J Neurol DOI 10.1007/s00415-016-8285-y
ORIGINAL COMMUNICATION
National validation and proposed revision of REM sleep behavior disorder screening questionnaire (RBDSQ) Sara Marelli1,2 • Paola M. V. Rancoita3 • Federica Giarrusso2 • Andrea Galbiati1,2 Marco Zucconi1 • Alessandro Oldani1 • Clelia Di Serio3 • Luigi Ferini-Strambi1,2
•
Received: 13 July 2016 / Revised: 6 September 2016 / Accepted: 7 September 2016 Ó Springer-Verlag Berlin Heidelberg 2016
Abstract We validated the Italian version of the rapid eye movement sleep behavior disorder (RBD) screening questionnaire (RBDSQ) and calculated its cut-off value for discriminating RBD group from other sleep disorders and healthy controls (HC). 380 patients with sleep disorders and 101 HC were enrolled. RBDSQ achieved an acceptable Cronbach’s a value of 0.787 and item 10 was the only one with a very low item-total biserial correlation (0.141). At ROC analysis, we obtained an AUC of 0.888, denoting a good performance of the RBDSQ total score for predicting the RBD status. The optimal cut-off value was 8 and it achieved good values of both sensitivity and specificity (0.842 and 0.780, respectively). Due to the poor performance of item 10 in our sample, we analyzed the RBDSQ without this item (called ‘‘revised RBDSQ’’). We obtained a good Cronbach’s a of 0.802. When evaluating the performance of the revised score in predicting the RBD status, we obtained an increased value of AUC (0.899). The optimal cut-off value was still 8 (sensitivity = 0.829; specificity = 0.820). The Italian version of RBDSQ is a
Electronic supplementary material The online version of this article (doi:10.1007/s00415-016-8285-y) contains supplementary material, which is available to authorized users. & Luigi Ferini-Strambi
[email protected] 1
Department of Clinical Neurosciences, Neurology, Sleep Disorders Center, San Raffaele Scientific Institute, Universita` Vita-Salute San Raffaele, Via Stamira D’Ancona, 20, 20127 Milan, Italy
2
Faculty of Psychology, Vita-Salute San Raffaele University, Via Olgettina, 58, 20132 Milan, Italy
3
University Centre of Statistics in the Biomedical Sciences (CUSSB), Vita-Salute San Raffaele University, Milan, Italy
sensitive tool for the identification of RBD patients. An improvement of the instrument could be obtained by removing item 10 and define a higher cut-off value of 8. The ‘‘revised RBDSQ’’ represents a reliable screening questionnaire for primary care physicians and neurologists and its employment may facilitate the choice of subjects that should undergo a PSG that confirms the diagnosis of RBD, thus avoiding polysomnographic exams when not needed. Keywords REM sleep behavior disorder RBDSQ Questionnaire Parasomnia
Introduction Rapid eye movement (REM) sleep behavior disorder (RBD) is a REM sleep parasomnia associated with the loss of muscle atonia in REM sleep stage and characterized by complex and vigorous behaviors with frequent episodes of nocturnal restlessness in conjunction with dreams typically with an aggressive content [1], in which the subject is being attacked, or chased by animals or other humans being [2]. The most frequently enacting behavior are screams, kicks, punches, crying and in some cases even falls out of bed [1]. The disorder mainly affects males [3] and the onset is typically fixed between 40 and 70 years (15–80 range) [2]. Prevalence of RBD is unknown. In United Kingdom, a telephone survey assessing the prevalence of violent behavior during sleep, suggested a prevalence of RBD of about 0.5 % [4]. Another study performed in Hong Kong, found a history of sleep-related injury in 0.8 % of the population, and an estimated RBD prevalence of 0.38 % [5]. A recent study examined the prevalence and the
123
J Neurol
clinical characteristics of RBD in the Korean elderly population, diagnosed by video-polysomnography (PSG). Authors found an RBD prevalence of 2.01 %, which was considerably higher than that found in the two previous studies [6]. PSG represents the gold standard tool in establishing the diagnosis of RBD, revealing loss of REM-related muscle atonia and increased electromyographic tone, associated with dream enactment behavior during REM sleep [7]. RBD is frequently associated with other sleep disorders including narcolepsy, sleepwalking and restless legs syndrome/Willis-Ekbom Disease (RLS/WED) [1, 8]. RBD can occur in idiopathic form or be secondary to neurological disorders (i.e. neurodegenerative diseases): its most common association is with a-synucleinopathies, such as Parkinson’s disease (PD) [9, 10], dementia with Lewy bodies (LBD) [11] and multiple system atrophy (MSA) [12, 13]. However, it has been commonly noted that RBD often precedes the onset of first motor symptoms of neurodegenerative disorders, as PD, and the duration of the followup of idiopathic cases seems to be crucial. Hence the importance of a timely assessment of the disorder is essential. Thus, a suitable and easy to administer screening tool would be useful both in the clinical practice and in research. For the screening of RBD different scales and questionnaires have been developed, in particular the REM sleep behavior disorder screening questionnaire (RBDSQ) has been proposed and validated [14]. The aim of this study is to validate the RBDSQ of Stiasny-Kolster and colleagues [14] in Italian language.
