Journal Pre-proof Multiple Sleep Latency Test and Polysomnography in Patients with Central Disorders of Hypersomnolence

A multiple sleep latency test (MSLT) with occurrence of sleep onset REM periods (SOREMP) is considered one of the central diagnostic criteria for narcolepsy according to the International Classification of Sleep Disorders, but its sensitivity and specificity have been questioned. This study aims to describe MSLT and polysomnography (PSG) findings, including frequency and distribution of SOREMP during the day, in a large cohort of patients with central disorders of hypersomnolence (CDH). We retrospectively analyzed electrophysiological data from MSLT and PSG in 370 consecutive patients with narcolepsy type 1 (NT1, n = 97), type 2 (NT2, n = 31), idiopathic hypersomnia (IH, n = 48), nonorganic hypersomnia (NOH, n = 116) and insufficient sleep syndrome (ISS, n = 78). NT1 and NT2 patients had a significantly shorter mean Sleep Latency (mSL) and REM-Latency (REML) in MSLT and PSG. SOREMP occurred more frequently in narcoleptic vs. non-narcoleptic patients in MSLT and PSG. Occurrence of 3 or more SOREMP in MSLT and a SOREMP in PSG had a very high specificity and positive predictive value (98%/96% and 100% respectively), however relatively low sensitivity (65% and 45% respectively). NT1 more than NT2 patients have shorter mSL and more frequent SOREMP in MSLT and shorter SL as well as REML during nocturnal PSG. Increasing numbers of SOREMP in MSLT and especially SOREMP during PSG increase specificity on the expense of sensitivity in diagnosing narcolepsy. Therefore, frequency of SOREMP in MSLT naps and PSG can help to discriminate but not clearly separate narcoleptic from non-narcoleptic patients.


Introduction
Excessive daytime sleepiness (EDS) was reported by up to 28% of the general adult US population in a study by Ohayon et al., but the prevalence depends on the exact definition [1]. According to the International Classification of Sleep Disorders (ICSD- 3), EDS for at least 3 months is the main feature of central disorders of hypersomnolence (CDH) [2]. EDS is often underrated by patients, leading to underdiagnosis and undertreatment of sleep-wake disorders [3]. Subjective sleepiness is usually measured by the Epworth Sleepiness Scale (ESS) [4], a simple self-administered questionnaire, whereas the multiple sleep latency test (MSLT) is the most often used objective method [5].
The occurrence of ≥ 2 sleep onset REM periods (SOREMP) during the MSLT including the nocturnal Polysomnography (PSG) is an essential criterion for the diagnosis of narcolepsy type 1 (NT1) as well as narcolepsy type 2 (NT2) according to the ICSD-3 criteria [2]. However, the sensitivity and specificity of SOREMP for narcolepsy have been questioned over the last few years, since several studies demonstrated the presence of ≥ 2 SOREMP in a series of other diseases. In a cohort of healthy adults multiple SOREMP were detected in 13% of males and 6% of females; some of these healthy subjects even had a mSL of ≤ 8 min formally fulfilling electrophysiological criteria for NT1 or NT2 [6]. Seneviratne et al. found that up to 14% of patients with sleep-disordered breathing (SDB) had 2 or more SOREMP in the MSLT with a mean sleep latency (mSL) of 4.5 min in a clinically sleepy patient group [7]. Marti et al. found that 15% of patients with behaviorally induced insufficient sleep syndrome (ISS) had an MSLT suggestive of narcolepsy (mSL ≤ 8 min and at least two SOREMP) [8]. In line with these findings Drakatos et al. found 1 SOREMP in the MSLT of IH (1%), ISS (7.14%) and periodic limb movement disorder (4.7%) patients [9]. However, in the same study sleep stage sequence analysis discovered, that SOREMP arising from stage N1 has a high sensitivity for NT1 diagnosis, since IH, ISS, PLMD and NT2 patients had mostly SOREMP arising from stage N2 [9]. The same study demonstrated that only narcoleptic patients present with SOREMP in the fourth (afternoon) nap of MSLT [9]. Occurrence of SOREMP during nocturnal PSG has been found highly specific for NT1 [10].
The aim of our present study was to test the current hypothesis of a specific temporal distribution pattern of SOREMP in MSLT and PSG in a large cohort of CDH patients comparing five diagnostic groups, including nonorganic hypersomnia patients.

