Short-Term Variability in Apnea-Hypopnea Index during Extended Home Portable Monitoring
ABSTRACT
Study Objectives:
Apnea-hypopnea index (AHI) is the primary measure used to confirm a diagnosis of obstructive sleep apnea (OSA). However, there may be significant night-to-night variability (NNV) in AHI, limiting the value of AHI in clinical decision-making related to OSA management. We examined short-term NNV in AHI and its predictors during home portable monitoring (PM).
Methods:
Single center prospective observational study of patients (n = 84) with newly diagnosed OSA by polysomnography (PSG) AHI ≥ 5/h. All participants underwent 2 to 8 consecutive nights of PM.
Results:
Participants (n = 84) were middle-aged (47 ± 8.3 y, mean ± standard deviation; SD), including 28 women, with mean AHI on baseline PSG (AHIPSG) of 30.1 ± 31.8. Mean AHI on PM (AHIPM) was 27.4 ± 23.7. Intraclass correlation coefficient (ICC) for AHIPM in the entire sample was 0.73 (95% CI 0.66–0.8), indicating that 27% of the variability in AHIPM was due to intra-individual factors. Mild severity of OSA, defined by AHIPSG 5–15/h, was associated with higher NNV (likelihood ratio, −0.4 ± 0.14; p = 0.006) and absence of comorbidity showed a trend towards higher NNV (−0.54 ± 0.27, p = 0.05) on AHIPM.
Conclusions:
The intraindividual short-term NNV in AHIPM is higher in mild versus moderately severe OSA, even in the home setting, where first-night effect is not expected. Larger studies of NNV focused on patients with mild OSA are needed to identify characteristics that predict need and timing for repeated diagnostic testing and treatment.
Commentary:
A commentary on this article appears in this issue on page 787.
Citation:
Prasad B, Usmani S, Steffen AD, Van Dongen HP, Pack FM, Strakovsky I, Staley B, Dinges D, Maislin G, Pack AI, Weaver TE. Short-term variability in apnea-hypopnea index during extended home portable monitoring. J Clin Sleep Med 2016;12(6):855–863.
INTRODUCTION
Apnea-hypopnea index (AHI) obtained from sleep studies is the primary measure of obstructive sleep apnea (OSA) utilized for diagnosis and treatment decisions by clinicians and third-party payers.1 AHI ≥ 5/h of sleep is a widely accepted criterion to diagnose OSA and qualifies patients for treatment reimbursement. Moreover, AHI provides a good indication of the severity of OSA and risk of adverse health outcomes and mortality.2–4 Nevertheless, AHI has several limitations as a sole diagnostic and prognostic indicator.5 One important limitation is the previously described intraindividual night-tonight variability in the expression of OSA (captured by AHI).6 High night-to-night variability may lead to intermittent or less cumulative exposure in individuals to sleep-disordered breathing events and consequences of OSA. Thus, night-tonight variability will not only influence diagnostic accuracy and management decisions, but may also affect symptoms, adherence and prognosis. However, studies examining night-to-night variability are limited by sample size, use of different technologies (polysomnography; PSG versus portable monitor; PM) and test-setting (attended/in-laboratory versus unattended/home) as well as monitoring time (number of nights tested, interval between tests).7–9 Although results across studies are conflicting, a higher night-to-night variability in patients with milder OSA severity has been reported.10 There is limited data regarding night-to-night variability with home PM performed in small samples or heterogeneous populations.11–13 Levendowski et al.14 showed that in a small clinical sample of mild to moderate OSA, home PM exhibits 50% less variability in AHI compared to in-laboratory PSG performed a few months apart.
BRIEF SUMMARY
Current Knowledge/Study Rationale: This study was undertaken to determine night-to-night variability in expression of obstructive sleep apnea (OSA) during extended home portable monitoring.
Study Impact: This study extends our knowledge of night-to-night variability in OSA by demonstrating that patients with mild OSA are more likely to exhibit night to night variability in their apnea-hypopnea index and time of recording in hypoxemia on home portable monitoring. This is important for management and prognostication of mild OSA.