Subjects and methods 380 patients with sleep disorders [Chronic Primary Insomnia, Obstructive Sleep Apnea Syndrome (OSA), REM sleep Behavior Disorder (RBD), Restless Legs Syndrome (RLS/ WED), NREM Parasomnias/Epilepsy and Narcolepsy/ Hypersomnia] and 101 healthy controls subjects (HC), consecutively assessed at the Neurology—Sleep Disorder Center of the San Raffaele Hospital in Milan, Italy were enrolled for this study, from 2013 to 2015. Patients with other sleep disorders were not affected by RBD. Subjects with RBD were not affected by other sleep disorder. All participants were monolingual native Italian speakers, had normal or corrected-to-normal visual acuity and a mini mental state examination (MMSE) score [24 to exclude subjects with cognitive deterioration. All participants provided written informed consent to the experimental procedure, which was previously approved by the local ethical committee. Only patients with sleep disorders underwent a nocturnal PSG recordings.
123
HC were examined through a clinical assessment to exclude any sleep disorder or other clinical conditions. Both patients and HC complete the Italian version of RBDSQ [14] (Supplemental material). The RBDSQ RBDSQ is a questionnaire composed by ten items investigating the patient’s sleep behavior with short questions and a dichotomic response (‘‘yes’’, ‘‘no’’). Dreams’ frequency, content and their relation to nocturnal movements and behavior are assessed by questions 1–4. Item five evaluates the presence of self-injuries and injuries to the bed partner. Item six is divided in four sub-items exploring, respectively: nocturnal vocalization, limb movements, complex movements, and objects that fall down around the bed (these sub-items will be called items 6.1, 6.2, 6.3 and 6.4 in following text and tables). Items seven and eight pertain to nocturnal awakenings. Disturbed sleep in general and the presence of any neurological disorder are investigated by items nine and ten, respectively. RBDSQ total score has a maximum of 13 points. Translation The RBDSQ has been translated in Italian by an experienced sleep researcher. The translation has been evaluated independently by two physicians expert in sleep disorders and the approved version has been translated back into English by one sleep specialist, fluent both in English and Italian. The final back-translated version has been compared with the original RBDSQ by a physician who was fluent in both languages and was blind to the purpose of the study. This procedure lead to the Italian version of the RBDSQ (used in the clinical practice) presented in the ‘‘Supplemental material’’. Statistical analysis Comparisons of median values between groups were performed using the Mann–Whitney test and, when needed, by accounting for multiple comparisons with the Bonferroni’s correction. Cronbach’s a was computed for evaluating the internal consistency of the RBDSQ. For assessing the reliability of each single item, the following indices were computed: the Cronbach’s a when the corresponding item was deleted, the item-total biserial correlation, and its sensitivity and specificity in predicting the RBD status. As suggested by Nunnally and Bernstein [15], values of the item-total biserial correlation greater than 0.3 were considered acceptable. Moreover, for a better understanding of the performance of the items, we computed the percentage of
J Neurol
yes-answers in the main subgroups of the sample: RBD, healthy controls, Insomnia, OSA and RLS/WED. The receiver operating characteristics (ROC) curve analysis was used to investigate the performance of the RBDSQ total score in predicting the RBD status. The diagnostic performance of the RBDSQ total score was evaluated by the area under the curve (AUC). The optimal cut-off of the RBDSQ total score was derived using the standard method, consisting in choosing that value corresponding to the point on the ROC curve nearest to the upper left corner of the ROC graph. p values less than 0.05 were considered significant. All statistical analyses were performed using R 3.2.0 (http:// www.R-project.org/).
