Selection of Endpoints and Sample Size Estimation in...
Transcript of Selection of Endpoints and Sample Size Estimation in...
![Page 1: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/1.jpg)
Selection of Endpoints and Sample Size Estimation in Clinical Trials
Jen-pei Liu, Ph.D. National Taiwan University
&National Health Research Institutes
atDepartment of Family MedicineNational Cheng-Kung University
May 7, 2005
![Page 2: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/2.jpg)
Selection of Endpoints and Sample Size Estimation in Clinical Trials
Selection of EndpointsSample Size EstimationAdjustment of Baseline
![Page 3: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/3.jpg)
Types of DataContinuous EndpointsNumerical discrete data
Heart beats per minutesTotal NINSSTotal Hamilton Rating Scale for DepressionTotal Alzheimer’s Disease Assessment Scale
Numerical continuous dataAge WeightALTPeak flow rate (liters per minute)FEV1 (% of predicted value)
![Page 4: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/4.jpg)
Types of DataCategorical EndpointsNominal scale dataClassification of patients according to their
attributesGenderRaceOccurrence of a particular adverse reactionOccurrence of ALT>3 times upper normal limit
![Page 5: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/5.jpg)
Types of DataOrdinal (Ordered) categorical dataA certain order among different categories
Symptom score0 = no symptom, 1 = mild, 2 = moderate, 3 = severeSeverity of adverse reactions
Censored EndpointsTime to the occurrence of a pre-defined event.The occurrence of the event may not observed for some patients. Then the time to the occurrence of the event for these subjects is censored
![Page 6: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/6.jpg)
Types of DataCross-sectional vs. longitudinal data
Cross-sectional data (snap shot at one time point)
Clinical data are collected and evaluated at a particular time point during the trialLongitudinal data (snap shots at several time points)
Clinical data collected and evaluated over a series of time points during the trial
![Page 7: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/7.jpg)
ExampleKnapp et al (JAMA 1994; 271: 985-991)
A multi-center trial with 33 centersDouble-blind, randomized, 4 parallel groupsForced escalation30 weeks of randomized treatment
6 visitsThe start of randomized treatment(baseline)6,12,18,24, and 30 weeks
Cross-sectional dataCIBI and ADAS-cog evaluated at the start of randomized treatment
LongitudinalA series of CIBI and ADAS-cog evaluated at the start of the study, the start
of randomized treatment,6,12,18,24, and 30 weeks
![Page 8: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/8.jpg)
Types of ComparisonWithin-group (patient) comparisonComparison of the changes within the same patients at different time points during the trial.Between-group (patient) comparison
Comparison between groups of patients under different treatments.
![Page 9: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/9.jpg)
Example: Major depression disorderStark and Hardison (VCP, 1985;46,53-58)Cohn and Wilcox (JCP,1985:46,21-31)
Double-blind, randomized, three parallel groupsOne-week placebo washout periodFluoxetine vs. imipramine vs. placebo6 weeks of randomized treatmentsPrimary efficacy endpoint
HAM-D score at the last follow-up visitWithin each groupChange from baseline in HAM-D score
Between groupsComparison of the change from baseline in HAM-D score between
groups
![Page 10: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/10.jpg)
![Page 11: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/11.jpg)
EndpointsRaw measurements at a time point.Change at a time point from baseline. Percent change at a time point from baseline.Clinically meaningful targeted value attained at a time point, i.e. sitting DBP <= 85 mm Hg Selection of time points should be able to measure the effect of the intervention.
![Page 12: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/12.jpg)
Selection of EndpointsEndpoints should reflect the change of clinical status caused by the intervention.Endpoints should be sensitive to the change of clinical status caused by the intervention.Endpoints should be validated.Raw measurements at a time point can only measure the static clinical status.Change at a time point from baseline can measure the magnitude of the change of clinical status caused by the intervention.Change from baseline has the same unit as the raw measurement
![Page 13: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/13.jpg)
Selection of EndpointsPercent change at a time point from baseline measures the relative magnitude of the change of clinical status caused by the intervention.Percent change from baseline is unitless.The same percent change may reflect different magnitudes of change20/100 = 2/10 = 200/1000 = 20%
![Page 14: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/14.jpg)
Selection of EndpointsOne of the key inclusion criteria for clinical trial in treatment of mild to moderate essential hypertension is sitting DBP being between 95-115 mm Hg.Three changes from baseline: 115 → 105, 105 → 95, 95 → 85.95 Changes from baseline: 8.7%, 9.5%, 10.5%Only 95 → 85 reaches the clinically meaningful targeted value.