Home PM is increasingly used as a first-choice diagnostic test by many health care systems, but the potential for missed diagnosis and misclassification of disease severity with home PM remains uncertain.15 This is important as home PM may be most relevant to patient's “true disease burden” and where “noise” such as variability due to first-night effect is minimized.16–18 We hypothesized that night-to-night variability assessed in patients' usual sleep environment (by home PM) would be higher in mild OSA and in patients without signs and symptoms of sleepiness (due to lower chronic exposure to intermittent hypoxia and sleep fragmentation). We prospectively examined night-to-night variability in AHI with home PM (AHIPM) over 2 to 8 nights in a sample of middle-aged newly diagnosed OSA patients. Our primary objective was to compare night-to-night variability in OSA expression during extended home PM as quantified by AHIPM and time (in minutes) of recording spent with oxygen saturation below 90% (O2 saturation < 90%) between groups with mild versus moderate to severe OSA and between groups with or without objective and subjective sleepiness. Additionally, we explored potential clinical predictors of night-to-night variability in AHIPM and O2 saturation < 90% and the effect of extended home PM on night-to-night variability. AHI and total time with hypoxemia were chosen as they are commonly used as complementary disease severity indices.19
METHODS
Participants
This study (protocol B) was performed on a subset of participants recruited for a larger study (protocol A, n = 398), which was aimed at examining determinants of sleepiness in treatment naïve patients with OSA. Details on participant recruitment and procedures are presented in Figure 1.
AHI, apnea-hypopnea index; ESS, Epworth sleepiness scale; PVT, psychomotor vigilance test; PM, portable monitoring. “SLEEPY,” PVT lapse ≥ 2 and ESS ≥ 11; “NON-SLEEPY,” PVT lapse < 2 and ESS < 11. *84 participants with ≥ 2 nights of home portable monitoring data with simultaneous actigraphy and sleep diary.
For protocol A, consenting participants were recruited from the University of Pennsylvania Center for Sleep Disorders, where they were referred for possible OSA. Inclusion criteria were: age 35–60 y and AHI on baseline PSG; AHIPSG ≥ 5/h. PSG was scored according to American Academy of Sleep Medicine recommended criteria, including 3% oxygen desaturation or arousal for hypopnea.20 Standard PSG criteria were used to diagnose and define severity of OSA (mild OSA: AHIPSG = 5–15/h and moderate to severe OSA: AHIPSG > 15/h). Middle-aged adults were chosen to avoid age-related alterations in sleep-wake schedules. The exclusion criteria were: (1) diagnosis of another primary sleep disorder on PSG (e.g., periodic limb movement index with arousal > 10/h; central sleep apnea index > 5/h; sleep-related hypoventilation syndrome); (2) diagnosis of another sleep disorder by history such as insomnia, parasomnia, or narcolepsy; (3) previous treatment of OSA, including home oxygen, uvulopalatopharyngoplasty, tracheotomy, or other surgery for OSA; (4) pregnancy; and (5) unable to perform tests—inability to write and read in English, less than a 5th grade reading level, visual impairment, hearing impairment, cognitive impairment, or motor deficit (e.g., previous stroke or spinal cord injury). All participants in protocol A had sleepiness assessments performed with the Psychomotor Vigilance Task (PVT, Ambulatory Monitoring Inc., Ardsley, NY) and the Epworth Sleepiness Scale (ESS) during protocol A. Two 10-min trials of PVT were performed by each participant, first in the morning and second in the afternoon, to account for circadian variation in alertness.21,22 Objective sleepiness was defined as number of PVT lapses (reaction time more than 500 msec) ≥ two per trial. Subjective sleepiness was defined as ESS ≥ 11.23
Protocol B aimed to examine night-to-night variability as a basis for differential manifestation of sleepiness in OSA patients with similar AHIPSG. It was designed to test the hypothesis that “sleepy” patients have more cumulative exposure to sleep-disordered breathing events with higher mean AHIPM due to less night-to-night variability, compared to “nonsleepy” patients. Therefore, only participants who were concordantly sleepy (PVT lapse ≥ 2 and ESS ≥ 11) or concordantly non-sleepy (PVT lapse < 2 and ESS < 11) were invited to continue in protocol B. A detailed description of protocol B is provided in the next paragraphs. This study was approved by the University of Pennsylvania, Philadelphia, Institutional Review Board.
