Ranjit B. Mani, MD, HFD-120 Medical Review
`NDA 21196, Xyrem , Orphan Medical, Inc.
`
`Page 25 of 89
`6/15/01
`
`0
`
`She applied an ANOVA the model for which included treatment, site and the
`treatment-by—site interaction. The assumption of normality for this model were
`however violated based on the Shapiro-Wilk test (p=0.017) and she therefore
`carried out a log transformation of the data. On re-applying the Shapiro-Wilk test
`to the log—transformed data the assumptions of normality was no longer violated
`(p=0.7365). An ANOVA on the log-transformed data revealed a p-value for the
`overall comparison of 0.0034. The subsequent comparison of each GHB group
`with olacebo revealed the followin-
`
`
`
`
`
`
`
`
`
`
`
`-rmm-
`
`. She had earlier also performed a Wilcoxon Rank Sum test. The p-value for the
`overall comparison of the 4 treatment groups, using the latter non-parametric test
`was 0.0101. She then compared each GHB group with placebo and the p-values
`for each of these com-arisons was as follows
`
`
`
`
`
`
`
`0 Thus, according to the protocol-specified primary efficacy analysis, andithe
`sponsor’s analysis, only the 9 g/day dose showed a statistically significant
`superiority to placebo in reducing the total number of cataplexy attacks,
`The evidence for efficacy at the 6 g/day dose appeared marginal and .
`analysis-dependent.
`
`0.1450
`
`There was no definite evidence that GHB was efficacious in treating complete
`cataplexy attacks, the most serious form of cataplexy. However the mean
`(and median) frequency of such attacks in both treatment groups was small
`as was the absolute change in frequency from baseline to endpoint; a trend
`to a treatment effect may have been seen.
`
`6. 14.2 Secondary Efficacy Measures
`
`In this application the sponsor has sought a claim for Xyrem® in treating
`daytime sleepiness accompanying narcolepsy.
`,
`The secondary efficacy measures used to assess daytime sleepiness
`included the Epworth Sleepiness Scale, the frequency of sleep attacks
`(inadvertent naps) during the day and the duration of daytime sleep attacks
`(inadvertent naps)
`On the sponsor’s analysis, a nominally statistically significant superiority (p <
`0.05) of GHB over placebo was seen on the Epworth Sleepiness Scale, the
`frequency of daytime sleep attacks and the duration of daytime sleep attacks,
`as measures of excessive daytime sleepiness. However given that there were
`12 secondary efficacy measures, only the analysis of the Epworth Sleepiness
`Scale was still statistically significant after adjustment for multiple
`comparisons.
`
`

`

`Fianjit e. Mani, MD, HFD-120 Medical Review
`NDA 21196, Xyrem , Orphan Medical, Inc.
`
`Page 26 of 89
`6/15/01
`
`0 On the sponsor’s analysis, the pairwise comparisons for the Epworth
`Sleepiness Scale indicated that only the 9 g/day dose of GHB was superior to
`placebo
`. Dr Sharon Yan, Agency Statistical Reviewer finds the analysis of secondary
`efficacy measures for this study problematical for the following reasons
`. There are many secondary efficacy measures
`- The methods of analysis were not stated in detail a priori
`.
`in specific reference to excessive daytime sleepiness, as measured by the
`Epworth Scale
`0
`She applied the protocol-specified ANOVA model to the original scale
`. After the treatment-by—site interaction was found not to be significant, if was removed
`from the model after which the residuals were no longer normally distributed, even
`after log transformation
`She therefore performed a Kruskall-Wallis test. The overall p-value obtained for the
`GHB-placebo comparison was then 00109. As noted earlier, the Epworth Sleepiness
`Scale was one of 10 secondary efficacy measures and the overall p-value for this
`measure did not achieve statistical significance when a Bonferroni adjustment was
`made.
`
`.
`
`. The results of this study, based on the sponsor’s analysis, nevertheless, do
`provide at least some support for the efficacy of GHB in a dose of 9 g/day in
`treating excessive daytime sleepiness in narcolepsy.
`
`6.14.3 Influence Of Stimulant Drugs On Efficacy
`
`6. 14. 3. 1 Background
`At the request of the Biopharmaceutics staff at the Agency, the following request
`was passed on to the sponsor on April 4, 2001
`
`"The clinical study database should be investigated further to ascertain potential pharmacodynamic
`interactions in narcoleptic patients with other commonly used drugs in this patient populations.”
`
`)
`
`.
`
`The structure of the sponsor-proposed analysis of these interactions was
`discussed between the Division and sponsor at a teleconference on 4/18/01. The
`sponsor suggested the following, which was acceptable to the Division:
`'
`
`o The analysis would focus on the differences in observed effects, as they related to
`both safety and efficacy in narcoleptic patients
`
`and would compare the following groups
`
`.
`
`Patients who received sodium oxybate alone
`
`Patients who received a selected concomitant medication alone
`
`Patients who received a combination of sodium oxybate and a selected concomitant
`medication
`
`

