A new design of a polysomnography-based multi-center treatment study for the restless legs syndrome

A new design of a polysomnography-based multi-center treatment study for the restless legs syndrome
of 8
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
  A new design of a polysomnography-based multi-center treatment studyfor the restless legs syndrome T. Penzel a, *, U. Brandenburg a , J.H. Peter a , R. Otto a , H.P. Hundemer b , A. Lledo b , T.C. Wetter c ,C. Trenkwalder d , for the PEARLS Study Group a Schlafmedizinisches Labor, Klinik fu¨ r Innere Medizin – Pneumologie, Klinikum der Philipps-Universita¨ t Marburg, Baldingerstrasse 1, D-35033 Marburg, Germany b Eli Lilly and Company, Indianapolis, USA c  Max Planck Institute of Psychiatry, Munich, Germany d  Abt. Klinische Neurophysiologie, Zentrum Neurologische Medizin, Georg-August-Universita¨ t, Go¨ ttingen, Germany Accepted 7 February 2002 Abstract Periodic limb movements (PLM) cause sleep disorders and daytime symptoms and are frequently associated with restless legs syndrome(RLS). Treatment of RLS with increased PLM during sleep (PLMS) has been evaluated in studies limited in size, methodology and studylength. This long-term, placebo-controlled, multi-center, study with polysomnography (PSG) recordings has been designed in order to assessefficacyandsafetyparametersofpergolidetreatmentinRLS.Thisnovelapproachforastudywascreatedtoassureconsistentlyhighqualityof sleeprecordingandanalysis.Usingdefinedcriteria,21sleepcenterswereapprovedforthestudyafterapilotphase.Seventeencenterswith16different PSG systems randomized 100 patients. Digital sleep recordings from 4 visits (baseline, 6 weeks, 6 months, 1 year) were submitted toone central evaluation center following previouslydefined standardized operating procedures. Visual scoring of allrecordings was performedby one independent scorer. Reliability of scoring was evaluated for 20 randomly selected baseline recordings. The mean epoch by epochagreement for sleep stages was 88% (range 81–96%), mean arousal re-scoring differed by 0.5 (range: 2 16 to 20), and mean PLM index re-scoring differed by 0.1 (range: 2 1.5 to 2.1). Using one scorer with a demonstrated high reliability in PSG scoring for all sleep recordings wasvery effective in terms of study cost, study duration, and data quality. q 2002 Elsevier Science Ireland Ltd. All rights reserved. Keywords : Restless legs syndrome; Periodic limb movement; Polysomnography; Reliability; Variability; Multi-center treatment study 1. Introduction Patients with the clinical diagnosis of restless legssyndrome (RLS) in more than 80% have an increasedindex of periodic limb movements during sleep (PLMS)(Montplaisir et al., 1998). Dopaminergic medication isconsidered the medication of first choice (Chesson et al.,1999), and is efficient in treating subjective symptoms andsleep disturbance in RLS patients and also reduces thenumber of PLMS substantially (Walters et al., 1988; Tren-kwalder et al.; 1995, Wetter et al.; 1999; Stiasny et al.,2000). Therefore, the number of PLMS and PLMS asso-ciated arousals is widely used to measure treatment effectsin RLS patients.Previous treatment studies of RLS with PLMS have beeneither conducted using actigraphy (Collado-Seidel et al.,1999; Benes et al., 2000) or polysomnography (PSG)(Allen and Earley, 1996; Walters et al., 1988; Trenkwalderet al., 1995; Earley et al., 1998; Montplaisir et al., 1999).There were certain limitations to these studies. During acti-graphic recording only information on motor activities canbe obtained, as we do not have sleep recordings as well. Inaddition, actigraphic studies lack bilateral anterior tibialiselectromyography (EMG), thus they do not meet thecommonly accepted criteria for PLMS assessment(ASDA, 1993). The PSG studies were either limited in thenumber of study centers (Wetter et al., 1999) or had alimited number of patients with limited follow-up data(Stiasny et al., 2001; Montplaisir et al., 2000). Furthermore,these studies included a maximum of about 25 patients(Stiasny et al., 2001; Montplaisir et al., 2000).In order to objectively evaluate efficacy and safety para-meters for pergolide, as a treatment of RLS, a new multi-centerstudydesign was needed. A controlled, blinded, studywith a 1 year follow-up and regular PSG recordings was Clinical Neurophysiology 113 (2002) 571–5781388-2457/02/$ - see front matter q 2002 Elsevier Science Ireland Ltd. All rights reserved.PII: S1388-2457(02) 2001676* Corresponding author. Tel.:  1 49-6421-2866-435; fax:  1 49-6421-2865-450. E-mail address: (T. Penzel).  