Study Protocol
Participants presented to the clinical research center and informed consent were obtained for protocol B. Urine pregnancy and drug screening tests were performed. They received instruction in the use of the EmblettaTM PDS system (Natus Neurology Inc., Middleton, WI) by trained study personnel. Embletta is a level 3 PM for diagnosis of OSA and is designed for home use.24,25 This system measures oxygen saturation, nasal-oral airflow by thermistor, snoring, nasal pressure, and abdominal and thoracic respiratory effort with inductance plethysmography belts. Participants were scheduled for consecutive nights of home PM the following week (PSG and home PM were performed 4 w apart). They also received an actigraph (Actiwatch, Mini Mitter Company, Bend, OR) and a sleep diary that they wore and used during the week of Embletta recordings. Data from the diary and actigraph were used to establish their sleep time during PM testing. A trained research technician went to the home of participants and attached the Embletta device to the participant, who recorded on the device during their major sleep period. The next evening, the technician returned to collect and replace the device with another Embletta device, and this process continued for 2–8 nights for all participants. The technician downloaded the Embletta data each day.
Two or more nights of PM data with simultaneous actigraphy and sleep diary data were successfully obtained in 84 participants (final sample). Participants who completed only 1 night of home PM were excluded from analysis. Following was the distribution of data from individual participants based on their completion of PM testing: 4 (4.8%) with 2 nights, 5 (6.0%) with 3 nights, 10 (11.9%) with 4 nights, 11 (13.1%) with 5 nights, 23 (27.4%) with 6 nights, 28 (33.3%) with 7 nights, and 3 (3.6%) with 8 nights. Data analysis was done using Somnologica software followed by manual scoring by a registered polysomnography technician to generate final reports for each night of recording. Respiratory events for PM recordings with Embletta device were scored according to American Academy of Sleep Medicine criterion of 3% oxygen desaturation for hypopneas.20
Statistical Analyses
Descriptive statistics and variable distribution plots were examined for extent of missing data and non-normal distributions. The Box-Cox method was used to determine the natural log as the optimal transformation to improve distributions of the key outcomes of interest; AHIPM and O2 saturation < 90%. Baseline characteristics were compared between mild (AHIPSG 5–15) versus moderate to severe OSA (AHIPSG ≥ 15) using two-sided Fisher exact or Wilcoxon tests. Overall severity across nightly measures was summarized accounting for the differing number of observations per person (2–8) and the correlation of repeated measures, using the fixed effect intercept (β0) in a mixed regression model with random intercepts for each participant. Intraclass correlation coefficients (ICC) were also calculated from these models. We assessed the a priori hypothesis of sleepiness as a predictor of night-to-night variability using two approaches: by comparing the ICC for sleepy and nonsleepy groups and adding ESS and PVT lapses as continuous predictors in the final model for multivariate analysis described in the next paragraphs.
To examine if night-to-night variability differed between mild and moderate to severe OSA groups, first the ICC approach was applied. Next, we applied mixed-effect regression models to jointly model OSA severity and AHIPM variability (standard deviations in log-linear form) and their association with patient characteristics using the location scale approach.26 In these models we regressed AHIPM (up to 8 nights per person) onto predictors to determine correlates of severity, the mean AHIPM per person or “location”. Within the same model, we then examined correlates of night-to-night variability for the AHIPM measures for each person. In our initial unadjusted location scale model we estimated a grand mean AHIPM across individuals (β0), an average intraindividual variance for the night-to-night variability in AHIPM (τ0), the inter-individual variance of the random intercepts (σ2υ), i.e., the variance of the intraindividual differences (u1) between each person's average and β0, and an interindividual variance (σ2ω) of the random intraindividual variances (u2), with τ and σ2υ components represented in log linear form to ensure positive variance estimates. In subsequent models we added predictors (βs) of disease severity and (τs) of night-to-night variability after checking for correlation between predictors to avoid collinearity. Likelihood ratio tests were used to compare nested models. The effect of extended home PM on night-to-night variability was assessed with mixed-effect regression model to jointly model number of nights of home PM intraindividual AHIPM data and assessment variability, similar to the previously described location scale approach. All analyses were conducted using the NLMIXED procedure within SAS software, version 9.3 (SAS Institute, Cary, NC).
RESULTS
Participant Characteristics
Table 1 shows baseline characteristics for the entire sample as well as groups defined by OSA severity (mild, AHIPSG < 15 and moderate to severe, AHIPSG ≥ 15). Participants were equally distributed by severity: mild, n = 42 and moderate to severe, n = 42. There were no differences between mild versus moderate to severe OSA groups in demographics, age, body mass index (BMI), and PVT lapses. Forty-six percent of the entire sample (39 of 84) reported comorbid medical conditions. The most common comorbid condition was diabetes (35%). Hypertension, obstructive lung disease, congestive heart failure, myocardial infarction, or stroke was reported by a minority of participants at ≤ 5%. There were no differences in prevalence of comorbidity between groups by severity of OSA (chi-square 0.04, p = 0.8) or sleepiness status (chi-square 1.08, p = 0.3). Although sleep efficiency by actigraphy was not significantly different in the moderate-to-severe OSA group (p = 0.24), this group did demonstrate less 24-h total sleep time (TST) and nocturnal sleep time (NST), and reported more daytime sleepiness.
