Impact of Hypopnoea Classifications on the Assessment of Sleep Disordered Breathing in Heart Failure

Heart failure (HF) is a significant public health issue, characterised by elevated rates of premature mortality, frequent hospital admissions, rising healthcare expenditure and diminished quality of life.¹ The estimated prevalence of HF exceeds 56 million individuals globally and is expected to increase due to ageing populations, improvements in treatment outcomes and improved survival rates.^1,2

Sleep-related breathing disorder (SRBD) is a prevalent comorbidity associated with HF, affecting up to 60% of patients.^3,4 Among comorbid SRBD in HF, both obstructive sleep apnoea (OSA) and central sleep apnoea (CSA) are common.^5,6 One ventilatory phenotype typically predominates, but variations in the clinical stability of HF patients can lead to day-to-day differences in the predominant ventilatory phenotype of sleep apnoea.^7,8 Left ventricular ejection fraction (LVEF) and the severity of HF play determinant roles in the prevalence and severity of CSA associated with HF. In patients with asymptomatic left ventricular systolic dysfunction, the prevalence of moderate-to-severe CSA reaches 55%, compared with only 11% for moderate-to-severe OSA.^9,10 In HF with preserved ejection fraction, the prevalence of CSA reaches 23–27%, increasing to 34–69% in patients with HF with reduced ejection fraction (HFrEF).^7,9,11,12 The predominance of CSA may also change over the course of the night, with both the frequency and duration of CSA increasing during the later stages of non-rapid eye movement sleep in patients with left ventricular dysfunction who predominantly experience CSA.¹³ Notably, if prevalence values are reported for stable, compensated HF, the severity and prevalence of CSA can be dramatically increased in the case of decompensation.¹⁴ Most notably, most of these previous studies did not differentiate between CSA and OSA, which may have contributed to an incomplete characterisation of the SRBD landscape in HF.

Longitudinal studies indicate that both OSA and CSA are relevant to long-term outcomes in HF patients. In a prospective cohort of 1,117 patients hospitalised for acute HF with LVEF ≤45%, untreated CSA or OSA was independently associated with mortality over the 3-year follow-up.¹⁴ In a study of registry-based longitudinal data from a cardiac centre including 1,547 HF patients (mean [± SD] LVEF 38.7 ± 16%) without SRBD (25%) or with CSA (28%) or OSA (46%), Khayat et al. showed that both types of SRBD were associated with increased readmission rates and mortality at 3 and 6 months after discharge.¹⁵ In the FACE study, among the six reported phenotypes of CSA patients with HF, male patients with HFrEF and predominant CSA had the worst outcomes.^16,17 The management of OSA and CSA in HF differs significantly.^4,18 Therefore, the accurate classification of these conditions is important for making informed treatment decisions. Achieving accurate classification necessitates a thorough and rigorous scoring of respiratory events of central origin, particularly in relation to hypopnoeas.

This review examines how the definition of hypopnoea affects SRBD classification in HF, outlines methods for accurately classifying hypopnoeas as central or obstructive and discusses the potential effects of this distinction on diagnosis, prognosis and treatment.

Effects of Hypopnoea Definitions on Sleep-related Breathing Disorder Severity and Classification

Hypopnoea Definitions and Sleep-related Breathing Disorder Severity

The International Classification of Sleep Disorders, third revised edition (ICSD-3), lists both OSA and CSA as part of SRBDs.¹⁹ Both are classified as syndromes that necessitate symptom identification for diagnosis, as well as a polysomnography (PSG), which is currently regarded as the standard method for classifying and assessing the severity of SRBD. In addition to its essential role in diagnosis, PSG serves as a critical examination for identifying phenotypical traits of OSA and CSA, providing a rationale for treatment selection. Moreover, a precise description of the PSG patterns of flow and respiratory effort is helpful for identifying archetypal respiratory patterns, such as Cheyne–Stokes breathing.²⁰ Cheyne–Stokes breathing is defined as follows: three or more consecutive central apnoeas and/or hypopnoeas, separated by a crescendo and decrescendo change in breathing amplitude (periodic breathing) with a cycle length of at least 40 seconds (typically 45–90 seconds); and five or more central events per hour associated with the periodic breathing pattern recorded over a minimum of 2 hours.²⁰ The European Respiratory Society recommends reserving the term ‘Cheyne–Stokes breathing’ for periodic breathing associated with congestive HF.²¹

