Abstract
Indices of sleep apnoea-related hypoxic burden and heart rate variability derived from full-night polysomnography might be useful for identifying sleep apnoea patients at risk for stroke https://bit.ly/3eIYVsc
To the Editor:
Obstructive sleep apnoea (OSA) is increasingly recognised as a risk factor for stroke [1]. However, the incidence of stroke in patients investigated for OSA has been assessed in a limited number of studies reporting conflicting results on the association between the apnoea–hypopnoea index (AHI) and stroke incidence [1–3]. As OSA is a very heterogeneous condition, the identification of subgroups of patients at high risk for stroke would be clinically desirable, in order to implement preventive actions [1]. Population-based studies have demonstrated that the sleep apnoea specific hypoxic burden (SASHB), an easily derived signal from polysomnography (PSG), predicts cardiovascular mortality and incident heart failure [4, 5]. A recent study has demonstrated that night-time heart rate variability (HRV) might play an important role in the association between OSA and cerebral small vessel disease, which is responsible for a substantial proportion of strokes [6]. Whether physiological markers of SASHB and HRV might be useful in a clinical setting for identifying those patients with OSA at risk for stroke remains to be determined. Within a large multicentre clinic-based cohort of patients investigated for OSA, we hypothesised that PSG-derived indices of SASHB and HRV could predict stroke incidence.
The study was conducted on the Pays de la Loire Sleep Cohort, linked with data from the French administrative health care database (SNDS) (details of the Pays de la Loire Sleep Cohort and the process linking it with the SNDS have been published previously [7]). We included stroke-free patients with available SNDS data investigated by PSG (CID102L8DTM, CIDELEC, Sainte-Gemmes-sur-Loire, France) for OSA between 15 May, 2007 and 31 December, 2017. Respiratory events were scored manually using recommended criteria [8]. As previously described, the SASHB was defined as the total area under the respiratory event-related desaturation curve [4]. Using standard recommendations [9], HRV was computed on 5-min segments of continuous non-overlapping PSG-derived ECG without ectopy or artefact. Time domain HRV measurements included the standard deviation for the mean value of all normal-to-normal R-R intervals and the root mean square of successive differences in normal-to-normal R-R intervals (RMSSD). Frequency domain measurements included the normalised low frequency (LF, from 0.04 to 0.15 Hz) and high frequency power (HF, from 0.15 to 0.4 Hz). The LF to HF ratio (LF/HF) estimated the sympathetic/parasympathetic tone.
The study endpoint was the first episode of stroke at any time between the PSG recording date and the end of December 2018. The first occurrence of stroke was identified from the national hospital discharge database (PMSI) and defined as the entry date of the first hospitalisation with a discharge diagnosis G45, G46, I60–I64 or I69 [10]. The accuracy of PMSI-based algorithm for stroke has been previously demonstrated with a positive predictive value at 90% [10]. As all physicians routinely contribute to PMSI data collection with annual quality control of coding, the accuracy increases with time [11].
Cox proportional hazard models were used to evaluate the association of stroke incidence with natural log transformed indices of OSA severity and HRV. Missing values were imputed using a multiple imputation method (MI procedure from SAS) [7]. Associations were considered statistically significant for a p-value <0.05. All statistical analyses were performed with SAS 9.4 software (SAS Institute, Cary, NC, USA).
The study population consisted of 3597 patients (median (interquartile range; IQR) age 58 (48–67) years), predominantly male (63%), obese or overweight (median (IQR) body mass index 28 (25–32) kg·m−2), frequently presenting cardiovascular and metabolic comorbidities (hypertension, 28.6%; diabetes, 10.4%; cardiac diseases, 8.7%), 85.4% of whom had mild-to-severe OSA (median (IQR) AHI 20 (8–35) events per h). During follow-up, 1159 patients were positive airway pressure (PAP) adherent (mean daily PAP use ≥4 h). After a median follow-up of 5.9 (3.5–8.4) years, 83 patients had been diagnosed with a stroke, including 70 ischaemic (29 transient ischaemic attack; TIA) and 13 haemorrhagic strokes (stroke incidence density rate 3.9 per 1000 person-years). Cox proportional hazard models (table 1) demonstrated an association between natural log transformed indices of OSA severity and stroke incidence (model 1), which remained significant after adjusting for confounding risk factors (model 2) for SASHB and the percentage of sleep time with oxygen saturation <90% (T90) (p=0.02 for both). Among natural log-transformed indices of HRV, stroke incidence was negatively associated with LF and LF/HF ratio (p=0.01 and 0.008, respectively) and positively associated with HF (p=0.01) after adjusting for confounders (model 2). The magnitude of the associations was unchanged after adjusting for PAP adherence (model 3) and controlling for a competing risk of death (not shown).
Cox proportional hazard models assessing the association of indices of obstructive sleep apnoea (OSA) severity and heart rate variability (HRV) with stroke incidence
When the analysis was restricted to 70 ischaemic cerebrovascular events, stroke incidence remained significantly associated with natural log-transformed SASHB (adjusted hazard ratio (HR) 1.30, 95% CI 1.05–1.61; p=0.02) and LF/HF ratio (adjusted HR 0.66, 95% CI 0.47–0.93; p=0.02) in the fully adjusted model (model 3). After exclusion of 29 TIA from the analysis, stroke incidence remained significantly associated with natural log-transformed LF/HF ratio (adjusted HR 0.60, 95% CI 0.41–0.90; p=0.01) but not with SASHB (adjusted HR 1.17, 95% CI 0.92–0.1.48; p=0.2).