required. As a high inter-rater variability in sleep scoringbetween different sleep laboratories has been reported(Norman et al., 2000), we used a design with a centralevaluation site. Only one scorer, who was blinded to allclinical and supplemental patient information, evaluatedthe recordings. As the European Data Format (EDF) allowsto convert data from different electroencephalographic(EEG) systems into a common file format, we used thisrecently developed data exchange format for transfer of data (Kemp et al., 1992). The duration of 1 year was chosenin order to assess long-term efficacy and safety of pergolidein RLS.Written standardized operating procedures (SOP) werecreated and distributed. We wanted to achieve maximumquality in the PSG data and evaluation in order to receivecomparable results for this long-term study. 2. Methods A steering committee was inaugurated in order to createthe design of this placebo-controlled multi-center study with1 year follow-up and regular PSG controls. This groupdefined a novel study approach with precise requirementsand a strict quality control defined in multiple proceduralstatements as is common in system and software develop-ment. 2.1. Study center requirements – pilot phase Minimal criteria for the acceptance of study centers weredefined. The initial contact was made if the centers wereexperienced in treating RLS patients and in performingsleep studies. The centers were interviewed in order to clar-ify their ability to participate in a multi-center study withdigital PSG recordings. The interview also assessed patientpopulation, number of patients seen with periodic limbmovements (PLM), and ability for follow-up studies. Thenext step was a pilot study during which each center had toprovide at least one digital sleep recording as specified bythe study protocol (see below). These recordings were eval-uated by the central evaluation center using previouslydefined SOP. If the recording or the accompanying informa-tion did not match the requirements, a revised pilot record-ing was required until it satisfied all proceduralrequirements. Finally, during a study group meeting, theparticipating centers were instructed about the proceduresto be followed, the subsequent quality control in the centralevaluation center, and received an addendum to the studyprotocol with all SOPs. 2.2. Study procedure2.2.1. Patient selection All patients had to fulfil the minimal diagnostic criteriafor an idiopathic RLS according to Walters and the theInternational Restless Legs Syndrome Study Group(1995). After a clinical examination each patient had toundergo a baseline evaluation. The recordings were evalu-ated by the central evaluation site in Marburg. Only after thepatients met inclusion criteria, defined prior to the studystart, they were randomized. The inclusion criteria aredescribed below. 2.2.2. Sleep recording PSGs were performed at baseline, after 6weeks, 6monthsand 1 year. At all steps, two sleep studies were performed.Each visit consisted of two consecutive nights with PSG.The second night was evaluated to avoid the first-night-effect.Sleep was recorded according to Rechtschaffen and Kales(1968) with a minimum of two EEG leads (C3–A2, C4–A1),two electrooculogram (EOG) leads and submental electro-myogram (EMG). Respiration was recorded using airflow,thoracic and abdominal respiratory movements and snoringsounds. Electrocardiogram (ECG) was also recorded. Theleg movements were recorded using separate anterior tibia-lis EMG from both legs. Minimal requirements for gain andfilter settings were specified as commonly used in sleeprecordings (Penzel et al., 1993). The specific additionalrequirements for digital sleep recordings also followedcommon recommendations (Penzel et al., 1998).In addition to signal requirements, procedural studyrequirements were also noted. Each recording had to includea documented electrode impedance check and a specified‘biological calibration test’. Several tests had to beperformed at the beginning of the recording in order tomake the visual signal evaluation more reliable. The legmovements consisted of repeated maximum dorsal flexionof each foot. No signal clipping should occur at maximumamplitude. A detailed recording protocol was required.Beside signal settings and impedance values, all eventsand their time of occurrence had to be recorded in thisprotocol. ‘Lights off’ and ‘lights on’ were most essentialbecause this specified the time period for evaluation andsleep latency. Time of interruptions during night recordingwere carefully noted in the protocol. Recommended ‘lightsoff’ time was 23:00 hrs. 2.2.3. Sleep evaluation