E TERF RDR
`
`E AL ATI NANDRE EAR H
`
`APPLIQATIQ 2N NIJMBER: NDA 20845
`
`STATISTIL :AL RE YIEEW S)
`
`
`
`

`

`Statistical Review and Evaluation
`.
`
`NOV
`
`£1 1999
`
`NBA: 20,845
`.
`,
`Applicant: INO Therapeutics, Inc.
`Drug Name: Inhaled Nitric Oxide (Nitric Oxide)
`Indication: Persistent pulmonary hypertension of newborn (PPHN)
`Document Reviewed: Vol. 9.1, 9.6, 9.7, 9.10, 9.11, 9.20, 9.21
`
`The sponsor’s resubmission of original NDA of Nitric Oxide (May 26, 1999) includes the
`results from a new randomized clinical trial, known as CINRGI. The original NDA was
`reviewed by FDA earlier but was withdrawn by the sponsor on September 16, 1997. This
`statistical review focuses on the sponsor's new study (CINRGI) in the resubmission.
`
`1. Outline of CINRGI study
`
`Design ofstudy
`
`Study CINRGI was a multicenter, placebo-controlled, double-blinded study and involved
`212 full-term and near-full term neonates with echocardiographic or clinical evidence of
`pulmonary hypertension. The objective of the study was to assess the safety and efficacy
`of inhaled NO added to the conventional therapy for PPHN as compared to conventional
`therapy alone.
`>
`
`According to the sponsor, the patients in this study were receiving diluted treatment gas
`with endotracheal tube NO concentrations of 0 ppm (for placebo patients) or 5-20 ppm
`(NO patients) according to the randomization. The randomization was stratified by
`patient disease status (CDH, MAS, Pneumonia, PPHN, RDS, and other).
`Inhaled NO
`was started at 20ppm. For all patients the ventilator settings were held constant over the
`first 30 minutes of treatment. Weaning of the treatment gas was done by decreasing the
`percent of treatment gas. Neonates who had a Pa02 (arterial partial presSure of oxygen)
`larger or equal to 60 mmI-Ig and a pH 7.35-7.55 after being in the study for 4 hours had
`the treatment gas concentration reduced to 5 ppm for the remainder of the treatment
`period. Treatment gas was continued at 5 ppm until the FiOZ was <0.7,
`the patient had
`received 96 hours of treatment, or the patient was 7 days old, whichever came first. The
`submitted data indicate that the gas treatment for the first patient started on 3/3/96 and for
`the last one on 12/18/98.
`
`Endpoints / hypotheses Yanalyses /sample size
`
`The primary efficacy endpoint was defined as the “need for ECMO” (N'FE) in the
`original protocol (December, 1995). In the protocol,
`the criterion for treatment with
`ECMO was defined (See Appendix 1). The sponsor later (Amendment 1) clarified the
`meaning of “need for ECMO" as actual “use of ECMO” (UOE) rather than “met ECMO -
`criteria" (MEC).
`For this reason, no detailed information on MEC was collected
`according to the sponsor (Appendix 2).
`'
`_
`
`