The apnoea–hypopnoea index (AHI) is the reference metric for classifying and assessing the severity of both OSA and CSA, both of which are characterised by recurring episodes of either complete cessation (apnoea) or partial reduction (hypopnoea) of ventilation. These episodes are accompanied by continuing respiratory effort in OSA, whereas in CSA the respiratory effort is diminished or absent. It is important to clarify that the classification of an event as apnoea does not necessitate the presence of desaturation or arousal criteria.²² In addition to categorising sleep apnoea as either OSA or CSA, some research has adopted a three-group model based on the proportion of central events within the total AHI for pathophysiological reasons as follows: OSA (AHI ≥15/h plus <20% of central events [apnoeas and hypopnoeas]); co-existing OSA–CSA (AHI ≥15/h plus ≥20–50% of central events); and CSA (AHI ≥15/h plus >50% of central events).^4,23 This classification has also been used in recent treatment recommendations.¹⁸

Currently, scoring of respiratory events on PSG is based on the American Academy of Sleep Medicine’s (AASM) rules (current version: v3.0), but has evolved several times since the seminal classification published in 1999, the so-called Chicago criteria.^22,24,25

If the apnoea index was introduced in 1978, the first description of hypopnoea using 4% oxygen desaturation dates back to 1979.^24,26,27 The importance of hypopnoeas was highlighted in 1988, and AHI was absent from the first edition of the ICSD.^24,28 Although the definition of an apnoea is well-established, the AASM’s guidelines for scoring hypopnoeas have been revised three times over the past two decades (Table 1).^{20,24–26,29,30} Finally, the AASM 2012 rules defined hypopnoea as a ≥30% airflow reduction associated with ≥3% oxygen desaturation or an arousal.^20,29

The rule used to score hypopnoeas may greatly influence SRBD severity and classification. As demonstrated by Hirotsu et al., the AASM criteria applied for scoring hypopnoeas significantly influenced the prevalence of OSA in the HypnoLaus cohort and its association with cardiometabolic outcomes.³¹ The median AHI based on AASM 1999, AASM 2007 and AASM 2012 scoring criteria was 10.9, 4.4 and 10.1 events/h, respectively.³¹ Hirotsu et al. demonstrated a twofold difference in the threshold above which an association with diabetes, hypertension and metabolic syndrome was observed using AASM 2007 versus AASM 1999 or AASM 2012 scoring criteria, indicating that scoring criteria affect risk stratification.³¹ In a cross-sectional analysis of a cohort of 1,400 veterans who underwent polysomnography for suspected SRBD, Won et al. compared the impact of scoring hypopnoeas using ≥4% desaturation (AASM 2007 guidelines) versus ≥3% desaturation or arousal (AASM 2012 guidelines).³² The use of the ≥3% desaturation or arousal definition of hypopnoea captured an additional 175 OSA diagnoses compared with the use of the ≥4% desaturation criterion. While this approach refines OSA classification, severe disease, defined by either ≥3% desaturation with arousals or ≥4% desaturation, continues to predict cardiac arrhythmias in this cohort.³² In a retrospective study, Park et al. found that the hypopnoea-predominant phenotype was associated with higher rates of coronary artery disease and HF.³³ Therefore, the group of OSA patients with a predominant hypopnoea presentation may represent a specific phenotype, rendering the precise scoring of hypopnoeas mandatory. Recently, the Hypopnoea Scoring Rule Task Force reaffirmed the need to create a strategy for the adoption and implementation of the AASM-recommended adult hypopnoea scoring criteria among AASM-accredited sleep centres, payers in the healthcare industry are organisations and device manufacturers.³⁴ The Hypopnea Scoring Rule Task Force stated that solely counting respiratory events associated with a ≥4% oxygen desaturation (current AASM scoring manual ‘acceptable’ hypopnoea definition for adults) may result in missing a diagnosis of OSA in symptomatic patients who would otherwise be diagnosed based on the recommended hypopnoea definition and would potentially benefit from treatment.³⁴ Thus, differences in hypopnoea scoring can lead to inconsistent cardiovascular risk estimates, affecting both individual patient management and epidemiological research.

