Compare nocturnal REM sleep without atonia (nRWA) and REM sleep behavior disorder (RBD) between pediatric patients with and without narcolepsy and determine if the nRWA index is a valid diagnostic biomarker for narcolepsy.
Retrospective cohort study of children ages 6 to 18 years who completed a nocturnal polysomnogram (PSG) and Multiple Sleep Latency Test (MSLT). Our study sample included 11 patients with narcolepsy type 1 (NT1), 6 with narcolepsy type 2 (NT2), 12 with idiopathic hypersomnia (IH), and 11 with subjective hypersomnia (sHS). We compared group nRWA indices (epochs of RWA/total stage R sleep epochs) from the nocturnal PSGs and analyzed nRWA index receiver operating curve (ROC) statistics for narcolepsy diagnosis.
The median nRWA index of patients with NT1 was 15 to 30 times higher compared to sHS and IH (Ps < .005) but similar to that of the NT2 group (P = .46). RBD was present in 25% of patients with narcolepsy (NT1 and NT2). In comparing those with and without narcolepsy, the nRWA index area under the curve was 0.87 (0.6), 95% confidence interval (CI) = 0.75 to 0.99, P < .001. The threshold of having ≥ 1% of stage R sleep epochs with nRWA yielded a sensitivity of 88.2%, 95% CI = 63.6–98.5 and specificity of 60.9%, 95% CI = 38.5 to 80.3 for diagnosis of narcolepsy. In contrast, a threshold of ≥ 8% yielded a specificity of 95.7%, 95% CI = 78.1 to 99.9 and sensitivity of 52.9%, 95% CI = 27.8 to 77.
The nRWA index is a very good diagnostic biomarker of pediatric narcolepsy. Depending on the clinical cutoffs utilized, this biomarker can identify more children/adolescents with narcolepsy using just the PSG or reduce false-positive diagnostic results.
Bin-Hasan S, Videnovic A, Maski K. Nocturnal REM sleep without atonia is a diagnostic biomarker of pediatric narcolepsy. J Clin Sleep Med. 2018;14(2):245–252.
Current Knowledge/Study Rationale: REM sleep behavior disorder (RBD) and REM sleep without atonia (RWA) are well-described comorbidities in adult populations with narcolepsy, but clinical characteristics in pediatric patients with narcolepsy are less known. We describe clinical characteristics of RBD and RWA in pediatric patients with narcolepsy and determine if a nocturnal RWA index can serve as a diagnostic biomarker for pediatric narcolepsy.
Study Impact: RWA and RBD are important to consider when scoring and interpreting diagnostic sleep studies in pediatric patients with possible narcolepsy. Recognition of the nRWA index as a clinically useful biomarker could help reduce the delay in diagnosis and misdiagnosis of pediatric narcolepsy.
Management of narcolepsy is plagued by diagnostic delays and misdiagnosis, highlighting a need for better awareness among health care providers and better diagnostic biomarkers.1 Recent attention has been placed on the presence of a nocturnal sleep onset REM sleep period (nSOREMP, REM sleep occurring ≤ 15 minutes after sleep onset) as a diagnostic biomarker on the nocturnal polysomnogram (PSG). The nSOREMP has high specificity (95% to 99%) for narcolepsy type 1 (NT1) in adults2 and children,3 potentially reducing the need for a daytime Multiple Sleep Latency Test (MSLT). However, the sensitivity of the nSOREMP is relatively poor at 54% to 57%. To enhance the diagnostic capabilities of the nocturnal PSG, clinical providers need additional and more sensitive nocturnal sleep biomarkers of narcolepsy.
Although the nSOREMP highlights the abnormal timing of REM sleep during the night, the high prevalence of nocturnal REM sleep without atonia (nRWA) reported among patients with narcolepsy4–7 provides evidence of abnormal REM sleep physiology. REM sleep is normally characterized by persistent atonia on chin electromyogram (EMG) on the PSG. The American Academy of Sleep Medicine (AASM) defines RWA as elevated chin EMG muscle tone in REM sleep that is manifest as either excessive tonic or phasic activity.8 REM sleep behavior disorder (RBD) is a distinct parasomnia characterized by RWA and complex motor behaviors and/or vocalizations during REM sleep.9 Up to 90% of adult patients with narcolepsy have RWA on nocturnal PSG,7 and these patients have 5 to 10 times higher RWA indices than controls without narcolepsy.4,6 Given the high frequency and severity of RWA among patients with narcolepsy, RWA has the potential to be a more sensitive diagnostic biomarker than the nSOREMP.