 Table 1 Baseline characteristics. |
Intraclass Correlation Coefficients
ICC represents the proportion of the variance related to interindividual differences, and the remaining variance is attributable to intraindividual differences.27 Both untransformed and transformed AHI and O2 saturation < 90% are presented for ease of interpretation in the clinical context and to improve data distribution, respectively. The ICCs for AHIPM and O2 saturation < 90% across nights in the entire sample as shown in Table 2 were 0.83 (95% confidence interval, CI 0.78–0.87) and 0.75 (0.68–0.81), respectively. The ICCs for natural log-transformed AHIPM and O2 saturation < 90% were similar (lnAHIPM = 0.73; 0.66–0.80 and lnO2 < 90 = 0.67; 0.59– 0.75, respectively), indicating good to fair overall intraindividual agreement in these measures.28 Nevertheless, the ICCs for lnAHIPM and lnO2 saturation < 90% indicate that 27% to 33% of the variance in these measures was due to intraindividual variation across nights. The ICCs for un-transformed AHIPM and O2 saturation < 90% as well as lnAHIPM and lnO2 saturation < 90% were not significantly different between sleepy versus nonsleepy groups with overlapping 95% confidence intervals.
 Table 2 Measures of intraindividual agreement for portable monitoring across nights. |
Table 2 further depicts the ICCs for untransformed as well as lnAHIPM and lnO2 saturation < 90% for mild and moderate to severe OSA groups. There was a trend toward poorer intraindividual agreement with lower ICCs in the mild versus moderate to severe OSA group, which reached statistical significance only for untransformed metrics (AHIPM and O2 saturation < 90%). The ICCs in “sleepy” compared to “nonsleepy” group were numerically lower for lnAHIPM and lnO2 saturation < 90%, but this difference was not statistically significant.
Intraindividual Night-to-Night Variability in AHIPM
As shown in Table 1, 42 participants had AHIPSG of < 15 (mild OSA group) and the remaining 42 participants had AHIPSG of ≥ 15 (moderate to severe OSA group). First, we assessed the number of participants that changed classification by AHIPM across nights of PM in each of these groups. OSA severity classification based on AHIPM was defined as normal (AHIPM < 5), mild (AHIPM 5–15) and moderate to severe (AHIPM > 15). Twenty-six of 42 participants (62%) changed classification across nights of PM in mild OSA group. Ten of 42 participants (24%) participants changed classification across nights of PM in moderate to severe OSA group. In order to further examine if intra-individual night-to-night variability was related to severity of OSA as indicated by mean AHIPM (unadjusted model), a plot was created of the empirical Bayes estimates of each individual's random intercepts for mean severity (y-axis) versus that for variance in AHIPM across nights (x-axis), on a standardized scale (unadjusted model, Figure 2). A positive value on the y-axis indicates an individual with higher mean AHIPM (averaged across nights) than the sample average (red dotted line), whereas a negative value on the y-axis indicates an individual with mean AHIPM below the sample average. A positive value on the x-axis indicates that an individual had higher night-to-night variability than the average sample night-to-night variability (green dotted line) across nights, and a negative value on the x-axis indicates less than the average sample night-to-night variability. This plot shows that individuals' means and variances were heterogeneous in this sample and that the two were inversely related, i.e., more night-to-night variability in AHIPM was observed among individuals with lower mean AHIPM (milder OSA; p = 0.0005).
*Individuals in the four quadrants have: Quadrant 1 = higher mean AHI and lower variance compared to sample average across nights of PM, Quadrant 2 = higher mean AHI and higher variance compared to sample average across nights of PM, Quadrant 2 = lower mean AHI and lower variance compared to sample average across nights of PM, Quadrant 4 = lower mean AHI and higher variance compared to sample average across nights of PM.