Patients and Methods
J o u r n a l P r e -p r o o f  [11,12]. REDCap® is a secure, web-based software platform designed to support data capture for research studies, providing 1) an intuitive interface for validated data capture; 2) audit trails for tracking data manipulation and export procedures; 3) automated export procedures for seamless data downloads to common statistical packages; and 4) procedures for data integration and interoperability with external sources.
Each original data set contained a final diagnosis (according to the third or second edition or earlier versions of the international classification of sleep disorders from the American Academy of Sleep Medicine (AASM)). After data transfer all records were reviewed and confirmed manually by two sleep specialists [2]. In cases of a documented change in diagnosis during a follow-up consultation (e.g., initially NT1 after electrophysiological workup, however changed to ISS after hypocretin measurement and due to clinical judgment during follow-up consultations), the final diagnosis was used for the analysis, irrespectively of the initial interpretation of diagnostic workup. The term nonorganic hypersomnia is used as an alternative name of hypersomnia associated with a psychiatric disorder in ICSD 3 and is also described in ICD-10 under the Code F51.1. Excessive daytime sleep or sleep attacks not accounted for by inadequate sleep and/or prolonged transition to the fully aroused state upon awakening (sleep drunkenness) are criteria for NOH under the ICD-10 [13]. A description of NOH classification in our cohort has been published elsewhere [14].
Exclusion criteria included incomplete clinical information, incomplete sleep-wake test datasets, as well as the presence of other sleep comorbidities with a potential effect on sleep architecture or unclear CDH diagnosis.
We analyzed electrophysiological data from PSG and MSLT, as well as HLA DQB1*0602 status and CSF hypocretin-1 concentration data, if available, from five selected diagnostic groups with CDH, namely NT1, NT2, IH, non-organic hypersomnia (NOH) and ISS. The sleep latency (SL) and REM-latency J o u r n a l P r e -p r o o f (REML) from PSG and the individual mean sleep latency (mSL) and individual mean REM-latency (mREML) from MSLT were compared between groups by comparing their median values. Proportions of patients with SOREMP occurring on PSG or MSLT naps were compared between groups. However, for SOREMP analysis only descriptive statistics were used since groups were -as expected -very heterogeneous. If no REM appeared in a single MSLT or PSG trial, subjects were not considered for REM-Latency or SOREMP analysis. CSF hypocretin-1 levels <20 pg/mL were referred to as undetectable hypocretin-1, levels between 20 and 110 pg/mL were considered as low, whereas levels >110 were considered as normal.
Statistical analyses and graphs were performed by Stata/MP 16.0. Data were described by medians (interquartile range (IQR). To compare continuous measures among the five groups, we used the Kruskal Wallis test. We used χ 2 test to compare categorical variables. Significance was accepted at p < 0.05.

Sensitivity, specificity and positive predictive value of SOREMP
As expected by definition, narcoleptic patients had significantly more MSLT and PSG SOREMP compared to non-narcoleptic patients (p < 0.001, table 3). Using ICSD3 criteria [2] of two or more SOREMP in either two of the MSLT naps or one SOREMP in PSG and at least one SOREMP in MSLT naps had a high sensitivity of 90% and specificity of 95% with a positive predictive value of 91% discriminating narcoleptic vs. non-narcoleptic patients. Testing criteria of SOREMP in 3 or more MSLT naps had a sensitivity of 65%, specificity of 98% and positive predictive value (PPV) of 96% in discriminating narcoleptic vs. non-narcoleptic patients. SOREMP in PSG had a lower sensitivity of 45%, higher specificity of 100% and PPV of 100%. Combined, 3 or more SOREMP in MSLT and SOREMP in PSG had an even lower sensitivity of 38%, specificity and PPV of 100%.

Discussion
The present study investigated diagnostic parameters of MSLT and PSG in a large cohort of 370 CDH patients, comparing five different diagnostic groups.

