Moving beyond an empiric trial to using combined physiology and anatomy to predict success of oral appliances in obstructive sleep apnoea

Both diagnosis and severity of obstructive sleep apnoea (OSA) have traditionally been assessed through the apnoea–hypopnoea index, a simple count of how often obstruction occurs. However, it is now widely accepted that this metric performs poorly in predicting the success of various non-continuous positive airway pressure (CPAP) therapies; the rate of obstructive events appears to be insensitive to the cause(s) of obstruction in a particular patient. This failure has led several groups to propose that an interaction of multiple pathophysiological factors underlie OSA and that specific therapies may work best in a subset of patients. The ‘holy grail’ has been to define such subsets using this ‘phenotypic’ approach. The problem is exemplified by response to oral appliances, as there is a substantial failure rate in unselected patients. As the time to titration is often prolonged and cost of an empiric therapeutic trial is non-trivial because of needing multiple titration visits and sleep studies, it would be advantageous to be able to select patients with a higher than average success rate. The simplest, but so far unsuccessful, approach to estimating the chance of success has been to use clinical, polysomnographic (PSG) and/or anthropomorphic measures of the patient at baseline. This assumes that anatomy drives OSA and that a relatively linear relation exists between external features, e.g. body mass index, neck size and awake visual assessment of the upper airway, and the causes of obstruction. The lack of success of these predictions in most published trials supports a competing assumption that sleep, airway mechanics and neural reflexes are as important as anatomy, and that their relative contribution is not evident without explicit testing. In this issue of The Journal of Physiology, Marques and colleagues test one such approach (Marques et al. 2019). Previous evidence, quoted by the authors, suggests that tongue position and measurement of collapsibility of the airway may individually modestly predict success of oral appliances, but their combined value has not been explored. In recent years, research techniques have been developed to assess the ‘phenotype’ of a given patient’s OSA. This is variably defined, and a recent trend is to focus on combinations of physiological ‘traits’ to define such a ‘phenotype’. Because the tools to measure the traits that contribute to the pathophysiology of obstruction require methodologies that are difficult and not suited to the clinical PSG environment, surrogates of the major components (anatomy, airway collapsibility or Pcrit, upper airway responsiveness and loop gain) are also being developed and tested for their ability to predict clinical outcomes. In their study Marques and colleagues test the utility of tongue position and Pcrit individually to predict the outcome of oral appliance therapy using state of the art techniques (sleep endoscopy and determination of Pcrit using CPAP drops). More importantly than the individual utility of these traits, they show that when combined in a prediction equation, an accuracy of 19/22 cases was achieved, with high positive and negative predictive value. If confirmed, this implies that it is possible to select patients prior to trying the therapy and predict success and failure of an oral appliance. However, as they themselves point out, the tools used – sleep endoscopy to define the role of the tongue in obstruction, and Pcrit determination using CPAP drops during a sleep study – are invasive, expensive and not well suited to clinical use. It remains to be tested if, as the authors suggest, non-invasive surrogates of these determinations, that are becoming available, will be equally predictive. The major contribution of this paper is the demonstration that it is possible to predict success of therapy from pre-therapy physiology, and thus there is reinforcement for trying to simplify the testing. Clinicians should anxiously await the follow-up data showing that parameters from the clinical PSG or other readily available clinical tests can perform equally well.