Hypopnoeas and Sleep-related Breathing Disorder Classification

In adults, according to ICSD-3, the PSG criteria defining CSA are an AHI >5 events/h of sleep, with more than 50% of these events being due to central apnoeas or hypopnoeas. Differentiation of central from obstructive hypopnoeas is strongly recommended, because it can greatly influence diagnosis and SRBD classification.²¹ However, the scoring of central versus obstructive hypopnoeas is not included in the routine practice of many sleep laboratories. Pépin et al. prospectively assessed symptoms, comorbidities, medications and treatment indications in a large single-centre real-life dataset of >2,400 patients referred for investigation of suspected SRBD.²³ These authors systematically distinguished central versus obstructive hypopnoeas to define OSA, CSA and co-existing CSA-OSA (AHI ≥15/h plus ≥20–50% of central events). When CSA was defined by the proportion of central apnoeas (all hypopnoeas considered obstructive), the prevalence of CSA was 4.59% (co-existing CSA-OSA: 11.03%; OSA: 84.37%); however, when the distinction between obstructive and central hypopnoeas was used, the prevalence of CSA was fourfold higher at 19.69% (co-existing CSA-OSA: 19.16%; OSA: 61.16%).²³ Most of the patients with SRBD were symptomatic after comprehensive evaluation, but the burden of cardiovascular and metabolic comorbidities was highest in the CSA and co-existing CSA-OSA subgroups.²³ Interestingly, the CSA group exhibited the most severe disturbances in sleep architecture on PSG.²³ These results outline the need to differentiate between central and obstructive hypopnoeas to avoid inappropriate therapeutic decisions that may limit improvements in quality of life and sleepiness that are expected in appropriately treated patients with CSA.

Effects of Hypopnoea on Sleep-Related Disordered Breathing Severity and Classification in Heart Failure

In HF patients, inconsistent hypopnoea criteria may misclassify SRBD severity and ventilatory phenotype (OSA/CSA).^4,18 In their seminal work, Ward et al. tested the hypothesis that the diagnosis of SRBD would be made in a significantly greater number of HF patients (median LVEF 41% [interquartile range 29–58%]) when hypopnoeas were scored using the 2007 AASM alternative rule (i.e. ≥50% reduction in airflow with an associated ≥3% oxygen desaturation or EEG arousal) compared with a more conservative hypopnoea definition (≥50% decrease in the amplitude of the nasal airflow signal for ≥10 seconds in association with a ≥4% oxygen desaturation).³⁵ With the alternative scoring rule, the prevalence of SRBD increased from 29% to 46% and median AHI increased from 9.3 to 13.8/h (median difference 4.6/h).³⁵ Classification of SRBD as OSA or CSA was not significantly altered by hypopnoea scoring rules or the categorisation of mixed apnoeas.³⁵ Table 2 presents a summary of respiratory event scoring criteria and SRBD characteristics in key randomised controlled trials focusing on CSA treatment in patients with HF. Overall, these findings indicate that inconsistent central hypopnoea scoring may lead to misclassification of SRBD, potentially affecting clinical decisions concerning the treatment and management of patients, particularly those with HF.