Predictors of Night-to-Night Variability
In this sample, we did not find significant correlations between body mass index (BMI), comorbidity (yes/no responses for self-reported medical, neurologic and psychiatric disorders) and AHIPSG. We examined age, sex, race, BMI, comorbidity, AHIPSG, nocturnal sleep time (NST), PVT lapses, and ESS as predictors of intraindividual night-to-night variability in lnAHIPM and lnO2 saturation < 90% within a series of nested models. Age, sex, and race did not improve the model fit (χ2(3) = 4.2, p > 0.20) and were removed from the intraindividual night-tonight variability portion of the model but remained as fixed effects. Our final model included all aforementioned fixed effects and incorporated BMI, AHIPSG, comorbidity, NST, PVT lapses, and ESS as predictors of night-to-night variability (Table 3). As shown in Table 3, lower AHIPSG was significantly associated with higher intraindividual night-to-night variability in lnAHIPM (parameter estimate −0.40 ± 0.14, p = 0.006) whereas absence of comorbid illness was associated with a trend toward higher night-to-night variability in lnAHIPM (parameter estimate −0.54 ± 0.28, p = 0.05). Intraindividual night-to-night variability in lnO2 saturation < 90% was also inversely associated with AHIPSG (parameter estimate −0.44 ± 0.15, p = 0.05), but no other significant predictors were found.
 Table 3 Predictors of intraindividual night-to-night variability in sleep apnea expression on portable monitoring. |
In total, 31% of the night-to-night variability in lnAHIPM and 17% of the night-to-night variability in lnO2 saturation < 90% was explained by the predictors in our final model. To examine the potential utility of extended home PM in reducing night-to-night variability, we evaluated the effect of the number of nights of home PM data (2 to 8 nights) on night-to-night variability. Increasing number of nights of AHIPM data had no effect on night-to-night variability (parameter estimate = −0.05, p = 0.6).
DISCUSSION
In this study, we show that short-term night-to-night variability in AHI measured during home PM is associated with severity of OSA. Clinical populations with mild OSA diagnosed by PSG exhibit greater night-to-night variability in AHI compared to those with moderate to severe OSA. Higher night-tonight variability may increase the potential for misdiagnosis and misclassification of disease severity in mild OSA. Inaccurate estimation of true severity, i.e., cumulative or chronic exposure to nocturnal hypoxia and sleep fragmentation, may explain the conflicting results of studies examining cardiovascular morbidity and the public health effect of mild OSA.2,29–31 Moreover, the efficacy and effectiveness of first-line treatment with continuous positive airway pressure, particularly in asymptomatic individuals, have not been established leading to clinical equipoise regarding treatment of mild OSA.32–35 To our knowledge this is the first study describing severity of OSA (defined by current gold-standard, AHIPSG) as a determinant of night-to-night variability in the home setting with extended level 3 PM.24 Similar results were previously reported in a small sample across 2 nights of PSG, with different respiratory scoring criteria.10 Data from the large Sleep Heart Health Study on home PSG performed months apart suggest a similar trend in mild OSA.36
Our overall night-to-night variability in AHIPM was slightly more than previously reported on 3 consecutive nights of home PM (ICC 0.83 versus 0.90).13 Unlike our investigation, the previous study was retrospective, evaluated all cases referred for diagnostic testing (range of AHI was 0 to > 100), did not include comparative PSG data and used a different hypopnea definition (4% oxygen desaturation) with automated scoring. Nevertheless, it should be noted that an ICC of 0.83 compares favorably to reproducibility of blood pressure measurements.37,38 Interestingly, similar to our findings of greater reproducibility in more severe disease, numerically higher ICCs have been reported in individuals with uncontrolled (higher) blood pressure.38
Our observation that increasing the number of nights of PM within individuals from 2 to 8 did not affect night-to-night variability suggests that extended home PM is not an effective strategy to reduce diagnostic errors in all patients with suspected OSA due to night-to-night variability. However, the utility of extended home PM should be prospectively examined in mild OSA, where we note significant night-to-night variability. In this regard, it should be noted from Table 1, that in the group with mild OSA (AHIPSG < 15) mean AHIPM was higher than AHIPSG (Δ 6.6; p < 0.0001). Conversely, mean AHIPM was lower than AHIPSG in the moderate to severe OSA group (AHIPSG < 15, Δ −12.1 p = 0.009). As observed in cases with moderate to severe OSA, it is expected that AHIPM will underestimate the severity of OSA compared to PSG, due to lack of electroencephalographic assessment of sleep time and arousals.24 The apparent “overestimation bias” in clinical populations with mild OSA on extended PM and its sources (technological versus short-term variation in expression of mild OSA) merits further investigation.