at the central reading site Visual evaluations were performed by the central evalua-tion center, which did not recruit patients. The task of theevaluation center was to evaluate and check all sleep record-ings and transfer completed protocols to the study sponsor.In addition, the evaluation center had to take care of thequality of the recordings by immediate feedback to thestudy sites. The scorer of the PSGs was blinded to anypatient information and exclusively scored the digitalrecordings. If more than 10% of the annotated time in bedwas missing or could not be evaluated, the recording wasdiscarded. Only the annotated time in bed was scored.Visual sleep evaluation was done strictly according toRechtschaffen and Kales (1968). Arousals were scored T. Penzel et al. / Clinical Neurophysiology 113 (2002) 571–578 572  according to ASDA criteria (1992). Two additions weremade to minimize scorer variability. After an arousal duringREM sleep, we continued to score REM sleep, if EEG andEMG indicated continuous rapid eye movement (REM).After an arousal in non-REM stage 2, we continued toscore stage 2, if the next epoch contained a K-complex ora sleep spindle and the background activity did not change.The leg movements were scored according to ASDA criteria(1993). A leg movement (LM) was defined as a burst of anterior tibialis muscle activity with a duration of 0.5–5 sand with an amplitude of at least 25% of the biologicalcalibration test. A PLM was scored if a sequence of 4 ormore LM, separated strictly by 4–90 s between movementonset was found. Visual scoring of PLM was done for bothlegs separately. In a second step, simultaneous PLM in bothlegs were counted as one event if the onset of both PLMoccurred within 4 s. The ASDA criteria were extended bysome specifications. For the detection of valid LM, werequired a drop of EMG amplitude between the eventsback to baseline amplitude in order to avoid PLM eventsdue to unspecific movement. We separated PLM with arou-sal and PLM with subsequent awakening for final analysis.In order to evaluate synchronicity between these events, atime window of 4 s was chosen. With regard to the charac-teristic sleep fragmentation and micro-sleep in patients withRLS, a PLM was scored even if the particular PLM occurredduring ‘WAKE’ according to Rechtschaffen and Kales(1968). The event had to be associated with an arousaland a preceding 10 s period of apparent sleep. To completethe scoring of the recording, the total number of apnea andhypopnea events were evaluated. This was done in order toexclude patients with sleep related breathing disorders.Study parameters for subsequent statistical analysis weretotal sleep time (TST), sleep efficacy (SE), sleep onsetlatency (SOL), duration of sleep stages in minutes, sleepstages expressed as percent of TST, number of PLM andPLM with arousal during sleep (PLMS-arousal) per hour.The visual evaluation of the sleep recording wasperformed by one experienced scorer using the certifiedWindows-based software, WINDAY, which generated areport file for later processing. 2.2.4. Patient inclusion criteria The PSG inclusion criteria consisted of:1. sleep efficiency less than or equal to 85% and/or sleeplatency to non-REM 2 of more than 25 min.2. PLMS-arousal index . 5/h during TST. 2.2.5. Data flow of the multi-center study The recording sites sent copies of the sleep recordings onCD-ROM to the central evaluation center using an overnightcourier service (Fig. 1). An initial check of the recordingincluded a check of file name conventions, of the submitteddata format, accompanying subject information, and exis-tence of the sleep protocol. Then the sleep recording wasconverted to the data format used for evaluation (WINDAYproprietary format, similar to the EDF (Kemp et al., 1992)).For this purpose, a conversion tool was created which couldread all data formats, adjust amplitude and offset of the 16systems (Table 1), resample signals to 100 Hz for EEG,EOG, ECG, and 200 Hz for EMG where required (ETOS). T. Penzel et al. / Clinical Neurophysiology 113 (2002) 571578  573Fig. 1. Data flow in the study. Each step of the process had standardizedoperating procedures.Table 1List of 16 systems used by the 21 participating sleep centersPolysomnography system (company) Data file formattransferredAlice3 (Healthdyne) EDF exportAstromed Grass (Astromed Inc.) ASCII conversionBrainlab (Schwarzer GmbH) EDF exportBrainstar (Schwind Medizintechnik) EDF exportSleepwatch (Compumedics Sleep PTY Ltd.) EDF exportED Sleep (Medizintechnik fu¨r Arzt und Patient) EDF exportEmbla (Flaga Inc.) EDF recordingEclipse (Glonner