> %(*'&`!Hl$:?K=lծ 'v(c8x_HeEsUv7nHV[5A܅E' y[Ȋ*>ҋCԂy5i7~}gΜϝ3ߙof~9s,Q"_hw"bf.#2kEq_CMi3+r?)P6\.GZ[[:;;h``
5M<8HKߙXnZ/$FQ@c
h`{C $P@u`8^
h` H
h`J
h`[
R@/^~(! @O
۱.vA
"#Aɔ[e4q/a=0}B$Pn}$[Q>qG@9`d*+Rv(7
h`!b;耭`Wm7vCI0FiM(7
h`[
Oayyac[L422BSSS433CsssHL+++JkkkTThss:]>xܶ8٣蜫ϝ)'K9
D8DEIIO].]z
B~(+4ɈN#7=lmv^CmZ%ǘc$ދW~<;hy7NT
Z9hm6hAZOzԺdnߋ~PIm5r>l}x<ȶֿG6O7OGQ;n1qqŘjhGQ:BFTgcKQlDBM:cvF6cWWU]et74ai,]i]A
Z.6A+m6hApzCչ\_ jɝnvоRZ9c?9k4maMէ>GPiyZ<ǕsIF=}gJ.YMEE_sE[)D'_}ԭgOFV?M'N_Q]ψxrDbKwt\zy+tPfQoyZˡǨDowe?>Ww9m˽7]ui_wחw(?`!+xxF4;[v(c8xMh]E%B?%ʹ7}BTLmlh;B
Kŷf6nMbBhjk;w{)$Y83so5DYu1D̔A.#4E;q/!afόBX"}:e2f(Sss3R{{;uvv
P?_!>uc23=ݼD_}Iv A40}# (I3pҙ}@
`?( @
`?x &}40%P@/
h`H_, 40Iaۗn5>E+#Aʔ,ʶCj#/a=0}@$P}$[>pүvc2.T@W8v:P@o Cv,[Gn&)솒@ov:P@o< NBaqqac[L#4::JSSS4;;KHL+++JkkkT*hcc:]m8٫+=!}qS}nm}%)?wשzRB;E/_096NI(Zosh̖vko7tNơj2;ox/~.9p!nd8iSi(ՊA+m.hsA Hкm9&3=x?>u+m\Ⱦf]!q,C8M9
t덆CGA:C_q+IGH{\tYg_;;:~JOcALqFZGJt.T+m$h#A
Z2NgD%9
ݠ}ZZ9c71k9M1/blwKO%_)O;xy~,W3qP{,KQ;VkVٮgɲ]O]KNu{fvm3&eiL#s3߸/,[Oϧ7Q\zy~*S
g^lk;ko^Cn?5!CfGQJMFw}T5 ]F5z_AvK֕3_v~Wo?ʶ;}W_N`!9«ɘVyu샄 `.PPxMhSAggKW&@.J<$%@Ȱ3xH5xH5(xH5PxH5xxH.AR#hxH5.xH5Y.RM $%$P`U03J N@u8PdP<.jc<ejjL>@P.зXl:@6<$IO"y9rIQ8pFFFٳtEWҍ7͛ts=x=zDt8
&P
)7fa#]AYa`Ѯ=vuN/)?/1cՎ}]Kϸ%䣖Ԓ[aKiK^gm!>
'X$y :jNp2<%:Jַb%ޢt0=<gi瓶^Fa:XiM/Wi1&UO/1%AYy9ĔѶ9Z2]9.j!kcEwbǝ4L30EnZC})
uFjяDGW7`9uV;fڟk*_J~joQ,ԒZaK>lӖU~0Ӳ"ў7A֞])yJ3M`LTy_R'3ws_Z˺gk#MV*}¿7NοX)seHxu/Ucp_kK%MDv?=UW"9_=MfcsXu&e\Ҟݪ{vmO3r.{jT_/ESS={7Ϳs8>P{sesS3fIf:3Sͥ[ ה#_+Sߌ:>ԯesFt=TluR߫w^sst:ޣIp{#<a
jCL2ք*_L^n[ CJ*g`ߙrQQE)n'~wzWzg$E?#](9iı }2`6{fˆ7^)eONMp
UG#M3SΛ}L+wK"a_/@=g<=]pf1vf8xOlTEobʟR@VFE!P/b!zqD_"QH{/MLLLBB<@PrB$¡ {7uLvo;{D6рƏCDK5"
gFrUDp>ST v PFM086"]MlŢ(!c.\GlrUWWSMM
Xꨩ֭[G]]]NLP$I7t
)D/P@s(AHD@
4AX"yM"F(A
h`d!"P<(C402!Z&&P@#402y!P@KcITPxU03GJ&;CyYDq(F)4\H50p̀ǤP@#(qx#D@
lD@
lDǂr`LVX
AR
LP"F6^"ЧB
k`y]1nk\ɇ:w0]FFF͛4::JwܡӃsΟ
qmoONdh%^ō'9n8Yu"bV:6 ~
hW^P~D]$n mX>x><{g^yϞų7zFzy
/B~/AƲt/y[36Uf+Ӧt>#!s7."i4k5=4N+i}iq믑zv^l(]7M.AP̋."ܗ3ݷ #Ӎ`B>F
cg) 4;3`Luytֆ̄~>ӆSj_ݳ=^dΦc#?
&{,_eKO'qGڃ G:Fv<`>l"}.t]~amM:h}'Y{\%}ܦD?[+%5c&JZƇK=MR>lƇ^BEWöi8 }M@>4aMaS>4O2%$K^T7َSߩbnZ8֠wH{ױM0VY`MUKѿZ2Q&LַP
NVķ߹OTmwpw2sTEY' FSxkߡ'ķ<'$8v%s4So`6ۏgƔu5{>^v+݅1[H$>vL~:f*·+srf܇3ώ]֑68ssϻ>[ m=*]˲ɀlݢg4FFVĦ'(
\
0zb
Chart MSGraph.Chart.804Microsoft Graph 2000 Chartd0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0bEquation Equation.30,Microsoft Equation 3.0/0DTimes New Romanbb(b0LbLb~0DSymbolew Romanbb(b0LbLb~0 DWP Greek Centurybb(b0LbLb~00DWP MathACenturybb(b0LbLb~0@DWP MathExtendedAbb(b0LbLb~0PDWP IconicSymbolsAb(b0LbLb~0
f@ @@``
@n?" dd@ @@``*
"
D4:;
"'. ?#&,'""A !#/!@($.;(9) 3
<>.14/==0C<2$Hl$:?K=lծ 'z2$+xxF4;[z2$9«ɘVyu샄9
z$z$z$z$z$z$z$z$z"$VLK8xunzl1z"$g<=]pfzS~1? f̙f333@>g4KdKdXb~0Pb
ppp@g4BdBdXb~0Pb
(ppp@<4ddddbbLC?O=jInference and Operational Conduct Issues with Sample Size Adjustment Based On Interim Observed Effect Size&kj(
H.M. James Hung (DB1/OB/OPaSS/CDER/FDA)
Lu Cui (Aventis Pharmaceuticals)
SueJane Wang (DB2/OB/OPaSS/CDER/FDA)
John Lawrence (DB1/OB/OPaSS/CDER/FDA)
Presented in Annual Symposium of New Jersey
Chapter of ASA, Piscataway, NJ, June 4, 2002 ILZ[Z)VXft(!H
Disclaimer
The views expressed in this presentation are not
those of the U.S. Food and Drug Administration,
nor of Aventis Pharmaceuticals.
Dr. Lu Cui was one of the primary investigators
of this research during his tenure in FDA.
0f,,ROG6Selected References in Adaptive Design/InterimAnalysis76'PH
Emerging Strategy$0ff6ApSelected References{l}m/l"7 !"#$%M ` ̙33` 3` 3333f` 999MMM` f` f3` 3>?" dd@,?" dd@ " @ `" n?" dd@ @@``PR @ ` `p>>;(
Zvgֳgֳ ?Pv
T Click to edit Master title style!
!:
T$vgֳgֳ ?v
RClick to edit Master text styles
Second level
Third level
Fourth level
Fifth level!
S
`vgֳgֳ ?``v
=*
`vgֳgֳ ?` v
?*
`D vgֳgֳ ?` v
?*H
0h? ? ̙330vn@(
TD#vjJjJ ?P.
v
_*
f
T#vjJjJ ? .v
a*
fp
01 ?*
v:
T$vgֳgֳ ?
@v
RClick to edit Master text styles
Second level
Third level
Fourth level
Fifth level!
S
Zd$vjJjJ ?nP v
_*
f
Z$vjJjJ ?n v
a*
fH
0hdS? ? ̙3300( X
H
0hdS ? ̙33F>P(
#lD&vgֳgֳ ? v
#l&vgֳgֳ ?` pv
"H
0h ? ̙33 ~v`( n48
#lbgֳgֳ ? p
"H
0h ? ̙33kp( ,
Zcgֳgֳ ?P0
OAcknowledgments$0ff
Ncgֳgֳ? Y
The research was supported by FDA/CDER RSR Funds, #96010A and #99/00008. Thanks are due to Dr. Lu Cui for sharing some of his slides
: 33 ,a H
0h ? ̙33
1)P(
Pr
P Se 0
P
`de1?
M *
P
Te1?vp
Bauer & Khne (1994, Biometrics)
Bauer & Rhmel (1995, Stat. In Med.)
Lan & Trost (1997, ASA Proceedings)
Fisher (1998, Stat. In Med.)
Posch & Bauer (1999, Biometrical J.)
Kieser, Bauer & Lehmacher (1999, Biometrical J.)
Lehmacher & Wassmer (1999, Biometrics)
Mller & Schfer (2001, Biometrics)
Berry (2002, ASA Biopharmaceutical Report)
Brannath, Posch & Bauer (2002, JASA)
& & & . etc$33f^&
?H
P0h ? ̙33D<(
Z$fgֳgֳ ?`
[ $0ff
Nfgֳgֳ?
1The materials of this presentation are selected from the main results of our RSR research work.
Cui, Hung, Wang (1997 ASA; 1999 Biometrics)
Lawrence & Hung (2002 ENAR talk)
2` O ,`JH
0h ? ̙33R( @
ZDdgֳgֳ ?P
M f;
Nfgֳgֳ?
S Background
Sample size (or amount of statistical information) is one of the design specifications vital to success of Phase II/III (confirmatory?) clinical trials
It relates directly and closely to the true effect size (treatment difference normalized by the measure of variability) of the targeted response variableNT
0f b H
0h ? ̙33}%(
ZDggֳgֳ ?P
M f
Nggֳgֳ?` b
( Background
Common recommendation
Make educated guess about the effect size
and plan sample size to detect this effect size (or a
range of plausible effect sizes) with sufficient
power [e.g., > 90%  Hung et al (1997 Biometrics)]
This is always good because the fixedinfo design
1) provides statistics that have important good
statistical properties
2) avoids datadriven adjustments that may induce
biases (statistical or operational) making the
results not interpretable F
( n
0f
,lH
0h ? ̙33x( (
Td1?`
RBiases use internal data to: change selection of patients, tuneup endpoints, drop/select centers, change patient mixture, do more data dredging or torturing to adjust analysis to get the desired conclusion, eliminate potential dropouts, change design to make treatmentrelated problems go away, tune up any design element to make the treatments easily differentiated, adjust or sample to a foregone conclusion & & & & ...
l f fH
0h ? ̙33,(
,
,
Zhgֳgֳ ?P
M f
,
Ndhgֳgֳ?`
Background
But & .
The effect size depends on a primary
parameter (e.g., mean treatment difference) and
nuisance parameters (e.g., standard deviation,
background event rate)
The effect size for detection may need to be clinically significant or meaningful (sometimes minimum clinically meaningful) benefit/risk assessment (subjective) that might be doable only in the late of the trial, hard to reach consensus F
F
0f
N )H
,0h ? ̙3307(
0
0
Z$igֳgֳ ?P
M f
0
Nigֳgֳ?`
Background
But & .