Results Description of the sample Seventy-six RBD patients, with a mean age of 71.86 years (standard deviation, SD, 7.94 and range 48–91) and including 84 % of male patients, were recruited in the study. The non-RBD group consisted of 405 subjects (54 % males), having a lower mean age than the RBD group (51.03 ± 16.72, range 8–89). Namely, the non-RBD group consisted of 101 HC and 304 patients with other sleep disorders: 123 with Chronic Primary Insomnia, 80 with OSA, 61 with RLS/WED, 20 with NREM Parasomnias/ Epilepsy and 20 with Narcolepsy/Hypersomnia. The descriptive statistics of the age and sex for all the subgroups are reported in Table 1. In all this subgroups the mean age was lower than the one of the RBD patients (p \ 0.001 for all subgroups) and almost all of them showed a balance between male and female subjects. Analysis of the RBDSQ total score and computation of its optimal cut-off The two groups of RBD and non-RBD subjects showed a significant different RBDSQ total score (p value \0.001). The mean RBDSQ total score of the RBD patients was Table 1 Descriptive statistics of the demographic characteristics of the subjects in all the subgroups
9.776 (SD = 2.266) and it was lower than the mean score of the entire group of non-RBD subjects, which was 5.168 (SD = 2.846). In particular, the mean RBDSQ total scores obtained in the different sub-groups of the non-RBD subjects were: 3.891 (SD = 2.603) for healthy controls, 5.049 (SD = 2.627) for Chronic Primary Insomnia, 6.279 (SD = 2.498) for RLS/WED, 5.200 (SD = 2.781) for OSA, 7.550 (SD = 3.502) for NREM Parasomnias/Epilepsy and 6.450 (SD = 2.781) for Narcolepsy/Hypersomnia. All the sub-groups had a significantly different score in comparison to the RBD patients (p value \0.001 for all groups except for OSA patients, whose p value = 0.040). The RBDSQ achieved an acceptable Cronbach’s a value of 0.787 in our sample. Items eight and ten were the only ones showing an increase of the Cronbach’s a when we removed the item from the questionnaire (see Table 2). Moreover, item ten was the only one with an item-total biserial correlation less than 0.3 (0.141). It had also a very low sensitivity (0.316), with the highest percentage of yesanswers in the RLS/WED group (73.77 %, while 31.58 % in the RBD patients and see Table 3 for the percentages in all the other subgroups). Instead, item eight showed a very low specificity (0.348) with high and similar frequencies of yesanswers in many sub-groups (see Table 3). Also the items 1, 4, 7 and 9 achieved a low specificity (0.284, 0.479, 0.472 and 0.314, respectively), because there was at least a nonRBD sub-group in which they got a high frequency of yesanswers. In particular, item one achieved a high percentage of yes-answers in all the sub-groups (see Table 3). Both the items seven and nine had the highest percentage of yesanswers in the RLS/WED group (68.85 and 88.52 %, respectively). For the item four both the RBD and the RLS/ WED patients showed a similar and very high percentage of yes-answers (81.58 and 80.33 %, respectively). Finally, the low sensitivity of item 6.4 (0.513) was due to the relatively low percentage of yes-answers (51.32 %) in the RBD group, although it was the group with the highest percentage. We evaluated the RBDSQ total score for predicting the RBD status with a ROC analysis (see Fig. 1). We obtained an AUC of 0.888, denoting a good performance. The optimal cut-off value was 8 and it achieved good values of both sensitivity and specificity (0.842 and 0.780, respectively).
