The shape of carbon-dioxide waveforms produced during ordinary breathing may reveal patterns associated with cardiopulmonary disease, opening the possibility of extracting more clinical information from a familiar monitoring tool. Daniel Neville, researchers at Portsmouth Hospitals University NHS Trust and collaborators at TidalSense describe the approach in the General Breathing Record study, published in the European Clinical Respiratory Journal.
Capnography plots the concentration of carbon dioxide across each breath. The study investigated whether detailed features within those waveforms, combined with analytical methods, could help distinguish cardiopulmonary conditions. If successful, the approach could move capnography beyond a single end-tidal carbon-dioxide measurement toward richer pattern recognition.
The technology is already familiar in anesthesia and emergency care, where the carbon-dioxide waveform helps clinicians confirm ventilation and detect sudden changes in breathing. Neville and his colleagues are asking a broader question: whether the precise shape and timing of many ordinary breaths contain patterns associated with chronic disease of the lungs, heart or circulation.
The General Breathing Record approach preserves more information than a single end-tidal carbon-dioxide value. Waveform features may reflect how different regions of the lungs empty, whether airflow is obstructed, how breathing is timed and how effectively the circulation delivers carbon dioxide back to the lungs. Analytical models can then test combinations of features that may be too subtle for visual inspection at the bedside.
The project is part of a wider movement toward signal-rich respiratory assessment. Spirometry records forced airflow during a coached maneuver. Pulse oximetry estimates oxygen saturation. A capnogram records carbon dioxide throughout the ordinary breathing cycle. None provides a complete diagnosis, but each offers a different view of ventilation, gas exchange and circulation.
Machine-learning methods make it possible to examine hundreds of waveform features at once. That creates both opportunity and risk. A model may uncover a reproducible physiological pattern, or it may learn artifacts associated with a particular device, clinic or patient population rather than the disease itself.
Prospective testing, calibration standards and clear reporting of sensitivity and specificity will therefore be essential. Clinicians will also need an explanation they can act on, rather than an unexplained score produced by a black box. The most credible future use may be triage: identifying patients who should receive confirmatory lung or cardiac testing while leaving the final diagnosis to established clinical evaluation.
The approach remains investigational and should not be confused with a consumer breath test. Algorithms trained in one clinical population can perform poorly in another, making independent validation and transparent reporting of false positives and false negatives especially important.
Any useful diagnostic signal would also need to survive ordinary real-world variation. Age, posture, medication, anxiety, recent physical activity and the quality of the recording can all alter a capnography trace. A model must distinguish disease from a poor-quality sample and be tested in populations that were not used to build it, including patients with overlapping heart and lung conditions.
The next hurdle is prospective validation among patients whose diagnoses are not already known. A waveform tool will be clinically useful only if it adds reliable information beyond the patient’s history, physical examination and existing tests without producing an unmanageable number of false alarms.
If those hurdles are cleared, capnography could develop into a low-burden screening or triage tool rather than remaining primarily a safety monitor. Its value will ultimately depend on whether waveform analysis can find important disease sooner, reduce unnecessary testing or direct patients toward the right confirmatory examination. The decisive test will take place in ordinary clinics, where the diagnosis is still uncertain and the next step is not always clear.