`

`Ranjit B. Mani, MD, HFD‘120 Medical Review
`NDA 21196, Xyrem , Orphan Medical, lnc.
`
`Page 27 of 89
`6/15/01
`
`The above analysis is the basis for an additional submission dated May 4, 2001
`which is summarized here as well as in my NDA Safety Review.
`
`6.14.3.2 Sponsor’s Methods
`
`6.14.321 General Observations
`
`. Narcoleptic patients commonly use the following classes of medications to
`treat that disorder
`
`0
`
`.
`
`Stimulants (e.g., methylphenidate, dextroamphetamine, methamphetamine,
`pemoline, modafinil) to treat excessive daytime sleepiness
`Tricyclic antidepressants and selective serotonin re-uptake inhibitors to treat
`REM dissociation phenomena: cataplexy, hypnagogic hallucinations and sleep
`paralysis
`. The entire NDA database did not include a trial specifically designed to
`investigate the potential pharmacodynamic interactions between Xyrem® and
`medications commonly used in patients with narcolepsy. Nevertheless, for
`analysis purposes all clinical trials in the database were examined
`. However, in only the OMC-GHB-Z, Lammers and Scrima trials was it possible
`to compare the following groups in a controlled setting
`~
`.
`Patients who received sodium oxybate alone
`.
`Patients who received a selected concomitant medication alone
`
`.
`
`Patients who received a combination of sodium oxybate and a selected
`concomitant medication
`
`Even in the setting of these 3 controlled trials
`.
`Stimulants were the only medication class on which such an analysis could be
`mme
`m
`. Both the Scrima and Lammers trials were not suitable for the analysis on
`account of a small sample size and variable use of stimulants
`
`_~
`
`6.14.822 Stimulant Use In OMC-GHB-2
`
`Of the 136 patients enrolled in this trial
`0
`115/136 (84.6%) maintained stable doses of stimulants during the trial
`0
`21/136 (15.4%) did not take stimulants
`o The distribution of these atients b treatment o-rou is below
`
`.
`
`.
`
`
`
`Of those taking stimulant drugs
`41 were taking amphetamines
`55 were taking methylphenidate
`25 were taking pemoline
`Some patients took more than 1 stimulant drug
`
`
`
`The distribution of these oatients b treatment oorou is in the followin-
`
`0 da
`
`_—__mm-
`Amphetamines
`Number not treated with amphetamine
`
`
`
`Number treated with am-hetamine _——_I_
`
`
`
`
`