Few studies have previously examined factors associated with night-to-night variability in primarily healthy volunteers or in community-based cohorts.36,39,40 Consistent with previous reports, we did not find age, sex, or BMI to be significant determinants of night-to-night variability. Individuals without medical comorbidity showed a trend toward higher night-tonight variability. In this regard, it is important to note that comorbidity was not correlated with other predictors, including AHIPSG and BMI. However, the association of comorbidity with night-to-night variability is difficult to interpret in this small sample, so unmeasured confounding cannot be excluded. We used simultaneous actigraphy and sleep diary with PM, thereby estimating and adjusting for effects of “nocturnal sleep time” in individuals' usual sleep environment on night-to-night variability. Nocturnal sleep time was higher in the mild OSA group and it did not appear to affect night-to-night variability.
This study provides novel insights into the relationship of night-to-night variability with sleepiness (objective with PVT and subjective with ESS) in OSA. It is acknowledged that daytime sleepiness in OSA populations exhibits only a modest correlation with AHI.41 However, this may represent an underestimate of the “true association” due to inherent instability of AHI as a disease severity measure, i.e., night-to-night variability. Our data indicate that night-to-night variability is not associated with objective or subjective sleepiness, as neither PVT lapses nor ESS significantly predicted night-to-night variability. Furthermore, the ICC's were not significantly different for sleepy and nonsleepy groups. Thus, night-to-night variability does not affect the association between daytime sleepiness and AHI. Nevertheless, as signs and symptoms of sleepiness are the primary indication for treatment of mild OSA, the effect of these signs and symptoms on night-to-night variability should be prospectively examined in populations with mild OSA and primary snoring.
Interestingly, the treatment naïve moderate to severe OSA group not only reported more daytime sleepiness, but also demonstrated shorter total and nocturnal sleep times by actigraphy. Risso et al.42 have reported similar findings, which suggests that self-reported sleepiness even in moderate to severe OSA is in part due to insufficient sleep. Both severity of OSA and short sleep duration have adverse effects on cardiovascular risk factors, for example hypertension and metabolic syndrome.43,44 The mechanisms underlying this association and the nature (additive versus synergistic) of its combined effect on daytime symptoms and cardiovascular health are critical gaps in our knowledge. We noted mean nocturnal sleep time of 6 h, which is not defined as short sleep duration in the general population, but may be insufficient time for sleep-induced physiologic restoration in clinical populations with moderate to severe OSA.
The strengths of this study include prospective design, definition of OSA based on current gold standard: AHIPSG ≥ 5 and up to 8 nights of data collection with home PM with simultaneous actigraphy. Specifically trained technologists were used for PSG and manual scoring of PM data to minimize technical errors and optimize interrater reliability. Nevertheless, there are limitations that merit consideration in interpreting our findings and for future research. First, this study included a sleep center referral population, where OSA and its severity were defined by AHIPSG. Although AHIPSG is the clinical gold standard for this purpose, its limitations as a single value are recognized. Second, only participants who were concordant on self-reported sleepiness and PVT determined vigilance were enrolled. Thus, potential referral and selection bias may exist and our findings may not be generalizable to all populations where PM is performed. Future studies should include clinical populations with primary snoring and all levels of sleepiness. However, our inclusion criterion of PSG determined AHI ≥ 5/h in clinical populations is consistent with the definition of OSA. Notably, AHIPSG was not significantly different (p = 0.49) in the groups included (concordantly sleepy or nonsleepy, 33.2 ± 31.5 and 27.8 ± 25.6, respectively) compared to those excluded due to discordant results on PVT and ESS (34 ± 31.9, 28.3 ± 27.9). Moreover, we did not find differences in night-to-night variability between the included sleepy and non-sleepy groups indicating a low likelihood of this bias. Therefore, we believe our findings are clinically relevant and applicable to the evaluation of OSA.
Due to the small sample size and missing data, we did not analyze the effect of body position on night-to-night variability, but time spent supine does not predict night-to-night variability according to prior reports.6,7 Although different AHI calculations (with 4% hypopnea criteria) were not assessed in this study because it does not significantly influence night-to-night variability, such expanded analyses in future studies may provide additional insights.36 Another aspect of night-tonight variability in mild OSA that should be examined in future studies is the effect of specific comorbid conditions such as cerebrovascular accident (CVA), chronic cardiopulmonary disease and sleep disorders such as insomnia.