J o u r n a l P r e -p r o o f
The main finding of our study is, as expected, significantly shorter mSL and REML in MSLT and PSG in NT1 compared to non-narcoleptic patients. Furthermore, distinguishing features for narcolepsy vs non-narcoleptic sleepy patients were up to three or more SOREMP in the course of the 5 MSLT naps (75% NT1 vs. 2% NOH, 2% ISS, 0% IH) and SOREMP in the PSG (53% NT1, 0% in IH, NOH, ISS). Applying AASM ICSD3 criteria [2] of 2 or more SOREMP in MSLT (including PSG) had a high sensitivity (90%), specificity (95%) and positive predictive value (91%) discriminating narcoleptic vs. non-narcoleptic patients. These findings are however most probably strongly biased by "circular reasoning" in such a retrospective study design. Increasing the limit to 3 or more SOREMP resulted in a higher specificity and positive predictive value (98%/96% and 100% respectively, however in a relatively low sensitivity (65% and 45% respectively) in discriminating narcoleptic vs. non-narcoleptic patients. A reliable differentiation between NT1 and NT2 with these parameter was however not possible, as expected, due to the similar electrophysiological diagnostic criteria [2].
These findings are perfectly in line with literature: Aldrich et al. reported a sensitivity and specificity of 3 or more SOREMP in MSLT of 48% and 98%, and SOREMP in PSG of 29% and 98% [15]. Andlauer et al. reported a sensitivity and specificity of 51% and 99% for PSG SOREMP in patients with NT1 vs. population based controls [10].
In a study by Drakatos et al. SOREMP occurred in NT1 patients in 63% of all MSLT naps, in NT2 in 60%, in IH in only 1% and in ISS patients in 7% [9]. Furthermore, a relatively homogeneous temporal distribution of SOREMP occurrence during the course of an MSLT day was shown for NT1 and NT2 [9]. In line with these results, we found proportions of narcolepsy, especially NT1 patients with SOREMP occurring in individual MSLT naps relatively evenly distributed throughout the course of the MSLT examination day (78% in MSLT nap 1 decreasing to 65% in MSLT nap 5). However, a statistical comparison of this temporal distribution to occurrence of SOREMP in non-narcoleptic groups was not rational, due to the very small total numbers of detected SOREMP in the non-narcoleptic groups.
Therefore, the hypothesis of a narcolepsy specific pattern of temporal distribution of SOREMP in MSLT can neither be confirmed nor rejected, due to a too small non-narcoleptic control group experiencing SOREMP in MSLT.
We found valuable MSLT and PSG criteria distinguishing narcoleptic from non-narcoleptic CDH patients. However, none of the investigated parameter allowed to distinguish NT1 from NT2 and the big overlap of all parameters such as sleep latency between the diagnostic groups in our cohort, supports literature criticism on the official electrophysiological gold standard criteria in diagnosing CDH patients [16]. Our data underline once again the importance of detailed history taking and clinical judgement in the diagnostic workup of CDH patients. Furthermore, a revision of current J o u r n a l P r e -p r o o f concept of CDH with better clinical, electrophysiological or biological markers would be needed [17,18].

Multiple studies on PSG and MSLT characteristics of patients with EDS have been published so far.
However, studies with very large cohorts of up to 2500 subjects [19][20][21] either describe overall population based results without going into detail of CDH diagnostic groups or are not especially focused on narcolepsy and narcolepsy related disorders. Other studies, comparing CDH patients and other sleepy patient groups, mostly have limited numbers of study population between 20 to 90 patients per diagnostic group [8,9,22], often comparing only 2 to 3 CDH diagnostic groups. Studies on NOH patients are rare [14]. Therefore, the main strength of this study is the relatively large number of included patients with mostly full data sets on PSG and MSLT data, including NOH patients. The major limitations of this study are the retrospective study design, and the lack of interpretation of other available data such as subjective measures of sleepiness (e.g. Epworth Sleepiness Scale), disease onset, treatment status and the lack of follow up data.

Conclusion
In conclusion, NT1 more than NT2 patients have shorter mSL and more frequent SOREMP in MSLT and shorter SL as well as REML during nocturnal PSG. The more SOREMP occur in MSLT and PSG the higher is the probability of discriminating narcoleptic from non-narcoleptic CDH patients, but increasing this limit resulting in greater specificity is at the expense of lower sensitivity. SOREMP occurred evenly distributed throughout a day of 5 MSLT naps in narcolepsy, including the PSG, especially in NT1 patients. Therefore, frequency and temporal distribution of SOREMP in MSLT naps and PSG can help to discriminate but not clearly separate narcoleptic from non-narcoleptic patients.
It remains essentially important to carefully phenotype patients clinically as well as electrophysiologically, besides research approaches to find novel biomarkers as well as investigation on pathophysiology of the different CDHs.

Disclosure statement
This was an industry supported study (financial Support by Jazz Pharmaceuticals). The authors have indicated no financial conflicts of interest.  J o u r n a l P r e -p r o o f      J o u r n a l P r e -p r o o f