Characterising dysfunctional breathing seen in post-acute sequelae of SARS-CoV-2 using approximate entropy.

Rationale: Dysfunctional breathing (DB) is a commonly identified abnormality in post-acute sequelae of SARS-CoV-2 (PASC) patients undergoing cardiopulmonary exercise testing (CPET), and is potentially a contributor to ongoing symptoms. Currently, this oscillating, irregular breathing pattern is identified by visual observation of CPET data by an experienced interpreter, which is subjective. We hypothesise that approximate entropy (ApEn), a regularity statistic that quantifies the unpredictability of time-series data can reliably distinguish DB from normal breathing states.

Methods: Breath-by-breath CPET data were obtained for 82 PASC subjects and 25 controls. CPETs were visually analysed for DB prior to analysis. Minute ventilation ('), tidal volume ( ) and breathing frequency (BF) over time data were normalised with 100% considered as the ventilation at anaerobic threshold (AT) and detrended before ApEn was calculated. Analysis was initiated at 25 W and ceased at AT.

Results: The ApEn ' of PASC subjects with visualised DB was 0.286±0.128 (mean±sd), which was significantly different from control subjects (0.143±0.081) and PASC subjects without visualised DB (0.183±0.104); p<0.05. Receiver operating characteristic curve analysis produced an optimal cut-off value of 0.17 for distinguishing DB, which resulted in a sensitivity of 81% and specificity of 72%. ApEn and ApEn BF were similar among all PASC patients despite visually recognised DB, but significantly greater than controls.

Conclusions: Identifying DB on CPET requires visual recognition, which has limitations. ApEn ' is an objective metric that can reliably differentiate DB from normal breathing patterns on CPET. This can be a valuable addition to keen visual scrutiny of CPET data.