Subgroup
n
Age [mean ± SD (range)]
Male [n (%)]
RBD Healthy controls
76 101
71.86 ± 7.94 (range 48–91) 45.81 ± 14.82 (range 24–79)
64 (84 %) 55 (54 %)
Insomnia
123
52.12 ± 17.76 (range 16–89)
54 (44 %)
OSA
80
55.46 ± 13.37 (range 8–78)
61 (76 %)
RLS/WED
61
59.79 ± 15.41 (range 20–84)
27 (44 %)
NREM Parasomnias/Epilepsy
20
37.41 ± 14.33 (range 20–76)
11 (55 %)
Narcolepsy/Hypersomnia
20
39.88 ± 16.15 (range 16–73)
9 (45 %)
123
J Neurol Table 2 Analysis of the RBDSQ items Item
RBDSQ
Revised RBDSQ
Sensitivity
Specificity
Cronbach’s a
Item-total biserial correlation
Cronbach’s a
Item-total biserial correlation
1
0.777
0.533
0.792
0.561
0.908
0.284
2
0.763
0.624
0.777
0.648
0.855
0.640
3 4
0.765 0.775
0.620 0.504
0.779 0.792
0.639 0.499
0.908 0.816
0.758 0.479
5
0.764
0.659
0.780
0.665
0.790
0.832
6.1
0.764
0.608
0.779
0.629
0.921
0.664
6.2
0.759
0.682
0.775
0.683
0.842
0.751
6.3
0.762
0.694
0.777
0.700
0.763
0.835
6.4
0.775
0.541
0.793
0.508
0.513
0.832
7
0.775
0.484
0.794
0.464
0.605
0.472
8
0.789
0.318
0.805
0.321
0.724
0.348
9
0.783
0.418
0.801
0.394
0.816
0.314
10
0.802
0.141
–
–
0.316
0.625
For each item, the Cronbach’s a was obtained by eliminating that item from the total score Table 3 Percentages of yes-answers for each RBDSQ items in the main subgroups of the sample Item
RBD (n = 76), %
Healthy controls (n = 101), %
Insomnia (n = 123), %
OSA (n = 80), %
RLS/WED (n = 61), %
1
90.79
74.26
69.92
65.00
73.77
2
85.53
37.62
37.40
32.50
21.31
3
90.79
20.79
20.33
26.25
21.31
4
81.58
38.61
42.28
55.00
80.33
5
78.95
8.91
10.57
23.75
27.87
6.1
92.11
30.69
27.64
36.25
34.43
6.2 6.3
84.21 76.32
16.83 8.91
21.95 12.20
26.25 16.25
27.87 18.03
6.4
51.32
11.88
13.82
15.00
24.59
7
60.53
43.56
46.34
56.25
68.85
8
72.37
62.38
70.73
56.25
67.21
9
81.58
33.66
83.74
77.50
88.52
10
31.58
0.99
47.97
33.75
73.77
Instead, the cut-off value of 5 used in the original work of Stiasny-Kolster and colleagues [14] corresponded to a very high sensitivity (0.973), but also to a low specificity (0.459), similarly to what reported by the authors.
When evaluating the performance of the revised score in predicting the RBD status, we obtained an increased value of AUC (0.899). The optimal cut-off value was still 8, corresponding to a sensitivity of 0.829 and to an increased specificity of 0.820 (see Fig. 1).
Analysis of the revised RBDSQ total score Since in the item analysis we found that item ten had a very poor performance in our sample, we analyzed the RBDSQ without this item (called ‘‘revised RBDSQ’’). We obtained a good Cronbach’s a of 0.802 and only the item eight showed a slight increase of this index (0.805) when eliminating the item from the questionnaire. No item had an item-total biserial correlation less than 0.30.