The effect size may depend on patient mixtures potential heterogeneous effects in subpopulations
For a hard clinical outcome endpoint, educated
guess about effect size is difficult
e.g., for composite event endpoint, require educated guess of where the potential signal lies and what noises may be F
p1P
0f 3 2 H
00h ? ̙3349(
4
4
ZDjgֳgֳ ?P
M f"
4
Njgֳgֳ?`
Background
But & .
The effect size for detection may depend on $$ benefit/risk/cost consideration
& & & & & & . etc
Practical considerations effect size for detection
can be a moving target and change as background
circumstances change and maximum amount of
statistical information one can commit to may also
change F
Zn
0f / 33 33H
40h ? ̙33<(
Zigֳgֳ ?P
M fe
Nkgֳgֳ?
; Background
Experiences:
Often oversimplify clinical trial designs and inferences and impose too many restrictions to the designs. If a trial fails, it is difficult to know whether it is because the treatment does not have an important effect or the study was underpowered for detecting it.n 2<
0f H
0h ? ̙338( ?p,
8
8
Zdkgֳgֳ ?P
M f
8
Nhgֳgֳ?
Background
Lan (2001, FDA/OB MiniSymposium)
If we know the values of design elements (e.g.,
effect size) a priori, No Need and Not Ethical to conduct a confirmatory trial
Bauer et al (2002, Method Inform Med)
& .. It does not make sense to apply uniformly most powerful test in an unchanged design even if we have convincing evidence that this best test in the preplanned design may be severely underpowered & & . 2Z
0f ! 33 + >"H
80h ? ̙33/' (
Z%gֳgֳ ?P
M f
ND%gֳgֳ?
*
Need to enhance flexibility in traditional clinical trial design/analysis strategy because
practical considerations may
change and often be
unpredictable at the design stageD S,[, H
0h ? ̙33(
0h( X
f$%gֳgֳ ?
%
n
N%gֳgֳ?
Midcourse modification of design specifications
 adjust sample size
 change tested hypothesis from superiority
to noninferiority or vice versa
 change from one prespecified primary
endpoint to another prespecified endpoint
 change test method
 drop a treatment arm
& .. etc$?0( ,{PjH
0h ? ̙33NF@( ̙33
NĐ%gֳgֳ?:
Type I error rate may greatly exceed the acceptable level
Statistical power may be compromised
Traditional estimate may be severely
biased
p;%/( 3f
T$%1?``
<Impact of Design Modification Based on Interim Observed Data(=P<0f
jH
0h ? ̙33!3+P<(
<
<
Nd%gֳgֳ?p*
_ (( 3f
<
T%1?``;
b Sample Size Reestimation(P0f
j
<
`%1?L
(Literature on sample size reestimation is abundant.
Increasing sample size (or amount of statistical
information) based on nuisance parameters without
breaking blind
 has little effect on type I error
 may preserve the intended power level
 needs little or mild statistical adjustment (e.g.,
estimate, CI)
Wittes & Brittain (1990), Gould (1992), Gould & Shih (1992)
Shih (1992, 1993, 1995), Birkett & Day (1994)
Jennison & Turnbull (1999, book), & & & etc5<
2 6 u fC H
<0h ? ̙33$`pC(
p
p
Nd%gֳgֳ?p*
_ (( 3f
p
T$%1?``;
b Sample Size Reestimation(P0f
j
p
`%1?Lq<
,HBut & ..
Lan (1997, ASA talk), Liu (2000, ICSA talk)
e.g., knowing the components of the variance can
lead to estimation of treatment difference; hence
sample size reestimation based on variance
might affect type I error depending on how it is
processed (e.g., by obtaining TSS & WSS)
(<5 `
p
c$A??0B
p@
ZD1?@pP
B
p@
ZD1?pH
p0h ? ̙33"D(
D
D
N%gֳgֳ?p*
_ (( 3f
D
T%1?``;
b Sample Size Reestimation(P0f
j
D
T%1?L
Increasing sample size (or amount of statistical
information) based on the internal data path
may substantially inflate type I error, bias the estimate, invalidate CI
 crude estimate of maximum amount of inflation
obtainable, at least by simulation
L %H
D0h ? ̙33
e
]
 (
Td%gֳgֳ?]
3
Nę%gֳgֳ?]
]
TDagֳgֳ?W
]
TU9gֳgֳ?
Q
T$V9gֳgֳ?
Q
TV9gֳgֳ?
Q
TV9gֳgֳ?
'
a *f
!
TW91?`g
\ Sample Size Reestimation&,0f
#
`X91?
v:
P f
$
TdX91?
ZQuestion:
At an interim time of a trial, if the observed treatment difference is far smaller than expected, we wish to increase sample size. Then, what adjustments are needed to
perform valid statistical testing?
Bn(n 3f 3f
H
0h ? ̙33
meL(
Lr
L SX9`@
%
L
`$Y91?
M f
L
TY91?p
Bauer & Khne (1994, Biometrics)
Proschan & Hunsberger (1995, Biometrics)
Lan & Trost (1997, ASA Proceedings)
Cui, Hung & Wang (1997 ASA Proceedings, 1999, Biometrics)
Fisher (1998, Stat. In Med.)
Shen & Fisher (1999, Biometrics)
Lehmacher & Wassmer (1999, Biometrics)
Mller & Schfer (2001, Biometrics)
Liu & Chi (2001, Biometrics)
Brannath, Posch & Bauer (2002, JASA)
Lawrence & Hung (2002, ENAR)33 D 13H
L0h ? ̙33H@( Ow@
pB
H8c?
NDZ9gֳgֳ?0P
W#Experimental (T) with N subjects$$
N[9gֳgֳ?
QControl (C) with N subjects
N[9gֳgֳ?Pp
<Baseline 3
T$\9gֳgֳ?]
3
N\9gֳgֳ?@
:
]
N\9gֳgֳ?@W:
7
TD]9gֳgֳ?
Q
T]9gֳgֳ?
7
T^9gֳgֳ?
Q
Td^9gֳgֳ?
'
J
*f
T^9gֳgֳ?
!
8
T$_9gֳgֳ?,
a
7
T_91?
Z ,
`_91?
v:
6 f
T`91?
HTest H0: D = 0 vs. H1: D > 0%n
3f,B
TD1? B
TD1?P B
TD1?PPB
TD1?
Tda91?
XD = T  C*
j 3f
Td91?opv
! Sample Size Reestimation B",0f,,
T91?P
N
s = 1( f fL
`$91?2A
To detect D = d at sig. level a and power 1b,
N (per group) = 2(za+zb)2/d21! % f f:H
0h ? ̙33x (
x
x
T9gֳgֳ?]
3
x
N9gֳgֳ?@
:
]
x
ND9gֳgֳ?@W:
7
x
T9gֳgֳ?
Q
x
T9gֳgֳ?
7
x
Td9gֳgֳ?
Q
x
T9gֳgֳ?
'
J
*f
x
T$9gֳgֳ?
!
8
x
T9gֳgֳ?,
a
7
x
T91?
Z ,
x
`D91?
v:
6 f
x
T91?opv
! Sample Size Reestimation B",0f,,
x
`d91?vO
`The process involves:
1) projection of effect size with internal data
2) determination of when to change sample size
3) reestimation of sample size
4) adjustment of statistical inference methods
 test statistic and/or critical value
 estimator and confidence interval
& fH
x0h ? ̙33Z
(
`91?i
> f
T$91?5
T Sample Size Reestimation
(nonsequential trial)
Plan to enroll N=100 subjects/group to detect
d = 0.46 at a = 0.025 and power 90%
After 40 subjects per group contribute data,
the estimate of D leads to d* = 0.37
Reestimate total sample size M = 150/group
>(,0f
,, f0 ,H
0h ? ̙33^(
`91?i
> f
P
T91?
r Sample Size Reestimation
(nonsequential trial)
At the end of the trial (M = 150) , compute the CHW adaptive test [Cui, Hung & Wang, 1999]
U = (40/100)1/2Z0.40 + (60/100)1/2W0.60
W0.60 : normalized test for the additional 110 subjects per group after the interim time t=0.4
Lan (2001 FDA/CDER/OB Minisymposium, 2002 FDA/ASA workshop)#P([i2(,nP`A=i,0f
,, f 3f( $$$$
$$$$
[ < >:H
0h ? ̙330(
`D91?i
> f
"
T91?B
Sample Size Reestimation
(nonsequential trial)
U is standard normal under H0
If U > 1.96 , then conclude D > 0
Significance level = 0.025
U is more powerful than
original Z w/o increasing NA(#i,0f
,, f$$$^$H
0h ? ̙33VN(
`91?i
> f
Td91?
T Sample Size Reestimation
(nonsequential trial)
Estimation & CI for D 
Lawrence & Hung (2002, ENAR talk)
Construct consistent estimator and valid CI for D
CHW test is Zratio of the consistent estimator
PDii,0f,, f ! 1 3 H
0h ? ̙33? !!( @
pB

