Scant evidence has been reported on what features clinicians currently use, and should use, to predict which patients with central apneas will benefit from an initial trial of CPAP, as opposed to proceeding to another modality such as ASV. We report on several demographic, etiologic, and sleep study parameters that independently associated with the prescription of an initial CPAP trial, the likelihood of an adequate response, or both:
Patient characteristics: Prior work suggests older age is associated with ventilatory control instability [25–27], which may explain why providers in this study were less likely to prescribe CPAP in older adults. However, this physiologic rationale did not translate to a higher CPAP failure rate in practice, perhaps because CPAP sufficiently stabilizes ventilatory control[28] or pharyngeal collapsibility has an even greater contribution to pathogenesis in older adults[29]. The discordance between provider-predicted response (represented by the likelihood of initial prescription for CPAP) and the likelihood of an adequate outcome suggests prescribing patterns might be improved for this group of patients.
Etiology: This study corroborates previously described low response rates to CPAP for patients with opiate use [30] and higher CPAP response rates when central apneas occur in the setting of OSA, as in TECSA [31; 32]. Prior data suggest roughly half of the patients with CSA related to CHF have normalization of the AHI with CPAP [33], but we are unaware of previous comparative data between etiologies. We found that central apneas cause by cardiac or neurologic conditions respond to CPAP at roughly average covariate-adjusted rates. Similar to prior reports [18; 34], we found that primary CSA is very rare (9 of 588), and thus conclusions about CPAP prescribing or responsiveness remain tentative.
The proportion of central vs. obstructive events: Like prior studies, we find that most patients with central apneas have substantial overlap with OSA (only 11 of 588 had > 90% central apneas)[21; 35]. Current management is implicitly dichotomized at the threshold of 50% central events that defines whether a patient is labeled as principally OSA or CSA[1]. We find a “dose”-response relationship, where patients in a higher percentage of central events category were less likely to receive a CPAP trial and less likely to have an adequate response when trialed. This suggests that the proportion of central events might be better conceptualized as a continuous spectrum. Future work utilizing improved event-level differentiation between central and obstructive events [36–38] or endotyping patient-level features [39; 40], such as predisposition for upper-airway collapse and ventilation control abnormalities, have the potential to refine this paradigm further.
Relatedly, one role of a disease definition is to delineate patients expected to respond to specific treatments. However, for patients with opiate-related central apneas, rates of an adequate CPAP trial were low even when central apneas constituted < 50% of total events. Therefore, opiate-associated central apneas might be considered “central sleep apnea syndrome” even when the proportion of central apneas is lower than 50%.
Care delivery parameters: Patients diagnosed with HSAT had a higher rate of initial CPAP trial and low rates of subsequent nonadherence or modality switch, even after controlling for etiology and other factors (see Supplement, S2). This suggests that unexpected identification of central apneas on home testing is often successfully treated with CPAP. This is reassuring, given trends toward HSATs due to insurance requirements, the COVID pandemic, and the convenience of HSATs that have led to more frequent identification of CSA on HSAT in recent years.
Our center is at a comparatively high elevation for the United States (4500ft above sea level), which may influence the likelihood [13] and physiology [14] of CSA. However, we did not find evidence that the elevation of residence influenced provider prescribing or rates of adequate response to CPAP.
Patterns of breathing: Lastly, Cheyne-Stokes breathing [41], ataxic breathing [42], or other information contained in polysomnograms [40] may reveal aspects of physiology that predict CPAP responsiveness, but they are not currently utilized in guideline-directed management algorithms [5; 6; 8]. The presence of these patterns was not systematically documented, and thus we cannot evaluate their current usage or predictive value. Furthermore, if providers accurately use these features, the patients who received CPAP may differ from those who did not, even after matching and propensity weighting by observed characteristics [43].
In sum, we observed that clinicians often match their frequency of initial CPAP prescription for patients with central apneas to characteristics that independently predict an adequate response to CPAP. However, clinicians were less likely to prescribe an initial CPAP trial in older patients and patients with CA-OSA than their response rate to CPAP would suggest. These findings should help clinicians better individualize their initial treatment recommendations for patients with central apneas.