Study compares cheap take-home sleep test to expensive polysomnography test

Sep 17 2004

Diagnosis of obstructive sleep apnea usually involves polysomnograpy, an overnight sleep test in a sleep clinic or lab. Results of a new study indicate that a take-home sleep test is just as effective as a polysomnography and is less expensive while providing timely results.

Obstructive sleep apnea (OSA) is a common medical condition that occurs in approximately nine to 24 percent of the population and can lead to hypertension, heart problems, and stroke. The effects of untreated OSA are responsible for a two-fold increase in traffic accidents, a decrease in the quality of life of affected patients, and billions of dollars of healthcare costs annually in the US. Obtaining an accurate and timely diagnosis is imperative, but is becoming more difficult as awareness of the disorder grows and more patients require overnight sleep lab tests.

Polysomnography, the gold standard for diagnosis of OSA, is an expensive test that can only be done in a sleep center that accommodates overnight testing. In addition to requiring an overnight stay, some patients may be limited by geographic accessibility to an appropriate sleep lab.

Researchers set out to determine the validity of the SNAP test, a take-home sleep test and whether or not it would produce effective results that could be used as an alternative diagnostic or screening tool for OSA. The study ¡°Validations of a Portable Home Sleep Study with 12-Lead Polysomnography: Comparisons and Insights into a Variable Gold Standard,¡± is authored by Peter G. Michaelson, MD, Patrick F. Allan, MD, John C. Chaney, MD, and Eric A. Mair, MD, of the Department of Otolaryngology¡ªHead and Neck Surgery and Department of Pulmonary/Critical Care and Sleep Medicine at Wilford Hall USAF Medical Center in San Antonio, Texas. Their findings are being presented at the American Academy of Otolaryngology-Head and Neck Surgery Foundation Annual Meeting & OTO EXPO, being held September 19-22, 2004, at the Jacob K. Javits Convention Center, New York City, NY.

Methodology:

This comparison study included 59 adult patients (49 men and 10 women) who presented at Wilford Hall USAF Medical Center (WHMC) Sleep Laboratory for polysomnography (PSG) evaluation between June and August 2003. Average male age was 37.8 with an average body mass index (BMI) of 27 while average female age was 50 years old with a BMI of 24.4. To mirror the population who commonly receive PSG, the only exclusion criteria were: those who did not wish to undergo trial enrolment, those presenting to the sleep laboratory only for a titration (determination of air pressure needed for effective control of OSA) of CPAP (continuous positive airway pressure), and all patients younger than 18 years of age.

Patients underwent a PSG and SNAP test simultaneously during the first half of the night to evaluate for OSA. Those who were determined to have OSA then underwent a CPAP titration for the remainder of the night. The PSG raw data were read in an independent, blinded fashion, by a separate, board-certified group of sleep physicians at WHMC, PSG1, and at an external center, PSG2. SNAP data were read by two independent readers (SNAP1 and SNAP2) at the SNAP laboratories. Since multiple variables are recorded during both tests, a determination was made to use the apnea/hypoxnia index (AHI), the most commonly used variable of sleep characteristics to test for OSA severity, for comparison between the two tests. Several other relationships were calculated, including Pearson correlation coefficients (CC), receiver operating characteristic (ROC) curve, sensitivity, specificity, positive and negative predictive value and Bland-Altman curves. To analyze inter-reader variability, multiple relationships were also calculated between PSG reads (PSG1 and PSG2) and SNAP reads (SNAP1 and SNAP2).

Results:

The average PSG recording time was 256 minutes; average SNAP recording time was 250 minutes. Due to the high correlation coefficient, ROC curve areas and Bland-Altman relationship, both SNAP reads (SNAP1 and SNAP2) were considered interchangeable and SNAP1 was used for further comparison against the PSG data. Comparison of both PSG reads indicated a weaker relationship between different reads.

• Correlation coefficients: The CCs calculated between the different testing modalities is a summary of the strength of the linear association between the variables, or in this instance, AHI. With a perfect, linear relationship being 1, the CC between SNAP1 and PSG1 and PSG2 were 0.882 and 0.916, respectively.
• Receiver operating characteristic: The ROC curve is a graphical representation of the trade off between the false negative and false positive rates for a given cut off, also serves as the representation of the tradeoffs between sensitivity and specificity, and helps to distinguish the accuracy of SNAP detecting individuals with OSA at the given AHI. The area under the curve for SNAP1 and PSG1 for AHI ¡Ý 5 was 0.916 and 0.911 for AHI ¡Ý 15. For SNAP1 and PSG2 it was 0.943 and 0.993 for AHI ¡Ý 5 and AHI ¡Ý 15, respectively.
• Sensitivity, specificity, positive and negative predictive value: Comparison of SNAP1 and PSG1 at AHI 5 or greater: sensitivity 75 percent, specificity 96.7 percent, PPV 95 percent, and NPV 81 percent. At AHI greater than 15: sensitivity 66.6 percent, specificity 100 percent, PPV 100 percent, and NPV 84.7 percent. Comparison of SNAP1 and PSG2 at 5 or greater: sensitivity 94 percent, specificity 86.6 percent, PPV 76 percent, and NPV 97 percent. At AHI greater than 15: sensitivity of 100 percent, specificity of 88.5 percent, PPV of 57 percent, and NPV of 100 percent.
• Bland-Altman curves: The Bland-Altman relationship examines the mean difference of the variables and provides an estimate of whether the two methods, on average, return a similar result. For SNAP1 and both PSG1 and PSG2, the majority of the data points fell within two standard deviations of the mean difference with minimal clustering of values.

Conclusion:

This study uses multiple styles of statistical analysis to determine that there is a solid correlation between SNAP and PSG in measuring AHI, a standard diagnostic measure of OSA. The authors believe these findings indicate that a take-home SNAP test may be proposed as an alternative to overnight PSG for the diagnosis of OSA, especially in selected populations. Furthermore, use of SNAP tests will expand the diagnostic and therapeutic prowess of the practicing otolaryngologist by offering an alternative OSA testing modality that is associated with not only less expense, decreased waiting time and increased convenience, but statistically proven accuracy. Interestingly, the researchers also confirmed that although the PSG remains the gold standard for diagnosis of OSA, it is plagued with inherent variability and problems with reproducibility.