H8c?@G

N$9gֳgֳ?
W#Experimental (T) with N subjects$$

N9gֳgֳ?p
QControl (C) with N subjects

N9gֳgֳ?
<Baseline 3pB

Hp?Y8pB

H8c?pB

H8c?b
b
pB

H8c?yypB

H8c?11pB

H8c?

To:gֳgֳ?]
3

No:gֳgֳ?]
]

TTp:gֳgֳ?W
]

Tp:gֳgֳ?
Q pB

Hp?vB

Np?p

Tq:gֳgֳ?
y=
70

Ttq:gֳgֳ?
: N/5

Tq:gֳgֳ?
`
:2N/5

T4r:gֳgֳ?
Q

Nr:gֳgֳ?qj
?
0

Ts:gֳgֳ?
'+
Q IA1
20%*f

Ttt:gֳgֳ?
a
O IA2
40%*
f

T4u:gֳgֳ?,
P
Final
100%*f

Tu:1?`
= Sample Size Reestimation
(group sequential trial)4>,0f",

`u:1?
v
S N4ff

Tv:1?
*HTest H0: D = 0 vs. H1: D > 0%n 3f 3f 3f 3f 3f 3f 3f 3f 3f,B

TD1?@ B

TD1? @B

TD1?B

TD1?
!
Ttw:1? #]
XD = T  C*
j 3fH
0h ? ̙330(
`
Tw:1?`
Sample Size Reestimation
(group sequential trial)
N is planned to detect D = d at level a and with power 1b
At interim time s, estimate )s 0 < d* < d
(say, based on conditional power)
Y increase sample size from N to M, approximately
M=N(* / d*)2
Total information changes from 1 to w = M/N Compute b = (w  s)/(1 s) E(BFRn72<O,f0f
,, f G 1 > (B
TD1?p0 0 pH
0h ? ̙33@?(
T4x:1?S
Sample Size Reestimation
(group sequential trial)
For interim analysis at time t when Mt subjects contribute information,
compute Nt = (Mt  Ns)/b + Ns and t = Nt/N.
Adapt the traditional repeated significance test:
Traditional: Zt = Zs(Ns/Mt)1/2 + Wts(1 Ns/Mt)1/2
New: Ut = Zs(Ns/Nt)1/2 + Wts(1 Ns/Nt)1/2
Cui, Hung, Wang (1999, Biometrics)xE(Nn(/n<2Z&27(92
#,f0f
,, f' 3f 3f! 3f f
m , 33#33mE B
TD1?p0 0 pH
0h ? ̙33D<P(
Tx:1?_
V Sample Size Reestimation
(group sequential trial)
{Ut} w/ N possibly changed to M
& {Zt} w/o change of N have identical distn.
Find critical value Ct at time t based on the initially selected alphaspending function
Reject H0 if Ut > Ct ; otherwise, trial continues
Cui, Hung, Wang (1999, Biometrics)D(nP5n2$n,f0f
,, f 3f 3f& 3f 3fA 3f 3fP 3f 3f 3f
j33#33K%#W B
TD1?p0 0 pH
0h ? ̙33`<(
Tx:gֳgֳ?@ @
Empirical type I error rate(Adaptive test; type I error = 0.0255 w/o N increase; Gaussian)(increase N by <= 4x; O BrienFleming boundary)>,
fdQ 4
ZA??X$
00H
0h ? ̙33p$( FpA<
TTy:gֳgֳ?
~Empirical power(Adaptive test; 1b = 0.587 w/o N increase; Gaussian)(increase N by <=4x; O BrienFleming boundary)Xu,
fR; 1
ZA??X$
00H
0h ? ̙33F( _
d
Ty:gֳgֳ?@ @
0Empirical type I error rate(Adaptive test; type I error = 0.0249 w/o N increase; Binomial, pC = 0.20)(increase N by <= 4x; O BrienFleming boundary)r,f2:\;
ZA??X$
00H
0h ? ̙332r(
P
Tz:gֳgֳ?
Empirical power(Adaptive test; 1b = 0.60 w/o N increase; Binomial, pC=0.20)(increase N by <= 4x; O BrienFleming boundary)
,f":D:
ZA
??X$
00H
0h ? ̙33VNl( ,
l
l
`tz:1?i
> f
l
Tz:1?!
T Sample Size Reestimation
(Example: group sequential trial)
Plan to have 100 subjects/group to detect
d = 0.46 at a = 0.025 and power 90%
After 50 subjects per group contribute data,
the estimate suggests to detect d* = 0.46/%2
Reestimate sample size M = 150/group
w = 1.5 b = (1.50.5)/(10.5) = 2
2,0f,!,,) W ,H
l0h ? ̙33pL( b
p
p
`4{:1?i
> f
>
p
`_<1?f
x Sample Size Reestimation
(Example: group sequential trial)
Suppose that an interim analysis will be done
when additional 50 subjects/group contribute
data (Mt = 100)
Nt = (100  50)/2 + 50 = 75 and t = 0.75
Suppose that OF alpha spending function is
originally used for interim analysis
Then the critical value for the adaptive test
at t = 0.75 is C0.75 = 2.36=n,0f,!,,a 3f 3f ) H
p0h ? ̙33s#t( B
t
t
``<1?i
> f
t
`d`<1?
I Sample Size Reestimation
(Example: group sequential trial)
The adaptive test at M0.75 = 100 is
U0.75 = T0.50(2/3)1/2 + W0.25(1/3)1/2
W0.25 is the normalized test performed on the
additional 50 subjects per group
If U0.75 > 2.36 , then stop the trial and conclude
that experimental treatment is superior to control@&ii,0f,!,, 3f 3f O_H
t0h ? ̙33@8x(
x
x
``<1?i
> f
x
`$a<1?
2f Sample Size Reestimation
(Example: group sequential trial)
If the trial continues to the end, then
the final adaptive test (i.e., at M1 = 150) is
U1 = T0.50(1/2)1/2 + W0.50(1/2)1/2
W0.50 is the normalized test performed on the
additional 100 subjects per group
If U1 > 2.01 , then conclude that experimental
treatment is superior to control@Zi*i,0f,!,,L 3f 3f
PMH
x0h ? ̙334t(
<
Ta<1?`.
p Sample Size Reestimation
CHW adaptive test has type I error rate attained at the targeted level and large power increase and its implementation is very easy
Consistent estimator and confidence interval compatible with CHW adaptive test are readily available
All the above discussions are based on asymptotic (i.e., sufficiently large sample size) theory
Cui, Hung, Wang (1999, Biometrics)
Lawrence & Hung (2002, ENAR talk)
((fcFF,f0f f b D 33sCH
0h ? ̙33#bZh( ,
h
h
Ta<1?`
6 Sample Size Reestimation
CHW adaptive test reduces to the conventional
test if sample size is not changed. So do the
consistent estimator and confidence interval compatible with CHW adaptive test.
Cui, Hung, Wang (1999, Biometrics)
Lawrence & Hung (2002, ENAR talk) (F,f0f f D 33CH
h0h ? ̙33%JB(