Variability in Sleep-Related Breathing Disorders Severity and Classification in Heart Failure

One parameter that is often overlooked is the night-to-night variability of CSA (for a review, see Orr et al.³⁶). In OSA, and due to the high night-to-night variability documented in the AHI, single-night sleep studies are estimated to misdiagnose and misclassify OSA severity in 20–50% of patients.³⁷ Accordingly, CSA can vary significantly within individuals over time, and the factors contributing to these variations, as well as their significance, are not yet fully understood.³⁶ However, increased CSA severity is known to be associated with HF decompensation.¹⁴

In stable chronic HF patients (n=50, New York Heart Association Class ≥II, LVEF ≤40%), Oldenburg et al., using cardiorespiratory polygraphy over two consecutive nights, showed that the reproducibility of SRBD diagnosis was dependent on SRBD severity, with low variability regarding SRBD classification (i.e., CSA and OSA) for an apnoea index ≥10/h.³⁸ Vazir et al. assessed the severity and type of SRBD in 19 male chronic HF patients (mean [± SD] age 61 ± 9 years and LVEF 34 ± 10%) over four consecutive nights.³⁹ In addition to minimal variations in AHI, these authors demonstrated a shift in the classification of SRBD from CSA to OSA and vice versa in 42% of the included patients.³⁹ Together, these results highlight the need for further studies and new tools and diagnostic pathways to assess night-to-night variability in SRBD and, more precisely, in CSA in the context of HF to improve CSA diagnosis and classification, as well as to refine prognostic evaluation, treatment indications and the right choice of ventilatory support.^4,18,40

Distinguishing Central and Obstructive Hypopnoeas

Although apnoeas are readily recognised in clinical settings using PSG and respiratory polygraphy, accurately distinguishing between central and obstructive respiratory events remains challenging (Figure 1). In an agreement analysis among nine international sleep centres, the mean (± SD) central apnoea index (CAI) ranged from 0.9 ± 1.4 to 7.3 ± 10.2, with an intraclass correlation coefficient of 0.46 (95% CI [0.27–0.70]), indicating poor reliability.⁴¹ In another study conducted under the supervision of the AASM, an agreement on specific events found an intraclass correlation coefficient of 0.52 (κ=0.41).⁴² These discrepancies are even higher for central hypopnoea, with hypopnoeas being frequently classified as obstructive without detailed evaluation.⁴³ Altogether, this underestimates the severity of respiratory disturbances of central origin and may hamper proper classification, evaluation of prognosis and therapeutic decisions.

The gold standard for the scoring of respiratory events of central origin is full PSG with oesophageal pressure measurement or diaphragm electromyography.^21,44 However, these procedures are not currently used in routine diagnosis. In clinical settings, surrogates of respiratory effort can be used, such as strain gauges, respiratory inductive plethysmography, pulse transit time (see below) or belts with piezoelectric or polyvinylidene fluoride sensors.^20,45 Moreover, diagnostic accuracy can be affected by the absence of a thermal sensor, which should be combined with nasal pressure prongs.⁴⁶ Thus, distinguishing between obstructive and central hypopnoeas remains challenging, and a combination of features derived from non-invasive signals should be used to infer the absence of pharyngeal obstruction and a reduction in or absence of respiratory effort (Figure 1). A hypopnoea can be considered as central if none of the following are present: snoring during the event; increased inspiratory flattening in nasal pressure or a positive airway pressure device flow signal compared with baseline breathing; and an associated thoracoabdominal paradox occurring only during the event, not during pre-event breathing.^6,21 Other PSG characteristics can be helpful for the identification of the central nature of a respiratory event. Thus, a crescendo–decrescendo shape in the inspiratory flow, limited desaturation following the event, arousal in the middle of the ventilatory upturns and the sleep stage (rapid eye movement versus non-rapid eye movement) are non-consensual but helpful signs of the central origin of ventilatory events.^6,47 Several studies have proposed algorithms or alternative automated approaches to properly distinguish between central and obstructive events.^47–49

