CENTER FOR DRUG EVALUATION AND RESEARCH
`
`APPROVALPACKAGEFOR:
`
`APPLICATION NUMBER
`
`21-372
`
`Statistical Review(s)
`
`

`

`STATISTICAL REVIEW AND EVALUATION
`
`CLINICAL STUDIES
`
`NDA/Serial Number:
`
`21- 372 /N000
`
`Drug Name:
`
`lndication(s):
`
`Palonosetron HCl 0.25mg/5ml
`
`Prevention ofacute and delayed nausea and
`vomiting associated with initial and repeated
`courses of emetogenic cancer chemotherapy
`
`Applicant:
`
`'
`
`Helsinn Healthcare S.A.
`
`Dates:
`
`Submitted: 26 September 2002
`
`Review Priority:
`
`Standard (10—month)
`
`Biometrics Division:
`
`Division of Biometrics 2 (HFD-715)
`
`Statistics Reviewer:
`
`Stella Grosser, PhD.
`
`Concurring Reviewers:
`
`Thomas Perrnutt, PhD.
`
`Medical Division:
`
`Division of Gastrointestinal and Coagulation Drug
`Products (HFD-l 80)
`
`Clinical Team:
`
`Narayan Nair, M.D.
`
`Project Manager:
`
`Brian Strongin, M.B.A.
`
`Keywords: minimization, active control tn'als, non-inferiority, meta-analysis.
`
`

`

`1. Executive Summary
`
`1.1 Conclusions and Recommendations
`
`There is sufficient evidence and reasonable certainty that palonosetron 0.25 mg is
`. efficacious in the prevention of acute nausea and vomiting following moderately and
`highly emetogenic cancer chemotherapy. There is also sufficient evidence that it is
`efficacious in the prevention of delayed emesis following moderately (but not highly)
`emetogenic chemotherapy. While the efficacy analyses are based on comparisons to
`approved anti-emetics (ondansetron and dolasetron), the efficacy conclusions and claims
`are relative to placebo; the label should reflect this distinction.
`
`1.2 Brief Overview of Clinical Studies
`
`The applicant proposes a single, intravenous injection of palonosetron 0.25 mg, given
`30 minutes prior to moderately or highly emetogenic chemotherapy. Eighteen clinical
`trials were conducted to study the safety and efficacy of palonosetron. Of these, four are
`presented in support ofthe applicant’s claim of efficacy ofpalonosetron 0.25 mg IV to
`prevent chemotherapy-induced nausea and vomiting (CINV) and are reviewed here. Two
`are for the prevention of CINV following moderately emetogenic chemotherapy (PALO-
`99-03 and PALO-99-04) and two are for the prevention of CINV following highly
`emetogenic chemotherapy (2330/PALO-00-01 and PALO-99-O3).
`
`Studies PALO-99-03, 99-04, and 99-05 were double—blind, multicenter, active—controlled
`
`studies enrolling 570, 592 and 680 patients respectively. They were conducted in
`Europe, including Russia, (99-03 and 99—05), and North America (99-04 and 99—05).
`Each study had three arms: 0.25 mg IV palonosetron, 0.75 mg IV palonosetron, and an
`active comparator (ondansetron 32 mg IV in 99-03 and 99-05, dolasetron 100 mg IV in
`99-04). Allocation to treatment was a mixture of algorithms primarily relying on
`minimization rather than randomization. That is, the assignment ofa new patient to a
`group was made to minimize differences among the treatment groups. Balance among
`the groups was in terms of the number of patients assigned to each stratum defined by
`prognostic criteria of gender, chemotherapy history (na'i've or not naive) and use of
`corticosteroids. This scheme does not correspond to what is usually thought of as
`randomization in a clinical trial.
`It most closely resembles a deterministic dynamic
`allocation procedure.
`
`It
`Study 2330 was designed as a phase 2 study using the IV formulation ofpalonosetron.
`was a randomized, double-blind, multicenter,.dose-ranging trial of palonosetron given to
`chemotherapy-naive patients 30 minutes before the administration of highly emetogenic
`chemotherapy. The enrolled population consisted of l 6] subjects. Palonosetron was
`administered at weight—based doses of 0.3, 1, 3, 10 or 30 ug/kg. Helsinn considers study
`2330 supportive. It was a dose-ranging study conducted by the drug innovator Syntex. It
`used a weight-based dosing regimen, which was roughly translated into the eventual
`(fixed) dosing regimen.
`
`