TDb<1?`y
Sample Size Reestimation
CHW adaptive test has another look using a combination of pvalues from the incremental group data [Lehmacher & Wassmer (1999), Brannath, Posch & Bauer (2002)]
(,f0f fd >33 P H
0h ? ̙33(
Tb<1?U
^ Sample Size Reestimation
Sample size reestimation criterion
After obtaining the observed )s at time s, one could recalculate sample size M such that conditional power
CP(d*)
= Pr{CHW rejects H0 at the end  )s , ) = d*} = 1b
for the new intended d*. Then, the power of CHW for detecting d* is at least 1b
Lawrence (2002, personal communication)6(
%2r<4n<{,f0f f # # V (
(H
0h ? ̙33og (
Tc<1?`Fd(
Cc Operational Conduct Issues
Sample size reestimation based on unblinded
data opens rooms for operational biases.
During adaptive change, only unblind data that
are necessary to be unblinded in order to avoid
operational bias. Standard Operation Procedure
(SOP) must be in place in the protocol and trial
conduct must comply with the SOP.
la P f0f,F PD
J H
0h ? ̙33>60(
Tdc<1?F
Operational Conduct Issues
Sample size reestimation based on unblinded
data opens rooms for multiple analyses that may
lead to more protocol amendments and other changes of design elements. These types of changes or amendments (potentially driven by current data) may introduce problems in the interpretation of the results. Attention should be given to this potential hazard with the design. B f0f,s D
EH
0h ? ̙33rj@d(
d
d
Tc<1?F
F Operational Conduct Issues
But & .
Most of the operational conduct issues related to sample size reestimation based on unblinded data are also encountered in traditional designs with interim analyses
` f0f,
KH
d0h ? ̙33P*(
T$d<1?`Fd
n Operational Conduct Issues
Recommend that sample size modification be done, if needed, by an independent third party that has no conflict of interest issue
What took place after sample size (or any design) modification needs to be documented fully
P f0f, fH
0h ? ̙33[`(
c
Td<1?Fd
U Other Issues
Estimation following datadriven sample size change is an important issue, particularly the effect estimate may be used to plan future superiority trials or activecontrol noninferiority trials.
Careful consideration of the benefit and risk of such sample size modification is needed in practical applicationsn<Fs f0f,: H
0h ? ̙33r"pT(
Tz
T
Td<1?Fd
B Other Issues
Bauer et al (2002, Method Inform Med)
& .. Clearly in such designs more logistics must be put in to properly handle all problems of
interim analyses including their consequences for the design. They need rigid rather than flexible planning modalities. & &
v< f0f, % ?H
T0h ? ̙33X!(
X
X
TDe<1?`
e Summary
Conventional fixedinformation design has made tremendous contributions to clinical science. The statistics have good statistical properties.
This design needs to permit sample size adjustment when it falls short in many applications, e.g., in studying endpoints for which
 prior data are poor or hard to provide reasonably
good educated guess about effect size
 minimum clinical meaningful effect not availablen<P f0f, _ZH
X0h ? ̙33]
\(
\e
\
Te<1?`
Summary
Conventional design properly adapted for necessary sample size adjustment can be very useful with
 proper planning to avoid any operational conduct
change that may lead to bias
 proper adjustment of statistical analysis method
estimation issue needs attention and research
Price paid for datadriven adaptation: lose
good statistical properties [Jennison & Turnbull (2001), Tsiatis & Mehta (2002)]. <S f0f,f 2 b H
\0h ? ̙33\T `(
`
`
T"v1?`j
0^ Summary
Adaptive design with proper planning is very attractive with proper caveat
 change sample size or randomization allocation
 change study hypothesis (e.g., superiority vs. non
inferiority or equivalence)
 change test method
 change the primary endpoint from one pre
specified endpoint to another prespecified one
 drop futile or unsafe treatment arms and more& & ( More work to do (< f0f, ( ((H
`0h ? ̙3308(
H1 ?*
<p
ff<gֳgֳ ?
@<
The subject of adaptation of design./analysis is closely related to modification of design specifications& The following are the most frequently seen modifications of design specifications in NDA/IND submissions.H
0hdS ? ̙33
A0Z(
N1 ?*
<
#ldf<gֳgֳ ?
@<
In our previous work, we have shown that type I error rate may be inflated or statistical power& ..H
0hdS ? ̙3320(
^
S*
=
#lg<__
@=
0In that project, a simple approach was proposed to properly adjust & ..
This adaptive test has the type I error rate maintained at the desired level and its power can be substantially improved as a result of sample size increase. The test can be implemented very easily in the sense of preserving the flexibility of interim analysis schedule Lan and DeMets procedure has.,WH
0hdS ? ̙3320(
^
S*
=
#lDh<__
@=
0In that project, a simple approach was proposed to properly adjust & ..
This adaptive test has the type I error rate maintained at the desired level and its power can be substantially improved as a result of sample size increase. The test can be implemented very easily in the sense of preserving the flexibility of interim analysis schedule Lan and DeMets procedure has.,WH
0hdS ? ̙3320(
^
S*
=
#lh<__
@=
0In that project, a simple approach was proposed to properly adjust & ..
This adaptive test has the type I error rate maintained at the desired level and its power can be substantially improved as a result of sample size increase. The test can be implemented very easily in the sense of preserving the flexibility of interim analysis schedule Lan and DeMets procedure has.,WH
0hdS ? ̙3320@(
^
S*
=
#li<__
@=
0In that project, a simple approach was proposed to properly adjust & ..
This adaptive test has the type I error rate maintained at the desired level and its power can be substantially improved as a result of sample size increase. The test can be implemented very easily in the sense of preserving the flexibility of interim analysis schedule Lan and DeMets procedure has.,WH
0hdS ? ̙330p!(
N1 ?*
9S
#l$%gֳgֳ ?
@9
Our RSR projects cover two subjects: adaptation of design/analysis and quantification of strength of evidence in Phase III clinical trialsH
0hdS ? ̙330!(
N1 ?*
9S
#l%gֳgֳ ?
@9
Our RSR projects cover two subjects: adaptation of design/analysis and quantification of strength of evidence in Phase III clinical trialsH
0hdS ? ̙33
0@Z(
@
@ N1 ?*
<
@ #lf<gֳgֳ ?
@<
In our previous work, we have shown that type I error rate may be inflated or statistical power& ..H
@0hdS ? ̙33
0HZ( 9<.
H
H N1 ?*
<
H #lg<gֳgֳ ?
@<
In our previous work, we have shown that type I error rate may be inflated or statistical power& ..H
H0hdS ? ̙3320 l( m 0N
l^
l S*
=
l #li<__
@=
0In that project, a simple approach was proposed to properly adjust & ..
This adaptive test has the type I error rate maintained at the desired level and its power can be substantially improved as a result of sample size increase. The test can be implemented very easily in the sense of preserving the flexibility of interim analysis schedule Lan and DeMets procedure has.,WH
l0hdS ? ̙33
0tZ(
t
t N1 ?*
<
t #l$g<gֳgֳ ?
@<
In our previous work, we have shown that type I error rate may be inflated or statistical power& ..H
t0hdS ? ̙33200( ,
^
S*
=
#ldi<__
@=
0In that project, a simple approach was proposed to properly adjust & ..
This adaptive test has the type I error rate maintained at the desired level and its power can be substantially improved as a result of sample size increase. The test can be implemented very easily in the sense of preserving the flexibility of interim analysis schedule Lan and DeMets procedure has.,WH
0hdS ? ̙33xWML\U>JoX(2:
EScҒRөuȮD`bꆘnX4BkhM&$hLLx@[OQ!\~\KmmĘnKDN]PyMyy]o=NT/gZw
<)bP[,Seddjhش=I=P*tked7Q
6d*Ӗ\8`=G)
,ziDkViEN3#aeM8ًI[n"T,;~R~qA!J$䛸0??Sа@&qyX
e]4)
\Bs.;t*]>//SA~"}z>z+B穿sF
5UCYʭ48B힠H$B=Ojj]BǢj@L@%T#BkzTI&1i
i5HZ!=u/#8۩b!v{gJ.HyS6ġ.+Ω^