In adults, Khalil et al. reported that nRWA indices of phasic activity and tonic activity have respective sensitivities of 90% and 80% and specificities of 100% and 90% for diagnosing narcolepsy compared to controls.4 Similar evaluation of nRWA in the pediatric narcolepsy population has not been conducted, but it is of critical interest as narcolepsy usually begins before age 20 years. Furthermore, nRWA and RBD have been recently characterized among the pediatric population with NT15 but it is unknown whether these findings are also present among pediatric patients with narcolepsy type 2 (NT2) and idiopathic hypersomnia (IH). Determining the nRWA and RBD characteristics in the pediatric population with hypersomnia is important for accurate scoring of REM sleep and for managing potential nighttime disturbances.
In this study, we determined the prevalence and characteristics of nocturnal RWA and RBD in drug-naïve pediatric patients with narcolepsy (NT1 and NT2) and compared findings to populations without narcolepsy (patients with IH and subjective hypersomnia [sHS]). Furthermore, we examined if the nocturnal RWA index (nRWA index) is a valid diagnostic biomarker for pediatric narcolepsy. Given that RWA inversely correlates with hypocretin levels,10 we hypothesized that (1) patients with NT1 have a higher nRWA index than the other groups, (2) the nRWA index predicts the severity of objective daytime sleepiness and number of daytime SOREMPs across all groups, and (3) nRWA index serves as a valid diagnostic biomarker for pediatric narcolepsy. Last, we examined characteristics of RWA and RBD during daytime REM periods captured on MSLT among pediatric patients with narcolepsy to show that REM sleep abnormalities extend beyond nocturnal sleep.
We conducted a retrospective, cross-sectional study by reviewing medical charts and consecutive diagnostic PSGs and MSLTs of patients evaluated for central nervous system hyper-somnias at Boston Children's Hospital (BCH) between January 2015 and March 2017. The study was approved by the BCH Research Ethics Board (IRB-P00006669).
We identified a total of 40 children and adolescents who had PSG/MSLT studies for initial evaluation of hypersomnia and reviewed their medical charts to establish diagnostic classification based on the International Classification of Sleep Disorders, Third Edition.9 Subjects were excluded from this study if they were older than 18 years, taking any REM sleep-suppressing medications (eg, stimulants, SSRIs, SNRIs) or sedating drugs at the time of the study, and/or had an obstructive apnea-hypopnea index of ≥ 2 events/h11 on the nocturnal PSG. Participants who reported subjective excessive daytime sleepiness but had normal PSG and MSLT studies are referred to as sHS in this study.
PSGs and MSLTs were performed at the AASM-accredited BCH pediatric sleep laboratory and manually scored in accordance with the AASM guidelines.12 The studies were recorded using Natus SleepWorks system (Natus Medical Inc., San Carlos, California, United States). During the PSG and MSLT recordings, electroencephalogram (EEG) activity was recorded from an 11-channel EEG (including 2 ground electrodes), sub-mental muscle tone was recorded from 3 separate chin EMG leads, eye movements were recorded from right and left electrooculogram leads, and limb movements were recorded from EMG leads on the right and left tibialis anterior muscles. In addition, the PSG recordings included the following signals: electrocardiogram, thoracoabdomnial excursions (plethymography), nasal airflow (pressure transducer and thermistor), snoring sounds (tracheal microphone), pulse oximetry, end tidal carbon dioxide, and body position. The MSLT consisted of five 20-minute nap opportunities done the day after the PSG study.