`

`1.3 Statistical Issues and Findings
`
`A primary concern from a statistical point of view is the minimization allocation
`procedure used in studies PALO-99-03, PALO-99-04 and PALO-99~05. It is not
`randomization, but rather a deterministic allocation with the occasional random
`
`assignment. Several drawbacks of using minimization have been cited in the literature
`(Scott et al., 2002). The concern in this application is that standard statistical tests, or,
`equivalently, confidence interval calculations, make the assumption of random allocation:
`more generally, “the correct statistical analysis is complex and not yet clearly worked
`out.” (Scott et al., 2002) Permutation methods can be used to check the results of
`standard analyses. The two approaches are likely but not guaranteed to yield similar
`conclusions; there are situations where the standard methods are very misleading. These
`situations have not been completely characterized and a permutation test is a good way to
`know whether the trials in this application fall into the problematic case. Apparently they
`do not: The results ofthe permutation analysis are in accordance with the primary,
`standard analysis.
`
`None ofthe efficacy trials done as part ofthis application included a placebo control. To
`assess trial validity and justify the value of delta used to declare non-inferiority of
`palonosetron to ondansetron or dolasetron, an examination and meta-analysis of results
`from the anti-emetic literature was carried out. In the few studies where ondansetron or
`
`dolasetron was directly compared to placebo, the active treatment reliably out-performed
`placebo to a greater extent than seen between treatments in the trials in'this application.
`A less direct comparison of the effects of setron treatments and placebo, achieved
`through logistic regression modeling by the applicant, yielded similar results and similar
`confidence in the assay sensitivity ofthe NDA studies. The magnitudes ofthe
`differences found or modeled in the meta-analysis also were large enough to justify a
`conclusion of non-inferiority ofpalonosetron in the current trials.
`
`In studies PALO-99-O3, 99-04, and 99-05, a higher proportion ofthe patients responded
`to palonosetron than to the comparator anti-emetics. Response rates ranged from a low
`of 57%, for ondansetron 32 mg following the administration of highly emetogenic
`chemotherapy, to a high of 81% for palonosetron 0.25 mg following moderately
`emetogenic chemotherapy
`
`The applicant calculated the two-sided 97.5% confidence interval of the difference
`between the proportions of complete response in each dose of palonosetron and
`comparator (calculated as palonosetron minus comparator) to demonstrate non-inferiority
`of palonosetron to the comparators. In all cases, the lower boundary of the interval was
`above -] 0%, implying a reasonable certainty that the proportion of complete responders
`to palonosetron was no less than 10% less than the proportion among the comparators.
`Results of the permutation test confirmed these conclusions.
`
`The applicant wishes to include a secondary outcome as part of the labeled indication,
`namely that palonosetron is effective for prevention of delayed nausea and vomiting.
`Following highly emetogenic chemotherapy (PALO-99-05), the rates of complete
`
`

`

`response are consistently numerically higher for palonosetron 0.25 mg; however, there is
`no time period for which palonosetron is statistically significantly higher than the
`comparator ondansetron (as judged by the lower limit ofthe confidence interval ofthe
`difference). Following moderately emetogenic chemotherapy, the rates of complete
`response again are consistently numerically higher for palonosetron 0.25. It is
`statistically significantly higher than ondansetron at all time periods other than the final
`96-120 hours, when there are high response rates in all three treatment arms; its
`performance against dolasetron is mixed, but is statistically significantly higher than for
`the overall time period 24-120 hours.
`
`The results for the primary efficacy outcome for study'PALO-OO—Ol (essentially the same
`as study 2330) support the choice of 0.25 mg as a threshold efficacy dose and confirm the
`results of 99-05 for highly emetogenic chemotherapy.
`
`

`

`2. Introduction
`
`2.1 Overview
`
`Palonosetron is a 5-HT3 (serotonin) receptor antagonist, described by the applicant as
`structurally unrelated to other currently available 5—HT3 receptor antagonists. This
`application was filed in support of palonosetron for “prevention of acute and delayed
`nausea and vomiting associated with initial and repeated courses of emetogenic cancer
`chemotherapy, including highly emetogenic chemotherapy.” (Other commercially
`available 5-HT3 receptor antagonists used as anti-emetic therapies include ondansetron,
`granisetron, and dolasetron).
`
`The applicant proposes a single, intravenous (IV) injection of palonosetron 0.25 mg,
`given 30 minutes prior to moderately or highly emetogenic chemotherapy. (Moderately
`emetogenic chemotherapy includes carboplatin, cisplatin_< 50 mg/m2, cyclophosphamide
`< 1500 mg/mz, doxorubicin > 25 mg/m2 epirubicin, irinzotecan, methotrexate > 250
`niog/m2',high1y emetogenic includes cisplatin > 50 mg/m2, cyclophosphamide >1500
`mg/m2, and dacarbazine.) Eighteen clinical trials, including healthy volunteer,
`pharmacokinetic, dose--ranging, controlled or open label studies, were conducted to study
`the safety and efficacy of palonosetron. Of these, four controlled studies are presented in
`support ofthe applicant’s claim of efficacy of palonosetron 0.25 mg IV to prevent
`chemotherapy-induced nausea and vomiting (CINV) and are reviewed here. Two are for
`the prevention of CINV following moderately emetogenic chemotherapy (PALO-99-03
`and PALO-99-04) and two are for the prevention of CINV following highly emetogenic
`chemotherapy (2330/PALO-00-Ol and PALO-99—03). Helsinn considers study 2330
`supportive. It was a dose-ranging study conducted by the drug innovator Syntex, and
`used a weight-based dosing regimen, which was roughly translated into the eventual
`(fixed) dosing regimen.
`
`2.2 Data Sources
`
`Materials reviewed included NDA paper volumes 1, 273- 371, amendment #006 dated 24
`April 2003, amendment #009 dated 16 June 2003, and data sets in CDER’5 electronic
`document room
`
`3. Statistical Evaluation
`
`3.1 Evaluation of Efficacy
`
`Because-studies PALO-99-03, ~04, -05 were very similar in purpose, design and analysis,
`these studies will be discussed together. Study 2330/PALO 00-01 is discussed separately
`
`3.1.1 Studies PALO-99-03, PALO-99-04, PALO-99-05
`
`

`

`Design
`
`Studies PALO-99-O3, 99-04, and 99-05 were double—blind, multicenter, active-controlled
`studies enrolling 570, 592 and 680 patients respectively. They were conducted in
`Europe, including Russia, (99-03 and 99—05), and North America (99-04 and 99—05).
`Each study had three arms: 0.25 mg IV palonosetron, 0.75 mg IV palonosetron, and an
`active comparator (ondansetron 32 mg IV in 99-03 and 99-05, dolasetron 100 mg IV in
`99-04).
`In 99-03 and 99-04, these treatments were administered following moderately
`emetogenic chemotherapy and in 99-05 following highly emetogenic chemotherapy.
`
`Dexamethasone 20 mg IV administered once prior to chemotherapy was allowed as a
`concomitant medication. However, due to late implementation of this practice as a
`protocol change, none of the patients in 99-03 and only 6% in 99-04 received
`concomitant steroids, while about 60% in 99-05 did. Other corticosteroids - 20 mg oral
`dexamethasone and 125 mg IV methylprednisolone — were also allowed due to a
`dexamethasone shortage in the United States.
`
`Allocation to treatment was a mixture of algorithms primarily relying on minimization
`rather than randomization. That is, the assignment of a new patient to a group was made
`to minimize differences among the treatment groups. Balance among the groups was in
`terms of the number of patients assigned to each stratum defined by prognostic criteria of
`gender, chemotherapy history (na'r‘ve or not naive) and use of corticosteroids. The initial
`allocation and final allocation schemes differ in their calculations of imbalance, with the
`first following the method outlined by Taves and the final that of Pocock and Simon.
`Scott et al., 2002; minutes of 22 May 2003 telecon with applicant).
`
`Ifa new patient could equally well be assigned to either oftwo treatment arms, or in the
`case of on-site shortage of assigned treatment, an assignment was made at random. All
`assignments took into account the stratification criteria of gender, chemotherapy history
`(na'r've or not naive) and use of corticosteroids.
`
`This scheme does not correspond to what is usually thought of as randomization in a
`clinical trial. It most closely resembles a deterministic dynamic allocation procedure,
`described in ICH E9 as an approach that “should be avoided.” (ICH E9, p.10)
`
`The applicant’s description of the allocation procedure can be found in the Appendix of
`this review.
`
`The evaluable, intent-to-treat (ITT) population was defined as all randomized subjects
`who received chemotherapy and study medication, and consisted of 563 patients in
`PALO-99-03, 569 in PALO-99—03, and 667 in PALO-99-05. The numbers in each of the
`prognostic strata are given in Table l. The applicant attributes the high proportion of
`female patients to the type of cancer for which moderately emetogenic chemotherapy is
`most frequently given, namely breast cancer (v. 273, p 237)
`
`