:H{P^ ;,?RU,l%AZt<{>'p,݆fo"BѥGC͐t$4
Ps:mޙi@hfè[>&qaņa.q<6iиB/E>smlS0WKܨ=(55JE15ef*Ywe3떉JNrxS/UT=<930^ZĎ@2ЏOZ}143CwrXYn@h5b!x7Ni=q3ӉӕUGH!u!}`ϓ.) ^WV%eJ[9;`t 6~v(nt}L:"ѡ2;xDqGܠ'9DxWML\U>Jon(R5UwWy2U䛩Rʴ%'1MPCʠB5K'n~t#9X_.ߌ}9)7m[.
Jp::L!u 49DB>xphaYRWNӐ(M1KWѵ.ѥ~esk;[42H[> "hj+M6
S&)R'=e0(#j=$tck)Tz=.r"ڐITai2j@fZǰfl/"Bn$)?q߶t >ß?#0ҳGFq+T?Q ?"U,l%PCX8n]>>7oEzfJi!=bJ:&MBԜݧOAwf30114ULkI之Ȱ0m8vMd
{ШEsn9>.+AnJ7i]ͺf5URgy"?+kؑz5gBt꣼\,tFeMM<#UŴEggi8_3viy&LmeݻF(D!8dEaJו*;^iko'CV^u'>!Y9&dg>4lnнCZP
0xZ}l{O{Koi)Xz{`ˇU
X (*ZEԡ5G2Ku1qfb\Xl
&~,[p&cQIS2b9罷esyxy7sURNbtR%0 :D~R)?Jyb&[ʀ@9PL*:ߦzOIZFۑ]`T 74?/͜uE
UtuIf4!m
/o
3Z\"rFc61˟+<lOFۀ?8
Lsj&<:0 fs@=0\4@Eb"XBR=@X+en@;}3mpNx2^բZY355#5~9llS+Ϊ{MUx5>1!sؘ:9=yb4u2;7^7DDj^Jva=ݜVMr/ּ8**[V*"%`wc o4]#N>Xuo5\Q6_"F_ NauYк ghnx0p~B飧n4SޯLF&~@gǦt!Q~z@_PAܖ7G= cǎQˁѣ.F6W=e\;m@[e :v7oM6Q1%&Ť4l0?5st:scڈ
{mNRo{TNؓ\+x5&݀;9M=><> :jOŭI?Cp:l)/"=J_D2pNtN_'4pyscƾޤՉ',r:^qzBA1Q
~NlDHG&Dpw_.6nW"s[҇djqh
Mֺ6$ximTAf#H}!T+šE M%357'E&B~
FpA(
*o}PAINy3t:}xK1U_
= +وPgՊaRQLLMFN'po"
GNFsU/\IW$ʹٯkG3r# w/G̱L1N\4ZyP&DQ1Qב^#DFߦ$7yokb()W=q_I܋:g%%!2j5?K(=FNxh?&x2ОD㲤DJY]]S3g\n,dssTdjyq%?>#OXCOL`,@d,hIpp!?
iOC~Z*~ t(,(l((8@Dt],bA,@d,lp(
8HlQ,ˁL>lƻ
lȷޑ+,p;smn8B6x<Rby9Ёur0yPYXPx&<`6
]6
PA9y^ /V(,CX
8X
"8ȴ;:(^c~r%:)w)#CIDޒG~(O8!?s}Б(?^p]ׅc%OC;{ǆ4 Hޙ\ӻk;=]ǽ{m"A/3(7wۙTi'#u];v^,><=UJY3i.F2(S/{xӨ4_aDsnKQ9:.w0ހk:p5592>&u0>M$}0jp{PF&9,.q+^(he3JڬQ.fjv8ˌrqP0vFE=_ct٢QΘ?pTHGZT;L!e14[CUX)O;p{no\_\{\].)p~*i}psk41
#19k;Ukwc쯇9wy4KiѫT{OSK_J2CavavU~\MrWDtxX2M
3( zWyk_`WWt&vo$;D\E<_%JaN]WϰxN /ߥl1g},7ro֨AϮ#=dB/D_ۮݔ}O}zt%:3fK#[{r_cqpK?z}wҏZhSEJ
[c
0xZ}lUs߶)k[o/ \cjũE=(ԕ*ʚ쒉>#!,sbdfMdmJ2osmK÷{?z9*^5R^"*'dYTHUk(%@i!nTUT:yOk?ډ7j옰/!hOo[_[`hsMedxڮ*o lU>$wНʹO>_lτFۀ? 8L.jKZ&<0\9\
`0h@H`)h,'ۈڀJ=@Xk7*:K
iӄw*)]TD>WR/k%g&g̛1kU×
seYlݘ[%WM}L);gasuCc 2$tEöm$rl+rei"ڲ,n1MڵQSVv#/NK=V{=h/ea׳a7yJ2Õj鳸7:Y9cXه{a0m]ןHN$nkqvnr*.i6;1l1}'L1pjBY>S}){~2vD]qY$3fwK߮>g m֥ڵ[8y.iz[w@2z7FVJ6h핰/ގ=&zeX
rʇd+f R
;hb?hc WعzKCP#PX(wЭ>l>88nPjlG/Vzhf5M?_Vs?B䏾;jQKsɂgGi۔vXz$h5ظvG^?w)<3`"yâkIe3Ƥq{'?hB/ߠ}
!щ+4Qzq8l]8˾s֑Qu\m߳oZˁ\ɉ.ىDO4N4]4/zn3/zyߦbe2bxQ~N,'>"$3ei&g"8ݎ/
߈]`eH$ŕk~mPDJSmZW GFQ ?>T*V0lYVآv}aur3F
IófnЁcÄ_ %AԟB5aqjkq:SǩwIAhjp7`E?5J?Z?L!^%4l¸)#g`pőӊNQ?LGUCnC1Avݴ]4t
qU~9)ÉFB1a4ɳŧ+QaTdMud(i,IǍD㿇pV+SqE_IX܋u&KʭDsek~P{72Nxh?&m<iC"qY^^.+**duudRΙ3G_^vvvJxUT7$?LAIp"<@8;28x8ђ
<.11. JA~
ST,t((\(<(8<@Dtq]@<@<\p(<xH\S,˃L>\Ż\wޑ+p{s
mnxB6x<RbyyЁu0yP98Px&<.
].
pPA9qy^ /R(C8x8 "8t{:(^c~r%%w~ ^;vL"'N'Oo)O>?C>СCu$F0*bb,u1Y {'Ɔ4 JٝX۷{'·{m"AYl*]=;h_/fώe~` u瞪%[G
,gc+%ҔP^IYdG=]d
A7BTE˯su
7hMd6õp
kkwżA7Ul%>_UFk:Ϡ)jZDRӳ4g`ofzRRj[2GUv6+Vrp~[xj+ōj*EG,<攇 Ý0Q5A4ܠ
pf65psh
j"\XFcw_\rMsYcW}%:}#77SLTk_A)l4LyrZZlZXhZsfbYp͆_>='gq_bY6
g4w,3_^ӭk%qjQ"wi^bPϥKD]6y\טL>pO7tTz&g=rz//L_ǩoR}fY17Zdox 6kAڏ]H9¹#w&ZlQMt\/PWgxWMl[EuĎۉJj5
A6Q.rCl7X\"WPŁ3^ʹR#!EHp!H=%[&NEA2ofvfvvv۟
ZHh4 Êbhpl=Rt8d0@b^N#kå9(ѯLC+Pm=
J/mw_}9tϔXܡb)(iw\?_2q?A9/w4ԯ}2">9fZ? D]5wG{{A} \>5M
`xx8Fď;ď{3$hdX^8}T.]8˫JTtNOVzqޓrA4ys\}0~$E0:d)%wRiӻ)]@IbHI&Qs`/7GanoSz߰M7VX^U;G0TēB{m7vӄ983eN
v!aWʨ+c<]tYqM]!Wv.vJl}P'A&%8c61N{cֿ7l
O}]VԏV
J_g8Zk[!Tw۫nJLxxf=Ok].
#$q?g_N[Łk3G(z!?@UEO'n;6>;[L[k\[Ꟊ3˻8o{yM}lomOt݂WKKOxf<z[3ϻmQ9?ژo]_Q
65a0_,CLSnY J:g Lc*͟lgg\zBH=k㐸RI+yB
z"!MCZ40+mm1Q@\:y)v5j1xL2BnyNvں;T=}!Łfma(?mE\?܄d;{x=w#z(Orh0(B//
u"i6 {mO 4&Cz@*%)PK!_&+W Kr@G&gNX)cr1p vZ{
?PE?IQT@0΅2pET&[%ajk~>ܞ8hW[_0dIM73ޖAh;2o(
!Onscreen ShowV41=Times New RomanSymbolWP Greek Century WP MathAWP MathExtendedAWP IconicSymbolsADefault DesignMicrosoft Graph 2000 ChartMicrosoft Equation 3.0kInference and Operational Conduct Issues with Sample Size Adjustment Based On Interim Observed Effect SizeNo Slide TitleNo Slide Title7Selected References in Adaptive Design/InterimAnalysisNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleEmerging StrategyNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleSelected ReferencesNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleNo Slide TitleFonts UsedDesign TemplateEmbedded OLE Serb
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0bEquation Equation.30,Microsoft Equation 3.0/0DTi MSGraph.Chart.804Microsoft Graph 2000 Chart0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0bEquation Equation.30,Microsoft Equation 3.0/0DTimes New Romanbb(b0LbLb~0DSymbolew Romanbb(b0LbLb~0 DWP Greek Centurybb(b0LbLb~00DWP MathACenturybb(b0LbLb~0@DWP MathExtendedAbb(b0LbLb~0PDWP IconicSymbolsAb(b0LbLb~0
f@ @@``
@n?" dd@ @@``:
2
F4:;
"'. ?#&,'""A !#/!@($.;(9) 3
<>.14/==0CEF<2$Hl$:?K=lծ 'z2$+xxF4;[z2$9«ɘVyu샄9
z$z$z$z$z$z$z$z$z"$VLK8xunzl1z"$g<=]pfzS~1? f̙f333@>g4KdKdXb~0Pb
ppp@g4BdBdXb~0Pb
(ppp@<4ddddbbLC?O=jInference and Operational Conduct Issues with Sample Size Adjustment Based On Interim Observed Effect Size&kj(
H.M. James Hung (DB1/OB/OPaSS/CDER/FDA)
Lu Cui (Aventis Pharmaceuticals)
SueJane Wang (DB2/OB/OPaSS/CDER/FDA)
John Lawrence (DB1/OB/OPaSS/CDER/FDA)
Presented in Annual Symposium of New Jersey
Chapter of ASA, Piscataway, NJ, June 4, 2002 ILZ[Z)VXft(!H
Disclaimer
The views expressed in this presentation are not
those of the U.S. Food and Drug Administration,
nor of Aventis Pharmaceuticals.
Dr. Lu Cui was one of the primary investigators
of this research during his tenure in FDA.
0f,,ROG6Selected References in Adaptive Design/InterimAnalysis76'PH
Emerging Strategy$0ff6ApSelected References{l}m/"7 !"#$%&X7(
X
X
TDe<1?`
{ Summary
Conventional fixedinformation design has made tremendous contributions to clinical science. The statistics have good statistical properties.
This design needs to permit sample size adjustment when it falls short (perhaps by surprise) in many applications, e.g., in studying endpoints for which
 prior data are poor or hard to provide reasonably
good educated guess about effect size
 minimum clinical meaningful effect not availablen<P( f0f, uZH
X0h ? ̙33ro(
\
0zb
Chart MSGraph.Chart.804Microsoft Graph 2000 Chartd0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0bEquation Equation.30,Microsoft Equation 3.0/0DTimes New Romanbb(b0LbLb~0DSymbolew Romanbb(b0LbLb~0 DWP Greek Centurybb(b0LbLb~00DWP MathACenturybb(b0LbLb~0@DWP MathExtendedAbb(b0LbLb~0PDWP IconicSymbolsAb(b0LbLb~0
f@ @@``
@n?" dd@ @@``
!"#$%&'()*+,./0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^_`abcdefghijklmnopqrstuvwxyz{}~mes New Romanbb(b0LbLb~0DSymbolew Romanbb(b0LbLb~0 DWP Greek Centurybb(b0LbLb~00DWP MathACenturybb(b0LbLb~0@DWP MathExtendedAbb(b0LbLb~0PDWP IconicSymbolsAb(b0LbLb~0
f@ @@``
@n?" dd@ @@``:
2
61:;
PowerPoint Document(VDocumentSummaryInformation8= \
0zb
Chart MSGraph.Chart.804Microsoft Graph 2000 Chartd0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0bEquation Equation.30,Microsoft Equation 3.0/0DTimes New Romanbb(b0LbLb~0DSymbolew Romanbb(b0LbLb~0 DWP Greek Centurybb(b0LbLb~00DWP MathACenturybb(b0LbLb~0@DWP MathExtendedAbb(b0LbLb~0PDWP IconicSymbolsAb(b0LbLb~0
f@ @@``
@n?" dd@ @@``:
2
F4:;
"'. ?#&,'""A !#/!@($.;(9) 3
<>.14/==0CEF<2$Hl$:?K=lծ 'z2$+xxF4;[z2$9«ɘVyu샄9
z$z$z$z$z$z$z$z$z"$VLK8xunzl1z"$g<=]pfzS~1? f̙f333@>g4KdKdXb~0Pb
ppp@g4BdBdXb~0Pb
(ppp@<4ddddbbLC?O=jInference and Operational Conduct Issues with Sample Size Adjustment Based On Interim Observed Effect Size&kj(
H.M. James Hung (DB1/OB/OPaSS/CDER/FDA)
Lu Cui (Aventis Pharmaceuticals)
SueJane Wang (DB2/OB/OPaSS/CDER/FDA)
John Lawrence (DB1/OB/OPaSS/CDER/FDA)
Presented in Annual Symposium of New Jersey
Chapter of ASA, Piscataway, NJ, June 4, 2002 ILZ[Z)VXft(!H
Disclaimer
The views expressed in this presentation are not
those of the U.S. Food and Drug Administration,
nor of Aventis Pharmaceuticals.
Dr. Lu Cui was one of the primary investigators
of this research during his tenure in FDA.
0f,,ROG6Selected References in Adaptive Design/InterimAnalysis76'PH
Emerging Strategy$0ff6ApSelected References{l}m/"7 !"#$%&S&P( 4
[
Tr>1?
Sample Size Reestimation
Sample size reestimation criterion
Better to look for more stable signal via
examination of sample path over time
 reduce the chance of being misled by possible
aberration of early data(
