Sleep apnea treatments not working? New findings could inform better treatment

Lead researcher

Dr Scott Sands

Main finding

Under normal circumstances, breathing is a monotonously repetitive sequence of inspiratory and expiratory efforts that ensure little change to arterial O2 and CO2 levels over time. However, humans commonly exhibit an irregular breathing pattern, called periodic breathing (PB), in which a small number of breaths alternate with a prolonged period of apnea. PB occurs at sea level in most preterm and term infants and in some adults with an idiopathic form. PB also occurs in almost all adults on ascent to high altitude. Our interest in PB was sparked by the potential dangers of repetitive apneas, which cause blood oxygen levels to plummet, potentially exposing patients to severe hypoxaemia and its adverse impact on brain metabolism and the cardiovascular system.

We chose to study patients with heart failure, many of whom exhibit an exaggerated form of PB known as Cheyne-Stokes respiration (CSR). In such patients it is relatively easy to measure all the variables of interest and to test the idea one can predict the level of inspired CO2 that will restore continuous breathing in each patient.

In the work published in the European respiratory Journal, we used a mathematical method developed in our laboratory to calculate what is known as loop gain of the respiratory control system: engineering control theory states that CSR will persist should loop gain continue to exceed a value of 1.0 after treatment, but if it falls below 1.0, continuous breathing should be restored. We confirmed that it is possible to calculate an effective dose of CO2, but in some patients it proved necessary to raise inspired CO2 concentration until loop gain fell to 0.8 or less to convert CSR to continuous breathing.

By demonstrating the predictive value of control theory, our study provides a firm theoretical basis for designing a treatment that will resolve CSR in an individual patient and for understanding why particular treatments fail. Such a finding takes us one step closer to personalizing the treatment of this very common sleep disorder.


The Ritchie Centre

Research group

Respiratory Control Laboratory


Scott A. Sands1,2,3,4*, Bradley A. Edwards3,5,6, Kirk Kee1,2, Christopher Stuart-Andrews1, Elizabeth M. Skuza4, Teanau Roebuck1, Anthony Turton7,
Garun S. Hamilton7,8, Matthew T. Naughton1,2, Philip J. Berger4

Journal and article title

European Respiratory Journal

Control Theory Prediction of Resolved Cheyne-Stokes Respiration in Patients With Heart Failure

Most surprising

Our work proves the concept that the level of inspired CO2 intervention required to restore continuous breathing in CSR is predictable knowing LGbaseline and Ftherapy. That is, a higher loop gain requires a stronger, yet predictable, intervention to achieve stable breathing. By demonstrating the predictive value of control theory, our study provides a firm theoretical basis to judge which treatments will resolve CSR in an individual patient and for understanding why particular treatments fail.

Future implications

Knowledge that a treatment such as continuous positive airway pressure overnight will not prevent repetitive apnea during sleep provides a clinician with means to spare patients from what are often unpleasant and poorly tolerated treatments. Conversely, knowing the size of a treatment that will work to prevent apnea provides a clinician with means to limit treatment size and to improve adherence with therapy.

Disease/health impact

Sleep apnea and heart failure

Other points of interest

Effective treatment of CSR remains a major challenge, and patients with severe ventilatory control instability—who do not respond well to CPAP—require novel solutions. To date we have only empirical approaches to therapy: we try them, not knowing whether they will work, or why they fail. As an alternative, we demonstrate proof-of-principle for the first time that the resolution of CSR can be predicted when we have a measure of the strength of the intervention being given to lower loop gain and we know the baseline loop gain. Application of a therapeutic dose that leaves the residual loop gain above 1 is futile, and can be readily predicted. We envisage that employing control theory, alongside loop gain measurement, will enable clinicians to estimate the magnitude of therapeutic intervention needed to resolve the underlying instability driving CSR in an individual patient. A quantitative approach to selecting sufficiently potent individual therapies, or combinations of therapies, may provide the means to restore continuous respiration and sleep even in patients with the most severe and currently intractable instability.