Diagnostic Accuracy of Nocturnal Oximetry for Detection of Sleep Apnea Syndrome in Stroke Rehabilitation
Background and Purpose—Sleep apnea syndrome (SAS) is a common sleep disorder in stroke patients and is associated with decreased recovery and increased risk of recurrent stroke and mortality. The standard diagnostic test for SAS is poly(somno)graphy, but this is often not feasible in stroke rehabilitation settings. This study investigated the diagnostic value of nocturnal oximetry for screening SAS in stroke rehabilitation.
Methods—Fifty-six stroke patients underwent nocturnal polygraphy and oximetry. Sensitivity, specificity, and positive and negative predictive values for the oxygen desaturation index were calculated. Patient and sleep characteristics were used to develop a predictive model of apnea–hypopnea index.
Results—Forty-six percent of the stroke patients had SAS. The majority of SAS patients was male, older, and had a higher body mass index than patients without SAS. Sensitivity, specificity, and positive and negative predictive values for the oxygen desaturation index ≥15 were, respectively, 77%, 100%, 100%, and 83%. Oxygen desaturation index predicted 87% of the variance in the apnea–hypopnea index. Patient characteristics did not add significantly to the prediction model.
Conclusion—Nocturnal oximetry is an accurate diagnostic screening instrument for the detection of SAS in stroke patients.
Sleep apnea syndrome (SAS) is much more common in stroke patients than in the general population, with prevalence ranging from 38% to 78% (2%–4% in the general population).1,2 SAS increases the risk of a stroke independent of other risk factors such as age, obesity, gender, and hypertension.3 In turn, SAS in stroke is associated with decreased functional recovery, increased risk of recurrent stroke, and mortality.4
The standard diagnostic test for SAS is overnight polysomnography or multichannel polygraphy. Full sleep examination, however, is costly, requires technical expertise, and forms an additional burden for the patient. Nocturnal oximetry has been proposed as a screening instrument.5 In the general sleep population, sensitivity and specificity have been found to vary widely (sensitivity 31%–100% and specificity 41%–100%).5 The aim of this study was to validate the use of nocturnal oximetry for screening SAS and to identify possible clinical predictors of SAS in stroke rehabilitation.
Subjects and Methods
We conducted a retrospective study of stroke patients admitted to Heliomare, a rehabilitation center in the Netherlands, during an 18-month period. All patients received active rehabilitation and underwent polygraphy within 2 months after admission. Diagnosis of stroke was confirmed by neurologists based on a history of sudden onset of a neurological deficit lasting >24 hours and a brain lesion compatible with the neurological deficit on computerized tomography or magnetic resonance imaging. Exclusion criteria were: (1) previous diagnosis of SAS or treated for SAS; (2) >6 months after stroke; (3) concomitant central nervous system diseases; (4) traumatic brain injury; and (5) paraplegia. Body mass index (BMI; kg/m2) was calculated for all patients, and patients were classified as overweight (25.0–29.9) and obese (≥30) according to World Health Organization criteria. Stroke was classified by subtype (ischemic or hemorrhagic) and location (supratentorial or infratentorial). Degree of disability attributable to stroke was classified with the modified Rankin Scale. The study was approved by the Institutional Review Board of Heliomare.
Polygraphy included recordings of airflow, oxygen saturation, heart rate, and respiratory effort. The data were recorded with a multichannel digital polygraph, which has been validated against hospital-based polysomnography (POLY-MESAM; Martinsried, Germany).6
Polygraphic records were analyzed automatically and checked manually by an experienced physician (T.V.B.). Apnea was defined as a reduction of airflow of 90% to 100% from baseline lasting >10 seconds and hypopnea was defined as a reduction of airflow of 50% to 90% from baseline lasting >10 seconds associated with an oxygen desaturation of ≥4%. Baseline was determined as an average value over the previous 10 seconds. Apneas with thoracic motion, without thoracic motion, or with initial lack of motion followed by respiratory effort were classified as obstructive, central, or mixed, respectively.