`

`Table 1.1 Mean methemo_lobin b treatment -
`
` Z
`o 95% C!
`_ lacebo 95%C1
`
`m 0.73 0.66090
`.79 0.67. 0.91
`
`Im— 0.78 0.63 0.91
` 0.92 0.77, 1.07
`0.58 0.44, 0.72
`
`
`
`
`
`‘
`
`
`
`Fallow-up duration / Patient withdrawal
`
`No maximum follow-up duration was clearly specified for the primary variable,_the use
`of ECMQ. The conditions of exiting the trial were specified in the protocol (Appendix 4).
`v
`
`Sponsor’s result
`
`A total of 248 neonates were entered into the study. Of these, 36 were enrolled into the
`pilot study phase of the trial that was randomized but not blinded. A total of 212 patients
`were randomized in the blinded trial. Among them, 26 patients (with disease status as
`CDH 0! other) had an enrollment diagnosis of lung hypoplasia and were analyzed
`separately from the other patients. The efficacy population for this trial consisted of
`remaining 186 patients.
`'
`
`The two treatment groups seemed to be comparable with reSpect to demographic factors:
`age, gender, weight, Apgar score, and race. The two groups seemed to be comparable
`with respect to most baseline prognostic characteristics except for a few,
`including
`baseline airleak, arterial pressure, Pa02 (arterial partial pressure of oxygen), Sa02
`(percent of oxygen saturation of the arterial blood), and OI (oxygen index). The
`imbalance with respect to these factors were generally in favor of N0 group. For
`instance, a numerically higher mean value of 01 (43.9) was observed in the placebo group
`as compared to that (35.0)
`in the NO group.
`The sponsor's explanation 'for the
`imbalances is that for some patients, the baseline oxygenation measurements were taken
`after the treatment started.
`
`
` table and some baseline characteristics
`Table 1.2 Demo
`_——IM_
`
`-—E_EM_ 0.95
`
`Inert—m..- 0.08
`_———m_ .
`_n—m-m-
`___II—IE—
`__--IE_
`III—M
`PaO (mmH-
`E_—IE_
`SaO %
`'
`III—_—
`W—
`it
`
` —IE—
`
`by physical examination
`
`
`
`
`
`
`
`
`
`
`The sponsor’s analysis (Cohran—Mantel-Haenszel adjusting for underlying disease) based
`on the intent-to-treat patient population indicated a statistically significant group
`difference in use of ECMO (31/98 forNO and 50/88 for placebo, p=0.001). The
`
`
`
`

`

`_. difference was still statistically significant when adjusted for baseline difference
`controlling by P302 or by 01 categories (p=0.007 from both adjusted analyses). The
`difference in use of ECMO between the two treatment groups was not statistically
`signifith in the 26 patients with lung hypoplasia (p=1.000).
`
`'
`
`
`
`
`
`Table 1.3 Number of use of ECMO b n'eatment
`
`N0 %
`-
`
`
`30/97 30.9 IE!-
`AH 186 aticntsJTT
`
`
`
`418 50.0
`0.007
`All 135 suj., adjusted for baseline PaOZ
`
`5m 83.3
`(mmHslwcsorics
`
`mamas
`405 16.0
`
`—-- 3/13 23.!
`
`2114043
`
`
`
`All
`l86 suj., adjusted for baseline Ol
`411506.?
`
`
`(cm H20/mmHg) categories
`‘
`9/43 20.9
`
`
`
`411506.?
`
`
`
`6/11 (54.5
`
`
`8/14 57.1
`
`
`
`9/1: (81.8
`‘
`
`
`
`0.00?
`
`Patients
`
`ith lune hv-o-lasia
`
`4r
`
`The sponsor compared the outcomes in oxygenation status between the two treatment
`groups and claimed a statistically significant difference between the-groups with respect
`to several indicators of oxygenation (Tables 34-37, the Sponsor's study report). However,
`the sponsor’s analyses used, instead of an ITT patient population, only the information
`from the completers. These analyses might introduce a selection bias and thus not
`preferred.
`='-'-
`
`The sponsor's analyses based on the 6-month or 12-month follow-up data showed no
`statistically significant difference in hospitalization. Six month death rates in the two
`groups were not statistically significantly different (5/89 for placebo and 4/97'for NO,
`p=0.738).
`
`2. Reviewer’s results and comments
`
`This reviewer compared the numbers of use of ECMO between the two treatment groups
`using all 212 randomized or-using only the 186 patients without lung hypoplasia (LH) at
`the time of enrollment (Table 2.1). There was a statistically significant difference in rate _
`of use of ECMO between the two treatment groups. According to the sponsor, it was
`assumed in the protocol that all ECMO therapy would be captured by recording ECMO
`use during the initial hospitalization.
`The submitted data indicated that durations of
`patient hospitalization (calculated by this reviewer as the difference between the time of _
`discharge to home and the time of the initiation of treatment gas) range from 6 days to
`several months. Majority of use of ECMO occurred within 6 days after an initiation of
`treatment gas except one case (7 days after the initiation of the treatment gas) and several
`cases with missing time ofdischarge to home ( 10 subjects in placebo and 3. subjects in
`the NO group). The patientfollow-up seemed to be complete and comparable between
`the groups (Table 2.2).
`‘
`
`

