Measuring cough from proven foundations to intelligent automation

Objective cough monitoring has become a cornerstone of chronic cough drug development, providing a reliable and clinically meaningful endpoint for evaluating treatment efficacy. VitaloJAK is the most widely used objective cough monitoring system in respiratory clinical trials and, as of March 2026, has been used in 58 commercial clinical studies, generating over 57,000 24h recording sessions and contributing to the counting of approximately 33 million coughs.

Over the past decade, VitaloJAK has supported multiple drug development programmes across chronic cough, idiopathic pulmonary fibrosis (IPF), and chronic obstructive pulmonary disease (COPD), spanning Phase 2a, Phase 2b and Phase 3 clinical trials. These include landmark studies such as Merck/MSD COUGH-1, COUGH-2, Bellus Health/GSK SOOTHE and Bayer PAGANINI, among others (see key trial summary at the end of this article).

The breadth of experience generated through these programmes has created a unique expertise in the field of cough. This experience has not only supported the delivery of regulatory-quality cough endpoints but has also enabled the development of next-generation technologies, including automated cough detection systems such as ABACAS and a new cough monitoring device.

From manual analysis to automated cough detection

Although cough counting by humans remains the benchmark for accuracy, increasing study sizes and the growing adoption of cough endpoints have created a need for more scalable approaches. To address this challenge, Vitalograph have developed ABACAS, an automated cough analysis algorithm trained using large, clinically representative datasets (Taylor et al. Performance evaluation of ABACAS AI cough analysis system in a chronic cough cohort. ERS Cough Conference 2026 Poster PS11).

Several aspects of the training dataset are critical to ABACAS performance:

• Use of real-life 24-hour recordings that capture environmental variability encountered during daily life.

• Disease-specific chronic cough data from 61 subjects containing 191,332 cough events.

• Device-specific algorithm development using 300 VitaloJAK recordings.

These principles were adopted to ensure that ABACAS was developed using data representative of the conditions encountered in real clinical trials rather than highly controlled laboratory environments. As a result, the algorithm was designed to perform across a broad range of recording conditions, background noises, and patient behaviours.

Large-scale validation in chronic cough

The performance of ABACAS was evaluated using one of the largest chronic cough validation datasets reported to date. Validation was performed on 563 independent 24-hour recordings collected from 85 subjects with chronic cough, comprising 368,125 cough events (NCT04866563, the data shared with Vitalograph as a special agreement with the trial sponsor). Importantly, all validation recordings were independent of the development dataset, reducing the risk of performance inflation.

Using analyst-generated cough timestamps as ground truth, ABACAS achieved:

• Median sensitivity: 90.5% (IQR 81.1%–95.3%)

• Median positive predictive value (PPV): 95.2% (IQR 88.7%–97.6%)

One of the key themes of the presentation was the importance of understanding variability in algorithm performance. While ABACAS demonstrated strong median performance, meaningful differences were observed both between subjects and, in some cases, between recording days within the same subject. Lower performance was typically associated with factors such as very low cough frequencies, acoustically challenging environments, or the presence of cough-like sounds that can be difficult even for sophisticated algorithms to distinguish from true cough events. Although these situations occurred in a minority of recordings, they are important to recognise when automated cough counting is intended for use in clinical trials.

No automated cough detection system is expected to achieve identical sensitivity and PPV across all subjects and all recording days. Consequently, both between-subject and within-subject performance variability should be considered when assessing endpoint reliability and when planning clinical studies. This is particularly relevant for sample size estimation, where the impact of measurement variability needs to be appropriately accounted for. The findings also reinforce the importance of evaluating automated systems using clinically meaningful endpoints, such as cough frequency, rather than relying solely on event-level performance measures. Strong sensitivity and PPV are important but understanding how those metrics translate into cough count accuracy is ultimately what determines the utility of an automated endpoint.

Beyond endpoint monitoring, ABACAS was evaluated for subject screening applications. Using screening recordings from 84 chronic cough participants and a threshold of at least 10 coughs per hour, the algorithm achieved 96.4% accuracy in study inclusion/exclusion decisions (Taylor et al. Exploring the feasibility of using ABACAS® as an automated cough frequency screening tool. ERS Cough Conference 2026 Poster PS114). This finding suggests that automated cough detection may have practical utility in accelerating study recruitment and reducing operational burden during clinical trial screening.

At the ERS Cough Conference 2026, Vitalograph also previewed a forthcoming cough monitoring device intended to satisfy the full spectrum of requirements across endpoint measurement, exploratory research, regulatory applications, and clinical utility in healthcare facilitating both human and automated cough counting. The goal is to combine patient-centric design with the scientific rigor required for objective respiratory monitoring. The company will be releasing more information about the new device at the ERS Conference in Barcelona.

Conclusion

The development of ABACAS reflects a broader shift in respiratory research toward scalable objective cough measurement. However, automated cough counting is not simply a replacement for manual analysis but a continuously evolving technology. Insights gained from large-scale validation datasets are already informing the development of future versions of ABACAS designed to improve performance in challenging acoustic environments and patient populations[LC1.1]. Understanding subject-level variability, evaluating endpoint-level accuracy, and validating algorithms under real-world conditions are all essential if automated tools are to support pivotal clinical trials and future clinical applications. As cough therapeutics continue to advance, objective measurement technologies will remain critical to demonstrating treatment benefit and enabling evidence-based patient care.

List of VitaloJAK clinical trial highlights

Phase 3

McGarvey et al. 2022. Efficacy and safety of gefapixant, a P2X3 receptor antagonist, in refractory chronic cough and unexplained chronic cough (COUGH-1 and COUGH-2): results from two double-blind, randomised, parallel-group, placebo-controlled, phase 3 trials. Drug trial sponsored by Merck Sharp & Dohme/ MSD.

Phase 2b

Dicpinigaitis et al. 2023. Efficacy and safety of eliapixant in refractory chronic cough: The randomized, placebo-controlled Phase 2b PAGANINI study. Drug trial sponsored by 5Imageaddexpandmore-dotsLogo Image checkCompacts the image to logo display. Block imageENAttach image Hide For LocalesNone (show in all locales)DEENBayer.

Smith et al. 2025. Camlipixant in refractory chronic cough: A Phase 2b, randomized, placebo-controlled trial (SOOTHE). Drug trial sponsored by GSK.

Maher et al. 2023. Nalbuphine Tablets for Cough in Patients with Idiopathic Pulmonary Fibrosis. Drug trial sponsored by Trevi Therapeutics.

NCT06504446. Study to Assess the Efficacy, Safety, and Tolerability of NOC-110 in Adults With Refractory or Unexplained Chronic Cough (ASPIRE). Drug trial sponsored by Nocion Therapeutics.

Phase 2a

NCT05660850. A Study To Evaluate The Efficacy, Safety, Pharmacokinetics, And Pharmacodynamic Effects Of GDC-6599 In Patients With Chronic Cough. Drug trial sponsored by Genentech/Roche.

Singh et al. 2022. The novel bronchodilator navafenterol: A Phase 2a, multicentre, randomised, double-blind, placebo-controlled crossover trial in COPD. Drug trial sponsored by Astra Zeneca.

Niimi et al. 2022. Randomised trial of the P2X3 receptor antagonist sivopixant for refractory chronic cough. Drug trial sponsored by Shionogi.

NCT04866563. A Study of Efficacy and Safety of AX-8 in Chronic Cough. Drug trial sponsored by Axalbion.