Standard RIPE for Drug-Sensitive Pulmonary Tuberculosis - A Phase II RCT

Tuberculosis (TB) kills 1.3 million people annually and remains the world's deadliest bacterial disease. The standard four-drug RIPE regimen achieves only 85% cure rates and causes drug-induced hepatotoxicity in 25-37% of patients. Hydroxychloroquine (HCQ), an FDA-approved antimalarial, has been shown to synergise with pyrazinamide (PZA) by inhibiting the BCRP-1 efflux pump and raising phagolysosomal pH, increasing intracellular PZA concentrations (FICI 0.38 in vitro). The AIPH-TB computational framework (Artificial Intelligence Physicochemical Harmonisation for Tuberculosis) uses multi-objective reinforcement learning, Gaussian process regression, and a digital twin macrophage simulator to identify an AI-optimised dosing schedule that maximises this synergy (PZA 1,500 mg + HCQ 200 mg at 0800 and HCQ 200 mg at 2000), maintaining phagolysosomal pH within 5.2-5.8 for 18 of 24 hours. The computational model predicts FICI 0.28 (strongly synergistic), 9.4-fold increase in intracellular PZA concentration, 99.5% cure rate, and \<1.5% hepatotoxicity. This Phase II randomised controlled trial will test whether the AI-optimised PYZ-HCQ protocol is superior to standard RIPE in 200 newly-diagnosed drug-sensitive pulmonary TB patients over 6 months of treatment with 6 months of follow-up.

BACKGROUND AND RATIONALE:

Pyrazinamide (PZA) is the only first-line agent active against dormant intracellular MTB, making it irreplaceable for sterilising activity. Its clinical utility is limited by BCRP-1-mediated efflux - after entering the phagolysosome, PZA is rapidly expelled before it can be protonated to its active form, pyrazinoic acid (POA). Hydroxychloroquine (HCQ) inhibits BCRP-1 and raises phagolysosomal pH. The AIPH-TB AI framework identified that an oscillating HCQ schedule (0800/2000) maintains optimal pH 5.2-5.8 for 18 h/day - a 125% improvement over unoptimised dosing - and predicts a novel second mechanism: reduction of mycobacterial cell wall zeta potential from -18 mV to -8 mV, increasing membrane permeability to POA by 340%.

STUDY DESIGN OVERVIEW:

This is a Phase II, open-label, randomised, parallel-group, active-controlled superiority trial conducted at two tertiary TB treatment centres in Riyadh, Saudi Arabia. Participants will be randomised 1:1 to receive either the AIPH-TB protocol (Arm A) or standard RIPE therapy (Arm B) for 6 months, with 6 months post-treatment follow-up (total study duration per participant: 12 months).

RANDOMISATION:

Block randomisation (block size 4 and 6, randomly varied), stratified by site and HIV status. Centralised web-based randomisation via REDCap (MOH Research Directorate Biostatistics Unit).

BLINDING:

Open-label study. Laboratory personnel processing sputum cultures and liver enzyme results are blinded to treatment arm (assessor-blind for primary outcomes). The DSMB will conduct unblinded interim analyses.

SAMPLE SIZE:

Total 200 participants (100 per arm). Based on 80% sputum culture conversion at Week 8 for standard RIPE vs 95% for AIPH-TB (15 percentage point difference), alpha=0.05, power=80%, with 25% dropout inflation.