%2,f0f f # Y QH
0h ? ̙3320`(
^
S*
>
#lq>__
@>
0In that project, a simple approach was proposed to properly adjust & ..
This adaptive test has the type I error rate maintained at the desired level and its power can be substantially improved as a result of sample size increase. The test can be implemented very easily in the sense of preserving the flexibility of interim analysis schedule Lan and DeMets procedure has.,WH
0hdS ? ̙33r =o(
\
0zb
Chart MSGraph.Chart.804Microsoft Graph 2000 Chartd0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0b
Chart#. ?"(#"A
Oh+'0hp
(4<3Modification of Trial Conduct in Clinical Studies*iFDCDER User Jim Hunger1740ungMicrosoft PowerPoint Co@g@{,@@BgI@`G
oM H& &&#TNPP0D
v
&
TNPP &&TNPP
 " ! "&y&
 Times New Roman . 2
e1 .! 0 Times New Roman 33.92
T:!Inference and Operational Conduct
*
'. 33.:2
:"Issues with Sample Size Adjustment
'
,&. 33.72
: Based On Interim Observed Effect$)
,)
$. 33.
2
:Size.i( 33Times New Roman .+2
jH.M. James Hung (DB1/OB/!($!!. .2
j%OPaSS!. .2
j
/CDER/FDA). .2
Lu Cui. .
2
(. .2
Aventis. .!2
> Pharmaceuticals)
$. .(2
SueJane Wang (DB2/OB/*
!. .2
OPaSS". .2
v
/CDER/FDA). .2
John Lawrence (. .2
LDB1/OB/". .2
OPaSS!. .2
r
/CDER/FDA).Times New Roman f.H2
k+Presented in Annual Symposium of New Jersey
!!
. f.2
Chapter of ASA,
. f.2
G
Piscataway. f."2
, NJ, June 4, 2002
."System&TNPP &
՜.+,D՜.+,
:
2
F4:;
"'. ?#&,'""A !#/!@($.;(9) 3
<>.14/==0CEF<2$Hl$:?K=lծ 'z2$+xxF4;[z2$9«ɘVyu샄9
z$z$z$z$z$z$z$z$z"$VLK8xunzl1z"$g<=]pfzS~1? f̙f333@>g4KdKdXb~0Pb
ppp@g4BdBdXb~0Pb
(ppp@<4ddddbbLC?O=jInference and Operational Conduct Issues with Sample Size Adjustment Based On Interim Observed Effect Size&kj(
H.M. James Hung (DB1/OB/OPaSS/CDER/FDA)
Lu Cui (Aventis Pharmaceuticals)
SueJane Wang (DB2/OB/OPaSS/CDER/FDA)
John Lawrence (DB1/OB/OPaSS/CDER/FDA)
Presented in Annual Symposium of New Jersey
Chapter of ASA, Piscataway, NJ, June 4, 2002 ILZ[Z)VXft(!H
Disclaimer
The views expressed in this presentation are not
those of the U.S. Food and Drug Administration,
nor of Aventis Pharmaceuticals.
Dr. Lu Cui was one of the primary investigators
of this research during his tenure in FDA.
0f,,ROG6Selected References in Adaptive Design/InterimAnalysis76'PH
Emerging Strategy$0ff6ApSelected References{l}m/"7 !"#$%&r;5o(
\
0zb
Chart MSGraph.Chart.804Microsoft Graph 2000 Chartd0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0bEquation Equation.30,Microsoft Equation 3.0/0DTimes New Romanbb(b0LbLb~0DSymbolew Romanbb(b0LbLb~0 DWP Greek Centurybb(b0LbLb~00DWP MathACenturybb(b0LbLb~0@DWP MathExtendedAbb(b0LbLb~0PDWP IconicSymbolsAb(b0LbLb~0
f@ @@``
@n?" dd@ @@``:
2
F4:;
"'. ?#&,'""A !#/!@($.;(9) 3
<>.14/==0CEF<2$Hl$:?K=lծ 'z2$+xxF4;[z2$9«ɘVyu샄9
z$z$z$z$z$z$z$z$z"$VLK8xunzl1z"$g<=]pfzS~1? f̙f333@>g4KdKdXb~0Pb
ppp@g4BdBdXb~0Pb
(ppp@<4ddddbbLC?O=jInference and Operational Conduct Issues with Sample Size Adjustment Based On Interim Observed Effect Size&kj(
H.M. James Hung (DB1/OB/OPaSS/CDER/FDA)
Lu Cui (Aventis Pharmaceuticals)
SueJane Wang (DB2/OB/OPaSS/CDER/FDA)
John Lawrence (DB1/OB/OPaSS/CDER/FDA)
Presented in Annual Symposium of New Jersey
Chapter of ASA, Piscataway, NJ, June 4, 2002 ILZ[Z)VXft(!H
Disclaimer
The views expressed in this presentation are not
those of the U.S. Food and Drug Administration,
nor of Aventis Pharmaceuticals.
Dr. Lu Cui was one of the primary investigators
of this research during his tenure in FDA.
0f,,ROG6Selected References in Adaptive Design/InterimAnalysis76'PH
Emerging Strategy$0ff6ApSelected References{l}m/"7 !"#$%&a \(
\i
\
Te<1?`
Summary
Conventional design properly adapted for necessary sample size adjustment can be very useful with
 proper planning to avoid any operational conduct
change that may lead to bias
 proper adjustment of statistical analysis method
estimation issue needs attention and research
Price paid for datadriven adaptation: may lose
good statistical properties [Jennison & Turnbull (2001), Tsiatis & Mehta (2002)]. <1S f0f,j 2 b H
\0h ? ̙33PH `(
`
`
T"v1?`j
$R Summary
Adaptive design with proper planning is very attractive with caution
 change sample size or randomization allocation
 change study hypothesis (e.g., superiority vs. non
inferiority or equivalence)
 change test method
 change the primary endpoint from one pre
specified endpoint to another prespecified one
 drop futile or unsafe treatment arms and more& & ( More work to do (< f0f, ( ((H
`0h ? ̙33ro5 TOYK5^o(
\
0zb
Chart MSGraph.Chart.804Microsoft Graph 2000 Chartd0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0bEquation Equation.30,Microsoft Equation 3.0/0DTimes New Romanbb(b0LbLb~0DSymbolew Romanbb(b0LbLb~0 DWP Greek Centurybb(b0LbLb~00DWP MathACenturybb(b0LbLb~0@DWP MathExtendedAbb(b0LbLb~0PDWP IconicSymbolsAb(b0LbLb~0
f@ @@``
@n?" dd@ @@``:
2
F4:;
"'. ?#&,'""A !#/!@($.;(9) 3
<>.14/==0CEF<2$Hl$:?K=lծ 'z2$+xxF4;[z2$9«ɘVyu샄9
z$z$z$z$z$z$z$z$z"$VLK8xunzl1z"$g<=]pfzS~1? f̙f333@>g4KdKdXb~0Pb
ppp@g4BdBdXb~0Pb
(ppp@<4ddddbbLC?O=jInference and Operational Conduct Issues with Sample Size Adjustment Based On Interim Observed Effect Size&kj(
H.M. James Hung (DB1/OB/OPaSS/CDER/FDA)
Lu Cui (Aventis Pharmaceuticals)
SueJane Wang (DB2/OB/OPaSS/CDER/FDA)
John Lawrence (DB1/OB/OPaSS/CDER/FDA)
Presented in Annual Symposium of New Jersey
Chapter of ASA, Piscataway, NJ, June 4, 2002 ILZ[Z)VXft(!H
Disclaimer
The views expressed in this presentation are not
those of the U.S. Food and Drug Administration,
nor of Aventis Pharmaceuticals.
Dr. Lu Cui was one of the primary investigators
of this research during his tenure in FDA.
0f,,ROG6Selected References in Adaptive Design/InterimAnalysis76'PH
Emerging Strategy$0ff6ApSelected References{l}m/"7 !"#$%&r7__~S(
\
0zb
Chart MSGraph.Chart.804Microsoft Graph 2000 Chartd0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0bEquation Equation.30,Microsoft Equation 3.0/0DTimes New Romanbb(b0LbLb~0DSymbolew Romanbb(b0LbLb~0 DWP Greek Centurybb(b0LbLb~00DWP MathACenturybb(b0LbLb~0@DWP MathExtendedAbb(b0LbLb~0PDWP IconicSymbolsAb(b0LbLb~0
f@ @@``
@n?" dd@ @@``:
2
E4:;
"'. ?#&,'""A !#/!@($.;(9) 3
<>.14/==0CEF<2$Hl$:?K=lծ 'z2$+xxF4;[z2$9«ɘVyu샄9
z$z$z$z$z$z$z$z$z"$VLK8xunzl1z"$g<=]pfzS~1? f̙f333@>g4KdKdXb~0Pb
ppp@g4BdBdXb~0Pb
(ppp@<4ddddbbLC?O=jInference and Operational Conduct Issues with Sample Size Adjustment Based On Interim Observed Effect Size&kj(
H.M. James Hung (DB1/OB/OPaSS/CDER/FDA)
Lu Cui (Aventis Pharmaceuticals)
SueJane Wang (DB2/OB/OPaSS/CDER/FDA)
John Lawrence (DB1/OB/OPaSS/CDER/FDA)
Presented in Annual Symposium of New Jersey
Chapter of ASA, Piscataway, NJ, June 4, 2002 ILZ[Z)VXft(!H
Disclaimer
The views expressed in this presentation are not
those of the U.S. Food and Drug Administration,
nor of Aventis Pharmaceuticals.
Dr. Lu Cui was one of the primary investigators
of this research during his tenure in FDA.
0f,,ROG6Selected References in Adaptive Design/InterimAnalysis76'PH
Emerging Strategy$0ff6ApSelected References{l}m/"7 !"#$%&r~~=S(
\
0zb
Chart MSGraph.Chart.804Microsoft Graph 2000 Chartd0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0bEquation Equation.30,Microsoft Equation 3.0/0DTimes New Romanbb(b0LbLb~0DSymbolew Romanbb(b0LbLb~0 DWP Greek Centurybb(b0LbLb~00DWP MathACenturybb(b0LbLb~0@DWP MathExtendedAbb(b0LbLb~0PDWP IconicSymbolsAb(b0LbLb~0
f@ @@``
@n?" dd@ @@``:
2
E4:;
"'. ?#&,'""A !#/!@($.;(9) 3
<>.14/==0CEF<2$Hl$:?K=lծ 'z2$+xxF4;[z2$9«ɘVyu샄9
z$z$z$z$z$z$z$z$z"$VLK8xunzl1z"$g<=]pfzS~1? f̙f333@>g4KdKdXb~0Pb
ppp@g4BdBdXb~0Pb
(ppp@<4ddddbbLC?O=jInference and Operational Conduct Issues with Sample Size Adjustment Based On Interim Observed Effect Size&kj(
H.M. James Hung (DB1/OB/OPaSS/CDER/FDA)
Lu Cui (Aventis Pharmaceuticals)
SueJane Wang (DB2/OB/OPaSS/CDER/FDA)
John Lawrence (DB1/OB/OPaSS/CDER/FDA)
Presented in Annual Symposium of New Jersey
Chapter of ASA, Piscataway, NJ, June 4, 2002 ILZ[Z)VXft(!H