In summary, the frequency of respiratory disturbance and exposure to pathophysiologic events in clinical populations with mild OSA exhibits short-term night-to-night variability. This may explain why many patients with mild OSA do not experience improvement in daytime function, are less adherent to prescribed treatment, and why some studies in mild OSA fail to demonstrate measurable improvement in health outcomes. Adequately powered prospective studies are needed to better understand the determinants and clinical significance related to long-term health implications of night-to-night variability in mild OSA. These data will allow targeted treatment of mild OSA to improve health outcomes.
DISCLOSURE STATEMENT
This was not an industry supported study. Funding was provided by NHLBI, 5P50HL060287-10. The project described was supported by 5P50HL060287-10, National Heart, Lung, and Blood Institute (Drs. Pack and Weaver). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Heart, Lung, and Blood Institute or the National Institutes of Health. Dr. Van Dongen is a Consultant and received grant support from Pulsar Informatics, Inc. and FedEx Express. Dr. Weaver has received grant support for an investigator-initiated grant from Teva, Inc. and royalty fees for the Functional Outcome of Sleep Questionnaire from Philips Respironics, Inc., Revent, Inc., and ResMed, Inc. Dr. Dinges is compensated by the Associated Professional Sleep Societies, LLC, for serving as Editor in Chief of SLEEP and has received compensation for serving on a scientific advisory council for Mars, Inc. Dr. Dinges recuses himself from all decisions related to SLEEP manuscripts on which he has a conflict of interest. Dr. Prasad acknowledges her current research support (1IK2CX001026-01; VA CSR&D), which allowed preparation of this manuscript. The other authors have indicated no financial conflicts of interest. This study does not involve the use of off-label or investigational use of drugs or devices. The work was performed at the University of Pennsylvania, Philadelphia, PA
ABBREVIATIONS
| AHI | apnea-hypopnea index |
| BMI | body mass index |
| CI | confidence interval |
| CVA | cerebrovascular accident |
| ESS | Epworth sleepiness scale |
| ICC | intraclass correlation coefficient |
| NST | nocturnal sleep time |
| OSA | obstructive sleep apnea |
| PM | portable monitor |
| PSG | polysomnography |
| PVT | psychomotor vigilance task |
REFERENCES
1 Updates on definition, consequences, and management of obstructive sleep apnea. Mayo Clin Proc; 2011;86:549-54, 21628617.
2 Prospective study of the association between sleep-disordered breathing and hypertension. N Engl J Med; 2000;342:1378-84, 10805822.
3 Sleep apnea and 20-year follow-up for all-cause mortality, stroke, and cancer incidence and mortality in the busselton health study cohort. J Clin Sleep Med; 2014;10:355-62, 24733978.
4 Prevalence of sleep disordered breathing in paroxysmal and persistent atrial fibrillation patients with normal left ventricular function. Eur Heart J; 2008;29:1662-9, 18515807.
5 Health Quality OPolysomnography in patients with obstructive sleep apnea: an evidence-based analysis. Ont Health Technol Assess Ser; 2006;6:1-38.
6 Mild to moderate sleep respiratory events: one negative night may not be enough. Chest; 2000;118:353-9, 10936124.
7 The variability of the apnoeahypopnoea index. J Sleep Res; 2001;10:245-51, 11696078.
8 Assessment of the test-retest reliability of laboratory polysomnography. Sleep Breath; 2009;13:163-7, 18766393.
9 Clinical diagnosis of sleep apnea based on single night of polysomnography vs. two nights of polysomnography. Sleep Breath; 2009;13:221-6, 19067010.
10 Night-tonight consistency of apneas during sleep. Am Rev Respir Dis; 1984;129:244-6, 6696325.
11 Night-to-night variability of disturbed breathing during sleep in an elderly community sample. Sleep; 1991;14:252-8, 1896727.
12 Measurement of sleep-related breathing disturbances in epidemiologic studies. Assessment of the validity and reproducibility of a portable monitoring device. Chest; 1991;100:1281-6, 1935282.
13 Nightly variability of sleep-disordered breathing measured over 3 nights. Otolaryngology; 2004;131:837-43.
14 The impact of obstructive sleep apnea variability measured in-lab versus in-home on sample size calculations. Int Arch Med; 2009;2:2, 19121211.
15 Home sleep testing for obstructive sleep apnea: one night is enough!. Chest; 2013;143:291-94, 23381307.
16 Paradoxes of the first-night effect: a quantitative analysis of antero-posterior EEG topography. Clin Neurophysiol; 2004;115:1178-88, 15066543.
