A longitudinal study of the accuracy of positive pressure therapy machine-detected apnea-hypopnea events

The association between the effective apnea-hypopnea index and blood pressure reduction efficacy following CPAP/oxygen treatment

BACKGROUND

The effect of sleep apnea treatment on reducing cardiovascular disease risk remains inconclusive. This study aims to assess if the effective apnea hypopnea index (eAHI), a measure of residual sleep apnea burden post-treatment, is a factor in determining blood pressure (BP) response to continuous positive airway pressure therapy. The eAHI integrates time on therapy, residual apnea, and % of sleep time untreated.

METHODS

A secondary analysis of the Heart Biomarker Evaluation in Apnea Treatment (HeartBEAT) study, a randomized, controlled, parallel group assessment of continuous positive airway pressure (CPAP), oxygen and sleep hygiene. The Delta-AHI (▲AHI) was defined as the difference between baseline AHI and effective AHI at 12 weeks. Logistic and linear regression models estimated the predictors for nocturnal systolic BP change following sleep apnea therapy.

RESULTS

One hundred and sixty-nine subjects with a mean age of 62.82 ± 6.99 years were included in the final analysis. Fifty subjects had ▲AHI ≤8/hour of sleep and 119 subjects were higher. After adjustment, baseline mean nighttime systolic blood pressure (OR 1.036, 95% CI 1.015-1.058, p: 0.001) and ▲AHI ≥8/hour (OR 2.406, 95% CI 1.116-5.185, p:0.025) were independent predictors for mean nighttime systolic blood pressure change >3 mm Hg. The higher effective AHI was negatively related with BNP (β: -2.564, SE: 1.167, p: 0.029) and positively related with troponin change (β: 0.703, SE: 0.256, p: 0.007).

CONCLUSION

The ▲AHI was an independent predictor of the blood pressure response to sleep apnea treatment.

REGISTER NUMBER

NCT01086800.

Accuracy of residual respiratory event detection by CPAPs: a meta-analysis

PURPOSE

Most continuous positive airway pressure (CPAP) machines have built-in manufacturer-specific proprietary algorithms for automatic respiratory event detection (AED) based on very specific respiratory events scoring criteria. With regards to the accuracy of these data from CPAP machines, evidence from the literature seems conflicting, which formed the basis for this meta-analysis.

METHODS

A meta-analysis was performed on studies that reported Bland-Altman analysis data on agreement (mean bias and limits of agreement [LoA]) of CPAP-determined apnea-hypopnea index (AHI) at therapeutic pressures (AHIFLOW) with that determined from simultaneously conducted polysomnograms (AHIPSG).

RESULTS

In six studies, ResMed CPAPs were used, and in another six studies, Respironics CPAPs were used, while only one study used Fisher & Paykel (F&P) CPAPs. The pooled mean AHI bias from ResMed CPAP studies was - 1.01 with pooled LoAs from - 3.55 to 1.54 (I2 = 17.5%), and from Respironics CPAP studies, pooled mean AHI bias was - 0.59 with pooled LoAs from - 3.22 to 2.05 (I2 = 0%). Pooled percentage errors (corresponding to LoAs) from four ResMed CPAP studies, four Respironics CPAP studies, and the F&P CPAP study were 73%, 59%, and 112%, respectively. A review of the literature for this meta-analysis also revealed lack of uniformity not only in the CPAP manufacturers' respiratory events scoring criteria but also in that used for PSGs across the studies analyzed.

CONCLUSIONS

Even though the pooled results of mean AHI bias suggest good clinical agreement between AHIPSG and AHIFLOW, percentage errors calculated in this meta-analysis indicate the possibility of a significant degree of imprecision in the estimation of AHIFLOW by CPAP machines.

Temporal Feature Extraction and Machine Learning for Classification of Sleep Stages Using Telemetry Polysomnography