Electronic GmbH) Binary conversionMedatec (Medatec) EDF exportNeurofile 4421 Binary conversionUltrasom (Nicolet) Binary conversionNightingale (Judex) EDF exportOxford Medelec (Oxford Medical Instruments) EDF exportSomnostar (Sensormedics Co.,) Binary conversionSirius Galileo (EB Neuro) ASCII conversionVitaport 2 (TEMEC Instruments) EDF export  The converted recording was checked for signal quality(Schlo¨gl et al., 1999) and forwarded to the scorer. Afterimmediate scoring, the evaluation was again checked formissing values and inconsistencies. For the baseline record-ing, a short note with the decision whether the patient wasincluded in the study was sent to the recording site within 2days. Detailed sleep and PLM results remained hidden fromthe recruiting site. Exclusion criteria were checked at thispoint of the study. None of the subsequent resulting reportswere transmitted to the recording site. For all recordings, acomputer generated Excel table with the study parametersdefined above was sent to the study sponsor. 2.2.6. Quality management of the study During the planning phase of the study, the data flow wasdiagrammed in a formal report and a risk analysis wasundertaken to find out at which steps errors could occur.The results of the risk analysis were stated in formalSOPs. The SOPs did specify every single step in the record-ing and evaluation process. Subject data, technical data (i.e.file name), completeness of the sleep report, were describedin separate statements. Procedural data (i.e. date of receiptof recording, date of scoring, and of sending the report tostudy sponsor) had to be noted in formalized protocols. Allchecks according to the SOPs were scored from 1 to 4 (1 –good recording through 4 – recording too bad to be evalu-ated). The formalized protocols with the scores andcomments were entered in an Access database. The partiallyautomatically generated quality check reports wereimported to this database (i.e. signal labels in the digitalPSG recording). The visually assessed results of the signalquality evaluation were also entered in the database. Theresults of the sleep scoring provided by the scorer, who wasblind to subject information, were imported in the databaseby reading a computer generated report file containing theresult of the visual sleep scoring. This result file was auto-matically generated by the WINDAY sleep scoring programonce visual scoring was completed. The database containedforms for each step of the data flow. The steps of the dataflow had been described in the SOPs. A potential violationof the SOPs was detected very early and corrective actionswere initiated. For each new sleep recording, a record wasadded to the database.A pilot phase was introduced in order to check thecompliance of the study sites with the SOPs. The pilotphase did apply all described quality checks for pilot record-ings prior to study start. Only if the pilot phase was success-fully completed, a sleep center was approved for the study.In order to assure the quality of the visual scoring, aquality control was introduced for the scorer. Six monthsafter the end of the study, 20% of the baseline recordingswere randomly selected and re-scored to determine intra-rater variability. As target parameters for this comparison,we chose parameters being relevant for patient inclusion, forefficacy, and safety assessment. Thus, epoch by epochagreement of sleep stages was calculated. Based on sleepstage re-scoring, SE and SOL were determined. In addition,PLM index and number of arousals were re-scored. Theinter-class correlation coefficient (ICC) was calculatedusing the two-way mixed model (SPSS 10.0). Bland andAltman diagrams (1986) were plotted to assess differences. 3. Results Initially 30 sleep laboratories in 8 countries were selectedand contacted. On the basis of the above described proce-dure, 25 laboratories were willing and able to participate inthe study. These laboratories entered the pilot phase of thestudy and did submit digital sleep recordings. The record-ings from 23 centers could be read immediately after havingclarified technical conversion problems. For 2 centers, theconversion problems could not be solved. The recordingsfrom 8 sleep centers could be evaluated according to theSOPs and were immediately sufficient. The recordingsfrom 15 centers did not follow all specifications required.The reasons were undocumented EEG gain, undefined EOGleads, insufficient submental EMG signal with dominantline interference, wrong signal labels, too low amplificationin anterior tibialis EMG, and missing annotations in the PSGprotocol. After feedback, 13 out of the 15 centers were ableto solve the problems. Two centers could not deliver suffi-cient pilot study recordings. Finally21 sleep centers, locatedin Australia, Belgium, Finland, Germany, Italy, Nether-lands, and Spain, were approved to participate in thestudy. Sixteen different digital PSG systems were used bythe centers. We were able to convert all digital sleep record-ings to our data format to do sleep scoring. Ten systemssupported the EDF format for digital sleep recordings. Byproviding a conversion tool, two systems stored the record-ings in ASCII format, and 4 systems provided binary datawhich required conversion in the central evaluation center. 