`

`Table 2.1 Com-arison of incidence in use of ECMO
`MIME—_ nvaiudnozsuatifiem
`— #of‘ECMO/n % #ofECMO/n %
`exclude Ll-I
`50/88 56.8
`31193 31.6
`include LH
`62/102 60.8 m-
`mn—
`‘x -tcst, " Cohmn-Mantcl-Haenszel controlling by strata (underlying disease)
`
`
`
`
`
`
`
`
`
`
`
`
`
`Table 2.2 time to dischar- e da
`
`The question now is whether or not the observed group difference in use of ECMO can be '
`attributed to the effect of nitric oxide treatment. To answer it, this reviewer focused on
`the following issues‘.
`
`(i) Unblinding of treatment Code
`
`As explained before, there are sufficient grounds to suspect that the ClNRGI study was
`largely unblinded.
`This is especially worrisome knowing that the primary endpoint,
`UOE (use of ECMO) can be subjective and initiation of ECMO depends on an
`investigator’s judgement and discretion.
`It is possible for biases of the investigators to
`be introduced to the trial. For this study, extracautions must be taken in examining and
`interpreting the trial results.
`
`(ii)
`
`Potential bias due todelaying initiation of ECMO
`
`To evaluate the effect of the NO treatment on use of ECMO, it is important to examine
`whether or not there was a delay of initiation of ECMO in the NO group as compared to
`placebo. The delay could be a result of more aggressive initiation of ECMO for placebo
`patients and/or more reluctant initiation of ECMO for the NO treated patients.
`If there
`was a delay in the N0 group, the observed lower rate of use of ECMO in the NO group
`could be a direct consequence of the delay since with the delay of initiation of ECMO, a
`patient might'pass the episode of need for ECMO and never needed ECMO again.
`In
`this case, one cannot relate the effect of the nitric oxide treatment to the lowered rate of
`ECMO use in the NO group without knowing the causes of the delay.
`If there were
`causes unrelated to the effect‘of the NO treatment, the observed treatment effect would be
`confounded partially or completely with biases attributed to the delay from these causes.
`
`A significant delay in initiation of ECMO in the NO group as compared to placebo was
`suggested by the data (Table 2.3a). The median duration from initiation of treatment gas
`to initiation of ECMO was 3.6 hours for placebo and 10.4 hours for the NO group. There
`
`

`

`_ __ is a shift in distribution of the time to initiation of ECMO for the NO group as compared
`to the placebo group. As discussed above, such a delay can contribute to the observed
`lower ECMO rate in the NO group. A similar pattern can be seen among the patients
`who had OI>40 at baseline and received ECMO later. These patients might need ECMO
`urgently, but ECMO was initiated later in the NO group as compared to the placebo
`(Table 2.31)).
`In order to conclude a treatment effect of NO on use of ECMO, one must
`demonstrate that the only cause 'of the delay was the effect of the nitric oxide
`treatment, or to demonstrate a negligible impact of the delay due to other treatment
`unrelated reasons.
`'
`-.
`
`'
`Table 2.3a Time to initiation of ECMO our
`
`
`
`
`. Table 2.3b Time to‘initiation of ECMO (hour) in patients with ECMO use and with
`baseline OI>40
`Treatment / # of patients with > 40 Ol at baseline
`
`
`
`' Placebo I 25
`
`
`
`“mflmlmflh-‘H«nil..‘1
`
`Many factors could cause the delay of initiation of ECMO in the NO group in ClNRGI
`study. As discussed in (i), the investigators in this study were very likely to know patient
`treatment assignments, had ability to determine the time of initiation of ECMO, and
`might consciously or unconsciously delay initiation of ECMO for qualified patients in
`the NO group or accelerate initiation of ECMO for patients in placebo.
`If this was the ‘
`case, the true effect of the NO treatment would be confounded with the bias introduced by '
`the differentiatial initiation of ECMO between the two treatment groups. The delay of
`initiation of ECMO could also relate to a treatment effect. For instance, such a delay
`might be due to a patient‘s improvement in oxygenation.
`In this case, the observed
`lowered ECMO rate in the NO group could be viewed as a reflection of the effect of NO
`on oxygenation.
`'
`
`The CINRGI study does not provide sufficient information to answer the question
`regarding delay of initiation of ECMO in the N0 group. For example, the sponsor
`did not collect information on the date and time when a patient met ECMO criteria at the
`first time (Appendix 2). This information is important to explore the nature and the cause
`of the delay, since MEC (met ECMO criteria) is more objective than use of ECMO.
`
`this reviewer
`To explore the difference in initiation of ECMO further indirectly,
`indicators of
`compared the two treatment groups with respect
`to two important
`oxygenation: oxygenation index (OI) and P302. It was noted that there was an imbalance
`between the groups with respect to baseline 01 and PaOZ. In this reviewer’s analysis, lTT
`patient population was used and thecomparisons were made based on change from
`baseline, using LOCF imputation and t-test with unequal variances, rather than an
`
`
`
`

`

`endpoint measurement. The following table summarizes the mean changes in 01 and
`PaOZ during the 24 hour period
`.
`-
`
`‘
`
`Table 2.4 Ana]
`
`is of chan- e in OI and Pa02
`
`Placebo n.A
`—l-—IEBEE_EEI_
`MEIE—
`
`Wm
`
`um—
`
`mun-l.— (n=82 —15.1
`P302
`_mm-—_
`—mrzs-l-—lms—
`M—
`WEEKE—
`Mm
`_m-IE_
`nmnr-m— (n=89 46.5
`’ for group difference
`
`
`
`
`
`
`
`.3
`
`
`
`
`there is a sustained treatment effect of NO on
`to conclude that
`The analyses fail
`oxygenation. Without any p-value adjustment for the multiple time points, the group
`differences in 01 and P302 were significant only at Hour 0.5 (OI and Pa02) and Hour 4
`(01 only). With Boferroni adjustment of p-values across the different time points, only
`the group difference in Pa02 at 30 minute was statistically significant. The results of the
`analysis might suggest that the NO treatment affected patient oxygenation more rapidly as
`compared to the conventional
`therapy used (as the background therapy) in placebo. The
`observed reduction in rate of use of ECMO in the NO group might be a consequence of
`this rapid effect.
`'
`
`_
`
`.__
`
`.
`
`IQ
`
`:
`
`(ii)
`
`Unbalanced baseline
`
`As pointed out, there were significant imbalances in favor of the study drug with respect
`to
`baseline 01 (p=0.011), Pa02 (p=0.007) and Sa02 (p=0.0]8). Fitting a logistic
`regression model with treatment and these baseline oxygenation indicators as covariates,
`a significant effect of baseline OI on ECMO rate was found (p=0.0001). The p.—value for
`the treatment effect on _ECMO rate adjusted for the baseline 01 is p=0.0443, a large
`increase from p=0.001 without baseline adjustment. The lack—of-fit test did not indicate a
`misspecified model. When the patients‘ disease status (the strata) was included in the
`model, the p—value' for the treatment effect was 0.0460. Two similar analyses were done
`using the -2 hour and —4 hour OI measurement to impute missing baseline 01. The
`resulting p-values'were 0.0147 (w/o stratain the model) and 0.0158 (with the strata in the
`model).
`
`_
`
`:
`;
`
`
`
`
`
`