`

`Ranjit B. Mani, MD, HFD-12O Medical Review
`NDA 21196, Xyrem ,Orphan Medical, Inc.
`
`Page 28 of 89
`6/15/01
`
`Methylphenidate
`
`Pemoline
`
`
`
`
`
`Number not treated with
`
`math l-henidate
`
`Number treated with meth Ihenidate _-_
`Number not treated with emoline
`Number treated with emoline __—
`
`
`
`
`
`
`
`
`
`
`6.14.323 Analysis Of Effects Of Stimulant Drugs On Efficacy
`
`. The 2 outcome variables chosen for the analysis were
`. The frequency of all cataplexy attacks
`. Daytime sleepiness as measured by the Epworth Sleepiness Scale
`- Descriptive statistics were calculated for each of the outcome variables, for
`each stimulant and for patients not taking stimulants, by treatment group
`. Analysis of the total number of cataplexy attacks (after log transformation)
`and the change in Epworth scores was accomplished using ANCOVA: the
`model included baseline value of the variable being analyzed (the covariate)
`and site and treatment as terms.
`
`0 Separate analyses were performed for each treatment group of patients,
`based on type of stimulant used, and any stimulant use
`. Adjustments were made for multiple comparisons using the Dunnett-Hsu
`procedure
`. An additional analysis was performed to assess the possible interaction
`between stimulant use and GHB treatment. This used the same ANCOVA
`model as above with 2 additional terms: stimulant use (yes or no) and the
`stimulant-by-treatment interaction. This analysis was performed for each of
`the stimulants above and for the stimulant group as a whole
`
`6.14.3.3 Results: Effects Of Stimulants On Efficacy Of GHB
`
`The differences among treatment groups were consistent between those
`patients taking stimulants and those not taking stimulants; this was determined
`using the additional ANCOVA model which included the stimulant-by-treatment-
`group interaction. For each stimulant and each efficacy variable this interaction
`..
`was not statistically significant.
`'
`
`Full tables describing the analysis are in the submission. I have not reproduced
`them here but they appear to confirm the sponsor’s conclusions
`
`6.14.3.4 Results: Effects Of Stimulants On Safety Of GHB
`
`These results are described in the NDA Safety Review
`
`6.14.3.5 Sponsor’s Conclusions
`
`0
`
`o
`
`In the analysis of cataplexy and daytime sleepiness there was no evidence of
`pharmacodynamic interaction between sodium oxybate treatment and .
`concomitant stimulants
`In the analysis of adverse events there was only one body system (Digestive)
`in which a very weak signal of difference between those treated with GHB
`and methylphenidate and GHB alone was detected
`
`

`

`Ranjit B. Mani, MD, HFD-120 Medical Review
`NBA 21196, Xyrem , Orphan Medical, Inc.
`
`Page 31 of 89
`6/15/01
`
`. Sleep patterns identified on the polysomnogram
`. Average number of REM onsets by Multiple Sleep Latency Test
`
`7.3 Design
`
`Randomized, double-blind, placebo-controlled, single-center, cross-over study
`comparing the effect of GHB 50 mg/kg total daily dose in with placebo in 20
`patients with narcolepsy.
`
`A schematic outline of the study design is presented in the figure below which I
`have copied from this submission.
`
`”human! tenor: 9}
`$29 Days!
`
`
`
`so @515
`
`aaiunur. 1:;
`
`:31 v: defi acct.
`bfi {I
`
`
`
`flashed:
`(6 days!
`
`trauma-:3. 2113.;
`5145: : Iii:
`
`:.
`
`'Ironxmnt Vienna 92
`i259 Days;
`
`
`
`——>
`Submaroncvnup
`maximum 5 days!
`
`
`
`SCEEEII'I!}3.’E.\E ELENE
`
`FARE}? " ‘rzmz
`
`
`
`
`
`transom: x'
`
`ts: 6.13:5}
`
`(Kznxmn: 5 013:7»
`
`—-—>
`Nashua: . FvLJ‘Jv - 4.
`
`Randomization was to be such that half of the men and half of the women
`
`participating in the study would receive GHB during the first 29-day double-blind
`treatment period, and placebo during the second. The remaining patients were to
`receive GHB first and placebo later.
`'
`
`7.4 Duration
`
`4 weeks of double-blind treatment during each cross-over period
`
`7.5 Dosage
`
`During each period of double-blind treatment, each participating patient was to
`take
`
`GHB 25 mg/kg at bedtime, and about 3 hours later (total dose: 50 mg/kg/day)
`OR
`
`Matching placebo
`
`

`

`Page 33 of 89
`Ranjit B. Mani, MD, HFD-120 Medical Review
`
`NBA 21198, Xyrem , Orphan Medical, Inc. 6/15/01
`
`Study Schedule
`
`
`
`MSLT Day 1
`93 Day -1
`’1. 2-231“; Day—3.4 to 7‘3 Labs Day 1
`P513 DA}! 29
`
`“BL? Day 30
`Lens say 30
`
`PSG Bay 35 fl 6‘
`KSLT L251! 37
`Labs Day 37
`
`REL? Day 65
`Lacs may 65
`
`In the above schematic
`MSLT stands for Multiple Sleep Latency Test
`PSG stands for polysomnogram
`
`Screening assessments consisted of a medical history, sleep disorders
`interview, physical examination, an overnight polysomnogram and Multiple
`Sleep Latency Test (the last 2 tests were for diagnostic purposes only)
`Polysomnograms (overnight) were performed on the last day of the baseline
`period, the first night (Day 1) of each treatment period and the last night (Day
`29) of each treatment period (note that this differs somewhat from what is
`stated in the above schematic)
`Multiple Sleep Latency Tests were performed at some time during the
`baseline period, on the morning and afternoon after the Day 1 overnight
`polysomnogram and on Day 29 of each treatment period. On each day the
`test was performed 5 times, each time over a 20 minute period. The timing of
`these epochs was as follows: 0800, 1000, 1200, 1400 and 1600 hours.
`A nighttime sleep log and daytime questionnaire were maintained by each
`patient throughout each study period (baseline, treatment period 1, washout,
`treatment period 2 and washout/follow-up). These assessed the following
`Nighttime sleep onset latency, nighttime arousals, total sleep time, feelings on awakening, number of
`sleep attacks, number of cataplexy attacks, number of naps, methylphenidale use, sleep/awake
`patterns, and mood
`Pre-sleep questionnaires were administered before each overnight
`polysomnogram. These assessed the following:
`General physical and mental health, food intake, activities on test days, sleep quality, and
`feelings
`‘
`Post—sleep questionnaires were administered on the morning after each
`overnight polysomnogram, after completion of the Multiple Sleep Latency.
`Test. They assessed the following
`Sleep quality, and feelings
`
`

`

`Ranjit B. Mani, MD, HFD-120 Medical Review
`NBA 21196, Xyrem , Orphan Medical, Inc.
`
`Page 34 of 89
`' 6/15/01
`
`0 Adverse events were to be recorded continually based on patient interviews,
`diaries and telephone contacts. The period of observation for these began at
`the time of obtaining informed consent and extended through at least the first
`5 days of the washout/follow-up period at the conclusion of Treatment Period
`2
`
`. Standard safety laboratory tests (hematology, chemistry and thyroid
`functions) were checked at the beginning and end of each treatment period
`- During each polysomnogram the following additional items were monitored:
`- Cardiac rhythm by electrocardiogram
`Respiration using either intercostal EMG or diaphragm bellows
`Ventilation by nasal 02 and/or thermocouples
`Blood arterial 02 saturation by ear oximetry
`Cataplexy by EEG, chin-EMG, ECG and electrocardiogram
`0 Women of child-bearing potential were to have a urine pregnancy test
`performed 30-45 days and 7 days prior to the baseline polysomnogram
`
`7.11 Efficacy Outcome Measures
`
`The efficacy outcome measures below are those listed in the study report-The
`primary outcome measures match those in the final analysis plan submitted by
`Dr Scrima in December1986, clarified further in March 1987. The secondary
`efficacy measures only partly match those listed in the final analysis plan.
`
`7. 1 1. 1 Primary Efficacy Measures
`
`9
`
`. Number of cataplexy attacks per day
`Objective daytime sleepiness as measured by the Multiple Sleep Latency
`Test* Sleepiness Index
`
`7.11.2 Secondary Efficacy Measures Based On Patient Diaries
`
`. The number of awakenings from sleep at night
`0 The numberof sleep attacks during the day
`.
`- Methylphenidate use (mg/day)
`. Patient sense of alertness on awakening on a scale from 1 (feeling active,
`alert, vital and wide awake) to 8 (asleep): this is also referred to as the
`Stanford Sleepiness Scale
`Feeling active, vital, alert, or wide awake: 1
`Functioning at high levels, but not at peak; able to concentrate: 2
`Awake, but relaxed; responsive but not fully alert: 3
`Somewhat foggy, let down: 4
`Foggy; losing interest in remaining awake; slowed down: 5
`Sleepy, woozy, fighting sleep; prefer to lie down: 6
`No longer fighting sleep, sleep onset soon; having dream-like thoughts: 7
`Asleep: 8
`. Mood in the morning recorded on a scale from —10 (extremely negative) to
`+10 (extremely positive) with 0 representing a neutral mood
`o Mood in the evening recorded on a scale from —10 (extremely negative) to
`+10 (extremely positive) with 0 representing a neutral mood
`
`7.11.3 Secondary Efficacy Measures Based On Polysomnograms
`
`. Sleep efficiency (%)
`. Sleep latency (minutes)
`
`