![Page 15: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/15.jpg)
Selection of EndpointsEndpoints should reflect clinically meaningful interpretation and applicability.Clinically meaningful targeted value > change from baseline > percent change from baseline.Clinical investigators should have responsibility for determination of the efficacy endpoints used in the clinical trials.
![Page 16: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/16.jpg)
Selection of EndpointsLDL HDL TG
Targeted Value < 100mg/dL 40-60 mg/dL < 150 mg/dL
Bile acid Binding Resin ↓15-30% ↑3-5% no change
Nicotinic acid ↓ 5-25% ↑15-35% ↓15-25%
Fibric acid ↓ 5-20% ↑10-20% ↓ 20-50%
HMG-CoA ↓18-55% ↑ 3-5% ↓ 7-30%Inhibitor
![Page 17: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/17.jpg)
Measures for Comparison in Proportions between Groups
Difference in proportionsRelative risk
The ratio of the proportions of the test group to the control.
Odds ratioThe ratio of the odds of the test group to the control.
OddsThe number of patients with the attribute to that without
the attribute.
![Page 18: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/18.jpg)
The US Physicians’ Health Study (NEJM 1989; 321: 129-35)
Aspirin PlaceboN 11037 11034MI 139 (1.26%) 239 (2.17%)No MI 10898 (98.74%) 10795 (97.83%)
Difference in proportion of MI = 1.26% - 2.17% = -0.91%(average of fewer 91 MIs per 10,000)
Relative risk of MI for aspirin = 1.26% / 2.17% = 0.581(the risk of MI in aspirin reduces 42%)
Odds ratio of MI for aspirin = (139 / 10898) / (239 / 10798)= 1.275% / 2.214% = 0.576
(the odds of MI in aspirin reduces 42%) Difference in proportions and relative risk can only be used in prospective studies while odds can be used in both prospective as well as retrospective studies.
![Page 19: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/19.jpg)
Categorical EndpointsDifference in proportions provides the absolute magnitude of difference.Both relative risk and odds ratio gives the relative magnitude of difference.50% → 25% and 0.05% → 0.025% both yield a relative risk of 50% but differences in proportion are 25% and 0.025% respectively.Relative risk and odd ratio are appropriate when the proportion of the event for control group is small (<5%).When the proportion of the event is small (<5%), the relative risk ≈ Odds ratio.
![Page 20: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/20.jpg)
Censored DataMedian survival
The time to the pre-defined event (e.g. death) occurring in 50% of the patients.
Survival rate at a particular time pointHazard ratioThe hazard of the occurrence of a pre-defined
event of the test group to the control groupSurvival rate and hazard ratio can be exchanged under statistical models
![Page 21: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/21.jpg)
681 676 675 673 670 611669 665 655 651 648 594677 675 672 668 667 612
Months sincefirst dose
Sample size
C 200 BIDC 400 BIDPlacebo
Log-rank statistic 8.74 (p= 0.013)
E:\Proj\APC Cox2\Programs\f KMplot.sas v.010 death MI stroke CHF 005 01.cgm (last run: 02/14/2005, 10:46)CEC adjudication data (received 02/03/2005) RAND data (received 11/23/2004) CRF data (received 02/03/2005)
Est
imat
ed p
roba
bilit
y of
CV
dea
th, M
I, st
roke
, or C
HF
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
6 12 18 24 30 36
Kaplan-Meier Estimates of the Risk of Serious CV Events in the APC Trial by Treatment Arm*
![Page 22: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/22.jpg)
Kaplan-Meier Estimates of the Risk of Serious CV Events in the APC Trial by Treatment Arm*
Solomon SD, et al: N Engl J Med 352, 2005*In this analysis, “serious CV events” include death from CV causes, MI, stroke, or heart failure
671
![Page 23: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/23.jpg)
Sample Size EstimationInvestigators must determine the expected difference you want to detect in the trial and variability associated with clinical endpointsExpected difference should be clinically meaningful.Expected difference should not be over-exaggerated nor over-conservative.Variability should be realistic.