Adding interaction terms in the analyses showed no significant interaction of gender and PAP adherence in the relationship between natural log-transformed PSG-derived indices and stroke incidence. However, stroke incidence was more strongly associated with SASHB and HF among non-obese subjects compared with those with obesity. The association of stroke incidence with SASHB and LF/HF was also stronger in patients aged ≥60 years compared with those aged <60 years (all p-values for interactions <0.05).
In a large multicentre, clinic-based cohort, we demonstrated an association of stroke incidence with PSG-derived indices of OSA-related hypoxaemia and HRV. Patients with higher SASHB and lower sympathetic/parasympathetic tone (LF/HF ratio) were at higher risk of stroke after adjusting for confounding risk factors and PAP adherence.
A strong association between untreated OSA and incident strokes (n=17 events) has been previously reported in women [2]. Conversely, Kendzerska et al. [3] reported that stroke in OSA was not associated with AHI. In the present study, we found no relationship between stroke incidence and common metrics of OSA severity, except for T90, which characterises not only OSA-related intermittent hypoxaemia, but also persistent hypoxaemia, such as that due to obesity hypoventilation or heart failure. Conversely, we demonstrated a dose–response relationship between stroke incidence and SASHB. Altogether, our findings and those from recent reports [4, 5] suggest that SASHB might predict stroke and cardiovascular morbidity more consistently than AHI, ODI and T90.
Previous studies have demonstrated a relationship between HRV and stroke [12]. In subjects with OSA, the sympathetic activation that occurs toward the end of obstructive events is accompanied by vagally mediated bradycardia due to activation of the diving reflex [13]. In older men with OSA, sleep-related reduced sympathetic/parasympathetic tone is associated with a higher risk of atrial fibrillation [14], which has been consistently associated with stroke risk. In the present study, a one-unit increase in natural log-transformed LF/HF ratio was associated with a 44% decrease in stroke risk, suggesting a contribution of low sympathetic/parasympathetic tone to stroke occurrence.
The strength of the current study includes a multicentre design, the adjustment for multiple stroke risk factors, and the assessment of different OSA severity and HRV indices. Our study also has some limitations. Its observational design does not allow any conclusions to be drawn regarding the causal pathway of the associations. Potential unmeasured confounding factors cannot be excluded. The use of a composite outcome combining different types of cerebrovascular events is also a potential limitation. We acknowledge that our study did not have sufficient statistical power to examine the association of OSA severity and HRV indices with the different stroke subtypes. All patients were investigated by in-laboratory overnight PSG using the same device. Further studies are required to evaluate the reproducibility of PSG-derived indices of SASHB and HRV. Furthermore, the high prevalence of OSA has led to an increasing use of simplified home sleep apnoea testing, with no ECG signal [15]. Whether oximetry-derived indices of pulse rate variability could provide an accurate estimation of HRV and predict stroke incidence should be also evaluated.
In conclusion, within a large clinic-based cohort of patients with suspected OSA, we demonstrated an association of SASHB and nigh-time sympathetic/parasympathetic tone with stroke risk. PSG-derived indices of hypoxic burden and heart rate variability may provide an opportunity to allow for stroke risk stratification in patients with OSA.
Shareable PDF
Supplementary Material
This one-page PDF can be shared freely online.
Shareable PDF ERJ-04022-2020.Shareable
Acknowledgements
We thank the French National Health Insurance for giving us access to the French administrative health care database. We thank the ERMES study group: Centre Hospitalier Universitaire, Angers: Christine Person, Pascaline Priou; Centre Hospitalier, Le Mans: Olivier Molinier, Audrey Paris. We thank Christelle Gosselin and Jean-Louis Racineux from the Institut de Recherche en Santé Respiratoire des Pays de le Loire. We thank Julien Godey, Laetitia Moreno and Marion Vincent, sleep technicians in the Dept of Respiratory and Sleep Medicine of Angers University Hospital.
Footnotes
Conflict of interest: M. Blanchard has nothing to disclose.
Conflict of interest: C. Gervès-Pinquié has nothing to disclose.
Conflict of interest: M. Feuilloy has nothing to disclose.
Conflict of interest: M. Le Vaillant has nothing to disclose.
Conflict of interest: W. Trzepizur has nothing to disclose.
Conflict of interest: N. Meslier has nothing to disclose.
Conflict of interest: F. Goupil has nothing to disclose.
Conflict of interest: T. Pigeanne has nothing to disclose.
Conflict of interest: F. Balusson has nothing to disclose.
Conflict of interest: E. Oger has nothing to disclose.
Conflict of interest: A. Sabil has nothing to disclose.
Conflict of interest: J-M. Girault has nothing to disclose.
Conflict of interest: F. Gagnadoux reports grants and personal fees from Resmed, personal fees and non-financial support from Sefam, Novartis and Air Liquide Sante, personal fees from Cidelec and Actelion, non-financial support from Boehringer Ingelheim and Asten, outside the submitted work.
Support statement: This study was supported by a grant from the Pays de la Loire Respiratory Health Research Institute (IRSR), Beaucouzé, France. Funding information for this article has been deposited with the Crossref Funder Registry.
- Received September 23, 2020.
- Accepted November 2, 2020.
- Copyright ©ERS 2021