`

`Ranjit B. Mani, MD, HFD-120 Medical Review
`NBA 21196, Xyrem , Orphan Medical, Inc.
`
`Page 35 of 89
`.
`6/15/01
`
`Stage I (%)
`Stage 2 (%)
`Stage 3 (%)
`Stage 4 (%)
`Stage shifts (the number of times during sleep that sleep transitions from one
`sleep stage to another sleep stage)
`0 REM sleep (%)
`. REM latency (minutes)
`0 Number of awakenings
`
`7.11.4 Secondary Efficacy Measures Based On Multiple Sleep Latency Test*
`
`7.11.5 Number of REM onsets
`
`7.11.6 Safety Outcome Measures
`Adverse events
`
`Changes in clinical laboratory parameters (adverse events and safety laboratory
`parameters)
`
`‘The Multiple Sleep Latency Test consisted of 5 separate sessions on a single day, each
`session lasting 20 minutes. Measurements in minutes derived from the test included the following:
`.
`Latency to sleep onset (time from lights out to the first epoch of sleep)
`Latency to REM sleep (time from the beginning of sleep onset to the first epoch of REM sleep)
`Average latency to sleep onset (total latencies to sleep/number of naps)
`Number of naps with REM across five naps
`Sleepiness Index = 100 —— (5 x average latency to sleep)
`
`7.12 Analysis Plan
`
`Defining the analysis plan actually used for the study took considerable effort
`
`7.12.1 General Considerations
`
`7. 12. 1. 1 Efficacy Analyses
`
`. All randomized patients were to be included in the primary and secondary
`efficacy analyses, except that patients missing an entire treatment period
`were not included in the post-treatment analysis.
`. Patients included in the analysis of the effects of GHB/placebo withdrawal
`during the washout and washout/follow-up periods must have had both
`baseline data and at least 3 of the 5 days following treatment periods fer
`each period
`. Missing data were substituted for 1 or 2 weeks during the treatment periods
`or 1 or 2 days during the washout or washout/follow-up periods. Missing data
`substitutions were made only for the diary data
`. All efficacy analyses were to be based on change of data from baseline to
`post-treatment for each study period
`
`7. 12. 1.2 Safety Analysis
`
`Any patient who received study medication was to be included in the safety
`analysis
`
`

`

`Ranjit B. Mani, MD, HFD-120 Medical Review
`NBA 21196, Xyrem , Orphan Medical, lnc.
`
`7.12.1.3 Study Periods
`
`These were defined as follows
`
`Page 36 of 89
`6/15/01
`
`The baseline period was the 2-week pre-treatment period over which patient
`diaries were recorded. The 5-days preceding the first dose of study
`medication were considered an adequate baseline for data collected on
`patients’ daily subjective assessments
`GHB/placebo were administered in Treatment Periods 1 and 2 (see
`schematic in Section 7.3)
`The washout period was the 5-day period after Treatment Period 1
`The follow—up period (also referred to as the washout/follow-up period ) was
`the 5-day period after Treatment Period 2
`
`7.12. 1.4 Others
`
`Sleep stages were converted to percentages of sleep time for analysis
`purposes
`Diary dates were converted to treatment day as follows
`. The date in the diary for the last day of the baseline period was the first
`treatment day
`0 The date in the diary for the last day of treatment for each treatment period was
`the day after treatment was stopped
`The criterion for a statistically significant difference was p < 0.05 (2-sided?)
`
`7.12.2 Demographic And Baseline Characteristics
`
`:
`
`Age, weight, age at diagnosis, and number of sleep and cataplexy attacks
`were analyzed using a 2—factor ANOVA. The effects in this model were
`sequence group, gender and the interaction of gender and sequence group
`The distributions of patients with or without histories of hypnagogic
`hallucinations or sleep paralysis was tested for independence from gender
`and sequence group using contingency table methods
`Only patients with baseline data who were included in the post-treatment
`analysis were analyzed for baseline comparability (of narcolepsy-related
`parameters).
`Note that the above methods for analyzing baseline and demographic
`characteristics were not specified prospectively.
`
`‘
`
`7. 123 Primary Efficacy Parameters
`
`The primary efficacy parameters consisted of the change from baseline
`scores for the following: the number of daily cataplexy occurrences and the
`sleepiness index measured from the multiple sleep latency test.
`'
`Change from baseline for the frequency of cataplexy attacks was to be
`measured as per the final prospectively-designated Scrima analysis plan
`submitted 12/4/86 as follows
`
`0 The mean daily number of cataplexy attacks was to be calculated from patient
`diaries for the following periods: the last 2 weeks of the baseline period, the last
`2 weeks of the GHB treatment period and the last 2 weeks of the placebo
`treatment period
`-
`
`