17 The effect of in-laboratory polysomnography on sleep and objective daytime sleepiness. Sleep; 2000;23:1109-13, 11145325.
18 Evaluation of first night effect using ambulatory monitoring and automatic sleep stage analysis. Sleep; 1988;11:273-6, 3399780.
19 Hour-to-hour variability of oxygen saturation in sleep apnea. Chest; 1998;113:719-22, 9515849.
20 Sleep-related breathing disorders in adults: recommendations for syndrome definition and measurement techniques in clinical research. the report of an American Academy of Sleep Medicine Task Force. Sleep; 1999;22:667-89, 10450601.
21 Reliability of a single objective measure in assessing sleepiness. Sleep; 2012;35:149-58, 22215929.
22 Sleep-disordered breathing and psychomotor vigilance in a community-based sample. Sleep; 2007;30:1309-16, 17969464.
23 Sleepiness in different situations measured by the Epworth Sleepiness Scale. Sleep; 1994;17:703-10, 7701181.
24 Practice parameters for the use of portable monitoring devices in the investigation of suspected obstructive sleep apnea in adults. Sleep; 2003;26:907-13, 14655928.
25 Validation of Embletta portable diagnostic system for identifying patients with suspected obstructive sleep apnoea syndrome (OSAS). Respirology; 2010;15:336-42, 20199644.
26 Modeling between-subject and within-subject variances in ecological momentary assessment data using mixed-effects location scale models. Stat Med; 2012;31:3328-36, 22419604.
27 The design and analysis of clinical experiments: reliability of measurement. New York, NY: John Wiley & Sons, Inc.1-32; 1999.
28 Agreement in the scoring of respiratory events and sleep among international sleep centers. Sleep; 2013;36:591-6, 23565005.
29 Obstructive sleep apnea and the risk of sudden cardiac death: a longitudinal study of 10,701 adults. J Am Coll Cardiol; 2013;62:610-6, 23770166.
30 Prospective study of sleep-disordered breathing and hypertension: the Sleep Heart Health Study. Am J Respir Crit Care Med; 2009;179:1159-64, 19264976.
31 Sleep apnea & automobile crashes. Sleep; 1999;22:790-5, 10505825.
32 Mild obstructive sleep apnea syndrome should not be treated. Con. J Clin Sleep Med; 2007;3:263-4, 17561592.
33 Effect of CPAP on blood pressure in patients with OSA/hypopnea a systematic review and meta-analysis. Chest; 2014;145:762-71, 24077181.
34 The Role of Severity of Obstructive Sleep Apnea Measured by Apnea-Hypopnea Index in Predicting Compliance With Pressure Therapy, a Meta-analysis. Am J Ther; 2014;21:260-4, 22407199.
35 Continuous positive airway pressure treatment of sleepy patients with milder obstructive sleep apnea: results of the CPAP Apnea Trial North American Program (CATNAP) randomized clinical trial. Am J Respir Crit Care Med; 2012;186:677-83, 22837377.
36 Short-term variability of respiration and sleep during unattended nonlaboratory polysomnography--the Sleep Heart Health Study. Sleep; 2002;25:843-9, 12489889.
37 Reproducibility of resting blood pressure and heart rate measurements. The HERITAGE Family Study. Ann Epidemiol; 2000;10:271-7, 10942874.
38 Reliability and reproducibility of clinic and home blood pressure measurements in hypertensive women according to age and ethnicity. Blood Press Monit; 2009;14:49-57, 19305185.
39 Factors associated with nightly variability in sleep-disordered breathing in the elderly. Chest; 1991;100:973-6, 1914614.
40 Nightly variability in the indices of sleep-disordered breathing in men being evaluated for impotence with consecutive night polysomnograms. Sleep; 1996;19:589-92, 8899939.
41 Daytime sleepiness and polysomnography in obstructive sleep apnea patients. Sleep Med; 2008;9:727-31, 18482866.
42 The impact of sleep duration in obstructive sleep apnea patients. Sleep Breath; 2013;17:837-43, 23099474.
43 Obstructive sleep apnea: a cardiometabolic risk in obesity and the metabolic syndrome. J Am Coll Cardiol; 2013;62:569-76, 23770180.
44 A review of evidence for the link between sleep duration and hypertension. Am J Hypertens; 2014;27:1235-42, 24778107.
