Development, Validation, and Assessment of an Ischemic Stroke or Transient Ischemic Attack-Specific Prediction Tool for Obstructive Sleep Apnea

https://doi.org/10.1016/j.jstrokecerebrovasdis.2017.03.042Get rights and content

Background

Screening instruments for obstructive sleep apnea (OSA), as used routinely to guide clinicians regarding patient referral for polysomnography (PSG), rely heavily on symptomatology. We sought to develop and validate a cerebrovascular disease-specific OSA prediction model less reliant on symptomatology, and to compare its performance with commonly used screening instruments within a population with ischemic stroke or transient ischemic attack (TIA).

Methods

Using data on demographic factors, anthropometric measurements, medical history, stroke severity, sleep questionnaires, and PSG from 2 independently derived, multisite, randomized trials that enrolled patients with stroke or TIA, we developed and validated a model to predict the presence of OSA (i.e., Apnea-Hypopnea Index ≥5 events per hour). Model performance was compared with that of the Berlin Questionnaire, Epworth Sleepiness Scale (ESS), the Snoring, Tiredness, Observed apnea, high blood Pressure, Body mass index, Age, Neck circumference, and Gender instrument, and the Sleep Apnea Clinical Score.

Results

The new SLEEP Inventory (Sex, Left heart failure, ESS, Enlarged neck, weight [in Pounds], Insulin resistance/diabetes, and National Institutes of Health Stroke Scale) performed modestly better than other instruments in identifying patients with OSA, showing reasonable discrimination in the development (c-statistic .732) and validation (c-statistic .731) study populations, and having the highest negative predictive value of all in struments.

Conclusions

Clinicians should be aware of these limitations in OSA screening instruments when making decisions about referral for PSG. The high negative predictive value of the SLEEP INventory may be useful in determining and prioritizing patients with stroke or TIA least in need of overnight PSG.

Introduction

Obstructive sleep apnea (OSA) occurs commonly in the post-ischemic stroke or transient ischemic attack (TIA) population, affecting 60%-80% of persons after their cerebrovascular event.1, 2 Untreated OSA is an independent risk factor for future vascular events (e.g., stroke, myocardial infarction), may complicate management of vascular risk factors (e.g., hypertension, atrial fibrillation), and increases the risk for mortality in patients with cerebrovascular disease.3, 4, 5, 6 Despite the potential treatment and prognostic implications of discovering whether patients with cerebrovascular disease have comorbid OSA, the condition frequently goes undiagnosed.5, 7 This situation may be due, in part, to the observation that hallmark features of OSA (e.g., excessive daytime sleepiness) can occur less commonly among patients with cerebrovascular disease than in the general population.5 Several well-validated sleep instruments used in the general population to screen for OSA and perceived somnolence, such as the Berlin Questionnaire (BQ)8 and Epworth Sleepiness Scale (ESS) score,6 rely heavily on symptomatic features of OSA, but have not been predictive of OSA (as defined by an Apnea-Hypopnea Index [AHI] ≥5 on polysomnography [PSG]) in mixed stroke populations (ischemic and hemorrhagic).5, 7 A modified version of the Snoring, Tiredness, Observed Apnea, high blood Pressure-Body mass index, Age, Neck circumference, and Gender (STOP-BANG) was only moderately predictive of OSA compared with home sleep testing equipment among patients with a cerebrovascular event,9, 10 whereas the Sleep Apnea Clinical Score (SACS) has not been studied among patients with cerebrovascular disease. Given the suboptimal performance of commonly used OSA screening instruments within the stroke or TIA population, authors have suggested that the development of models based on medical comorbidity should be pursued.10

The most recent American Heart Association/American Stroke Association ischemic stroke or TIA prevention guidelines provide new recommendations addressing OSA, noting that PSG might be considered for patients with an ischemic stroke or TIA and, that once diagnosed, treatment of OSA might be considered, given its association with improved post-cerebrovascular event outcomes.2 The guidelines do not, however, specify which patients should be considered for PSG referral. Because universal OSA screening with PSG may not be feasible based on the worldwide prevalence of stroke, we sought to (1) develop and validate a cerebrovascular disease-specific instrument that would be less reliant on patient symptomatology and anthropometric features, and; (2) determine how well the new instrument, and the BQ, ESS, SACS, and STOP-BANG, predicted the presence and absence of OSA in an exclusively post-ischemic stroke or TIA population (rather than a mixed ischemic and hemorrhagic stroke population). These analyses could then help identify which patients are most (i.e., high positive predictive value [PPV]) or least (i.e., high negative predictive value [NPV]) in need of PSG referral.

Section snippets

Overview

Participants in 2 separate, multisite, 1-year randomized controlled trials examining the utility of unattended PSG to identify and treat OSA in the post-ischemic stroke or TIA populations with continuous positive airway pressure (CPAP) were used as the study population; the methods of each trial are described elsewhere.11, 12 A clinical prediction model was developed and validated. Results across different OSA prediction instruments were compared.

Patient Populations

From 1 study,11 Veterans with an ischemic stroke

Results

OSA occurred frequently in the development (119 of 194; 61%) and validation (84 of 109; 77%) cohorts, as shown in Table 1. Associations between predictor variables and OSA are shown in Table 2. The final model included an indicator for female (sex), congestive or left heart failure, the ESS, an indicator for having an enlarged neck, weight (in pounds) and weight squared, insulin sensitivity or diabetes, and NIHSS score; these factors can be combined as the SLEEP INventory, with the NIHSS score

Discussion

This study demonstrates that the BQ, ESS, STOP-BANG, and SACS did not strongly predict the presence of OSA on formal PSG in an exclusively post-ischemic stroke or TIA population. We also demonstrated that a clinical prediction rule combining patient symptomatology with readily available and routinely collected patient demographic, anthropometric, medical history, and stroke severity data could be derived from and applied to a population of patients with ischemic stroke or TIA. Although the

Conclusions

This study confirmed the high prevalence of previously undiagnosed OSA and reported the limited utility of the BQ, ESS, STOP-BANG, and SACS as OSA screening instruments for patients with an ischemic stroke or TIA. Given the high prevalence and potential treatment implications in discerning whether a patient with a cerebrovascular event has OSA, if clinicians presume that their patient has OSA, the SLEEP INventory (with its NPV) may help make determinations regarding which patients are least in

References (30)

  • W.N. Kernan et al.

    Guidelines for the prevention of stroke in patients with stroke and transient ischemic attack: a guideline for healthcare professionals from the American Heart Association/American Stroke Association

    Stroke

    (2014)
  • R. Kanagala et al.

    Obstructive sleep apnea and the recurrence of atrial fibrillation

    Circulation

    (2003)
  • M. Arzt et al.

    Dissociation of obstructive sleep apnea from hypersomnolence and obesity in patients with stroke

    Stroke

    (2010)
  • M.W. Johns

    A new method for measuring daytime sleepiness: the Epworth Sleepiness Scale

    Sleep

    (1991)
  • N.C. Netzer et al.

    Using the Berlin Questionnaire to identify patients at risk for the sleep apnea syndrome

    Ann Intern Med

    (1999)
  • Cited by (19)

    • Prediction of sleep-disordered breathing after stroke

      2020, Sleep Medicine
      Citation Excerpt :

      Although physiologic testing is more time-consuming, requires more resources and more expense, it remains necessary for accurate SDB determination after stroke. Other SDB inventories and prediction models such as the Berlin questionnaire, Epworth Sleepiness Scale, STOP-BANG and its variations, SLEEP INventory, and Sleep Apnea Clinical Score also have had poor to fair performance in stroke patients [5,6]. Our prediction model, despite inclusion of demographic, clinical, anthropometric, and questionnaire data, and application of sophisticated statistical methods, did not perform substantially better.

    • Application of Neck Circumference in Four-Variable Screening Tool for Early Prediction of Obstructive Sleep Apnea in Acute Ischemic Stroke Patients

      2019, Journal of Stroke and Cerebrovascular Diseases
      Citation Excerpt :

      Other researcher also demonstrated the majority of stroke patients with OSA do not present with excessive daytime sleepiness, and are statistically less likely to have a high ESS (10% vs 38%; P < .001) versus OSA cases of the general population.12,26 Most recently, ESS has proven to only weakly predict the presence of OSA in ischemic stroke/TIA patients, with one study using a "SLEEP INventory" tool that performed better than the ESS.21 Based on the literature and our results, it seems clear that the ESS is a poor predictor of OSA in acute ischemic stroke/TIA patients.

    • Using a modified version of the “STOP-BANG” questionnaire and nocturnal oxygen desaturation to predict obstructive sleep apnea after stroke or TIA

      2019, Sleep Medicine
      Citation Excerpt :

      When comparing AUCs for different cut-offs of STOP-BAG-O, the greatest AUC was achieved with a cut-off of 5 (AUC = 0.792). This value was comparable to previous literature [20,21] and superior to other studies [16,22]. The value obtained suggests that this screening tool had good accuracy in identifying mild cases of OSA (AHI ≥5).

    View all citing articles on Scopus

    Grant support: This study was supported by the U.S. Department of Veterans Affairs (HSR&D CDA 11-262 and NCT00984308) and the National Institutes of Health, National Heart, Lung, and Blood Institute (NIH/NHLBI U34 HL105285-01) sponsored the study.

    J.J.S., H.K.Y., and D.M.B. had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

    J.J.S. has served as a consultant to Acorda Therapeutics; all other coauthors have no conflicts of interest or disclosures to report.

    View full text