Revolutionizing sleep apnea care: Mount Sinai's automated tool predicts mortality risk

By Pooja Toshniwal PahariaReviewed by Sophia CoveneySep 13 2023

In a recent study published in the American Journal of Respiratory and Critical Care Medicine, researchers proposed ventilatory burden (VB) as an automated measure of obstructive sleep apnea (OSA) severity and a predictor of cardiovascular (CVD) and any-cause mortality, attempting to overcome apnea-hypopnea index (AHI) limitations.

OSA is a chronic condition causing daytime sleepiness and long-term CVD consequences. The AHI, used for OSA diagnosis, only captures the frequency of respiratory events defined by apneas and hypopneas. OSA severity is determined by the frequency or immediate physiological consequences of these events.

AHI has limitations, as it cannot segregate ventilatory disturbances from hypopnea consequences. Therefore, a measure for reliably identifying the ventilatory load of the illness needs to be developed.

About the study

In the present study, researchers investigated whether VB could predict cardiovascular and any-cause mortality among OSA patients.

Data obtained from two groups (Sleep Heart Health Study [SHHS] and Sao Paolo Epidemiological Study [EPISONO]) and two retrospective clinical cohorts (New York University Center for Brain Health and Relating Sleep Disordered Breathing to Daytime Function [DAYFUN]) were used to (i) derive the normal VB range, (ii) determine the association between VB and the extent of obstructive sleep apnea, and (iii) test the utility of VB in predicting cardiovascular (CVD) and any-cause deaths alone as well as when evaluated in combination with the automated hypoxic burden (HB).

In addition, the team evaluated the associations between VB, hypertension, and daytime sleepiness. The desired signal was overnight airflow measured using nasal cannulas or pressure transducers. For SHHS data, airflow was determined using digital differentiators applied to abdominal and thoracic respiratory inductance plethysmography (RIP) signals.

The researchers measured each breath’s amplitude as the mean value mid-flow during an overnight investigation. Breaths were assigned into zero amplitude based on the then-current rate of respiration for apneas. A moving "mean" of breathing amplitude during the middle third of flow was calculated from 10 preceding breaths with amplitudes below 150 percent of the priorly computed average and no indication of inspiratory airflow restriction assessed by flow shape.

The breath amplitudes were then calculated using the amplitude mid-flow adjusted to moving "mean" values.

VB was defined as the percentage of overnight breaths with less than 50% amplitude. Polysomnography (PSG) and mortality data were provided by the National Sleep Research Resource (NSRR).

Linear mixed effect-type modeling was performed to evaluate the impact of continuous positive airway pressure (CPAP) on VB, and Cox proportional modeling was performed to determine the association between ventilatory burden and cardiovascular and any-cause deaths. Covariates included age, sex, race, body mass index, smoking habits, hypertension, time spent in bed, and prior history of stroke or congestive cardiac failure. The team constructed four models for the analysis: Model 1 (AHI), Model 2 (HB), Model 3 (VB), and Model 4 (VB and HB).

Results

In total, 34,446,696 inhalations across all 5,182 individuals were analyzed. The 95th percentiles of VB among healthy asymptomatic participants from the EPISONO and DAYFUN cohorts were 25% and 27%, respectively. VB showed a dose-response association and exhibited low night-to-night variability.

The dose-response association between VB and OSA severity was altered by therapeutic and sub-therapeutic CPAP. In the DAYFUN_Symptomatic cohort, at baseline, OSA subjects had a VB of 32%, which reduced to 7.2% after three months of CPAP therapy, increased to 18% on sub-optimal CPAP, and increased to 42% after two nights off CPAP.

VB was predictive of CVD-related and any-cause mortality for SHHS participants, before and after adjusting for covariates including HB. VB values for the EPISONO_Normal and DAYFUN_Normal cohort individuals were 5.5% and 9.8%, respectively. At baseline, VB morphology in DAYFUN_Symptomatic participants revealed a greater number of hypopneic and apneic inhalations and a lower peak for normalized amplitude.

With CPAP therapy, the ventilatory pattern started to return to normal, becoming more comparable to that seen in the DAYFUN_Normal and EPISONO_Normal groups. Concerning the SHHS and DAYFUN_Symptomatic groups, ventilatory burden showed moderate correlations with AHI and HB, whereas HB was strongly correlated with AHI among SHHS individuals. VB was also correlated with body mass index and age.

Among SHHS participants, the ventilatory burden was higher compared to that among DAYFUN_Normal and EPISONO_Normal group individuals (28%). In addition, the ventilatory burden was statistically significantly higher among sleepy individuals (n=1,467) than non-sleepy individuals and among hypertensive individuals (2,042 individuals) than normotensive individuals.

Ventilatory burden quintiles 3, 4, and 5 (vs. quintile 1), AHI statistical quintiles 4 as well as 5, (vs. quintile 1), and Hb quintile 5 (vs. quintile 1) showed associations with increased likelihood of sleepiness. Likewise, ventilatory burden quintiles 3-5 (vs. quintile 1), AHI quintiles 4 and 5 (vs. quintile 1), and HB quintiles 3-5 (vs. quintile 1) showed correlations with an increased likelihood of hypertension.

After covariate adjustment, AHI quintiles 3-5 (vs. quintile 1), all HB quintiles (vs. quintile 1), and VB (vs. quintile 1) showed significant associations with an increased risk of any-cause mortality, whereas only VB quintiles 3-5 (vs. quintile 1) were significantly associated with a greater risk of CVD mortality. The fourth model was the most appropriate fit for all model predictions.

Conclusion

Overall, the study findings suggested that ventilatory burden, as a measure of obstructive sleep apnea severity, was predictive of cardiovascular and any-cause mortality.

Clinicians now have access to a better and validated alternative to the AHI in managing care of sleep apnea patients. The researchers are sharing this tool/metric to all sleep labs as a software program, which does not require any technical expertise in operating.

Although research is ongoing, researchers hope that this systematically investigated sleep apnea severity metric will aid clinicians in selecting the type of treatment that is likely to be most effective to patients, which is one of the biggest limitations of AHI, the current sleep apnea severity metric.” -Corresponding author Ankit Parekh, PhD, Assistant Professor of Medicine (Pulmonary, Critical Care and Sleep Medicine) at the Icahn School of Medicine at Mount Sinai