![Page 24: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/24.jpg)
Sample Size EstimationAll formulas for sample size estimation provide the minimal number of patients required for trials. Sample size increases in square as the expected difference decreases or the standard deviation increases.Sample size increase four times as the expected difference decreases 50% or standard deviation doubles.Sample size increases if the significance level decreases or power increases.
![Page 25: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/25.jpg)
Sample Size EstimationIntroduction
E = experimental treatment groupC = control treatment group
We considerMean differenceDifference in proportionsRelative riskEquivalence trialTime to event
![Page 26: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/26.jpg)
Notation α= typeⅠerrorβ= typeⅡerrorZα= theαth normal quartile: Level of SignificanceP{ Z > Zα} = α one sidedP{|Z| > Zα/2} = α two sidedZβ= the βth normal quartile: Level of PowerP{Z < Zβ} = 1 - β
![Page 27: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/27.jpg)
References Donner, A. (1984) Approaches to sample size estimation in the design y clinical trials. Statistic in Medicine, vol. 3, 198-214Lachin, J.M. (1981) Introduction to sample size determination and power analysis for clinical trials, Controlled Clinical Trials, vol. 2, 93-114Chow, S. C., and Liu, J.P. (2004) Design and Analysis of Clinical Trials: Concepts and Methodologies, 2nd Ed. Chapter 11, Wiley, New York, New York.
![Page 28: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/28.jpg)
Assume Equal Allocation Mean difference: μE – μC
2 2
2 2
0
2 2 2/ 2
2
2 2/ 2 2 2 2
2
( , ) ( , )
: 0:
( )[ ]
2 [ ]
C C E E
C E
C E
A C E
C E
C E
Assume N and N
and are knownHH
Z Zn per group
Z Zif
α β
α β
μ σ μ σ
σ σμ μμ μ
σ σ
σσ σ σ
− =− = Δ
+ +=
Δ+
= = =Δ
![Page 29: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/29.jpg)
Example
562.558
]84.096.1[)15(2
84.0,8.0196.1),(05.0
15
8:0:
2
22
22
0
orn
ZZsidedtwo
HH
ECA
EC
=+
=
==−==
=
=−=−
β
α
βασ
μμμμ
![Page 30: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/30.jpg)
Risk difference: PE - PC
0
22
exp::
2 (1 ) (1 ) (1 )
2
C
E
E C
A E C
E E C C
E C
P event rate of control group
P event rate of erimental groupH P PH P P
n Z P P Z P P P P per group
P Pwhere P
α β
=
==
− =Δ
⎡ ⎤= − + − + − Δ⎢ ⎥⎣ ⎦+
=
![Page 31: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/31.jpg)
Example
[ ]978.96
)4.06.0()4.0(6.0)6.0)(4.0(84.0)5.0)(5.0(296.1
84.0,8.0196.1),(05.0
4.0,6.0::
2
2
0
orn
ZZsidedtwo
PPHPPH
ECA
EC
=−
++=
==−==
===
β
α
βα
![Page 32: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/32.jpg)
Unequal Allocation
22
n experimental groupsn: control group
( 1) (1 ) (1 ) (1 ) ( )
( )( 1)
s s E E C C
E Cs
n Z s P P Z sP P P P s
P sPwhere Ps
α β σ
→
⎡ ⎤= + − + − + −⎢ ⎥⎣ ⎦+
=+
![Page 33: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/33.jpg)
Relative Risk: PE / PC
0
22 2
,/
: 1: /
2 (1 ) [1 (1 )] [ (1 )]
(1 )2
Compare with results for risk difference
C E
E C
A E C
R C C c
CR
P P and Paredefined asbeforeR P PH RH R P P r
n Z P P Z P r P r P r
P rwhere P
α β
=
=
= =
⎡ ⎤= − + + − + −⎢ ⎥⎣ ⎦+
=