123
Discussion The aim of this study was to evaluate the psychometric properties and the diagnostic value of the Italian version of the RBDSQ [14]. Regarding the internal consistency of the RBDSQ, Cronbach’s alpha (0.787) showed a satisfactory overall
J Neurol
Fig. 1 ROC curves of the RBDSQ total score (on the left-hand side) and of the revised RBDSQ total score, obtained by eliminating item 10 (on the right-hand side)
reliability [16]. Biserial item-total correlation was good for all items ([0.3) except for the item 10 (0.141), that also presents the lowest sensitivity. Focusing more specifically on the analysis of the sensitivity and the specificity of each item we can make further remarks. In particular, low specificity was found for items 1, 4, 7, 8 and 9 as reported by Stiasny-Kolster and colleagues [14]. All these items showed a low specificity despite a high rate of positive responses in RBD. Items four and seven investigate sleep-related movements that represent the core features of other sleep disorders such as RLS/WED. In patients with RBD, sleep movements are linked with dream enacting behavior [2], while in RLS/WED patients positive responses are due to sensory and motor symptoms like feelings of numbness, tingling and leg cramps that result in an irresistible urge to move legs, or to the presence of PLM (periodic limb movements) that occur in 80 % of patients affected by this condition [17]. Conversely item nine assesses disturbed sleep that is common to all pathologic conditions in our sample that reported a positive response rate [75 %. Neither the assessment of dream recall and content seems to be adequate to discriminate between pathological and healthy conditions, as highlighted by items one and eight. The highest specificity was found in item 5, 6.3 and 6.4 in analogy with the results obtained by Stiasny-Kolster and colleagues [14]. These results confirm that these features are prominent in RBD.
Regarding the sensitivity of the items, we can affirm that they have a good reliability to correctly identify RBD patients and their content adequately reflects the main characteristics of the disorder. The highest value of RBDSQ total score was found in the RBD group (9.776 ± 2.266) with significantly lower scores in the HC (3.891 ± 2.603). Looking at the averages and standard deviations for the other sub-groups, we reported relatively high scores that anyway are still significantly lower from those of RBD. ROC curve analysis show that the area under the curve is equal to 0.8880, suggesting that RBDSQ has good predictive properties. We found the total score ‘‘8’’ as an optimal cut-off value with higher degree of sensitivity and specificity (84 and 78 %, respectively). In their original study, Stiasny-Kolster and colleagues [14] proposed a different cut-off value ‘‘5’’, corresponding to a sensitivity of 96 % and a specificity of 56 %. Our study confirms that a similar cut-off presents a high sensitivity (97 %) but also a very low specificity (46 %) thus being not suitable to discriminate RBD clinical characteristics from those of other sleep disorders. According to our results, by eliminating item 10 from the analysis (as this item had a very poor performance), we got a total score with a higher predictive performance since the area under the curve was equal to 0.899. The optimal value of the cut-off remained equal to 8 with a sensitivity of 83 % and a greater specificity (82 %) compared with the previous analysis.
123
J Neurol
One consideration about the decision to eliminate the item ten is the following. Although from a statistical point of view, its exclusion entails an increase in the value of Cronbach’s a (0.802) compared to the value obtained when item 10 was included (0.787), from a clinical point of view, it may provide important information about the influence of concurrent diseases (head trauma, stroke, epilepsy, parkinsonism, RLS/WED, narcolepsy, depression and other inflammatory brain diseases) on RBD. Thus, it might be useful to keep it in the questionnaire but without considering it for the computation of the score. In line with the results obtained we can assume that the Italian version of RBDSQ is found to be a sensitive tool for the identification of RBD patients. An improvement of the instrument could be obtained by removing item ten and defining a higher cut-off value ‘‘8’’ rather than that which was identified by Stiasny-Kolster and colleagues [14]. In conclusion the ‘‘revised RBDSQ’’ represents a reliable screening questionnaire for primary care physicians and neurologists and its employment may facilitate the choice of subjects that should undergo a PSG to confirm the diagnosis of RBD, thus avoiding polysomnographic exams when not needed. Compliance with ethical standards Ethical standards All participants provided written informed consent to the experimental procedure, which was previously approved by the local ethical committee. Conflicts of interest The authors declare that they have no conflict of interest.