3.1. Management of patient inclusion Altogether 17 study centers submitted 100 sleep record-ings of patients who fulfilled clinical baseline criteria. Fourapproved centers did not recruit patients after the start of thestudy. Two out of 17 centers submitted recordings, but thesedid not fulfill the patient inclusion criteria. The recruitmentof patients varied from 1 to 15 patients per site.Out of the 100 enrolled patients, 16 did not meet therequirements for inclusion by PSG criteria and wereexcluded from the study. Of these 16 patients, 15 had aPLMS-arousal index below or equal to 5 and one patienthad both, too high SE and too low SOL.Once patients met the baseline criteria, only a fewsubjects did not complete the study. After 6 weeks, 81patients revisited the labs, after 6 months 70 patientsreturned, and 66 patients completed the study with the 1-year follow-up examination. Out of the patients that finishedthe study, 65 had 4 PSG recordings each, which satisfied the T. Penzel et al. / Clinical Neurophysiology 113 (2002) 571578 574  requirements for evaluation. One patient finished the studybut had only 2 PSG recordings meeting the requirements. 3.2. Quality control of PSG protocol and evaluation Once the pilot phase was completed, very few problemswere found in the sleep recordings. These were mainlyrelated to the sleep protocol. Inconsistencies between anno-tated ‘lights off’ time and obvious start time of the record-ing, differences between the sleep protocol and the sleeprecording for interruptions of the night were found andhad to be clarified by contacting the recording site. Allproblems could be solved.To control quality of visual sleep evaluation, we assessedthe intra-rater variabilitybased on 20 randomly chosensleeprecordings. We compared epoch by epoch agreement of sleep stages. The mean match was 88.2% (range: 81.5–96.5%). The mean SE in the re-test recording set was72.5% (range: 1.5–88.3). The mean difference for SE was0.1% (range: 2 1.8 to 2.9%). The correlation between initialscoring of SE and second scoring was ICC ¼ 0.998( P , 0 : 01). The deviations are depicted in a Bland andAltman diagram (Fig. 2). The mean SOL in the re-testrecording set was 66.6 min (range: 6.0–478.0 min). Themean difference for SOL was  2 3.6 min (range:  2 112 to36 min). The correlation between initial and second scoringof SOL was ICC ¼ 0.965 ( P , 0 : 01) (Fig. 3). Fig. 3 showsthat a remarkable difference of 112 min was found for sleeplatency in one subject. This is due to an oversight, agraphoelement in the sleep EEG was missed and in conse-quence onset of sleep stage 2 had been scored much later.One subject out of the rescored group did not sleep at all(SOL: 478 min, SE: 1.5%). This extreme value influencedthe Pearson correlation coefficient for the two variables. Themean PLM index in the re-test recording set was 37.7 (range3.7–88.8). The mean difference for PLM index was 0.1(range: 2 1.5 to 2.1). The correlation between initial scoringof PLM and re-scoring was ICC ¼ 0.998 ( P , 0 : 01) (Fig.4). The mean number of arousals in the re-test set was 181.6(range 20–479). The mean difference for arousals was 0.5(range 2 16 to20). The correlation betweeninitial scoring of arousals and re-scoring was ICC ¼ 0.996 ( P , 0 : 01) (Fig.5). 4. Discussion This multi-center study with one independent centralevaluation center allowed the evaluation of efficacy andsafety parameters of pergolide in patients with RLS andsleep disturbances. The structured design of the study, T. Penzel et al. / Clinical Neurophysiology 113 (2002) 571578  575Fig. 2. Comparison of sleep efficiency differences using a Bland andAltman diagram.Fig. 3. Comparison of sleep onset latency differences using a Bland andAltman diagram.Fig. 4. Comparison of PLM index (total number of PLM during TIB perhour) differences using a Bland and Altman diagram.
Similar documents
View more...
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks