CENTER FOR DRUG EVALUATION AND
`
`RESEARCH
`
`APPLICA TION NUMBER:
`
`21-196
`
`STATISTICAL REVIEW! S}
`
`

`

`corrmrran MAR 2 3 2901
`
`Statistical Review and Evaluation
`
`MAR 2 3 200!
`
`
`NDA#:
`
`21-196
`
`DRUG COMPANY:
`NAME OF DRUG:
`INDICATION:
`STUDIES REVIEWED:
`DOCUMENTS REVIEWED:
`NIEDICAL REVIEWER
`
`Orphan Medical, Inc.
`Xyrem (Sodium Oxybate)
`Narcolepsy
`OMC -GHB—2, Scrima, Lammers, and SXB—21
`Sponsor’s original NDA submission
`Ranjit Mani, MD.
`
`1. Introduction
`
`Narcolepsy is a chronic neurological disorder characterized by excessive daytime
`sleepiness, disturbed nocturnal sleep, cataplexy, sleep paralysis, and hypnagogic
`hallucinations (Aldrich, 1990). The usual treatment for narcolepsy includes symptomatic
`treatment of daytime sleepiness with stimulants. The symptoms of cataplexy, hypnagogic
`hallucinations, and sleep paralysis are typically treated with tricyclic antidepressants
`(TCAs) or selective serotonin reuptake inhibitors (SSRIS). The effectiveness of these
`treatments is limited by frequent undesirable adverse events and tolerance.
`
`As part of the sodium oxybate development project, Orphan Medical, Inc. acquired the
`rights to the Scrima and Lammers data for this NDA. Results from these two trials and
`along with Orphan Medical's own trial OMC-GHB—Z were submitted in this NDA to form
`the data sets to establish the efficacy of sodium oxybate in the treatment of narcolepsy
`symptoms. After the submission of the NDA, another long—term efficacy trial SXB—021
`was completed. Results of the study SXB-021 were later submitted to be included in this
`NDA.
`
`2. Specifications and Findings of Study OMC-GHB-2
`
`2.1 Objectives
`
`The primary objective of this study was to evaluate and compare the efficacy of 3 doses
`(3g, 6g, and 9g) of GHB and placebo in the treatment of the symptoms of narcolepsy.
`
`The secondary objective of this study was to evaluate and compare the safety of GHB
`and placebo when used in a narcoleptic patient population.
`
`2.2 Study Design
`
`This was a randomized, double blind, placebo controlled, parallel—group, multi-center
`study of 3 doses of GHB and placebo in the treatment of patients with narcolepsy. The
`study was conducted in 16 centers and a total of 136 patients were enrolled. The study
`was divided into 5 periods as follows:
`
`

`

`Table 1. Study flow
`
`
`
`Screening
`Washout
`Treatment
` lda —4wks
`
`
`
`2—3wks
`4wks
`5—28days
`Visit 3
`Visits 4, 5, 6
`Visit 2
`
`
`
`
`
`Withdrawal
`of TCAs
`
`No treatment for cataplexy
`
`Baseline
`
`Double—blind
`
`
`Follow—up
`
`
`3—5das
`
`
`
`
`No Treatment for
`cataplexy
`
`Placebo or GHB
`3g, 6g, or 9g
`
`1. Screening period - lasted 1 day to 4 weeks. Tricyclic antidepressants (TCAs) or other
`drugs used to treat cataplexy were gradually withdrawn from patients on these drugs.
`Patients not on TCAs proceeded directly to the next study period if they met entry
`criteria. Patients were permitted to continue taking stable doses of stimulant
`medication throughout the study.
`2. Washout period - lasted 5 - 28 days. This period allowed time to eliminate any
`clinical effects of TCAs, for rebound cataplexy (cataplexy that is spontaneous,
`unprovoked, and with greater frequency and severity than usual) to abate, and to train
`patients on the use of diary.
`3. Baseline period — lasted 2 to 3 weeks. This period was an opportunity to assess the
`patients' cataplexy attacks and to establish a stable number of attacks. Eligibility for
`admission into the double-blind treatment period required patients to report an
`average of 3 or more complete and/or partial cataplexy attacks per week, during the
`last two weeks of the baseline period.
`4. Double—blind treatment period - lasted 4 weeks. Eligible patients were randomly
`assigned to receive each night 3g, 6g, or 9g GHB or placebo in 2 divided doses.
`Patients returned approximately every 2 weeks during this period to assess safety and
`efficacy.
`5. Follow-up period - a visit for final assessment 3 - 5 days after study medication was
`discontinued.
`
`Approximately 104 patients (26 in each of the 4 treatment groups) were planned to be
`enrolled and 136 patients were actually enrolled and assigned to receive 4 weeks of
`treatment with study medication.
`
`The study was initiated on February 7, 1997 and completed on February 9, 1998.
`
`2.3 Main Inclusion Criteria
`
`Patients were included in the study if they met the following criteria:
`0 Were 18 years of age or older;
`0 Had not received investigational therapy within 30 days prior to study entry;
`0 Had a history of excessive daytime sleepiness;
`0 Had a history of sudden loss of voluntary muscle control or muscle weakness
`(cataplexy) localizable to a specific muscle group(s) or part(s) of the body during
`which the patient was lucid and not experiencing an inadvertent nap or micro sleep;
`
`