We identified all REM sleep periods during the PSGs and MSLTs and reviewed them in detail. As per the scoring manual published by the AASM,12 we scored REM sleep tonic activity (sustained muscle activity) if 50% or more of the stage R sleep epoch (30 seconds) had chin EMG amplitude higher than that demonstrated in NREM sleep epochs. We identified epochs of REM phasic activity by first dividing a 30-second epoch of REM sleep into 10 sequential 3-second mini-epochs and determining if at least 5 (50%) of the mini-epochs contained bursts of transient muscle activity (chin EMG activity lasting 0.1 to 0.5 seconds that was 4 times higher in amplitude than the background EMG activity). If an epoch had phasic and tonic activity meeting both these criteria, we scored the epoch as phasic REM sleep activity. We did not score sub-mental EMG activity that was associated with arousals from sleep, gross body movements induced by the technician or co-sleeping parent, obstructive or central events, or bruxism. To assess reliability of scoring, two authors (SBH and KM) independently scored 43 stage R sleep epochs from 5 patients (3 NT1, 1 NT2, and 1 IH). The interclass correlation between scorers was 0.86 (P = .001).
We calculated nRWA index using the following formula:
(total number of stage R sleep epochs with RWA [tonic and phasic]
on the PSG / total stage R sleep epochs on the PSG) × 100
We repeated this formula to obtain indices for both REM sleep tonic and phasic activity separately.
Two authors (SBH and KM) reviewed the video recordings (synchronized PSG and MSLT recordings) of all REM sleep periods for assessment of RBD to have consensus on RBD identification. RBD was identified if a subject had RWA during the PSG and/or the MSLT and had complex behavioral movements on video, such as talking, smiling, laughing, singing, whistling, shouting, crying, and gesturing.9
We analyzed data using SPSS version 19 (IBM Corp, Armonk, New York, United States).
We first reviewed the distribution of data to determine normality. We analyzed categorical data using the Fisher exact test. We report continuous data with normal distribution as means (and standard deviations) and nonparametric data as median (and minimum, maximum values). For group comparisons, we used analysis of variance and independent t tests or the non-parametric equivalents (Kruskal-Wallis). For significant group differences, we conducted post hoc tests with t tests or Mann-Whitney U tests. To identify the strength of relationships between continuous variables (ie, nocturnal RWA index, mean sleep latency [MSL] on MSLT, number of MSLT SOREMPs), we used Pearson correlation coefficient tests. We performed linear regression to assess (1) main effect of group for nRWA index, (2) main effect of nRWA index on MSL, (3) main effect of nRWA index on number of MSLT SOREMPs and included a priori confounders of age, sex, and body mass index in these analyses. Receiver operating characteristic (ROC) statistics were used to determine the validity of nRWA indices to identify diagnosis of narcolepsy (NT1 and NT2) compared to patients without narcolepsy (IH and sHS). We report significance at P ≤ .05.
Our sample included patients with NT1 (n = 11), NT2 (n = 6), IH (n = 12), and sHS (n = 11). Table 1 demonstrates study cohort demographic and sleep architecture data. Overall, the mean age of subjects was 13.7 years and 47.5% were female. We found no group differences for age, body mass index, ethnicity, or sex (Ps > .05). As expected, human leukocyte antigen (HLA) DQB1*06:02 was more frequently positive in the NT1 group compared to other groups (P = .006) and all 11 patients with NT1 were positive for this haplotype. HLA typing was available only in 5 of the 6 patients with NT2 and was positive in 60% of these NT2 cases. HLA testing was inconsistently performed in the IH and control groups: 25% of tested patients (2/8) with IH and 40% of those with sHS (2/5) were positive for HLA DQB1*06:02. At the time of testing, all patients with NT1 had daily cataplexy triggered by emotion (especially laughter), four patients with NT1 had cataplectic facies, and one had status cataplecticus.
Sleep Architecture and Mean Sleep Latency Findings
In terms of nocturnal sleep architecture (Table 2), the four groups demonstrated comparable total sleep time, sleep efficiency, stage N1 sleep, stage N3 sleep, and stage R sleep percentages (Ps > .05). Only mean periodic limb movement indices, stage N2 sleep percentage, and REM sleep onset latency were significantly different across groups (Ps ≤ .05). Underlying this finding, patients with IH had more mean stage N2 sleep than patients with NT1 (52.1% [7.8] versus 40.9% [6.9] P = .001), as well as compared to patients with NT2 (42.1% [8.5] P = .01). Mean stage N2 sleep time in patients with IH was comparable to patients with sHS (47.2% [7.5] P = .13). As expected, nocturnal REM latency was shorter for patients with NT1 compared to patients with IH (52 [67.4] versus 114.3 [60.7] minutes, P = .02) and sHS (124.2 [43.5] minutes, P = .01) but not compared to patients with NT2 (86.8 [73.2] minutes, P = .26). In contrast, the mean nocturnal REM latency of patients with NT2 did not differ from that of those with IH (P = .37) or sHS (P = .23). Nocturnal SOREMPs were three times more frequent in the patient group with NT1 compared to those with NT2 and never occurred in the groups with IH or sHS. Last, periodic limb movements of sleep index was highest in the group with NT1 compared to groups with IH and sHS and also differed between the groups with NT2 and IH (P ≤ .05).