`

`Table 1. Numbers of patients in prognostic subgroups
`
`_
`
`Gender
`Female Male
`
`Chemotherapy
`Na'r've
`Non-naive
`
`Corticosteroid use
`Yes
`No
`
`N161
`
`N (‘70)
`
`N (‘70)
`
`N (‘70)
`
`N (W
`
`N (°/o)
`
`N (%)
`
`189
`189
`185
`
`189
`189
`191
`
`223
`223
`221
`
`135 (71)
`138 (73)
`133 (72)
`
`54 (29)
`51 (27)
`52 (28)
`
`76 (40)
`80 (42)
`78 (42)
`
`113 (60)
`109 (58)
`107 (58)
`
`0
`0
`0
`
`189
`189
`185
`
`155(82)
`156(83)
`156(82)
`
`34(18)
`33(17)
`35 (.18)
`
`124(66)
`131(69)
`125 (65)
`
`65(34)
`58(31)
`66(35)
`
`11 (6)
`12(6)
`8
`(4)
`
`178 (94)
`177 (94)
`183 (96)
`
`115 (52)
`113(51)
`113 (51)
`
`108 (48)
`110(49)
`108 (49)
`
`133 (60
`129 (58)
`131 (59)
`
`90 (40)
`94 (42)
`90 (41)
`
`150 (67)
`150 (67)
`147 (67)
`
`73 (33)
`73 (33)
`74 (33)
`
`Study and
`treatment arm
`
`PALO-99-03
`
`Palonosetron 0.25
`Palonosetron 0.75
`Ondansetron 32
`
`PALO—99-04 ,
`Palonosetron 0.25
`Palonosetron0.75
`DolasetronlOO
`PALO—99-05
`
`Palonosetron 0.25
`Palonosetron 0.75
`Ondansetron 32
`
`Analysis
`
`Primary efficacy outcome
`
`The primary efficacy outcome in all three studies was the proportion of subjects
`considered to have achieved a complete response (CR), defined as no emetic episode and
`no rescue medication, during the first 24 hours after administration of chemotherapy.
`Subjects with partially or completely missing data for the primary outcome were
`classified as not having a complete response.
`
`The primary efficacy hypothesis was that at least one dose of palonosetron was non—
`inferior to the comparator dose, using a maximum delta of 15%. To demonstrate this
`non-inferiority, the lower bound of the two-sided 97.5% confidence interval ofthe
`difference between the proportions of complete response at 24 hours in each dose of
`palonosetron and comparator (calculated as palonosetron minus comparator) was
`compared to the pre-set threshold of —1 5%.
`
`In addition, the applicant calculated 95% confidence intervals for the proportion of
`responders in each treatment group and compared these intervals between the two doses
`of palonosetron.
`
`The applicant assessed assay sensitivity and “confirm[ed] the value of delta” (v. 297,
`p.61) using the results of study PALO-Ol-23, a meta-analysis of historical data. This
`meta-analysis is described separately, in section 3.1.2 below.
`
`The applicant “check[ed] if the treatment allocation procedure described
`[above]. . .worked correctly ” by performing permutation tests on the primary outcome.
`
`

`

`Specifically, in the original NDA submission the applicant performed the following
`procedure: (v. 297, p.61)
`
`...the proportion of complete responders in the observed trial was compared between
`treatment groups by a one-sided Fisher‘s exact test taking delta into account. For
`Fisher’s permutation test a random sample (n=30,000) of all possible permutations was
`used for construction of the permutation distribution. The probability attached to the null
`hypothesis (i.e., p-value) was calculated as follows: (number of the same or more extreme
`outcomes as that observed)/30,000. This probability was compared to the p—value of
`Fisher’s exact test.
`
`In review, the nature ofthe allocation procedure was identified as a crucial element ofthe
`validity ofthe efficacy analysis in these trials (see discussion ofthis issue in section 5.1
`below). Moreover, the permutation test as originally performed by the applicant did not
`adequately address this issue. The appropriate permutation test should take into account
`the actual allocation scheme used; the set of potential permutations of outcome values
`should be restricted to those that result from simulations ofthe trial as it actually
`occurred. That is, potential permutations are those that result from the observed
`enrollment sequence of patients, with their fixed values of prognostic factors. A potential
`permutation would correspond to an allocation sequence based on a random assignment
`ofthe first enrollee, deterministic assignment of subsequent enrollees based on prognostic
`factors and the calculation of imbalance between treatment arms, until a tie occurs, then a
`random assignment of the patient who creates that tie, then detenninistic assignment with
`imbalance scores recalculated until a tie occurs, etc. This revision ofthe confirmatory
`permutation test was agreed upon in a teleconference with the applicant on 22 May 2003.
`
`Secondary eflicacy outcomes
`
`Secondary outcomes in these studies included
`
`-
`
`-
`
`-
`
`-
`
`-
`
`-
`
`-
`
`The proportion of subjects with a complete response evaluated on a daily basis and
`during the overall 0 to 48, 0 to72, 0 to 96, 0 to 120, and 24 to lZO-hour time periods;
`these were analyzed using the same statistical methods as for the primary efficacy
`parameter.
`The proportion of subjects with complete control (complete response and no more
`than mild nausea) evaluated daily and for the overall 0 to lZO-hour interval; analyzed
`with a Pearson chi-squared test.
`number of emetic episodes daily for the 0 to 120-hour interval and for the overall 0 to
`lZO-hour interval; analyzed with a Wilcoxon or Kruskal-Wallis test.
`time to first emetic episode; Kaplan-Meier estimates were calculated and a log-rank
`test used to compare treatments.
`severity of nausea measured on a Likert scale daily for the O to lZO-hour interval;
`analyzed with a Wilcoxon or Kruskal-Wallis test.
`need and time of administration of rescue medication; proportion receiving rescue
`medication analyzed using a chi-squared test, and time to administration was
`analyzed with Kaplan-Meier estimates and a log-rank test.
`time to treatment failure (emetic episode or rescue medication); Kaplan-Meier
`estimates were calculated and a log~rank test used to compare treatments.
`
`

`

`-
`
`subject VAS of global satisfaction with anti-emetic therapy daily for the 0 to 120-
`hour interval; and quality of life measured (twice) by the Functional Living Index-
`Emesis (FLIE) questionnaire; analyzed with a Wilcoxon or Kruskal-Wallis test.
`
`Results
`
`All results presented below are from analyses done by the applicant, unless otherwise
`indicated.
`
`Primary eflicacy outcome
`
`The proportion of patients achieving a complete response during the first 24 hours after
`chemotherapy is shown below in tables 2a-c. Following that, in table 3, are the 97.5%
`confidence intervals for the difference between the palonosetron doses and the
`comparators in complete response rates during the first 24 hours after chemotherapy.
`
`Table 2a. Proportion of patients achieving a complete response (CR) during the
`first 24 hours after chemotherapy, study PALO-99—03
`
`Palonosetron 0.25
`I Ondansetron 32 Palonosetron 0.75
`
`
`
`Proportion
`Cl
`Proportion
`Cl
`Proportion
`Cl
`
`75,86
`153/189
`139/189
`67, 80
`127/185
`61,75
`(81%)
`(74%)
`(69%)
`
`
`
`
`
`Table 2b. Proportion of patients achieving a complete response (CR) during the
`first 24 hours after chemotherapy, study PALO—99-04
`
`Palonosetron 0.25
`
`Dolasetron 100
`
`Palonosetron 0.75
`
`
`
`Proportion
`Cl
`Proportion
`Cl
`
`119/189
`56, 70
`(63%)
`
`.108/189
`(57%)
`
`
`
`50,64
`
`CI
`
`46, 60
`
`Proportion
`
`101/191
`(53%)
`
`
`
`Table 2c. Proportion of patients achieving a complete response (CR) during the
`first 24 hours after chemotherapy, study PALO-99-05
`
`Palonosetron 0.25 mg
`
`Palonosetron 0.75 mg
`
`Ondansetron 32 mg
`
` Proportion
`Cl
`Proportion
`C1
`Proportion
`Cl
`132/223
`52, 66
`146/223
`59, 72
`126/221
`50, 64
`(59%)
`(66%)
`(57%)
`
`

`

`A higher proportion of the patients responded to palonosetron than to the comparator
`anti-emetics. The response rates in 99-04 were higher in all treatment arms than in 99-03,
`although both sets ofpatients received moderately emetogenic chemotherapy. The two
`studies differed primarily in the geographic location of the centers; response rates were
`higher in Eurépe. Also, in both of 99-03 and 99-04, response rates at the lower dose of
`palonosetron were higher than at the higher one.
`I was not able to find an explanation for
`this.
`
`Table 32!. Confidence intervals for the difference between the palonosetron doses and the
`comparators in complete response rates during the first 24 hours after chemotherapy,
`standard analysis
`
`Palonosetron
`0.25 vs.
`Ondansetron
`
`Palonosetron
`0.75 vs:
`Ondansetron
`
`Palonosetron
`0.25 vs.
`Dolasetron
`
`Palonosetron
`0.75 vs.
`Dolasetron
`
`2, 23
`
`-9, l3
`
`-6, 16
`
`-2, 19
`
`-2, 22
`
`-8, 16
`
`Study
`
`99-03
`99—04
`99-05
`
`In all cases, the lower boundary ofthe 97.5 % CI was above -10% , implying a
`reasonable certainty that the proportion of complete responders to palonosetron was no
`more than 10% less than the proportion among the comparators. The lower boundary of
`the confidence interval for the difference between palonosetron 0.25 and ondansetron in
`99-03 is above zero, which the applicant takes as evidence of the superiority of
`palonosetron 0.25 to ondansetron in the treatment of acute nausea and vomiting following
`moderately emetogenic chemotherapy.
`
`Confirmatory permutation analysis
`
`For each of studies PALO-99-03, -04 and —05, the applicant computed a permutation
`distribution of the difference in response rates between each dose ofpalonosetron and the
`comparator. This distribution was based on 30,000 simulations of the trial; each
`simulation represented treatment assignments possible under the allocation scheme used
`and patient arrival sequence observed. The 2.5‘h and 97.5lh percentiles of this
`distribution, added to the point estimate (the observed difference), represent the ends of
`the 95% confidence interval around the observed difference.
`I calculated these intervals,
`given in table 3b. The results are in accordance with the primary, standard analysis.
`
`10
`
`

`

`Table 3b. Confidence intervals for the difference between the palonosetron doses and
`the comparators in complete response rates during the first 24 hours after chemotherapy,
`permutation analysis
`
`Palonosetron
`‘ 0.25 Vs.
`
`Palonosetron
`0.75 vs.
`
`Palonosetron
`0.25 vs.
`
`Palonosetron
`0.75 vs.
`
`Study -
`
`Ondansetron
`
`Ondansetron
`
`Dolasetron
`
`Dolasetron
`
`99—03
`99-04
`99-05
`
`3, 20
`
`-7, ll
`
`—4, 3
`
`0, 18
`
`0, 20
`
`—6, 14
`
`Secondary efficacy outcomes
`
`A large number of secondary outcomes were recorded and analyzed. In general,
`palonosetron, particularly the 0.25 dose, compared favorably to the comparators.
`Complete response afier 24 h is discussed below separately from the other secondary
`outcomes, since it is closely related to the primary efficacy outcome as well as a part of
`the basis for the proposed indication.
`
`Delayed complete response
`
`The applicant carried out an analysis for complete response rates for the time periods 24-
`48, 48-72, 72-96, and 96-120 hrs identical to the primary analysis. The results are shown
`below in tables 4 and 5.
`
`APPEARS THIS WAY
`0N ORlGlNAL
`
`ll
`
`

`

`Table 4. Complete response rates, 24-120 hours. The N for a time period refers to the number of patients
`with complete response.
`
`Palonosetron 0.25
`
`Palonosetron 0.75
`
`Ondansetron 32
`
`Dolasetron 100
`
`12
`
`N
`
`%
`
`191
`
`'74
`
`85
`107
`137
`156
`
`39
`
`45
`56
`72
`82
`
`N
`
`185
`
`102
`
`122
`124
`145 '
`161
`
`221
`
`86
`
`109
`118
`142
`156
`
`%
`
`55
`
`66
`67
`78
`87
`
`39
`
`49
`53
`64
`71
`
`%
`
`65
`
`70
`78
`85
`89'
`
`57
`
`62
`73
`82
`86
`
`48
`
`58
`62
`74
`76
`
`N 1
`
`89
`
`122
`
`132
`147
`161
`169
`
`189
`
`107
`
`118
`138
`155
`162
`
`223
`
`107
`
`129
`139
`164
`170
`
`%
`
`74
`
`82
`85
`89
`93
`
`54
`
`62
`68
`79
`88
`
`45
`
`57
`61
`67
`74
`
`N 1
`
`89
`
`140
`
`154
`161
`168
`175
`
`Study
`and
`time
`
`period
`
`99-03
`
`24-120
`
`24-48
`48-72
`72-96
`96-120
`
`99-04
`
`189
`
`24—120
`
`'102
`
`24-48
`48-72
`72-96
`96-120
`
`99-05
`
`24-120
`
`24-48
`48-72
`72-96
`96-120
`
`118
`128
`149
`
`_167
`
`223
`
`101
`
`127
`137
`149
`165
`
`

`

`Table 5. Confidence intervals for the difference between the palonosetron doses and the
`comparators 'in complete response rates during 24 —120 hours after chemotherapy
`
`Palonosetron
`0.25 vs.
`
`Palonosetron
`0.75 vs.
`
`Palonosetron
`0.25 vs.
`
`Palonosetron
`0.75 vs.
`
`ondansetron
`
`ondansetron
`
`dolasetron
`
`dolasetron
`
`8, 30
`
`5, 26
`8, 28
`2, 20
`—2, 13
`
`-5, 17
`
`-3, 20
`-3, 19
`-8, 13
`-7, 13
`
`-2, 21
`
`-8, 15
`-0. l , 22
`-3, 16
`-6,10 '
`
`—2, 20
`
`—3, 20
`-2, 20
`-1, 20
`-4, 16
`
`3, 27
`
`6, 30
`0.1, 23
`-3, 18
`
`-2, l5
`
`6, 30
`
`6, 30
`6, 28
`0.1, 20
`
`-5, 13
`
`Study
`
`99-03
`
`24-120
`
`24-48
`48-72
`72—96
`96-120
`
`99-04
`
`24-120
`
`24-48
`48-72
`72-96
`
`96-120
`
`99-05
`
`24-120
`
`24-48
`48-72
`72-96
`96-120
`
`'
`
`On the basis ofthis analysis, the applicant claims that palonosetron is effective for
`prevention of delayed nausea and vomiting. There are two issues with including this as
`part of the indication for palonosetron. The first is that neither ondansetron nor dolasetron
`are approved for prevention of delayed emetogenesis, and thus may not be valid
`comparators in this situation. However, there is‘no reason to think that they would be
`less efficacious than placebo over a 5-day period, given their efficacy in prevention of
`emesis in the first 24 hours following chemotherapy. Thus a finding of greater efficacy
`of palonosetron relative to ondansetron and dolasetron could be taken as evidence of the
`general efficacy of palonosetron over the extended time period. Following highly
`emetogenic chemotherapy (PALO—99-05), the rates of complete response are consistently
`numerically higher for palonosetron 0.25 mg; however, there is no time period for which
`palonosetron is statistically significantly higher than the comparator ondansetron (as
`judged by the lower limit of the confidence interval of the difference). Following
`moderately emetogenic chemotherapy, the rates of complete response'again are
`consistently higher for palonosetron 0.25 mg.
`It is statistically significantly higher than
`ondansetron at all time periods other than the final 96-120 hours, when there are high
`
`13
`
`

`

`Its performance against dolasetron is mixed,
`response rates in all three treatment arms.
`but is statistically significantly higher for the overall time period 24-120 hours as well as
`for each ofthe first two days in the extended time period.
`
`The second issue is that delayed response was not pre-specified as a primary endpoint,
`and “delayed” is itself not precisely defined (it could be the overall five day time period,
`the overall post-24 hour four-day time period, or each of the four post 24-hour days).
`The outcome analyzed, however, is the primary one of complete response. The most
`sensible definition of delayed response is the overall post-24 hour four-day value: it does
`not include the acute response but does include information from all subsequent days, in
`particular the final day of observation when the emetogenicity of the chemotherapy is
`most muted. The lower limit of the confidence interval for the difference between
`
`palonosetron 0.25 mg and the comparator is well above zero for this time period in both
`99-03 and 99-04.
`
`Other secondary outcomes
`
`The treatment groups were generally comparable on secondary outcomes. Differences
`that were found tended to be in favor of palonosetron relative to the comparators.
`
`Prognostic subgroups
`
`Gender, chemotherapy history (naive or non-naive) and corticosteroid use were
`prognostic factors taken into consideration in the allocation algorithm to balance the
`treatment groups. Complete response rates for 0-24 hours for each prognostic subgroup
`are shown below in table 6.
`
`A greater proportion of the men than of the women responded to anti-emetic treatment.
`The differences between the genders were mostly on the order of 15-20%. Men
`responded markedly better than women to palonosetron 0.25 mg in study 99—04, and
`women to palonosetron 0.75 mg relative to the other treatments in 99-05, but there were
`no consistent differences between men and women among the treatments (i.e., no
`interaction between treatment group and gender). Both male and female subjects
`responded better to palonosetron than to the comparators.
`
`There were no consistent or significant differences in complete response rates based on
`chemotherapeutic history or corticosteroid use.
`
`14
`
`