Compared with oesophageal pressure, the algorithm proposed by Randerath et al., based on flattening of the inspiratory airflow curve, paradoxical breathing, arousal position, sleep stages and breathing pattern at the end of the hypopnoea, showed an overall accuracy of 68% and correct definition of 76.9% of central hypopnoeas.⁴⁷ These findings were substantiated by Dupuy-McCauley et al., who also found similar results for another non-invasive differentiation based on the AASM criteria.⁵⁰ Parekh et al. developed an automated algorithm to determine the probability of obstruction on a breath-by-breath basis.⁴⁹ The targeted parameter proved to correlate with upper airway resistance on an individual basis with low night-by-night variability.⁴⁹ Javaheri et al. added prolongation of the inspiratory duty cycle as a marker of obstructive hypopnoeas to the previous findings.⁴⁸

In addition, complementary techniques embedded in some PSG devices can be helpful to distinguish between events that are obstructive or central in nature, such as pulse transit time (PTT). The integration of PTT in PSG offers a non-invasive method that is both specific and sensitive in quantifying the respiratory effort during the four stages of sleep and in analysing sleep fragmentation.^51,52 Measuring PTT involves using ECG leads for the recognition of the R wave and oximetric photoplethysmography (usually a fingertip probe or an ear probe) to assess the arrival of the pulse wave at the periphery. PTT is measured as the time interval between the ECG R wave and the subsequent arrival of the pulse wave at the photoplethysmography probe (defined as the point on the pulse waveform corresponding to 50% of the height of the maximum amplitude). The speed at which the pulse wave spreads is inversely proportional to systolic blood pressure. Consequently, a modest elevation in blood pressure leads to heightened vascular tone, thereby increasing arterial wall stiffness and resulting in a reduction in PTT, which manifests as a brief dip in the PTT baseline.

Obstructive events or upper airway resistance episodes are associated with a steady increase in the amplitude variation of PTT terminated by a microarousal and then the normalisation of the trace.⁵¹ During a central event, the amplitude of the PTT reduces in a similar manner to the reduction in respiratory effort. It should be noted that central hypopnoeas can be judged on the reduction in amplitude of PTT proportional to the reduction in respiratory effort, rendering PTT helpful for the identification of central hypopnoeas. PTT can also help detect microarousals (e.g. identified by a transient dip in PTT from baseline), enabling assessment of sleep fragmentation from respiratory or non-respiratory disturbances such as periodic limb movements.⁵³

In PSG performed under continuous positive airway pressure (CPAP) treatment, PTT represents an interesting tool to ease and increase the accuracy of scoring residual respiratory events. The recent identification of treatment-emergent central sleep apnoea, and the identification of high residual AHI variability to which HF can contribute, has led to a wider use of PSG recordings under CPAP to improve the identification of central events (both apnoeas and hypopnoeas).^54–57 At a patient level, this may lead to a change in the type of ventilatory support (usually from CPAP to adaptive servo-ventilation in the case of compatible cardiac ejection fraction). PTT also has limitations. Medical conditions, such as cardiac arrhythmias, or the use of certain medications, such as β-blockers, may affect PTT measurements. In severe HF, the relationship between PTT and blood pressure may be altered and may indicate impaired ventricular–arterial coupling.^58,59

Recently, sleep mandibular movement signals recorded with a triaxial gyroscopic chin sensor (Sunrise) were shown to be a reliable surrogate of oesophageal pressure in patients with suspected OSA.⁶⁰ Based on machine learning approaches, the device was able to accurately distinguish between obstructive and central events, including hypopnoeas. The device is approved by the US Food and Drug Administration and is recommended by the National Institute for Health and Care Excellence in the UK for home sleep apnoea diagnosis. Because the device relies exclusively on mandibular movements, which are generally not affected in HF, its accuracy within the context of HF should be preserved but warrants further examination.⁶¹

Implications for Diagnosis, Prognosis and Treatment

In HF, a PSG should be ordered in case of SRBD symptoms because it is the most precise means for SRBD classification and severity evaluation to date (Figure 2).⁴ In HF, CSA symptoms may overlap with those of the underlying medical condition, and existing screening tools lack adequate sensitivity and specificity when applied to individuals with HF.^62,63 Despite lacking evidence, home sleep apnoea testing or single-channel screening tests can be considered in populations at high risk of CSA but without symptoms.⁴

Effect of Predominant Central Sleep Apnoea on Outcome in Patients with Heart Failure

There is increasing evidence from longitudinal studies indicating the relevance of both OSA and CSA to long-term outcomes in chronic HF.^7,15,64 The randomised controlled SERVE-HF trial demonstrated altered survival in patients with a higher degree of CSA with Cheyne–Stokes breathing, whereas cardiovascular mortality was negatively affected by lower LVEF values.⁶⁵

The FACE study focused on the heterogeneity of CSA patients with chronic HF while scoring hypopnoeas of central origin.⁶⁶ Six phenotypes, differing in terms of survival and severity of morbidity at the 3-month and 2-year follow-up, were described.^16,17 Adding cardiorespiratory variables and the use of latent class analysis improved phenotyping and helped individualise health trajectories characterised by different prognoses: male patients with HFrEF and predominant CSA had the worst outcomes.^16,17,66 Together, these findings reinforce the need for clear identification of CSA severity in HF patients through precise scoring of hypopnoeas to improve risk stratification and prediction.

Effect of Central Hypopnoea Scoring on the Choice of Treatment in Patients With Heart Failure

As previously mentioned, precise classification and severity assessment of SRBD are key to treatment selection (e.g. CPAP, adaptive servo-ventilation).^4,18 If the basis of CSA treatment in HF is optimisation of HF medical therapies, some patients may require ventilatory support. The initial clinical presentation may be key because it may influence response to ventilatory support.^17,65,67,68 Evidence indicates that treating CSA in HF patients positively affects outcomes.^4,18 To the best of our knowledge, there are currently no data addressing the effect of hypopnoea scoring on treatment selection, or any studies involving randomisation of treatment allocation based on hypopnoeas, despite their recognised prognostic importance. However, an increasing number of studies, including the FACE study, use precise, standardised hypopnoea scoring to assess SRBD severity in HF patients.⁶⁶ As the field of CSA in HF continues to evolve, emerging care and treatment pathways should be based on accurate phenotyping and thorough characterisation of respiratory events.^18,40

Current Evidence and Gaps

Current evidence indicates that SRBD, including both OSA, CSA and coexisting OSA-CSA, are highly prevalent in HF, affecting up to 60% of HF patients. The prevalence of CSA increases with reduced ejection fraction and in decompensated HF states. Left untreated, SRBD significantly affects prognosis, in terms of morbimortality, quality of life and function. The severity and classification of these disorders are influenced by how hypopnoeas are scored, and distinguishing central from obstructive hypopnoeas is crucial, because failing to do so can lead to misclassification and inappropriate treatment indications. If PSG remains the gold standard for SRBD diagnosis in HF, non-invasive diagnostic tools, such as mandibular movement sensors, show promise for improving classification accuracy through repeated night assessment. Despite these advances, several gaps remain. Standardised hypopnoea definitions are still lacking in HF-specific research, and the differentiation between central and obstructive hypopnoeas is not routinely made in many sleep laboratories, leading to misclassification and suboptimal treatment. In addition, night-to-night variability in SRBD severity and classification is frequently overlooked. Finally, although personalised treatment strategies based on hypopnoea characteristics show potential, current data supporting their impact on long-term outcomes, including symptoms and adherence, remain limited.

Conclusion

The precise identification and categorisation of hypopnoeas, particularly in differentiating central from obstructive events, is essential for the effective management of SRBD in HF. Emerging research underscores the significance of hypopnoeas in cardiovascular risk assessment and patient outcomes. Variability in definitions can affect the reliability of epidemiological data, hinder accurate diagnosis and lead to inappropriate treatment indications. Standardisation is required, particularly within HF populations, with consistent hypopnoea scoring criteria being implemented across clinical and research environments. Innovative approaches, such as mandibular movement analysis, present promising, non-invasive methods to enhance accuracy in diagnosis. In addition, tailored personalised interventions informed by advanced hypopnoea metrics may improve therapeutic outcomes, patient adherence and overall quality of life.