PSG and MSLT data with post hoc comparisons.
PSG and MSLT data with post hoc comparisons.
Despite having similar Epworth Sleepiness Scale scores (P = .22), the MSL scores on the MSLT revealed group differences (P < .001). Patients with NT1 had the shortest MSL compared to patients with IH (2.9 [2.2] versus 8.5  minutes, P < .001) and patients with sHS (15.4 [2.8] minutes, P < .001). Patients with NT2 had an MSL of 4.4 (2.8) minutes, and this too was shorter than that of the patients with IH and sHS (Ps < .005) but not statistically different from that of the patients with NT1 (P = .29).
RWA and RBD Findings on PSG and MSLT
Using AASM criteria, 80% of all children/adolescents studied demonstrated nRWA in at least one epoch during the PSG, but groups had significantly different median nRWA indices (P = .001, Figure 1). These group differences in nRWA index retained significance after adjusting for age, sex, and body mass index (F = 4.05, P = .02). The median nRWA index in the group with NT1 was 30 times greater than those with IH (9.2% [1.5, 24.1] versus 0.3% [0, 8.2], P < .001) and 15 times greater than those with sHS (0.6% [0, 4.5], P < .001). However, the median nRWA index for the group with NT1 did not differ from that of the group with NT2 (3.9% [0, 30.3], P = .46). Group differences between groups without narcolepsy and those with NT1 were greater with REM sleep tonic activity than with REM sleep phasic activity. Within the group with NT1, median nocturnal REM sleep tonic activity was more frequent than median nocturnal REM sleep phasic activity (6.4% [0.6, 19.4] versus 1.6% [0, 9.3], P = .006). Figure S1 in the supplemental material shows examples of REM sleep phasic and tonic activity in a patient with NT1. Additionally, nRWA seemed equally distributed across the night in the group with NT1 as the median nRWA index was comparable in the first quartile compared to the last quartile of the night (P = .74). In exploratory analyses, the group with NT2 did not show a higher median nRWA index than children with IH (P = .11) and sHS (P = .1).
Group comparison of nRWA index on the PSG.
Median nRWA indices are plotted with bar representing median values and whiskers representing minimum and maximum values. Outliers (> 3 standard deviations from median) are designated as circles. IH = idiopathic hypersomnia, nRWA = nocturnal REM sleep without atonia, NT1 = narcolepsy type 1, NT2 = narcolepsy type 2, PSG = polysomnography, sHS = subjective hypersomnia.
Group comparison of nRWA index on the PSG.
In addition, 82% of patients with NT1 and 50% of patients with NT2 had RWA during daytime REM sleep periods captured on the MSLT. There were too few daytime REM sleep periods in the group with IH and controls for comparison. The median daytime RWA index on the MSLT was higher in the group with NT1 than in those with NT2 (7.1% [0, 60.7] versus 1.1% [0, 21.7]), but results only trended toward significance (P = .08).
In terms of RBD, two patients with NT1 and two patients with NT2 demonstrated vocalization and/or complex motor movements consistent with RBD on the PSG. Movements could be best described as pantomime-like5 for one patient with NT1 (hitting a ball with a bat, writing) and the others were noted to talk, laugh, and gesture during REM sleep. During the MSLT, the same two patients with NT1 and one of the patients with NT2 who had nocturnal RBD demonstrated similar RBD during daytime REM sleep periods. RBD did not occur in the group with IH or controls.