`

`Analyses
`
`Lo_'stic re_ ession ad'usted for baseline 01
`Placebo
`NO
`-# of ECMO I n (%)
`it of ECMO I n (“/o)
`
`-
`
`
`
`
`
`
`
`
`
`Wm
`
`
`
`Odds ratio
`NO vs. control
`
`
`
`
`
`'
`
`g
`7'
`
`II
`
`The reviewer and sponsor’s analyses suggest a difference in the rate of use of ECMO
`between the two groups even adjusted for the imbalance of baseline 01.
`
`3. Reviewer’s Conclusion
`
`3
`
`Study CINRGI was to assess the efficacy and safety of inhaled NO added to the
`conventional therapy as compared to the conventional therapy alone in patients with
`PPHN. A beneficial effect of the N0 treatment was supposed to be indicated by a
`reduction in rate oTuse of BCMO (the printary endpoint).
`
`The protocol or protocol amendments of Study CIN'RGI did not clearly address many
`important issues like endpoint definition, blinding procedure, patient follow-up period,
`time of randomization and study termination. Since the protocol is a general guidance for
`the trial, a negative impact of the ambiguity in the protocol on the conduct of the study
`can be anticipated.
`
`The analyses of use of ECMO in this study indicate a group difference with respect to rate
`of use of ECMO. The p-values for such a difference range from 0.001 to 0.0460,
`depending on whether or not there is an adjustment for unbalanced baseline prognostic
`factors and the methods for the adjustment. To conclude a treatment effect based on the
`observed group difference, several aspects of the trial must be carefully examined.
`
`The important issues identified in this review include (i) possibly unblinding treatment
`codes and (ii) potential bias due to the difference in time of initiation of ECMO between
`the two treatment groups. The submitted data suggest a significant delay in initiation of
`ECMO in the NO group as compared to placebo. Since the observed lower rate of use of
`ECMO in the N0 group could be solely due to such a delay and the delay could be‘due to
`manycauses, relevant or irrelevant to the NO intervention, it is very difficult to'relate the
`observed group difference in rate of use of ECMO to the effect of the NO treatment in a
`clear-cut manner, knowing the potential unblinding in this trial and the possibility for an
`investigator to hold the initiation of ECMO temporarily for a patient. To conclude that
`there is a beneficial effect of the N0 treatment on use of ECMO, one must show that the
`' reduction in ECMO use in the NO group was solely attributed to the effect of the nitric
`oxide treatment. Unfortunately,
`the current database does not provide sufficient
`information for such a conclusion.
`
`Although it is impossible for this reviewer to conclude the effect of the nitric oxide
`treatment on lowering rate of ECMO use,
`some analyses of oxygenation seemed to
`
`

`

`_._. suggest that the N0 treatment yield a more rapid effect on oxygenation as compared to
`the conventional therapy. This might further suggest that the observed lower ECMO rate
`in the NO group as compared to placebo be a consequence of the rapid effect of the nitric
`oxide on oxygenation. Again, the observation about the possible relationship between the
`ECMO use and‘the improvement in oxygenation, by no means, is conclusive since the
`ambiguity of the efficacy outcome for the primary endpoint, and the post-hoe nature and
`the multiplicity involved for the secondary endpoints for oxygenation.
`
`Lu Cui
`
`-
`
`/
`“
`[.31
`
`Ph.D., Mathematical Statistician
`11/3/99
`
`~
`
`’
`
`"‘ This revievver has discussed the issues related to this NDA with Dr. Douglas Throcl-nnorton.
`Concur: Dr. H.M. Janjes Hung..—..
`EI més/
`“Z3/??
`gDr. George Chi __
`/S/
`u/xf/71
`
`V ’-
`
`cc:
`NDA $20,845 Nitric Oxide
`HFD—llo
`
`HFD-l-IOI Dr. Lipicky
`HFD-l 10 I Dr. Stockbridgc
`l-lFD—l 10 I Dr. Throckmorton
`HFD-l 10 I Ms. McDonald
`RFD-344 I Dr. Barton
`RFD-710 I Dr. Chi
`
`.
`
`RFD-710 I Dr. Mahjoob
`HFD-710 I Dr. Hung
`RFD—710 I Dr. Cui
`HFD-‘IIO I Chron.
`
`O.
`
`lcuiIZO.845m:\cuilreview‘mitridrpa.docI5945302‘
`
`

`

`
`
`-
`
`~
`
`.
`
`.
`'
`‘
`ystem disease {exceptions include small ASDNSD-
`
` 1; Weight >2000 g
`
`flg’estational age 334 weelm
`-
`‘53 ‘ cchanical ve‘ntiiation foi- < 10 days.
`
`-'_"é'versible lung disease
`.
`
`
`o hongenital heart, neurological, orrnultis
`
`‘.
`
`
`‘No intracranial hemorrhage >grade 1.
`
`
`1518 severe coagulopathy
`
`
`1::#1119 severe asphyxia (10 minute Apgar 02 or initial base excess greater than -20)
`
`
`1}? ffgiiure of maximal medical management (alkalosis, pressor support, vasodilator trial)
`V._
`.
`
`
`_
`"
`E #131115 one of the following:
`'1‘
`
`
`£5315 __0;;3;tgcr_1ation Imiex (MAP x FiO2 x IOOIPaoz) >40 for 3 of 5 blood gases at least 112 hour
`
`
`3p
`.
`-
`-
`.
`11L PaO2 <40 for 4 hours despite maximal medical management
`
`aficiaPaoz <35 for l hourficlespite maximal medical managemEnt
`
`
`$1751: Progressive hemodyaamic deterioratipn umEsp‘onsive to maximal medical management
`
`
`'
`' :i: (colloid infilsions, catechoiamine drugs, etc)
`'
`'
`
`
`
` a}!
`
`
`II! ‘
`
`
`
`
`