`

`0 Had a valid polysomnography (PSG) within the last 5 years and a current diagnosis of
`narcolepsy for at least 6 months according to the following 2 items of Criteria A as
`established by the American Sleep Disorder Association (ASDA):
`a) Recurrent daytime naps or lapses into sleep that occur almost daily for at least 3
`months;
`
`b) Sudden bilateral loss of postural muscle tone in association with intense emotion
`(Cataplexy).
`
`In addition, a patient must have'recorded an average of 3 or more complete and/or partial
`cataplexy attacks per week during the last 2 weeks of the baseline period to be eligible for
`entry into the randomized portion of the study.
`
`2.4 Efficacy Variables
`
`2.4.1 Primary Efficacy Variables
`
`The primary efficacy variable for this study as defined in the protocol was the change
`from baseline in the total number of cataplexy attacks, which is the sum of complete and
`partial cataplexy attacks that occurred.
`
`The endpoint was defined as the last 2-week period of the treatment (Visit 6) and the
`baseline was defined as the last 2—week period before the treatment (Visit 4).
`
`A cataplexy attack, episode, or event was defined as a sudden loss of voluntary muscle
`tone usually triggered by emotions such as those associated with laughter, anger, elation,
`fear or surprise. These events could range from a brief experience of partial muscle
`weakness to an almost complete loss of muscle control lasting for several minutes and
`resulting in total physical collapse during which time the patient was unable to move or
`speak. To be classified as cataplexy for this study the patient must have been aware of the
`time and place during the event. The event must have been of sudden onset and
`localizable to a specific muscle group(s) or part of the body.
`
`2.4.2 Secondary Efficacy Variables
`
`Secondary measures of efficacy include following variables:
`0 Change in the number of complete and number of partial cataplexy attacks;
`0 Daytime sleepiness as measured by the Epworth Sleepiness Scale and number and
`duration of inadvertent naps;
`0 Quality of nighttime sleep as measured by number of awakenings during the nights
`and total amount of sleep;
`Incident of hypnagogic hallucination;
`
`0
`
`Incident of sleep paralysis;
`o
`- Change in severity of the patients’ narcolepsy symptoms as measured by the Clinical
`Global Impression of Change
`
`

`

`2.5 Statistical Method
`
`2.5.1 Primary Efficacy Analysis
`
`The efficacy analyses were to be done on the intent—to—treat (ITT) population. The
`planned analyses called for an analysis of variance (ANOVA) on the change from
`baseline to endpoint including in the model the factors of treatment, site, and their
`interaction. The interaction term was then to be removed if found to be not statistically
`significant. In addition, an analysis of covariance (ANCOVA) was planned for the
`primary efficacy variable using the baseline value as a covariate. The factor of site was to
`be removed from this ANCOVA model, as indicated in the statistical analysis plan. An
`additional analysis would examine for a possible dose response relationship.
`
`It was stated in the Statistical Analysis Plan, documented July 1997, that the method of
`statistical analysis implied normal distribution assumption. If the data were not normal
`then the most suitable data transformation method (log, square root, etc.) was to be
`applied. If the assumption of normality and the data transformation did not appear to be
`satisfied, then appropriate non—parametric methods were to be used in the analyses.
`
`2.5.2 Analysis of Secondary Efficacy Variables
`
`All secondary measures of efficacy except CGI—c were to be analyzed in a similar way as
`to the primary efficacy variable.
`
`CGI-c was to be analyzed using Fisher's exact test and Cochran—Mantel—Haenszel (CMH)
`test for nonzero correlation.
`
`2.5.3 Center Grouping
`
`The study was to be conducted in 16 centers. In the event of any site(s) fail to reach a
`minimum of 8 patients, these sites were to be pooled and treated as a single site for
`purpose of statistical analysis.
`
`2.5.4 Interim Analysis
`
`No interim analyses were planned or performed for this study.
`
`2.6 Results: Sponsor’s Analysis
`
`2.6.1 Patient Disposition
`
`The sponsor reported that a total of 136 patients were enrolled from 16 centers. The
`number of patients entered by each center ranged from 1 to 21. Of the 136 patients
`enrolled, 16 withdrew from the study before completion. Disposition of patients is
`summarized in Table 2. The primary reason of withdrawal from the study was adverse
`
`

`

`experience (10 patients). Patient withdrawal for adverse events was more frequent in the
`9g GHB dose group than in the other 3 treatment groups.
`
`Table 2. Patient disposition
`
`Disposition
`
`Placebo
`
`Receive study medication
`Withdrew from the study
`Adverse event
`
`Protocol deviation
`
`Patient request
`Lost of follow—up
`Lack of efficacy
`Total Withdrawals
`
`Completed study
`
`34
`
`1
`
`1
`
`33
`
`GHB Dose
`6g
`
`33
`
`2
`
`1
`1
`4
`
`3g
`
`34
`
`1
`
`1
`
`1
`
`1
`4
`
`3O
`
`29
`
`9g
`
`35
`
`6
`
`1
`
`7
`
`28
`
`All Patients
`
`136
`
`10
`
`1
`
`2
`1
`2
`16
`
`120
`
`2.6.2 Patient Baseline and Demographic Characteristics
`
`Demographic characteristics of the 136 patients who received study medication are
`summarized in Table 3. Significant differences between groups in sex and height were
`noted by the sponsor. The 6g GHB group was predominantly male, while the placebo and
`3g GHB groups were predominantly femalet Consistent with the larger proportion of
`females in 3g GHB group, the mean height of this group was less than the other treatment
`groups. The sponsor reported that there were no difference between treatment groups in
`baseline blood pressures, pulse rates, respiration, or ECGs.
`
`AWE:5 52:3
`
`éllS‘éVAY
`
`0h; fitttGltlAL
`
`

`

`Table 3. Demographic characteristics by treatment group
`Characteristic
`Total
`Placebo
`GHB Dose
`6g
`
`3g
`
`34
`
`33
`
`47.06
`16.89
`
`43.52
`14.98
`
`40.91
`13.53
`
`7
`
`27
`
`33
`
`0
`
`0
`
`1
`
`33
`
`21
`
`12
`
`31
`
`1
`
`1
`
`0
`
`33
`
`17
`
`18
`
`31
`
`4
`
`0
`
`0
`
`34
`
`p—valueI
`
`0.2737
`
`9g
`
`35
`
`0.0027
`
`0.1379
`
`0.0283
`
`0.4847
`
`136
`
`43.06
`15.03
`
`34
`
`40.82
`14.33
`
`12
`
`22
`
`29
`
`4
`
`0
`
`1
`
`31
`
`57
`
`79
`
`124
`
`9
`
`1
`
`2
`
`131
`
`170.91
`9.53
`
`Age
`N
`
`Mean (year)
`SD
`
`Sex (n)
`Male
`
`Female
`
`Race (n)
`Caucasian
`
`African American
`
`Asian
`
`Other
`
`Height
`N
`
`Mean (cm)
`SD
`
`Weight
`N
`Mean (kg)
`SD
`
`171.97
`8.18
`
`166.7
`8.78
`
`173.1
`10.39
`
`171.9
`9.64
`
`p
`
`134
`82.87
`17.36
`
`34
`83.98
`18.89
`
`33
`78.86
`15.65
`
`33
`85.04
`15.54
`
`34
`83.56
`19.08
`
`1. p—value based on ANOVA (GLM)
`
`The severity of narcolepsy in the patient population was assessed by documenting the
`frequency of symptoms that were reported in the 3 months prior to screening. The
`sponsor reported that the distribution of patient's symptoms was similar across the 4
`treatment groups (see Table 4). Cataplexy, excessive daytime sleepiness, and inadvertent
`naps/sleep attacks were among the most common symptoms. When focusing on the key
`symptom of interest in this study, cataplexy, the sponsor reported that the 4 treatment
`groups were similar in the average number of cataplexy attacks per week over the last 3
`months reported by the patients at the screening visit. The mean numbers of the reported
`cataplexy attacks for the 4 treatment groups were 15.5 for the placebo group and 12.3,
`16.6, and 16.7 for the 3g, 6g, and 9g GHB group, respectively.
`
`Table 4. Number of patients reporting symptoms in the 3 months prior to screening
`
`Symptoms
`
`Number (%) of patient with symptoms
`Placebo
`3g GHB
`6g GHB
`9g GHB
`n=34
`n=34
`n=33
`n=35
`
`Cataplexy
`Excessive daytime Sleepiness
`Awakening at night
`Inadvertent naps/sleep attacks
`Sleep paralysis
`Hypnagogic hallucinations
`
`34 (100)
`34 (100)
`27 (79)
`32 (94)
`26 (76)
`27 (79)
`
`34 (100)
`34 (100)
`31 (91)
`33 (97)
`25 (74)
`29 (85)
`
`32 (97)
`32 (97)
`27 (82)
`31 (94)
`25 (76)
`26 (79)
`
`35 (100)
`34(97)
`30 (86)
`33 (94)
`24 (69)
`26 (74)
`
`

`

`A summary of baseline narcolepsy symptoms by treatment group is presented in Table 5.
`The sponsor noted that that the increase in cataplexy attacks at baseline (Visit 4) was due
`to the discontinuation of anti-cataplexy medication that was present at screening (Visit 1).
`The sponsor reported that the values of total, complete, and partial cataplexy attacks did
`not vary in a statistically significant manner across treatment groups, though it was noted
`that, in general, the median number of partial cataplexy attacks was almost 5 times
`greater than the median number of complete cataplexy attacks.
`
`Table 5. Summary of baseline (Visit 4) narcolepsy symptoms by treatment group
`Type of event
`Placebo
`GHB Dose
`p—value
`Kruskal-Wallis
`0.7749
`
`(n=34)
`
`3g (n=33)
`
`6g (n=33)
`
`9g (n=35)
`
`Total cataplexy/wk
`Mean (SD)
`Median
`
`Complete cataplexy/wk
`Mean (SD)
`Median
`
`Partial cataplexy/wk
`Mean (SD)
`Median
`
`Hypnagogic
`hallucination/day
`Mean (SD)
`Median
`
`Sleep paralysis/day
`Mean (SD)
`Median
`
`34.27 (46.63)
`20.21
`
`28.57 (30.53)
`20.00
`
`38.85 (55.04)
`23.00
`
`34.60 (33.92)
`23.50
`
`6.86 (12.37)
`1.12
`
`7.08 (8.50)
`4.50
`
`15.26 (27.53)
`4.85
`
`8.61 (14.01)
`2.00
`
`27.44 (42.08)
`15.03
`
`21.49 (28.30)
`15.00
`
`23.59 (29.01)
`16.15
`
`26.12 (26.14)
`18.79
`
`0.57 (0.74)
`0.23
`
`0.56 (0.68)
`0.43
`
`1.14 (3.72)
`0.33
`
`0.53 (0.70)
`0.29
`
`0.51 (0.74)
`0.26
`
`0.42 (0.55)
`0.14
`
`0.73 (1.84)
`0.08
`
`0.41 (0.60)
`0.10
`
`0.5151
`
`0.7289
`
`0.9766
`
`0.9597
`
`Inadvertent naps/day
`Mean (SD)
`Median
`Excessive day‘sleepI
`16.66 (4.07)
`17.26 (3.49)
`17.06 (3.71)
`18.47 (3.13)
`Mean (SD)
`
`
`0.7008
`
`Not reported
`
`1.71 (0.96)
`1.57
`
`1.91 (1.43)
`1.93
`
`1.70 (1.12)
`1.45
`
`1.72 (1.56)
`1.27
`
`1. By Epworth Sleepiness Scale
`
`2.6.3 Efficacy Evaluation
`
`The sponsor reported that at the time of analysis, each of the primary and secondary
`efficacy variables was assessed for normality and whether a log transformation would
`improve the distribution. This assessment was based on using Wilk—Shapiro test for
`normality on the residuals from the fitted model and a plot of the residuals against the
`predicted response, also from the fitted model. The following measures were analyzed
`using the log transformation:
`
`0 Total number of cataplexy attacks
`Partial cataplexy attacks
`Complete cataplexy attacks
`Duration of inadvertent naps/sleep attacks/day
`Sleep paralysis (episodes/day)
`Hypnagogic hallucinations
`
`

`

`0 Number of awakenings
`
`For each of these measures the value of 1 was added prior to the log transformation
`because 0 was possible. Therefore, the variable analyzed was log (endpoint+1)—log
`(baseline+1). The ANCOVA model used to assess overall treatment group comparisons
`include treatment, site, and log (baseline+1). The interaction of treatment and site and
`treatment with log (baseline+1) were included in the model and then removed when
`found to be not statistically significant. Comparisons of GHB dose to placebo were
`performed using least-square means with Dunnett's adjustment. The significance of the
`median change from baseline for each treatment group was assessed using a paired t-test.
`
`Several measures did show a normal distribution without a log transformation. They
`included
`
`0 Epworth Sleepiness Scale;
`
`_ 0 Total amount of sleep/night;
`
`0 Number of inadvertent naps/day.
`
`For these variables, the analysis procedures were consistent with those previously
`described, but were based on the untransformed values.
`
`The Clinical Global Impression of Change (CGI—c) was assessed by correlation analysis
`using Cochran-Mentel—Haenszel (CMH) test for Nonzero Correlation between the CGI—c
`score and treatment.
`
`2.6.3.1 Primary Efficacy Variable: Total Number of Cataplexy Attacks
`
`The primary efficacy variable was the change from baseline to the endpoint in the total
`number of cataplexy attacks per week. The sponsor reported that there was a significant
`difference (p=0.0021) among treatment groups in change from baseline to endpoint in
`total number of cataplexy attacks. The analysis was based on an ANCOVA model on log-
`transformed data including factors of treatment, site, and log (baseline+1).
`
`The sponsor reported that a dose response relationship in improvement was observed
`across all dosage groups. The change in the total number of cataplexy attacks exceeded
`placebo and was in a clinically meaningful range with all GHB treatment groups, was of
`marginal statistical significance in the 6g GHB group (p=0.0529) and was of
`unambiguous statistical significance in the 9g GHB group (p=0.0008). In a post hoc
`analysis, there was no evidence of rebound cataplexy on withdrawal of GHB at the end of
`the study.
`
`

`

`Table 6. Total number of cataplexy attacks
`
`Dose group
`Placebo
`
`Statistics
`N
`
`Observed
`Baseline
`Endpoint
`33
`33
`
`Change from
`baseline
`
`33
`
`p—value
`
`Mean (SD)
`Median
`
`35.1 (47.1)
`20.5
`
`24.0 (28.4)
`16.3
`
`-11.1 (27.7)
`-4.3
`
`3g
`
`6g
`
`9g
`
`p—value
`N
`Mean (SD)
`Median
`
`p—value
`N
`Mean (SD)
`Median
`
`p-value
`N
`Mean (SD)
`Median
`
`p—value
`
`33
`28.6 (30.5)
`20.0
`
`33
`19.5 (27.5)
`9.5
`
`31
`33.8 (45.6)
`23.0
`
`31
`24.6 (62.9)
`8.0
`
`33
`35.7 (34.5)
`2.5
`
`33
`14.4 (19.3)
`8.7
`
`0.028
`33
`—9.1 (22.4)
`—7.0
`
`0.026
`31
`-9.2 (27.3)
`-9.9
`
`0.070
`33
`—21.3 (29.8)
`—16.1
`
`<0.001
`
`0.5235
`
`0.0529
`
`0.0008
`
`The sponsor reported that placebo, 3g, 6g, and 9g GHB groups showed percent
`reductions in the median number of cataplexy attacks of -28%, —40%, -50%, and —55%
`after 2 weeks of treatment. There were no further reductions in the median number of
`
`cataplexy attacks from 2 to 4 weeks in the placebo and 6g GHB groups. The 3g and 9g
`GHB groups had further reductions in the median number of cataplexy attacks to -49%
`and —69% respectively. The overall pattern is that the greater part of the response in all
`groups was seen by 2 weeks with smaller changes by 4 weeks.
`
`2.6.3.2 Secondary Efficacy Analysis
`
`Complete cataplexy attacks
`
`The sponsor reported that complete cataplexy attacks were much less frequent than
`partial cataplexy attacks. At baseline the median number of complete cataplexy attacks
`was 1.2, 4.5, 4.7, and 2.0 in the placebo, 3g, 6g, and 9g GHB treatment groups,
`respectively. At the endpoint that number changed by 0, —1.00, -1.62, and -1.62 in the
`placebo, 3 g, 6g, and 9g GHB treatment groups, respectively. None of the decreases in the
`median number of complete cataplexy attacks reached statistical significance when
`compared to placebo.
`
`Partial cataplexy attacks
`
`The sponsor reported that from baseline to endpoint the median number of partial
`cataplexy attacks changed by —2.72, —3.69, -6.35, and -10.00 in the placebo, 3g, 6g, and
`
`

`

`9g GHB treatment groups, respectively. The difference between placebo and 9g GHB
`treatment groups carried a p—value of 0.0009 from the ANCOVA model.
`
`Excessive daytime sleepiness
`
`Excessive daytime sleepiness as assessed by Epworth Sleepiness Scale improved in all
`GHB treated groups, as reported by the sponsor. The overall treatment group comparison
`on the change from baseline to endpoint on excessive daytime sleepiness carried a p—
`value of 0.0006 while the comparison between 9g GHB group and placebo carried a p-
`value of 0.0001.
`
`Clinical Global Impression of Change gCGI—c)
`
`The CGI—c was a 7-point scaled measure of improvement/deterioration based on the
`clinical investigator's overall impression of the change in the patient's condition. The
`sponsor reported that a highly significant treatment effect was noted on the CGI-c scale
`(p:0.0010 from overall treatment group comparison based in Cochran—Mantel’Haenzel
`test). The following table presents the summary of Clinical Global Impression of Change
`at endpoint by treatment group.
`
`Table 7. Summary of Clinical Global Impression of Change at endpoint by treatment
`group
`GHB Dose
`
`Impression
`
`Placebo
`
`3g
`
`6g
`
`9g
`
`Very much improved
`
`3 (9%)
`
`3 (10%)
`
`5 (16%)
`
`11 (37%)
`
`Much improved
`
`8 (24%)
`
`11 (37%)
`
`11 (35%)
`
`13 (43%)
`
`Minimally improved
`
`8 (24%)
`
`9 (30%)
`
`9 (29%)
`
`3 ( 10%)
`
`No change
`
`12 (35%)
`
`6 (20%)
`
`5 (16%)
`
`Minimally worse
`
`2 (6%)
`
`1 (3%)
`
`Much worse
`
`0
`
`0
`
`0
`
`0
`
`1 (3%)
`
`2 (7%)
`
`0
`
`
`
`1 (3%) 0 1 (3%)Very much worse 0
`
`
`
`
`
`
`
`The CGI—c data was also analyzed by categorizing a patient as a responder or not. A
`responder was defined as a patient falling into the much improved or very much
`improved category. A p-value of 0.0014 was reported for the overall treatment group
`comparison based on Fisher's exact test. A p—value of 0.0002 was reported by comparing
`9g treatment with placebo group on Fisher's exact test.
`
`Other Secondary Measures of Efficacy
`
`10
`
`

`

`The sponsor reported that compared with placebo, a significant decrease in the number of
`inadvertent naps/sleep attacks was seen in both 6g and 9g GHB groups (p:0.0497 and
`p=0.0122 respectively). A significant decrease in the number of awakenings was seen in
`the 9g GHB group (p:0.0035). No significant differences between treatments were seen
`in the change from baseline of hypnagogic hallucinations, sleep paralysis episodes, total
`amount of sleep, and duration of inadvertent naps/sleep attacks.
`
`2.7 Reviewer’s Analysis
`
`2.7.1 Primary Efficacy Analysis
`
`The primary efficacy measure is the change from baseline to endpoint in the total number
`of cataplexy attacks. An ANOVA model including factors of treatment, site, and
`treatment-by-site interaction was applied. It was found that the assumption of normality
`was violated (p:0.0017), indicating that the ANOVA might not be appropriate for the
`analysis.
`
`As specified in the analysis plan, a log transformation was employed to endpoint as well
`as to baseline. The dependent variable of log (endpoint+1) - log (baseline+1) was used in
`the analysis. After the log transformation, the normal assumption was no longer violated
`(p:0.7365; Shapiro-Wilk). The treatment-by-site interaction was not statistically
`significant, and was removed from the model. The final model contained factors of
`treatment and site showed a significant treatment effect with a p-value of 0.0034. The
`effect of site was also significant with a p-value of 0.0321. The descriptive statistics of
`the change from baseline in cataplexy attacks were the same as obtained by the sponsor
`(Table 6) except for the 9g GHB group, where the patient #0824 was included in the
`reviewer's analysis. The mean, SD, and median number for the 9g GHB group obtained
`by this reviewer were —20.5, 29.7, and -15.6, respectively.
`
`Note that the p—value of 0.0034 for the treatment effect obtained by this reviewer is
`different from the one reported by the sponsor (p:0.0021), although both p—values
`reached significance level of 0.05. The difference is due to the following reasons:
`1) The data submitted in this NDA contained patient #0824, which was not contained in
`the data set in the original treatment IND. The reviewer's analysis was based on this
`new data set of 131 patients, while the sponsor's analysis was based on the data set of
`130 patients that did not include patient #0824.
`2) The protocol and analysis plan called for an ANOVA with factors of treatment and
`site. An additional analysis of ANCOVA would include baseline value as a covariate,
`but would exclude site as a factor. However, the sponsor used an ANCOVA model
`that included both site and baseline value in the primary analysis.
`
`Dose Response Relationship
`
`The dosage of GHB studied in this trial included 3 dose levels: 3g, 6g, and 9g. The
`significance level of the effect of each dose as compared to placebo was to be adjusted by
`Dunnett multiplicity adjustment. Among the 3 dose levels of GHB, 9g reached
`
`11
`
`

`

`significance level by Dunnett adjustment (p=0.0021). Dose groups of 3g and 6g did not
`reach the significance level (p=0.6358 for GHB 3g and p=0.0772 for GHB 6g). The 95%
`confidence interval for the mean difference of the dependent variable as compared to
`placebo was (—0.6658, 0.2755) for GHB 3g, (-0.9196, 0.0368) for GHB 6g, and (-1.1478,
`-0.2134) for GHB 9g, given by the Dunnett adjustment.
`
`Summary Statistics by Demographic and Baseline Characteristics
`
`Means and medians of the change in the number of cataplexy attacks were calculated for
`the subgroups of gender, age, and baseline total cataplexy attacks (Table 8). Median age
`of 42 years and median baseline cataplexy attacks of 21 were used as cutoff points. There
`were no substantial discrepancies between the subgroups of gender and age in the change
`of cataplexy numbers. Patients who had fewer cataplexy attacks at baseline appeared to
`have less change in the number of cataplexy attacks during the treatment as compared to
`patients who had more than 21 cataplexy attacks per week at baseline.
`
`Table 8. Means and medians of the change in the number of cataplexy attacks by
`subgroups of sex, age, and baseline number of cataplexy attacks.
`
`Characteristics
`
`Sex
`
`Male
`
`Female
`Age1 (years)
`<=42
`> 42
`Baseline # of attacks2
`
`Change in the Number of Cataplexy Attacks
`Mean (Median)
`3g GHB
`6g GHB
`
`9g GHB
`
`Placebo
`
`-10.8 (0.2)
`
`-11.3 (—5.5)
`
`—9.5 (—5.0)
`
`—8.9 (-7.1)
`
`-7.0 (-10.6)
`
`—27.7 (—16.8)
`
`-13.3 (—9.7)
`
`—13.4 (—8.8)
`
`'
`
`—13.0 (-3.7)
`—8.3 (—4.5)
`
`-8.7 (-7.0)
`—9.2 (-5.1)
`
`-7.9 (—6.4)
`—10.6 (—12.3)
`
`—18.2 (—11.3)
`—23.2 (—l6.8)
`
`<=21
`
`> 21
`
`-l.4 (—3.2)
`
`-4.1 (—3.7)
`
`-7.5 (—8.3)
`
`-5.6 (-4.3)
`
`-24.3 (-19.4)
`
`—15.1 (—17.5)
`
`-10.9 (-17.9)
`
`-33.8 (-31.0)
`
`1. Median age of 42.5 was used as a cutoff point;
`2. Median number of 21 was used as a cutoff point.
`
`2.7.2 Analysis of Secondary Efficacy Variables
`
`A total of 10 secondary efficacy measures were specified for this study. Among those 10
`efficacy variables, number of complete cataplexy attacks, number of partial cataplexy
`attacks, daytime sleepiness as measure by Epworth sleepiness scale, and CGI—c were
`analyzed by this reviewer. The results of these analyses are discussed in this section. A
`table of summary is presented at the end of this section.
`
`Number of Complete and Partial Cataplexy Attacks
`
`The number of complete and partial cataplexy attacks were analyzed separately using the
`same method as to the total number of cataplexy attacks. In both analyses of complete
`and partial cataplexy attacks, the normal assumption for the parametric ANOVA model
`
`12
`
`

`

`did not hold and log transformations were applied. Using log (endpoint+1) — log
`(baseline+1) as the dependent variable, the normal assumption was no longer violated in
`the ANOVA model for analyses of both complete and partial cataplexy attacks.
`
`From the ANOVA model based on the log transformed data the treatment-by-site
`interaction was not significant, and was removed from the analyses of complete and
`partial cataplexy attacks.
`
`For the number of complete cataplexy attacks, treatment effect carried a p—‘value of
`0.2362. A p-value of 0.0204 was obtained from the analysis of partial cataplexy attacks.
`In both analyses, the effect of center was significant.
`
`Daflime Sleepiness Measured by Epworth Sleepiness Scale
`
`Daytime sleepiness, as measured by Epworth Sleepiness Scale, was to be analyzed
`similarly to the primary endpoint. Original scale was used first. The interaction of
`treatment-by-site was found not significant. After removing the interaction term from the
`model, the normal assumption was no longer held by the residuals. The log
`transformation made no improvement in the normality of the distribution and non-
`parametric method of Kruskal—Wallis test was employed. A p-value of 0.0109 was
`obtained from the Kruskal—Wallis test for the treatment effect across 4 treatment groups.
`
`Clinical Global Impression of Change gCGI-cg
`
`CGI—c was to be analyzed by Fisher's exact test and Cochran-Mantel-Haenszel (CMH)
`test for nonzero correlation, as specified in the protocol. Due to the large size of the
`categorical table, the Fisher's exact test was not applicable, and CMH test for nonzero
`correlation was employed. A p—value of 0.322 was obtained, indicating that the
`hypothesis of no association between treatment groups and CGI-c values was not
`rejected.
`
`
`Table 9. Summary of secondary efficacy results (not including CGI-c)
`
`Variable
`
`Complete attacks
`
`Placebo
`-3.0 (0)
`
`Change from Baseline
`Mean (Median)
`3g GHB
`6g GHB
`—2.3 (-1)
`-3.9 (-1.6)
`
`9g GHB
`-6.5 (-1.6)
`
`p-value
`
`0.2362
`
`Partial attacks
`
`-8.1 (—2.7)
`
`—6.7 (-3.7)
`
`—5.3 (-6.3)
`
`—14.0 (—9.9)
`
`0.0204
`
`Epworth
`Slee iness Scale
`
`
`—2.4 (—2)
`~25 (-1)
`-1.2 (~1)
`
`
`—4.5 (—3)
`
`0.0109
`
`Note that the p-values obtained by this reviewer in the analyses of above secondary
`efficacy measures are different from those reported by the sponsor due to the differences
`in the data set and analysis method as described in Section 2.7.1. For results of
`inadvertent naps and number of awakenings, the p-values reported by the sponsor were
`
`13
`
`