Disclaimer
The views expressed in this presentation are not
those of the U.S. Food and Drug Administration,
nor of Aventis Pharmaceuticals.
Dr. Lu Cui was one of the primary investigators
of this research during his tenure in FDA.
0f,,ROG6Selected References in Adaptive Design/InterimAnalysis76'PH
Emerging Strategy$0ff6ApSelected References{l}m/"7 !"#$%&%
+@$ *;$6% 0
<>*1+==,C.F<2$Hl$:?K=lծ 'z2$+xxF4Le(
Lr
L SX9`@
%
L
`$Y91?
M
L
TY91?pX
JBauer & Khne (1994, Biometrics)
Proschan & Hunsberger (1995, Biometrics)
Lan & Trost (1997, ASA Proceedings)
Cui, Hung & Wang (1997 ASA Proceedings, 1999, Biometrics)
Fisher (1998, Stat. In Med.)
Shen & Fisher (1999, Biometrics)
Lehmacher & Wassmer (1999, Biometrics)
Mller & Schfer (2001, Biometrics)
Liu & Chi (2001, Biometrics)
Lan (2001, FDA/CDER/OB minisymposium)
Lan (2002, FDA/ASA workshop)
Brannath, Posch & Bauer (2002, JASA)
Lawrence & Hung (2002, ENAR)33: D 1#3H
L0h ? ̙33rq̽MS(
\
0zb
Chart MSGraph.Chart.804Microsoft Graph 2000 Chartd0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0bEquation Equation.30,Microsoft Equation 3.0/0DTimes New Romanbb(b0LbLb~0DSymbolew Romanbb(b0LbLb~0 DWP Greek Centurybb(b0LbLb~00DWP MathACenturybb(b0LbLb~0@DWP MathExtendedAbb(b0LbLb~0PDWP IconicSymbolsAb(b0LbLb~0
f@ @@``
.
% !"#$'()*+,/0123456789:;<=>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^_`abcdefghijklmnopqrstuvwxyz{}~ @n?" dd@ @@``:
2
1:;
#. ?"(#"A;[z2$9«ɘVyu샄9
z$z$z$z$z$z$z$z$z"$VLK8xunzl1z"$g<=]pfzS~1? f̙f333@>g4KdKdXb~0Pb
ppp@g4BdBdXb~0Pb
(ppp@<4ddddbbLC?O=jInference and Operational Conduct Issues with Sample Size Adjustment Based On Interim Observed Effect Size&kj(
H.M. James Hung (DB1/OB/OPaSS/CDER/FDA)
Lu Cui (Aventis Pharmaceuticals)
SueJane Wang (DB2/OB/OPaSS/CDER/FDA)
John Lawrence (DB1/OB/OPaSS/CDER/FDA)
Presented in Annual Symposium of New Jersey
Chapter of ASA, Piscataway, NJ, June 4, 2002 ILZ[Z)VXft(!H
Disclaimer
The views expressed in this presentation are not
those of the U.S. Food and Drug Administration,
nor of Aventis Pharmaceuticals.
Dr. Lu Cui was one of the primary investigators
of this research during his tenure in FDA.
0f,,ROG6Selected References in Adaptive Design/InterimAnalysis76'PH
Emerging Strategy$0ff6ApSelected References{l}m/"7 !"#$%&_`\(
\g
\
TD:1?`
Summary
Conventional design properly adapted for necessary sample size adjustment can be very useful with
 proper planning to avoid any operational conduct
change that may lead to bias
 proper adjustment of statistical analysis method
estimation issue needs attention and research
Price paid for datadependent adaptation: may lose
statistical efficiency [Jennison & Turnbull (2001), Tsiatis & Mehta (2002)]. <4N f0f,h 2 b H
\0h ? ̙33r5S(
\
0zb
Chart MSGraph.Chart.804Microsoft Graph 2000 Chartd0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0bEquation Equation.30,Microsoft Equation 3.0/0DTimes New Romanbb(b0LbLb~0DSymbolew Romanbb(b0LbLb~0 DWP Greek Centurybb(b0LbLb~00DWP MathACenturybb(b0LbLb~0@DWP MathExtendedAbb(b0LbLb~0PDWP IconicSymbolsAb(b0LbLb~0
f@ @@``
@n?" dd@ @@``:
2
E1:;
#. ?"(#"A+@$ *;$6% 0
78
<>*91+==,C.E<2$Hl$:?K=lծ 'z2$+xxF4;[z2$9«ɘVyu샄9
z$z$z$z$z$z$z$z$z"$VLK8xunzl1z"$g<=]pfzS~1? f̙f333@>g4KdKdXb~0Pb
ppp@g4BdBdXb~0Pb
(ppp@<4ddddbbLC?O=jInference and Operational Conduct Issues with Sample Size Adjustment Based On Interim Observed Effect Size&kj(
H.M. James Hung (DB1/OB/OPaSS/CDER/FDA)
Lu Cui (Aventis Pharmaceuticals)
SueJane Wang (DB2/OB/OPaSS/CDER/FDA)
John Lawrence (DB1/OB/OPaSS/CDER/FDA)
Presented in Annual Symposium of New Jersey
Chapter of ASA, Piscataway, NJ, June 4, 2002 ILZ[Z)VXft(!H
Disclaimer
The views expressed in this presentation are not
those of the U.S. Food and Drug Administration,
nor of Aventis Pharmaceuticals.
Dr. Lu Cui was one of the primary investigators
of this research during his tenure in FDA.
0f,,ROG6Selected References in Adaptive Design/InterimAnalysis76'PH
Emerging Strategy$0ff6ApSelected References{l}m/"7 !"#$%&b
`\(
\j
\
TD:1?`
Summary
Conventional design properly adapted for necessary sample size adjustment can be very useful with
 proper planning to avoid any operational conduct
change that may lead to bias
 proper adjustment of statistical analysis method
estimation issue needs attention and research
Price paid for datadependent adaptation: may not
be statistically efficient [Jennison & Turnbull (2001), Tsiatis & Mehta (2002)]. <3R f0f,k 2 b H
\0h ? ̙33r3S(
\
0zb
Chart MSGraph.Chart.804Microsoft Graph 2000 Chartd0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0Root EntrydO)"c&@PicturesCurrent User\/SummaryInformation($
!"#$
!"#$%&'()*+,./0123456789:;<>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[vers
Slide Titles4 6>
_PID_GUIDAN{9D8BB7C06A6C11D6B3000002A5598AD8}598AD8}'_2Matilde Sanchez+@$ *;$% 0
78
<>*91+==,C.E<2$Hl$:?K=lծ 'z2$+xxF4;[z2$9«ɘVyu샄9
z$z$z$z$z$z$z$z$z"$VLK8xunzl1z"$g<=]pfzS~1? f̙f333@>g4KdKdXb~0Pb
ppp@g4BdBdXb~0Pb
(ppp@<4ddddbbLC?O=jInference and Operational Conduct Issues with Sample Size Adjustment Based On Interim Observed Effect Size&kj(
H.M. James Hung (DB1/OB/OPaSS/CDER/FDA)
Lu Cui (Aventis Pharmaceuticals)
SueJane Wang (DB2/OB/OPaSS/CDER/FDA)
John Lawrence (DB1/OB/OPaSS/CDER/FDA)
Presented in Annual Symposium of New Jersey
Chapter of ASA, Piscataway, NJ, June 4, 2002 ILZ[Z)VXft(!H
Disclaimer
The views expressed in this presentation are not
those of the U.S. Food and Drug Administration,
nor of Aventis Pharmaceuticals.
Dr. Lu Cui was one of the primary investigators
of this research during his tenure in FDA.
0f,,ROG6Selected References in Adaptive Design/InterimAnalysis76'PH
Emerging Strategy$0ff6ApSelected References{l}m/"7 !"#$%&3+ (
Zh%gֳgֳ ?P
M f
NDi%gֳgֳ?
.
Need to enhance flexibility in traditional clinical trial design/analysis strategy because
practical considerations may
change and may often be
unpredictable at the design stageD S,_, H
0h ? ̙33r4+/S(
\
0zb
Chart MSGraph.Chart.804Microsoft Graph 2000 Chartd0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0bEquation Equation.30,Microsoft Equation 3.0/0DTimes New Romanbb(b0LbLb~0DSymbolew Romanbb(b0LbLb~0 DWP Greek Centurybb(b0LbLb~00DWP MathACenturybb(b0LbLb~0@DWP MathExtendedAbb(b0LbLb~0PDWP IconicSymbolsAb(b0LbLb~0
f@ @@``
@n?" dd@ @@``:
2
E0:;
". ?!'""A *$@#);#$ /
<>)90*==+C1<2$Hl$:?K=lծ 'z2$+xxF4;[z2$9«ɘVyu샄9
z$z$z$z$z$z$z$z$z"$VLK8xunzl1z"$g<=]pfzS~1? f̙f333@>g4KdKdXb~0Pb
ppp@g4BdBdXb~0Pb
(ppp@<4ddddbbLC?O=jInference and Operational Conduct Issues with Sample Size Adjustment Based On Interim Observed Effect Size&kj(
H.M. James Hung (DB1/OB/OPaSS/CDER/FDA)
Lu Cui (Aventis Pharmaceuticals)
SueJane Wang (DB2/OB/OPaSS/CDER/FDA)
John Lawrence (DB1/OB/OPaSS/CDER/FDA)
Presented in Annual Symposium of New Jersey
Chapter of ASA, Piscataway, NJ, June 4, 2002 ILZ[Z)VXft(!H
Disclaimer
The views expressed in this presentation are not
those of the U.S. Food and Drug Administration,
nor of Aventis Pharmaceuticals.
Dr. Lu Cui was one of the primary investigators
of this research during his tenure in FDA.
0f,,ROG6Selected References in Adaptive Design/InterimAnalysis76'PH
Emerging Strategy$0ff6ApSelected References{l}m/"7 !"#$%&{,(
,
,
Zd#gֳgֳ ?P
M f
,
Ną#gֳgֳ?`
~~ Background
But & .
The effect size depends on a primary
parameter (e.g., mean treatment difference) and
nuisance parameters (e.g., standard deviation,
background event rate)
The effect size for detection may need to be clinically significant or meaningful (sometimes minimum clinically meaningful) benefit/risk assessment (subjective) that might not be doable in designing the trial, hard to reach consensus F
F
0f
K )H
,0h ? ̙33rV/N2/US(
\
0zb
Chart MSGraph.Chart.804Microsoft Graph 2000 Chartd0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0bEquation Equation.30,Microsoft Equation 3.0/0DTimes New Romanbb(b0LbLb~0DSymbolew Romanbb(b0LbLb~0 DWP Greek Centurybb(b0LbLb~00DWP MathACenturybb(b0LbLb~0@DWP MathExtendedAbb(b0LbLb~0PDWP IconicSymbolsAb(b0LbLb~0