3.
4. 5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
References 1. Mahowald MW, Schenck CH (2011) REM sleep parasomnias. In: Kryger MH, Roth T, Dement C (eds) Principles and practice of sleep medicine, 5th edn. Elsevier-Saunders, Philadelphia, pp 1083–1097 2. Boeve BF (2010) REM sleep behavior disorder. Updated review of the core features, the REM sleep behavior disorder-
123
15. 16. 17.
neurodegenerative disease association, evolving concepts, controversies, and future directions. Ann N Y Acad Sci 1184:15–54 Olson EJ, Boeve BF, Silber MH (2000) Rapid eye movement sleep behaviour disorder: demographic, clinical and laboratory findings in 93 cases. Brain 123:331–339 Ohayon MM, Caulet M, Priest RG (1997) Violent behavior during sleep. J Clin Psychiatry 58:369–376 Chiu HF, Wing YK, Lam LC, Li SW, Lum CM, Leung T, Ho CK (2000) Sleep-related injury in the elderly—an epidemiological study in Hong Kong. Sleep 23:513–517 Kang SH, Yoon IY, Lee SD, Han JW, Kim TH, Kim KW (2013) REM sleep behavior disorder in the Korean elderly population: prevalence and clinical characteristics. Sleep 36:1147–1152 American Academy of Sleep Medicine (2014) The international classification of sleep disorders, 3rd edn. American Academy of Sleep Medicine, Darien Schenck CH, Boyd JL, Mahowald MW (1997) A parasomnia overlap disorder involving sleepwalking, sleep terrors, and REM sleep behavior disorder in 33 polysomnographically confirmed cases. Sleep 20:972–981 Comella CL, Tanner CM, Ristanovic RK (1993) Polysomnographic sleep measures in Parkinson’s disease patients with treatment-induced hallucinations. Ann Neurol. 34:710–714 Adler CH, Hentz JG, Shill HA, Sabbagh MN, Driver-Dunckley E, Evidente VG, Jacobson SA, Beach TG, Boeve B, Caviness JN (2011) Probable RBD is increased in Parkinson’s disease but not in essential tremor or restless legs syndrome. Parkinsonism Relat Disord 17:456–458 Boeve BF, Silber MH, Ferman TJ, Kokmen E, Smith GE, Ivnik RJ, Parisi JE, Olson EJ, Petersen RC (1998) REM sleep behavior disorder and degenerative dementia: an association likely reflecting Lewy body disease. Neurology 51:363–370 Plazzi G, Corsini R, Provini F, Pierangeli G, Martinelli P, Montagna P, Lugaresi E, Cortelli P (1997) REM sleep behavior disorders in multiple system atrophy. Neurology 48:1094–1097 Tachibana N, Kimura K, Kitajima K, Shinde A, Kimura J, Shibasaki H (1997) REM sleep motor dysfunction in multiple system atrophy: with special emphasis on sleep talk as its early clinical manifestation. J Neurol Neurosurg Psychiatry 63:678–681 Stiasny-Kolster K, Mayer G, Scha¨fer S, Mo¨ller JC, HeinzelGutenbrunner M, Oertel WH (2007) The REM sleep behavior disorder screening questionnaire—a new diagnostic instrument. Mov Disord 22:2386–2393 Nunnally JC, Bernstein IH (1994) Psychometric theory, 3rd edn. McGraw-Hill, New York Bland JM, Altman DG (1997) Cronbach’s alpha. BMJ 314:572 Walters AS (1995) Toward a better definition of the restless legs syndrome. The International Restless Legs Syndrome Study Group. Mov Disord 10:634–642