`

`Ranjit B. Mani, MD, HFD-12O Medical Review
`NBA 21196, Xyrem , Orphan Medical, Inc.
`
`Page 37 of 89
`6/15/01
`
`0 The change in mean daily cataplexy frequency from the last 2 weeks of the
`baseline period to the last 2 weeks of the GHB treatment period and to the last 2
`weeks of the placebo treatment period was then calculated
`The sponsor states thatIn Dr Scrima’3 final analysis plan, the “data called for
`analyzing only Weeks 1 and 4." WhatIS statedIn Dr Scrima’s submission of
`March 6, 1987, is that 2 within patient factors would be included in the
`ANOVA model: substance and time (15‘ week versus 4‘h week for the patient
`measures, and 15‘ day versus 28‘h day for the sleep study measures). if the
`intention was to use the mean change from baseline to Weeks 1 and 4 in
`cataplexy frequency as a primary efficacy measure, it is not clearly stated
`which of these 2 mean changes (i.e., baseline to Week 1 or baseline to Week
`4) would be considered the primary efficacy parameter, in the first of 2
`publications describing this study (Scrima et al 1989, see Section7.1) the
`sponsor states that the change in mean daily cataplexy frequency from
`baseline to Week 1 and baseline to Week 4 were both used.
`
`At any rate, the post-hoc ’
`. analysis (see Section 7.12.7) is what
`was described in the study report. In the study report the following appear to
`have been done in analyzing the change in frequency of cataplexy attacks
`
`. The baseline mean daily cataplexy frequency was calculated based on the last 5
`days of the baseline period (this is not specified in any prospective version of the
`analysis plan; all such versions called for the baseline frequency to be calculated
`using the last 14 days of the baseline period)
`‘
`o The mean daily cataplexy frequency was calculated for each of the 4 weeks of
`treatment with GHB or placebo.
`-
`. The overall mean daily frequency of cataplexy attacks was calculated for each
`treatment period in their entirety.
`-
`. The change from baseline for each treatment modality was then calculated
`based on the difference between the mean daily number of cataplexy attacks
`during the baseline period and the mean daily number of cataplexy attacks
`overall during the GHB and placebo treatment periods. The 2 treatments were
`then compared
`The change from baseline for the Multiple Sleep Latency Test Sleepiness
`index was to be calculated, as per the final prospectively-designated Scrima
`analysis plan submitted 12/4/86, based on the difference in scores from the
`Multiple Sleep Latency Test performed on the last day of the baseline period
`to the last day of the GHB treatment period, and from the last day of the
`baseline period to the last day of the placebo treatment period. However the
`analysis presented in the study report appears to imply that the primary
`analysis of the Sleepiness index was based upon the difference in mean
`scores between baseline and the entire period of treatment with either GHB
`or placebo.
`A multifactorial repeated measures ANOVA was used for this analysis; the
`“original” model submittedIn the final prospective Scrima analysis plan
`included one between-subject factor (treatment order) and 2 within-subject
`factors [treatment and time (week)]. In addition to the above factors the-
`following interactions were to be tested: treatment x order; treatment x time;
`and treatment x substance x time (null hypotheses are stated). The Mfr—~-
`
`_,.,_.,. ‘ V.
`
`,V. ,-V.,
`
`.__‘.-Av.
`
`...
`
`.. -
`
`