`

`j
`
`>
`
`Appendix 2-
`“
`(The sponsor’s response)
`
`
`
`.,-
`
`
`
`‘-
`
`facsimile tra nsrnlttal sheet I fax number (908) 238-6833
`To: Dr. Lu Cui
`.
`From: Richard N. \Miliams. Ph.D..
`Tei: sea-23845522
`'
`
`
`Company: FDA -
`-
`‘
`'
`Date: August 5. 1999
`‘
`Fax number: 301-594-5494
`No. of pages including cover. Two
`'
`Phone number: 301-594-5302
`66'
`Re: Written Confirmation Regarding ECMO Criteria
`notes/Comments:
`\_
`
`'
`
`_
`
`‘
`
`'
`
`.
`
`'
`
`- --
`
`,
`
`_
`
`NDA 20—845
`
`Nitric Oxide
`
`. '
`
`Dgér Dr. Lu Cui,
`
`
`
`

`

` -
`
`'f""'—" . ..
`
`

`

`Appendix 3
`(Sample data collecting sheet)
`
`.
`
`?
`
`12
`
`

`

`Appendix 4
`(Exiting criteria)
`
`_ _
`‘ted from the study if:
`
`'
`
`The patient will
`
`\
`
`1. The patient meets criteria for treatment with ECMO or is placed on ECMO.
`
`."
`
`a. Oxygenation Index (MAP x FiO; x 100fl°a0,) >40 on 3 of 5 blood gases drawn 30
`minutes apart-
`.
`_
`.
`.
`
`-w
`b. PaOz <40 for 2 hours despite maximal medical management
`
`c. PaCl, <35. for 1 hour despite maximal medical management
`
`d. Progressive hemodynamic deterioration (mean blood pressure <35) unresponsive to.
`maximal medical management (colloid infusions; vasopressors. etc).
`
`.
`‘3'
`
`OR
`
`2. The patient has an inadequate response to treatment gas or fails to tolerate being weaned
`from the study gas
`_
`.
`x
`
`
`
`a. PaO2 < 60 on Flo2 of 1.0 and optimized ventilator settings after 24 hours of
`treatment. Maximum dose =20 ppm for first 24 hours in study.
`
`b. PaO2 < 60 on FiO, of 1.0 and optimizedventilator settings 24—96 hours of
`treatment Target dose =5_ ppm during 24—96 hours in the study.
`
`c. Failure to tolerate a decrease'1n the treatment gas to 5 ppm after 24 hours (i.e.
`PaOz < 60 on FiO2 of 1.0 and optimized ventilator settings after 24 hours of
`treatment. Maximum dose =20 ppm for 0-24 hours'1n the study). See details1n NO
`inhalation protocol--7_C
`
`(1. Failure to tolerate discontinuation of treatment gas after 96 hours (i.c., P‘aO;I <$0
`on FiOz of 1.0 and optimized ventilator settings after 96 hours of treatment. Target
`dose =5 ppm for 24 to 96 hours'1n the study). See details'111 N0 inhalation protocol
`- 7c
`.
`
`OR‘
`
`'5.
`
`3.
`
`Initiation of the treatment gas is associated with a sustained deterioration in the patients gas
`exchange that is due to initiation of the treatment gas.
`.
`.
`
`a. PaOz decreases by more than 10 torr
`
`l). Pa02 falls below 40
`
`'10'
`
`
`
`
`
`

`

`-.'4'. ' Toxicity develops (MethEmogIobin > 4%. No2 >5 ppm)
`
`0R
`
`
`
`
`. Parents withdraw consent A
`. The patient is discovered to have met a patient exclusion criteria. (e.g.. diagnosis of
`congenital heart disease, or to have a lethal congenital anomaly)
`
`m—vqwm‘
`'VHII‘H
`
`715'u’.
`.37
`4..g”.
`
`-..e.I-..I["T"
`
`
`V".
`
`.
`
`fie
`
`i".-=.-T.=-.-:J
`
`-.
`
`