Accurate sleep stage detection is crucial for diagnosing sleep disorders and tailoring treatment plans. Polysomnography (PSG) is considered the gold standard for sleep assessment since it captures a diverse set of physiological signals. While various studies have employed complex neural networks for sleep staging using PSG, our research emphasises the efficacy of a simpler and more efficient architecture. We aimed to integrate a diverse set of feature extraction measures with straightforward machine learning, potentially offering a more efficient avenue for sleep staging. We also aimed to conduct a comprehensive comparative analysis of feature extraction measures, including the power spectral density, Higuchi fractal dimension, singular value decomposition entropy, permutation entropy, and detrended fluctuation analysis, coupled with several machine-learning models, including XGBoost, Extra Trees, Random Forest, and LightGBM. Furthermore, data augmentation methods like the Synthetic Minority Oversampling Technique were also employed to rectify the inherent class imbalance in sleep data. The subsequent results highlighted that the XGBoost classifier, when used with a combination of all feature extraction measures as an ensemble, achieved the highest performance, with accuracies of 87%, 90%, 93%, 96%, and 97% and average F1-scores of 84.6%, 89%, 90.33%, 93.5%, and 93.5% for distinguishing between five-stage, four-stage, three-stage, and two distinct two-stage sleep configurations, respectively. This combined feature extraction technique represents a novel addition to the body of research since it achieves higher performance than many recently developed deep neural networks by utilising simpler machine-learning models.

Acute and long-term effects of acetazolamide in presumed high loop gain sleep apnea

BACKGROUND

The acute effect during positive pressure titration and long term efficacy of acetazolamide (AZT) in high loop gain sleep apnea (HLGSA) is inadequately assessed. We predicted that AZT may improve HLGSA in both conditions.

METHODS

A retrospective analysis of polysomnograms from patients with presumed HLGSA and residual respiratory instability administered AZT (125 or 250 mg) about 3 h into an initially drug-free positive pressure titration. A responder was defined as ≥ 50% reduction of the apnea hypopnea index(AHI 3% or arousal) before and after AZT. A multivariable logistic regression model estimated responder predictors. Long term efficacy of AZT was assessed by comparing both auto-machine (aREI_FLOW) and manually scored respiratory events (sREI_FLOW) extracted from the ventilator, prior to and after 3 months of AZT, in a subset.

RESULTS

Of the 231 participants (median age of 61[51-68] years) and 184 (80%) males in the acute effect testing: 77 and 154 patients were given 125 mg and 250 mg AZT. Compared to PAP alone, PAP plus AZT was associated with a lower breathing related arousal index (8 [3-16] vs. 5 [2-10], p < 0.001), and AHI3% (19 [7-37] vs. 11 [5-21], p < 0.001); 98 patients were responders. The non-rapid eye movement sleep (NREM) AHI3% (OR 1.031, 95%CI [1.016-1.046], p < 0.001) was a strong predictor for responder status with AZT exposure. In the 109 participants with 3-month data, both aREI_FLOW and sREI_FLOWwere significantly reduced after AZT.

CONCLUSIONS

AZT acutely and chronically reduced residual sleep apnea in presumed HLGSA; NREM AHI3% is a response predictor. AZT was well tolerated and beneficial for at least 3 months.

Predictors and consequences of residual apnea during positive airway pressure therapy

STUDY OBJECTIVES

Determine the risk factors for, and consequences of, residual apnea during long-term positive airway pressure (PAP) therapy for obstructive sleep apnea (OSA).

METHODS

A prospective cohort study of 195 subjects after a split-night polysomnogram. Estimation of residual respiratory events on PAP were done by both automated and manual scoring of data in EncoreAnywhere™. Clinical and polysomnographic predictors of residual apnea were estimated.

RESULTS

There were 166 and 101 patients still on PAP at the 3 and 12 months, respectively. Seventy four (44.6%) and 46 (45.5%) had a residual scored respiratory event index-flow (sREI_FLOW) ≥ 15/hour of use and 46 (45.5%) at the 3rd and 12th month, respectively. Treatment phase central apnea hypopnea index (TCAHI), a surrogate of high loop gain, was the main predictor for residual sREI_FLOW (β = 0.345, p: 0.025) at the 3rd and 12th month (β = 0.147, p: 0.020). TCAHI also predicted unstable breathing (U) %. The body mass index (hazard ratio [HR] 1.034, 95% CI 1.008-1.062, p: 0.012) and effective sREI_FLOW>15/hour in the first month (HR 2.477, 95% CI 1.510-4.065, p < 0.001) were the key predictors for drop out of PAP use at the 12th month. Effective sREI_FLOW>15/hour in the first month was also a predictor for median usage duration >4 h for 70% of the night at both the 3rd month (odds ratio [OR] 0.947, 95% CI 0.909-0.986, p: 0.008) and 12th month (OR 0.973, 95% CI 0.951-0.994, p: 0.014).

CONCLUSIONS

Treatment-phase CAHI predicts long-term residual apnea on PAP. High residual disease adversely impacts adherence.