The apnea–hypopnea index (AHI) was defined as the mean number of apneas and hypopneas per hour. SAS was defined by an AHI of ≥15. Oxygen desaturation was determined by a pulse oximeter. The oxygen desaturation index (ODI) was defined as the mean number of desaturations of ≥4% from baseline per hour. Subjective daytime sleepiness was measured with the Epworth Sleepiness Scale.7
The patient sample was characterized with descriptive statistics. We visually checked for outliers in AHI and ODI and used Grubbs extreme deviate test to exclude the outliers from further analysis. Patients with and without SAS (SAS/no SAS) were compared with χ2, Mann-Whitney, or Student t tests. We calculated Pearson correlation coefficients for AHI and ODI. Sensitivity, specificity, positive predictive value, and negative predictive value with 95% confidence interval (CI) were calculated. The diagnostic accuracy of ODI was assessed with receiver-operator characteristic analyses. We developed a predictive model of AHI by calculating a correlation matrix and by entering significantly correlating variables into a simultaneous linear regression analysis. Statistical significance was set at 0.05. A Bonferroni-Holm correction for multiple testing was applied to polygraphic and sleepiness data.
We obtained sleep recordings of 67 stroke patients. Ten patients who met the exclusion criteria and 1 outlier were omitted from further analysis. Fifty-six patients were included in the final analysis (Table 1).
Sixty-two percent of the patients were male. Mean age was 55.6 years (±10.3; range, 26–74). The average duration from stroke onset was 30.3 days (±27.1). Seventy-three percent of the patients had an ischemic stroke and 88% had a stroke in the supratentorial region. Sixteen patients were moderately disabled, 26 had a moderately severe disability, and 14 had a severe disability. Twenty-three percent of the patients were overweight and 18% were obese.
Forty-six percent of the stroke patients had SAS diagnosed. All SAS patients had predominantly obstructive apneas. The majority of SAS patients were male (78% vs 22% female), older, and had higher BMI and ODI than patients without SAS. SAS patients did not differ from patients without SAS in stroke subtype, stroke location, or degree of disability. Excessive daytime sleepiness (Epworth Sleepiness Scale ≥10) was reported by one-third of patients with and without SAS.
The ODI correlated strongly with the AHI (r=0.92; P<0.01). The sensitivity and specificity of pulse oximetry are presented in Table 2. With ODI ≥15, the sensitivity for SAS was 77% (CI, 56%–90%), with 100% specificity (CI, 86%–100%). A lower ODI cut-off (≥5) increased the sensitivity to 96% (CI, 79%–100%), but the specificity declined to 43% (CI, 26%–62%). The diagnostic accuracy of ODI for SAS is represented by the receiver-operator characteristic curve in the Figure. Given a 46% prevalence of SAS in stroke, the positive predictive value of oximetry was 100% (CI, 80%–100%), with a negative predictive value of 83% (CI, 67%–93%).
The clinical variables age, gender, and BMI correlated significantly with AHI. The regression model showed that age and BMI were significant predictors, explaining 51% of the variance in AHI. When ODI was added to the regression analysis, 87% of the variance in AHI was explained, with ODI being the only significant predictor. The resultant regression equation is: predicted AHI=2.66+0.94*ODI.
In the current study, 46% of the stroke patients had SAS diagnosed. We found that nocturnal oximetry is an accurate predictor for SAS, with a sensitivity of 77% and a specificity of 100%. Given the high prevalence of SAS in stroke, a positive oximetry result increased the likelihood of SAS to 100%, whereas a negative result lowered the probability to 17%.
The majority of the SAS patients were male. In addition, SAS patients were older, with a higher BMI than patients without SAS.
Oximetry was the best predictor of SAS, explaining >80% of the variance in AHI. Clinical variables such as age, gender, and BMI did not significantly add to the predictive value of oximetry. Further validation of oximetry in larger samples is required to determine whether our findings can be generalized to other stroke samples.
The authors thank Brita Daniels and Irene Kos for their support with data collection.
- Received June 6, 2012.
- Accepted June 21, 2012.
- © 2012 American Heart Association, Inc.