Associations Between nRWA Index and MSLT Results
We further analyzed whether the nRWA index is associated with objective sleepiness scores as indicated by the MSL and number of SOREMPs on the daytime MSLT. Across all groups, we found that nRWA index inversely correlated with MSL (r = −.61, P < .001) and positively correlated with number of daytime SOREMPs (r = .66, P < .001). Even after controlling for age, sex, and body mass index, the nocturnal RWA index remained a robust predictor of MSL (F = 19.5, P < .001) and number of SOREMPs (F = 23.2, P < .001). Within groups, only the group with NT1 demonstrated an association between nRWA index with MSL (r = −.70, P = .02) and daytime SOREMPs (r = .6, P = .05).
Diagnostic Value of Nocturnal RWA Index for Pediatric Narcolepsy
We present the nRWA index ROC statistics—area under the curve (AUC), sensitivity, specificity, and positive predictive value—for diagnosis of pediatric narcolepsy (NT1 and NT2) versus patients without narcolepsy (IH and sHS) in Table 3. Overall, the AUC had very good accuracy (0.87 [0.6], 95% CI = 0.75–0.99, P < .001, Figure 2). To optimize diagnostic sensitivity, we selected the threshold of ≥ 1% of stage R sleep epochs with nRWA. This nRWA index of ≥ 1% yielded a sensitivity of 88.2%, 95% CI = 63.6 to 98.5 and specificity of 60.9%, 95% CI = 38.5 to 80.3. However, to optimize specificity for pediatric narcolepsy, we selected the nRWA index cutoff of ≥ 8% which provides a sensitivity of 52.9%, 95% CI = 27.8 to 77.0 and specificity of 95.7%, 95% CI = 78.1 to 99.9.
ROC statistics for nRWA index using different clinical cutoff values (≥ 1% and ≥ 8%).
ROC statistics for nRWA index using different clinical cutoff values (≥ 1% and ≥ 8%).
ROC of nRWA index for diagnosis of pediatric narcolepsy.
Diagonal segments are produced by ties. ROC curve for diagnosis of pediatric narcolepsy (NT1 and NT2) compared to those without narcolepsy (IH and sHS): AUC = 0.87 (0.6), 95% CI = 0.75–0.99, P < .001. AUC = area under the curve, CI = confidence interval, IH = idiopathic hypersomnia, nRWA = nocturnal REM sleep without atonia, NT1 = narcolepsy type 1, NT2 = narcolepsy type 2, ROC = receiver operating characteristic, sHS = subjective hypersomnia.
ROC of nRWA index for diagnosis of pediatric narcolepsy.
We also analyzed the ROC statistics of nRWA index for patients with NT1 versus patients without narcolepsy and patients with NT2 versus patients without narcolepsy (Table 3). Notably, the ROC statistics for the group with NT1 (AUC = 0.94 [0.4], 95% CI = 0.86 to 1.00, P < .001) were even higher than the combined pediatric narcolepsy group. Using the cutoff value of a total nRWA index ≥ 1%, the sensitivity for NT1 diagnosis was 100%, 95% CI = 71.5 to 100 and specificity was 60.9%, 95% CI = 38.5 to 80.3. The cutoff value of nRWA index ≥ 8% yielded a sensitivity for NT1 diagnosis of 63.6%, 95% CI 30.8 to 89.1 and specificity was 95.7%, 95% CI 78.1 to 99.9. In contrast, the AUC for the group with NT2 trended toward significance (AUC = 0.74 [0.13], 95% CI = 0.49 to 0.99, P = .07). Furthermore, the sensitivity and specificity values for the patients with NT2 have wide 95% CI values, suggesting that this study is underpowered to examine the nRWA biomarker for this group alone.
Last, we looked at the predictive values of nocturnal REM sleep tonic versus phasic activity for pediatric narcolepsy diagnosis (groups with NT1 and NT2 combined). REM sleep tonic index had a higher AUC (0.86 [0.07], 95% CI = 0.73 to 0.99, P < .001) than REM sleep phasic activity AUC (0.79 [0.07], 95% CI = 0.65 to 0.94, P = .002).