`

`not from comparisons across 4 treatment groups. Analyses for these two variables were
`not performed by this reviewer.
`
`2.7.3 Center Effect
`
`The study was conducted in 16 centers. The number of patients enrolled in each center
`ranged from 1 to 21. For the purpose of statistical analysis, eight centers that had less
`than 8 patients were pooled together to form a larger center with 30 patients altogether,
`which was planned in the protocol. In the analysis of primary efficacy variable and two
`secondary efficacy variables (number of complete and partial cataplexy attacks), center
`effect was significant. Summary statistics of change in the number of cataplexy attacks
`within each center were explored. No evidence was found that efficacy was driven by any
`individual center(s).
`
`2.8 Reviewer's Conclusion for Study OMC-GHB-Z
`
`In this study 136 patients were enrolled and randomized into 3g, 6g, and 9g GHB and
`placebo treatment groups. The primary efficacy variable was the change from baseline in
`the total number of cataplexy attacks. Based on the protocol defined primary analysis, a
`p-value of 0.0034 was obtained by this reviewer for the overall treatment effect.
`Therefore, this reviewer would conclude that the study has provided sufficient evidence
`that GHB, given in the dose of 3g, 6g, and 9g, was efficacious in treating patients with
`symptoms of narcolepsy.
`
`Although the study has shown the efficacy of GHB successfully, it was not clear which
`dose level would provide most benefit with limited risk. The 9g GHB was considered as
`the high end of tolerable dose, while the 6g GHB failed to show a significant treatment
`effect. The safe and effective dose has yet to be detemrined.
`
`A total of 10 secondary efficacy measures were specified in the protocol. None of the
`differences between treatment groups reached significance level when adjusted by
`Bonferroni procedure. Therefore, this reviewer would conclude that the evidence of the
`treatment effect with regard to any of the secondary efficacy measures was not sufficient.
`
`3. Specifications and Findings of Lammers’ Study
`
`The study was carried out by Dr. G}. Lammers, neurologist in training, together with
`other investigators. Dr. Lammers has prepared, in collaboration with the Dutch
`Narcolepsy Association, this double—blind study to investigate gammahydroxybutyrate.
`Polysomnography was analyzed at the I W Clinical monitoring was
`done in ”
`
`Orphan Medical analyzed and reported the Lammers trial results to the FDA in October
`1998 as part of the package for approval of a Treatment 1ND. The Lammers data were
`analyzed in two different ways. The analysis using investigator-designated methodology
`
`14
`
`

`

`failed to reach the significance level for cataplexy. However, the primary analysis method
`proposed in the protocol was not appropriate for the crossover design. The sponsor also
`analyzed data employing the same methodology used in study OMC—GHB—Z. Although it
`showed a significant treatment effect, the sponsor acknowledged that the analysis was not
`prospectively defined This reviewer‘s analysis using a similar model showed a contrary
`result. The Lammers study was submitted as part of this NDA.
`
`3.1 Trial Objectives
`
`The purpose of the trial was to investigator whether:
`
`0 Gammahydroxybutyrate, in a double—blind study, has an effect on the REM
`dissociation phenomena and possibly on the elevated inclination to sleep during the
`day;
`
`0 Gammahydroxybutyrate affects alertness during the day; and
`
`o Gammahydroxybutyrate has a mood-improving effect.
`
`It was noted that the second objective listed above was not investigated following the
`discussions among the investigators concerning the difficulties in defining and measuring
`the al