`

`Table 6. Complete response rates, 0-24 hours, for prognostic subgroups. Ncr
`patients with complete response
`
`is the number of
`
`Palonosetron 0.25
`
`Palonosetron 0.75
`
`Ondansetron 32
`
`Dolasetron 100
`
`Study
`and
`
`subgroup
`
`N
`
`Ncr %
`
`N
`
`Ncr %
`
`N
`
`Ncr
`
`%
`
`N
`
`Ncr
`
`%
`
`99-03
`
`189
`
`189
`
`Gender
`Male
`Female
`
`54
`135
`
`49
`104
`
`Chemotherapy
`Naive
`76
`Non-
`113
`nai've
`
`67
`86
`
`91
`77
`
`88
`76
`
`51
`138
`
`46
`93
`
`90
`67
`
`80
`109
`
`55 ' 69
`84
`77
`
`185
`
`52
`133
`
`78
`107
`
`41
`86
`
`58
`69
`
`79
`65
`
`74
`65
`
`--
`
`--
`
`--
`
`--
`
`--
`
`-
`
`99-04
`
`189
`
`_
`
`189
`
`191
`
`21
`87
`
`64
`56
`
`--
`
`--
`
`-—
`
`35
`156
`
`22
`79
`
`63
`51
`
`Gender
`Male
`Female
`
`34
`155
`
`30
`89
`
`Chemotherapy
`Nai‘ve
`124
`Non-
`65
`nai've
`
`75
`44
`
`88
`57
`
`61
`68
`
`33
`156
`
`131
`58
`
`73
`35
`
`56
`60
`
`--
`
`--
`
`--
`
`125
`66
`
`58
`43
`
`Corticosteroid use
`Yes
`1 l
`N0
`178
`
`8
`111
`
`73
`62
`
`12
`177
`
`' 6
`102
`
`50
`58
`
`8
`183
`
`5
`96
`
`46
`65
`
`63
`53
`
`99-05
`
`223
`
`223
`
`Gender
`Male
`Female
`
`108
`115
`
`Chemotherapy
`Nai’ve
`133
`Non-
`90
`na'ive
`
`72
`60
`
`75
`57
`
`67
`52
`
`56
`63
`
`110
`113
`
`75
`71
`
`129
`94
`
`87
`59
`
`Corlicosteroid use
`Yes
`150
`No
`73
`
`97
`35
`
`65
`48
`
`150
`73
`
`94
`52
`
`68
`63
`
`67
`63
`
`63
`71
`
`22]
`
`108
`113
`
`131
`90
`
`73
`53
`
`72
`54
`
`147
`74
`
`82
`44
`
`68
`47
`
`55
`60
`
`56
`60
`
`15
`
`

`

`3.1.2 Study PALO 99-01-23: Meta-analysis of historical data
`
`The applicant says (v.368, p.5)
`
`Effective antiemetics are currently available for use by patients undergoing emetogenic
`chemotherapy. For this reason it is considered unethical to include a placebo control arm
`in trials investigating treatments for the prevention of chemotherapy-induced nausea and
`vomiting.
`
`The efficacy trials done as part ofthis application did not include a placebo control. To
`assess trial validity and justify the value of delta used to declare non—inferiority of
`palonosetron-to ondansetron or dolasetron, a meta-analysis of results from the anti-emetic
`literature was carried out,
`to estimate both historical placebo response and the
`corresponding historical rates for active comparators. These rates were then adjusted for
`covariates, in an attempt to make the historical populations comparable to the patients
`under study in PALO-99-03, -04 and —05. Covariates included features of the study, such
`as endpoint and emetogenicity ofthe administered chemotherapy, as well as of the patient
`population, such as the percentage of males in the study.
`
`The applicant included information from 46 studies, with a total of 78 treatment arms.
`These studies were selected from the results of a literature search of Medline, along with
`studies suggested by Dr. Robert Prizont, FDA medical officer, or included in the FDA
`Summary Bases of Approval of ondansetron, dolasetron and granisetron. Study inclusion
`criteria are summarized in the Appendix.
`
`I present some descriptive summaries gleaned from the database of study findings
`analyzed by the applicant (v. 371, appendix 5). Ofthe included studies, only four
`compared either ondansetron or dolasetron to placebo, with the results shown below.
`(The remaining studies either had no placebo control or compared placebo to other anti-
`emetogenics.) The response in all four was defined as no emetic episodes in 24 hours.
`
`Table 7. Response rates in comparative trials
`
`Study
`
`Active
`treatment
`
`Response
`in placebo
`
`Response Difference
`in active
`in rates
`
`Emeto-
`genicity
`
`Ondansetron
`
`Beck et al, Ondansetron
`1993 x
`Cubeddu
`et al, 1990
`Cubeddu
`et al, 1990
`Cubeddu
`
`Ondansetron
`
`Ondansetron
`
`15/81
`(19%)
`0/10 (0%)
`
`0/14 (0%)
`
`9/73
`
`(12%)
`
`52/79
`(66%)
`7/10
`(70%)
`2/14
`(14%)
`47/7 1
`
`(66%)
`
`47%
`
`70%
`
`14%
`
`54%
`
`Moderate
`
`Moderate
`
`High
`
`Moderate
`
`et al, 1994
`
`The differences in response rates range from 14 to 70% overall, and from 47 to 70% for
`moderately emetogenic chemotherapy.
`In studies PALO-99-03, -04, -05, the-observed
`lower limits of the confidence intervals for the difference between palonosetron and
`
`16
`
`

`

`dolasetron or ondansetron (table 3