Quantifying the ventilatory control contribution to sleep apnoea using polysomnography

Abstract

Elevated loop gain, consequent to hypersensitive ventilatory control, is a primary nonanatomical cause of obstructive sleep apnoea (OSA) but it is not possible to quantify this in the clinic. Here we provide a novel method to estimate loop gain in OSA patients using routine clinical polysomnography alone. We use the concept that spontaneous ventilatory fluctuations due to apnoeas/hypopnoeas (disturbance) result in opposing changes in ventilatory drive (response) as determined by loop gain (response/disturbance). Fitting a simple ventilatory control model (including chemical and arousal contributions to ventilatory drive) to the ventilatory pattern of OSA reveals the underlying loop gain. Following mathematical-model validation, we critically tested our method in patients with OSA by comparison with a standard (continuous positive airway pressure (CPAP) drop method), and by assessing its ability to detect the known reduction in loop gain with oxygen and acetazolamide. Our method quantified loop gain from baseline polysomnography (correlation versus CPAP-estimated loop gain: n=28; r=0.63, p<0.001), detected the known reduction in loop gain with oxygen (n=11; mean±sem change in loop gain (ΔLG) -0.23±0.08, p=0.02) and acetazolamide (n=11; ΔLG -0.20±0.06, p=0.005), and predicted the OSA response to loop gain-lowering therapy. We validated a means to quantify the ventilatory control contribution to OSA pathogenesis using clinical polysomnography, enabling identification of likely responders to therapies targeting ventilatory control.

Conflict of interest statement

Conflict of interest: Disclosures can be found alongside the online version of this article at erj.ersjournals.com

Copyright ©ERS 2015.

Figures

FIGURE 1

Mathematical basis of the method. a) Schematic of the feedback loop controlling ventilation showing the influence of arousal and airflow obstruction. Ventilatory drive is the sum of chemical drive and the response to arousal (γ) (equation 1 in the main text). Airflow obstruction provides a disturbance that reduces ventilation from the intended level (i.e. ventilatory drive). In response, chemical drive rises as determined by the chemical control system (loop gain). b) Time course of chemical drive during a step reduction in ventilation (e.g. obstructive hypopnoea). The rise in chemical drive is governed by and the parameters that determine its gain (LG0), time constant (τ) and delay (δ) (equation 2 in the main text); these system characteristics are revealed in the time course of ventilation when the airway is reopened.

FIGURE 2

Mathematical model validation. a) Example simulation showing that loop gain is accurately recovered from ventilation in a model of obstructive sleep apnoea (loop gain, LG1 is the response to a 1-cycle·min−1 disturbance). Shaded regions denote periods of obstruction. The estimated chemical drive (solid smooth black line) is precisely superimposed on true chemical drive (dashed black line is not visible due to near-perfect overlap); likewise, estimated ventilatory drive (green staircased line) is closely overlaid upon the observed ventilation (blue staircase line) in the absence of obstruction. b) Group simulation data show that the method accurately reveals the true loop gain given to the model. Model parameters: delay 12 s, time constant 12.5 s and response to arousal 0.4 (40% eupnoeic ventilation). Obstructive events were imposed by halving the controller gain (doubling resistance) for three or more breaths at random times in a graduated manner. Arousals were imposed for two breaths at the termination of 80% of obstructive events and on 1% of unobstructed breaths. #: normalised such that 1=eupnoea.

FIGURE 3

Estimating loop gain using diagnostic polysomnography. Example traces illustrate epochs with a) relatively low loop gain (response to a 1-cycle·min−1 disturbance (LG1)=0.6) and b) relatively high loop gain (LG1=1.1). Note that ventilatory drive (chemical drive + response to arousal) closely fits ventilation during periods of unobstructed airflow. Loop gain determines the increase in chemical drive in response to the reduction in ventilation. EEG: electroencephalogram; RIP: respiratory inductance plethysmography. #: normalised.

FIGURE 4

Comparison of our method and the continuous positive airway pressure (CPAP) drop method for measuring loop gain. Agreement was observed across a range of frequencies including a) “mid-frequency” (1 cycle·min−1 (LG1)), b) “high frequency” (LG2) and c) “low frequency” (LG1/6; 6-min period). Note that loop gain (the chemical drive response to a reduction in ventilation) is a function of the frequency (e.g. timing) of the disturbance in ventilation.

FIGURE 5

Detecting the reduction in loop gain with oxygen. a) Reduction in loop gain (response to a 1-cycle·min−1 disturbance (LG1)) with oxygen versus baseline (B). b) Reduced ventilatory response to arousal (γ), as a fraction of mean ventilation, with oxygen. c) The feedback system’s natural cycling period (Tn) rose with oxygen (i.e. feedback was more sluggish). Data are presented as mean±SEM.

FIGURE 6

Detecting the reduction in loop gain with acetazolamide (ACZ). a) Reduction in loop gain (response to a 1-cycle·min−1 disturbance (LG1)) with ACZ versus baseline (B). b) Reduced ventilatory response to arousal (γ), as a fraction of mean ventilation, with ACZ. c) The feedback system’s natural cycling period (Tn) rose with ACZ (i.e. feedback was more sluggish). Note that in one subject, LG1 and other variables were not measured from the obstructive sleep apnoea pattern on ACZ due to insufficient obstructive events. The open circle represents a patient whose loop gain unexpectedly rose with ACZ as confirmed with the continuous positive airway pressure drop method. Data are presented as mean±SEM.

FIGURE 7

Predicting responses to lowering loop gain with oxygen and acetazolamide (ACZ). a) A larger reduction in sleep apnoea severity with oxygen or ACZ was seen when treatment induced a greater fall in loop gain in response to a 1-cycle·min−1 disturbance (LG1). b) The reduction in aponea–hypopnoea index (AHI) could be predicted a priori by a high baseline LG1 and c) a low baseline cycling period (Tn); that is, responders have a more sensitive and brisk feedback response than nonresponders. The outlier (open circle) whose LG1 rose greatly and unexpectedly (confirmed by the continuous positive airway pressure drop method) was excluded from associations in (b) and (c) because the intention was to examine the effectiveness of lowering loop gain on AHI. AHI: apnoea–hypopnoea index. #: Spearman rank correlation.