`

`Raniit B. Mani, MD, HFD-120 Medical Review
`NDA 21196, Xyrem , Orphan Medical, lnc.
`
`Page 38 of 89
`6/15/01
`
`'
`
`\__. analysis (see Section 7.12.7) added a between--patient factor of
`gender, and all possible interactions with gender; this analysis also pooled
`various interaction terms related to the within--patient error term
`. Since the term week was frequently significant either as part of a main effect
`or interaction, further repeated-measures analyses for the individual weeks
`were performed to support the overall analyses
`In the final prospective Scrima analysis plan, the effect of GHB withdrawal
`was to be evaluated using a paired t-test contrasting the washout week
`following withdrawal of GHB with the washout week following withdrawal of
`placebo. In the actual analysis the effect of GHB withdrawal was assessed as
`follows
`
`.
`
`. The analysis was confined to diary-derived data since only diary-based data
`were recorded during the washout and washout/follow-up periods
`. One analysis compared the 2 treatment sequences over the initial washout
`period, in regard to change from baseline. The analysis was based on a
`repeated-measures ANOVA. The model had a single within-patient factor, days,
`and two between-patient factors, sequence group and gender. Separate
`univariate supportive analyses were performed for washout days 1 to 5 with
`sequence group, gender and their interaction as factors. The intercept was ,
`tested in each model to identify departure from baseline
`0 A second analysis compared the washout from treatment in Period 1 and the
`washout/follow-up period after Treatment Period 2. A repeated-measures,-
`ANOVA was performed on the change from baseline data. There were two
`between-patient factors, treatment group and gender, and two within-patient
`factors, Day 1 to 5 of washout or follow—up
`'
`
`7.12.4 Secondary Efficacy Parameters
`
`. These were analyzed using methods largely similar to the primary efficacy
`analysis
`0 Conventional levels of statistical significance (i.e., p < 0.05) were retained for
`the comparisons (Dr Scrima, had in the final prospective analysis plan, 'stated
`that the a Bonferroni correction would not be used but the results would have
`
`to be interpreted with caution unless replicated)
`
`7.12. 5 Safety Parameters
`. The GHB and placebo treatment periods were comparedIn regard to the
`incidence of adverse events and changesIn clinical laboratory results
`. Treatmentrelated adverse events were grouped based on body system,
`severity and relationship to study drug.
`. The clinical laboratory parameters were examined with respect to reported
`values that were outside the normal range.
`-
`-
`
`7.12.6 Sample Size Rationale
`A formal a priori sample size calculation was not performed “as this was the first
`clinical trial examining GHBIn the treatment of narcolepsy”
`
`

`

`Ranjit B. Mani, MD, HFD—120 Medical Review
`NBA 21196, Xyrem , Orphan Medical, Inc.
`
`7. 12. 7 «seem Analysis
`
`Page 39 of 89
`6/15/01
`
`This analysis, performed, as noted earlier, in 1992, differed from the original
`analysis plan as follows
`. An analysis of data for Weeks 1, 2, 3 and 4 was performed for the frequency
`of cataplexy attacks, instead of what was specified prospectively.
`1
`. The original analysis plan addressed only the analysis of male patients (it
`stated that since the male narcoleptic group was reaching Completion at that
`time, the first reports were to based on a single gender group and gender
`
`was not to be included as a factorIn the analysis) At the time of the
`,/////* analysis, data from both the male and female patients were available
`and the model used at that time to analyze data from this study included a
`between--patient factor of gender, as well as all possible interactions with
`gender
`. The model usedIn the original analysis plan included various interaction
`terms related to the within--patient error term The“saga” analysis
`pooled these interaction terms so as to simplify the model.
`
`7. 13 Protocol Amendments
`
`They were no formal protocol amendments and no comprehensive final version
`of the protocol. Changes to the protocol have been inferred from
`correspondence between Dr Scrima and the Agency.
`
`7.14 Efficacy Results
`
`2141 Patient Disposition
`
`.
`
`20 patients (10 men and 10 women) were enrolled in and completed the
`study
`0 Of these 10 patients (5 men and 5 women) were randomized to the
`GHB:placebo sequence and 10 patients (5 men and 5 women) were
`randomized to the placebozGHB sequence
`. All 20 patients completed the study
`
`7.14.2 Protocol Deviations
`
`.
`
`.
`
`.
`
`.
`
`In the case of one 16-year old patient randomized to the placebozGHB ,
`sequence, neither the patient nor her parents signed the original informed
`consent document; her parents did sign the informed consent attachment for
`minors and/or fertile females
`
`3 patients (1 in the GHB:placebo sequence and the remaining 2 in the .
`placebozGHB sequence) had a baseline sleepiness index that fell below the
`minimum required for study entry
`1 patient randomized to the GHB:placebo sequence took a diuretic
`throughout the study
`1 patient randomized to the GHB:placebo sequence continued taking
`propranolol 40 mg daily for hypertension throughout the study Propranolol
`has apparently been used off-label for treating narcolepsyIn doses of 80-480
`mg/day
`
`

