Remote Monitoring and Future Models of Liver Disease Care

Overview

Prof. Jalan described how liver care can move from reactive, episodic clinic visits to continuous, closed-loop remote monitoring for decompensated cirrhosis and alcohol-related liver disease, with dynamic AI prediction as the next step. He disclosed that CirrhoCare and AlcoChange are products of CyberLiver, a company he co-founded.

Key Clinical Topics

• Why episodic care fails after decompensation
• Earlier telehealth and detection tools, and their limits
• CirrhoCare closed-loop monitoring: alert thresholds, pilot outcomes and clinician workload
• Measuring alcohol as a trajectory: PEth, urine EtG and verified breath sampling
• AlcoChange digital behaviour-change platform
• Design of two NIHR-funded UK randomised trials
• Barriers to implementation
• AI-based dynamic prediction of organ failure

Key Learning Points

• Reactive care: After decompensation, about half of patients (nearly 53%) are readmitted within 90 days. Most readmissions are for ascites, encephalopathy and infection, which develop over days at home and give a window to intervene.
• Closing the loop: Remote monitoring only matters if an alert triggers a clinical response while the patient is still at home.
• Earlier tools were open systems: Telehealth, EncephalApp and speech analysis showed that access and detection were possible, but none joined detection to treatment.
• CirrhoCare routine: Each morning patients spend 12-15 minutes recording smartwatch data, blood pressure, bioimpedance weight and body water, the CL-ART cognitive test and a self-report. Alerts are judged against the patient's own baseline and followed by text or voice contact.
• Alert thresholds: Systolic BP <95 mmHg or a fall >15%. Heart rate >90 bpm or a rise >10%. Weight change >1.5 kg over 3 days. A failed or slowed CL-ART. No data for >2 days.
• Thresholds prompt judgement: They are not automatic rules. Each alert was reviewed by a hepatologist who knew the patient, and responses were modest adjustments to diuretics, laxatives, rifaximin or fluids.
• Complementary alcohol markers: PEth reflects the past 2-4 weeks but is modulated by sex, haemolysis and cirrhosis. Urine EtG reflects recent days. Breath alcohol is event-level, quantitative and unaffected by cirrhosis. When breath and PEth disagree, that is a reason to talk to the patient.
• Verified breath sampling: Two scheduled tests and one random prompt daily, with identity confirmed by camera. Missed prompts are themselves a signal of likely drinking.
• AlcoChange: Delivers 34 validated behaviour-change techniques in nine categories in real time and links to a companion and a clinical dashboard.
• Implementation barriers: Reach, sustained engagement, staff workload and responsibility, and proof of value. The technology itself is no longer the bottleneck.
• AI performance: Discrimination is good, but calibration is imperfect and needs further work.

Clinical Data & Study Results

• CirrhoCare pilot (Kazankov, J Hepatol 2023): 20 patients vs 20 contemporaneous, non-randomised controls (CLIF-C AD score about 50). 85% met the primary feasibility endpoint over 10 weeks.
• Pilot outcomes: Total readmissions 8 vs 13. Median length of stay 5 vs 7 days. ICU days 5 vs 16. Unplanned large-volume paracentesis 1 vs 6. The study was not powered for these outcomes, so this is a signal to test rather than an effect to claim.
• Workload: 1,397 patient-days monitored, 61 clinical calls, 66 technical contacts and 147 automated reminders. Review of all 20 patients took <1 h/day. 2 of 8 readmissions were arranged early by the team from dashboard data.
• AlcoChange proof of concept (Mehta, JHEP Rep 2024): Open-label study, 65 recruited and 41 completed. Patients with ≥60 logins vs fewer logins had a median intake change of −100% vs −57% (p=0.029), abstinence of 57.1% vs 22.2% (p=0.025) and fewer alcohol-related readmissions at 12 months (p=0.008). There was no control arm.
• CirrhoCare-RCT: 214 patients at ≥12 UK centres. The primary endpoint is unplanned medical intervention for a new liver-related complication within 90 days, powered for 35% vs 60%. Final follow-up is due in December.
• AlcoChange RCT: 400 patients randomised 1:1. Follow-up is complete and results are expected shortly.
• AI model (Yeo, eGastroenterology 2026): Predicts ≥2 extrahepatic organ failures 12 hours ahead in ICU patients with cirrhosis. AUROC 0.83 internally and 0.77 on external validation.

Practical Takeaways

• Treat the post-discharge period after decompensation as high risk. Complications evolve at home between visits.
• Remote monitoring should be paired with a defined clinical response pathway, not used as passive data collection.
• Use personalised baselines and simple thresholds (BP, heart rate, weight, cognition) as prompts for clinician judgement.
• Combine PEth with frequent breath testing to track alcohol use over time, and explore discordant results with the patient.
• Plan for staffing, out-of-hours responsibility and technical support before adopting monitoring platforms.
• Await randomised trial and cost-effectiveness data before treating pilot signals as proof of benefit.