This study highlights the unique clinical characteristics of RWA and RBD among pediatric patients with narcolepsy and shows that nRWA is a promising diagnostic biomarker of pediatric narcolepsy. As hypothesized, the median nRWA index is highest in the group with NT1 and significantly greater than in the group with IH and controls. In contrast, the groups with NT1 and NT2 had comparable median nRWA indices. Similarly, both subjects with NT1 and NT2 had occasional RWA during the MSLT and were the only subjects to demonstrate RBD highlighting shared mechanisms that cause abnormal REM sleep physiology. Next, we show that the total nRWA index is a good predictor of objective daytime sleepiness and number of daytime SOREMPs across all groups even after controlling for age, sex, and body mass index. However, these results were largely driven by subjects with NT1, and thus, it is likely that hypocretin deficiency underlies the relationship between nRWA, daytime sleepiness, and dysregulated REM sleep latency. Last, we demonstrate that the nRWA index for diagnosis of pediatric narcolepsy has very good accuracy (AUC = 0.87). This accuracy further improves for NT1 diagnosis (AUC = 0.94). Overall, our study emphasizes the need to look for RWA and RBD during the PSG and MSLT of children and adolescents with possible narcolepsy, and offer the nRWA index as a promising biomarker of narcolepsy based on the nocturnal PSG alone.
Establishing a clinical cutoff for the nRWA as a diagnostic biomarker is a tradeoff between sensitivity and specificity. With this in mind, we present two different nRWA thresholds of ≥ 1% and ≥ 8% that emphasize sensitivity or specificity, respectively. The nRWA threshold of ≥ 1% yields a robust sensitivity of 88.2% for pediatric narcolepsy and 100% for NT1 subtype, but its specificity is poor (60.9%). Notably, the ≥ 1% nRWA index value has the potential to identify more pediatric patients with narcolepsy than the nSOREMP with its sensitivity < 60% i3. Likely, these two biomarkers could be used in conjunction to diagnose pediatric narcolepsy (and especially NT1 subtype) using the nocturnal PSG alone.
The modest specificity of the ≥ 1% nRWA index is due to the high prevalence of nRWA among children in general. In both our study and that of Antelmi et al.,5 most pediatric subjects demonstrated at least one epoch of nRWA meeting the AASM scoring criteria.12 Children show some degree of nRWA due to immaturity of ponto-medullary networks that inhibit muscle tone and activity during REM sleep that normalizes with age.13 Additionally, we found that subjects with NT1 have greater quantitative severity of nRWA compared to participants with sHS and IH. To more accurately define pathological nRWA, the AASM definition of RWA requires revision to reflect normative values in children or the nRWA clinical cutoff must be increased.
The nRWA index cutoff of ≥ 8% provides excellent specificity (95.7%) for the diagnosis of pediatric narcolepsy but has poor sensitivity (52.9%). In addition to its diagnostic value, the nRWA biomarker highlights the unique pathophysiology of narcolepsy, and its specificity is very comparable to that of nSOREMPs (specificity of 97%),23 suggesting shared effects of hypocretin loss. Hypocretin regulates REM sleep occurrence through excitation of REM-OFF neurons,14 and low hypocretin is strongly associated with the early appearance of REM sleep during sleep periods.14,15 Our findings further support prior animal and adult human findings that hypocretin deficiency disrupts the normal atonia of REM sleep.10,16 The resulting RWA can manifest as RBD in patients with narcolepsy during both nighttime and daytime REM sleep as shown in this study and others.5,10 Our study additionally shows that patients with NT2 have more intermediate levels of RWA and RBD, implying that even the presumed partial hypocretin loss of NT217,18 can result in abnormal REM sleep physiology. Of course, given that cataplexy symptoms can develop years after symptom onset19,20 and that we did not measure hypocretin in our study, it is plausible that some of our pediatric patients categorized as NT2 could be early presenters of NT1. Longitudinal studies are needed to determine if the nRWA index in NT2 predicts development of cataplexy.
The nRWA index is highest and has optimal ROC statistics in our group with NT1, supporting the dual roles of hypocretin in motor control10,16: maintaining muscle tone during wake and maintaining atonia during REM sleep. It would follow, then, that the patients with NT1 with the most severe cataplexy would have the highest nRWA indices. Antelmi et al.5 found that more complex RBD correlated with cataplexy severity (as defined by status cataplecticus/cataplectic facies) in their pediatric group with NT1, but it is unclear if the severity of cataplexy correlates with that of RWA. Our study had few patients with such severe cataplexy or RBD, but all of our patients with NT1 had daily cataplexy, another measure of severe cataplexy. As such, we did not have enough variability in our sample to explore the relationship between cataplexy severity and RWA.