f@ @@``
@n?" dd@ @@``:
2
E0:;
". ?!'""A *$@#);#$ /
<>)90*==+C1<2$Hl$:?K=lծ 'z2$+xxF4;[z2$9«ɘVyu샄9
z$z$z$z$z$z$z$z$z"$VLK8xunzl1z"$g<=]pfzS~1? f̙f333@>g4KdKdXb~0Pb
ppp@g4BdBdXb~0Pb
(ppp@<4ddddbbLC?O=jInference and Operational Conduct Issues with Sample Size Adjustment Based On Interim Observed Effect Size&kj(
H.M. James Hung (DB1/OB/OPaSS/CDER/FDA)
Lu Cui (Aventis Pharmaceuticals)
SueJane Wang (DB2/OB/OPaSS/CDER/FDA)
John Lawrence (DB1/OB/OPaSS/CDER/FDA)
Presented in Annual Symposium of New Jersey
Chapter of ASA, Piscataway, NJ, June 4, 2002 ILZ[Z)VXft(!H
Disclaimer
The views expressed in this presentation are not
those of the U.S. Food and Drug Administration,
nor of Aventis Pharmaceuticals.
Dr. Lu Cui was one of the primary investigators
of this research during his tenure in FDA.
0f,,ROG6Selected References in Adaptive Design/InterimAnalysis76'PH
Emerging Strategy$0ff6ApSelected References{l}m/"7 !"#$%&V(
^
T:1?`M
Sample Size Reestimation
CHW adaptive test has type I error rate attained at the targeted level and large power increase (relative to w/o reestimation) and its implementation is very easy
Consistent estimator and confidence interval compatible with CHW adaptive test are readily available
All the above discussions are based on asymptotic (i.e., sufficiently large sample size) theory
Cui, Hung, Wang (1999, Biometrics)
Lawrence & Hung (2002, ENAR talk)
fcFF,f0f f b D 33CH
0h ? ̙33rU#uU{S(
\
0zb
Chart MSGraph.Chart.804Microsoft Graph 2000 Chartd0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0bEquation Equation.30,Microsoft Equation 3.0/0DTimes New Romanbb(b0LbLb~0DSymbolew Romanbb(b0LbLb~0 DWP Greek Centurybb(b0LbLb~00DWP MathACenturybb(b0LbLb~0@DWP MathExtendedAbb(b0LbLb~0PDWP IconicSymbolsAb(b0LbLb~0
f@ @@``
@n?" dd@ @@``:
2
E0:;
". ?!'""A *$@#);#$ /
<>)90*==+C1<2$Hl$:?K=lծ 'z2$+xxF4;[z2$9«ɘVyu샄9
z$z$z$z$z$z$z$z$z"$VLK8xunzl1z"$g<=]pfzS~1? f̙f333@>g4KdKdXb~0Pb
ppp@g4BdBdXb~0Pb
(ppp@<4ddddbbLC?O=jInference and Operational Conduct Issues with Sample Size Adjustment Based On Interim Observed Effect Size&kj(
H.M. James Hung (DB1/OB/OPaSS/CDER/FDA)
Lu Cui (Aventis Pharmaceuticals)
SueJane Wang (DB2/OB/OPaSS/CDER/FDA)
John Lawrence (DB1/OB/OPaSS/CDER/FDA)
Presented in Annual Symposium of New Jersey
Chapter of ASA, Piscataway, NJ, June 4, 2002 ILZ[Z)VXft(!H
Disclaimer
The views expressed in this presentation are not
those of the U.S. Food and Drug Administration,
nor of Aventis Pharmaceuticals.
Dr. Lu Cui was one of the primary investigators
of this research during his tenure in FDA.
0f,,ROG6Selected References in Adaptive Design/InterimAnalysis76'PH
Emerging Strategy$0ff6ApSelected References{l}m/"7 !"#$%&&<4( 4
T:1?
Sample Size Reestimation
Sample size reestimation criterion
Better to look for more stable signal via
examination of sample path over time
 reduce the chance of being misled by
possible aberration of early data(
%2S<Q,f0f f # U S ,(#H
0h ? ̙33r{{S(
\
0zb
Chart MSGraph.Chart.804Microsoft Graph 2000 Chartd0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0bEquation Equation.30,Microsoft Equation 3.0/0DTimes New Romanbb(b0LbLb~0DSymbolew Romanbb(b0LbLb~0 DWP Greek Centurybb(b0LbLb~00DWP MathACenturybb(b0LbLb~0@DWP MathExtendedAbb(b0LbLb~0PDWP IconicSymbolsAb(b0LbLb~0
f@ @@``
@n?" dd@ @@``:
2
E0:;
". ?!'""A *$@#);#$ /
<>)90*==+C1<2$Hl$:?K=lծ 'z2$+xxF4;[z2$9«ɘVyu샄9
z$z$z$z$z$z$z$z$z"$VLK8xunzl1z"$g<=]pfzS~1? f̙f333@>g4KdKdXb~0Pb
ppp@g4BdBdXb~0Pb
(ppp@<4ddddbbLC?O=jInference and Operational Conduct Issues with Sample Size Adjustment Based On Interim Observed Effect Size&kj(
H.M. James Hung (DB1/OB/OPaSS/CDER/FDA)
Lu Cui (Aventis Pharmaceuticals)
SueJane Wang (DB2/OB/OPaSS/CDER/FDA)
John Lawrence (DB1/OB/OPaSS/CDER/FDA)
Presented in Annual Symposium of New Jersey
Chapter of ASA, Piscataway, NJ, June 4, 2002 ILZ[Z)VXft(!H
Disclaimer
The views expressed in this presentation are not
those of the U.S. Food and Drug Administration,
nor of Aventis Pharmaceuticals.
Dr. Lu Cui was one of the primary investigators
of this research during his tenure in FDA.
0f,,ROG6Selected References in Adaptive Design/InterimAnalysis76'PH
Emerging Strategy$0ff6ApSelected References{l}m/"7 !"#$%&E \(
\M
\
T&v1?`
Summary
Conventional design properly adapted for necessary sample size adjustment can be very useful with
 proper planning to avoid any operational conduct
change that may lead to bias
 proper adjustment of statistical analysis method
estimation issue needs attention and research
Price paid for datadependent adaptation: may not
be statistically efficient [Jennison (2001), Tsiatis & Mehta (2002)]. <3G f0f,k ' P H
\0h ? ̙33rCĞS(
\
0zb
Chart MSGraph.Chart.804Microsoft Graph 2000 Chartd0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0bEquation Equation.30,Microsoft Equation 3.0/0DTimes New Romanbb(b0LbLb~0DSymbolew Romanbb(b0LbLb~0 DWP Greek Centurybb(b0LbLb~00DWP MathACenturybb(b0LbLb~0@DWP MathExtendedAbb(b0LbLb~0PDWP IconicSymbolsAb(b0LbLb~0
f@ @@``
@n?" dd@ @@``:
2
E0:;
". ?!'""A *$@#);#$ /
<>)90*==+C1<2$Hl$:?K=lծ 'z2$+xxF4;[z2$9«ɘVyu샄9
z$z$z$z$z$z$z$z$z"$VLK8xunzl1z"$g<=]pfzS~1? f̙f333@>g4KdKdXb~0Pb
ppp@g4BdBdXb~0Pb
(ppp@<4ddddbbLC?O=jInference and Operational Conduct Issues with Sample Size Adjustment Based On Interim Observed Effect Size&kj(
H.M. James Hung (DB1/OB/OPaSS/CDER/FDA)
Lu Cui (Aventis Pharmaceuticals)
SueJane Wang (DB2/OB/OPaSS/CDER/FDA)
John Lawrence (DB1/OB/OPaSS/CDER/FDA)
Presented in Annual Symposium of New Jersey
Chapter of ASA, Piscataway, NJ, June 4, 2002 ILZ[Z)VXft(!H
Disclaimer
The views expressed in this presentation are not
those of the U.S. Food and Drug Administration,
nor of Aventis Pharmaceuticals.
Dr. Lu Cui was one of the primary investigators
of this research during his tenure in FDA.
0f,,ROG6Selected References in Adaptive Design/InterimAnalysis76'PH
Emerging Strategy$0ff6ApSelected References{l}m/"7 !"#$%&\ \(
\d
\
T&v1?`
4 Summary
Conventional design properly adapted for necessary sample size adjustment can be very useful with
 proper planning to avoid any operational conduct
change that may lead to bias
 proper adjustment of statistical analysis method
estimation issue needs attention and research
d< f0f, $H
\0h ? ̙33r'S(
\
0zb
Chart MSGraph.Chart.804Microsoft Graph 2000 Chartd0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0b
Chart MSGraph.Chart.804Microsoft Graph 2000 Chart0bEquation Equation.30,Microsoft Equation 3.0/0DTimes New Romanbb(b0LbLb~0DSymbolew Romanbb(b0LbLb~0 DWP Greek Centurybb(b0LbLb~00DWP MathACenturybb(b0LbLb~0@DWP MathExtendedAbb(b0LbLb~0PDWP IconicSymbolsAb(b0LbLb~0
f@ @@``
@n?" dd@ @@``:
2
E0:;
". ?!'""A *$@#);#$ /
<>)90*==+C1<2$Hl$:?K=lծ 'z2$+xxF4;[z2$9«ɘVyu샄9
z$z$z$z$z$z$z$z$z"$VLK8xunzl1z"$g<=]pfzS~1? f̙f333@>g4KdKdXb~0Pb
ppp@g4BdBdXb~0Pb
(ppp@<4ddddbbLC?O=jInference and Operational Conduct Issues with Sample Size Adjustment Based On Interim Observed Effect Size&kj(
H.M. James Hung (DB1/OB/OPaSS/CDER/FDA)
Lu Cui (Aventis Pharmaceuticals)
SueJane Wang (DB2/OB/OPaSS/CDER/FDA)
John Lawrence (DB1/OB/OPaSS/CDER/FDA)
Presented in Annual Symposium of New Jersey
Chapter of ASA, Piscataway, NJ, June 4, 2002 ILZ[Z)VXft(!H
Disclaimer
The views expressed in this presentation are not
those of the U.S. Food and Drug Administration,
nor of Aventis Pharmaceuticals.
Dr. Lu Cui was one of the primary investigators
of this research during his tenure in FDA.
0f,,ROG6Selected References in Adaptive Design/InterimAnalysis76'PH
Emerging Strategy$0ff6ApSelected References{l}m/"7 !"#$%&rm"Root EntrydO)s&@PicturesCurrent User\/SummaryInformation($
!"#$
!"#$%&'()*+,./0123456789:;<>?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[vers
Slide Titles4 6>
_PID_GUIDAN{9D8BB7C06A6C11D6B3000002A5598AD8}598AD8}'_2Matilde Sanchez)