`

`Ranjit B. Mani, MD, HFD-120 Medical Review
`NBA 21196, Xyrem , Orphan Medical, Inc.
`
`Page 40 of 89
`6/15/01
`
`2 patients ( #s 6 and 13; one in each treatment sequence) had no diary‘
`record for cataplexy attacks during the pre-treatment period; in one of these
`instances (a patient, # 13, in the placebo:GHB sequence) the technician
`noted that the patient had had frequent cataplexy attacks and recorded that
`she had had 10 cataplexy attacks during the 2—week pre-treatment period, the
`minimum required to qualify for the study. Another patient in the GHszlacebo
`sequence (# 11) had attacks that were too frequent to count during the
`baseline period, and was therefore recorded by the technician to have 10
`attacks so as to enable her to enter the study
`1 patient in the placebo:GHB sequence got drunk during the second
`treatment period
`Another patient in the placebo:GHB sequence had an extended baseline
`period because he was initially taking > 100 mg of methylphenidate daily and
`was slowly tapered to a dose of 30 mg daily. He also had a 4-week washout
`period between treatment periods. In Weeks 1 and 2 of Treatment Period 1
`his methylphenidate dose exceeded protocol guidelines
`Another patient in the placebo:GHB sequence had a 4-week washout period
`between treatments
`"
`
`5 patients in the GHszlacebo sequence and 4 patients in the placebo:GHB
`sequence took methylphenidate after 6 PM
`2 patients in the placebo:GHB sequence and 1 patient in the GHszlacebo
`sequence received their numbers out of sequence
`
`I],\\
`
`.5
`7.14.3 Baseline And Other Demographic Characteristics
`These are summarized in the following 2 tables which compares the 2 treatment
`sequences. There were 5 men and 5 women in each treatment sequence
`
`Variable
`
`
`
`
`
`GHB:placebo sequence
`Mean SD
`
`PlacebozGHB sequence
`Mean SD
`
`
`
`
`mn—l—l—
`
`
`m—_
`
`Number of natients
`
`Number of oatients
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`As the tables above indicate
`
`. The GHszlacebo sequence had an older mean age than the placebo:GHB
`sequence, although the sponsor points out that the difference was not statistically
`significant (p = 0.089)
`0 The placebo:GHB sequence had a mean lower weight than the GHB:placebo
`sequence, and the sponsor acknowledges that the difference was'statistically‘
`significant (p = 0.008)
`’
`
`Analyses of concomitant medications and illnesses are not provided
`
`

`

`Page 41 of 89
`Ranjit B. Mani, MD, HFD-120 Medical Review
`
`NBA 21196, Xyrem , Orphan Medical, Inc. 6/15/01
`
`7.14.4 Medication Compliance
`
`A single patient (#6 participating in the GHszlacebo sequence) failed to bring
`his supply of study drug to the site on the last day of Treatment Period 1. The
`pharmacy at the study site prepared the last dose for that study period.
`
`No data are available for the exact dose in grams that each patient received.
`
`7. 14.5 Primary Efficacy Analysis
`
`7.14.5. 1 Cataplexy Attacks
`
`7.14.5.1 .1 Baseline Comparability
`
`An analysis of baseline comparability was initially performed, excluding patients
`6 and 13 who had no diary records for the baseline period. An analysis of
`baseline comparability on the remaining 18 patients showed no significant
`differences between sequence groups, genders nor their interaction for the mean
`number of daily cataplexy attacks
`
`7.14.512 Active Treatment Period
`
`Mi .
`
`Mun numb-r cf Cataploxy Attacks Per Day
`
`Pr.-
`
`fimnumt than.
`
`f Treatment
`3:931?
`
`Baanlint
`