Recognition of RWA and RBD in children with narcolepsy may help reduce the delays in diagnosis that are common in pediatric patients with narcolepsy.21 However, this first requires health care providers to order a PSG to investigate the cause of hypersomnia among patients presenting with excessive daytime sleepiness. Thus, this biomarker may be more useful to prevent misdiagnosis of other sleep disorders such as insomnia or obstructive sleep apnea.22 Given that we excluded patients with obstructive sleep apnea (apnea-hypopnea index > 2 events/h), future research validating the nRWA biomarker needs to include children with a broader range of sleep disorders including obstructive sleep apnea to ensure the specificity of our findings.
Our study has additional limitations. First, we have a small sample size in all groups, particularly patients with NT2. We found the nRWA index in the group with NT2 ranged from no RWA to amounts comparable to patients with NT1. This wide variability in nRWA indices and small sample size suggests the study is likely underpowered to compare the patients with NT2 to our groups without narcolepsy, and accordingly, we combined both the groups with NT1 and NT2 in our final analyses. Second, we did not use an automated scoring tool to precisely identify nRWA in our sample as such tools have yet to be validated in pediatric populations. However, once validated, automated scoring of RWA would reduce the additional time burden on sleep study scorers or readers in marking epochs of REM sleep with RWA to calculate required indices. As mentioned earlier, we did not measure cerebrospinal fluid hypocretin levels in our sample nor have HLA test results for all participants. Hypocretin levels predict phasic REM sleep muscle activity in adult narcolepsy patients,10 but it is unknown if hypocretin predicts RWA as more broadly defined by the AASM.8 HLA positivity for the DQB1*06:02 haplotype was present for 40% of our group with sHS, higher than expected in a normative population.23 But because participants with sHS were inconsistently tested, these findings may reflect sample bias. Third, we adhered to the AASM scoring definitions of RWA12 and developed an nRWA index based on combined tonic and phasic RWA findings to make findings imminently useful. Although tonic activity in RWA showed modestly better ROC statistics than phasic nRWA activity, we did not parse respective sensitivities/specificities for these EMG designations as previously reported in the literature.4 Fourth, we were not blinded to the diagnoses of participants when we reviewed stage R sleep epochs. Thus, bias in our RWA scoring is plausible. Last, we did not establish RBD diagnosis using additional patient/parent reported history of dream enactment because such information was inconsistently documented in our subjects' medical records. Thus, it is possible that the RBD rates reported are an underestimation of the true prevalence.
Our results highlight the potential for false-negative testing of children with narcolepsy as their REM sleep could be misclassified as NREM sleep or wake state due to the presence of preserved chin EMG tone. Thus, it is critical for health care providers, sleep technicians, and sleep scorers to be aware of the phenotypic variability of REM sleep in the narcolepsy population.
Importantly, we find that the nocturnal RWA index is a sensitive or specific diagnostic biomarker of pediatric narcolepsy, depending on the clinical cutoff applied. Using the ≥ 1% nRWA index cutoff, this biomarker can complement the presence of an nSOREMP to diagnose narcolepsy on the nocturnal PSG. Conversely, using a ≥ 8% nRWA index as a clinical cutoff offers specificity comparable to the nSOREMP and should strongly raise suspicion for narcolepsy among sleep study scorers and interpreting clinicians. The strong association between nRWA and objective sleepiness and daytime SOREMPs across groups and a greater nRWA index in the group with NT1 compared to all others suggests that nRWA could serve as a proxy for reduced hypocretin levels. Certainly, our findings need to be replicated in larger cohorts of patients, ideally with cerebrospinal fluid hypocretin testing. Nevertheless, the nRWA index represents a clinically valuable biomarker that can aid in narcolepsy diagnosis in children and adolescents when narcolepsy often begins.
Dr. Maski receives consulting fees and research support from Jazz Pharmaceuticals, Inc. The other authors report no conflicts of interest.