![Page 34: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/34.jpg)
Example
978.96
84.0,8.0196.1),(05.0
:1:
32
6.04.0
0
orn
ZZsidedtwo
RHRH
A
=
==−==
==
=
β
α
βα
![Page 35: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/35.jpg)
Non-interiority Trial
0
2
2
E C
:: , 0
2 (1 ) (1 ) (1 )
( )0
In practice, set P P andθ the difference in treatment efficacy
C E
A C E
E E C C
C E
C E
H P PH P P
Z P P Z P P P Pn
P Pwhere P P and
α β
θθ θ
θθ θ
≥ +< + >
⎡ ⎤− + − + −⎢ ⎥⎣ ⎦=− −
< + >
==
![Page 36: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/36.jpg)
Example
0
2
2
E
0.75 0.8:: , 0.10.1( ), 1.282
1 0.8 0.84
1.28 2(0.775)(0.225) 0.84 (0.8)(0.2) (0.75)(0.25)624
(0.75 0.8 0.1)0
When P is assumed 0.8, then n reduces
E C
C E
A C E
C E
P PH P PH P P
one sided ZZ
n
where P P and
α
β
θθ θ
αβ
θ θ
= =
≥ +< + =
= − =− = =
⎡ ⎤+ +⎣ ⎦= =− −
< + >
2
2
to145Unequal allocation
( 1) (1 ) (1 ) (1 )
( )
s s E E C C
C E
Z s P P Z sP P P Pn
s P P
α β
θ
⎡ ⎤+ − + − + −⎢ ⎥⎣ ⎦=− −
![Page 37: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/37.jpg)
Time to Event ---Assume exponential Distribution
)(log)(2
:1:
)(,)(
2
0
θ
θμμ
μμ
βα
e
C
EA
C
E
ce
ZZn
failuretoupFollow
HH
etSetScurveSurvival uct
uet
+=
==
== −−
![Page 38: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/38.jpg)
Time to Event ---Assume exponential DistributionPatient’s enter the trial at a uniform rate over a T-year period. If the trial terminates at a time T, then
ECiTTTwhere
ZZn
iiii
CE
CE
,)exp(1)(
)()]()([)(
3
211
2
=⎥⎦
⎤⎢⎣
⎡−+−=
−
++= −−
μμμμφ
μμμφμφβα
![Page 39: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/39.jpg)
Time to Event ---Assume exponential DistributionPatients are recruited over the interval (0, T0), but with a follow up until T, then
{ } 1
0
02
)/exp(]/)(exp[11)(−
⎥⎦
⎤⎢⎣
⎡ −−−−−=
TTTT iii
ii
μμμμ
μφ
![Page 40: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/40.jpg)
Example
105)5.1(log
)282.1645.1(2
282.1,9.01645.1,)(05.0
5.1:
2
2
=+
=
==−==
=
en
failuretoupFollow
ZZsidedone
H CEA
β
α
βα
μμ
![Page 41: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/41.jpg)
Example
207105.0)(
0.184)(year1additionanforupfollowbutperiodyear4aoverrecruited
238)35.4(
)125.0217.0()282.1645.1(
125.0]1[)(
217.0]1[)(
3,5.1:,periodyear5aoverStudy
E
C
211
2
15.4
55.45
135
35
5.45
3
35
2
===
−
=−
++=
=+−=
=+−−=
==−
−−
−−
−−
n
n
e
e
H
E
C
CCEA
μφμφ
μφ
μφ
μμμ
![Page 42: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/42.jpg)
Account for Patient dropoutDrop-out rate in group E
2
**
*
)1(*
*))(1(
)1(
dnn
forsubstitutePPdPP
dPdPP
CECE
CEE
−=
ΔΔ−−=−=Δ
+−=
![Page 43: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/43.jpg)
Two-sided equivalenceFor two parallel groups
δθ0where,δμμandθμμ
whenvariancethebeσandσand
δδδletandθμμDefine
RLRL
2a
20
ULRL
≤≤=−=−
==−=−
![Page 44: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/44.jpg)
Two-sided equivalenceGeneral Formulas for sample size per group
[ ]
[ ] 220
220
]/[)2/()(
/)2/()(
0
θδσβσα
δθμμδσβσα
μμ
−+=
≤=−+=
=−
a
RT
a
RT
ZZn
ForZZn
For
![Page 45: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/45.jpg)
Two-sided equivalence
[ ]
[ ] 222
222
22a
20
)/()()(2
,/)2/()(2
0)1995,;1992,(2
thatassume,endpointscontinuousFor
θδβασ
δθμμδβασ
μμσσσ
−+=
≤=−+=
=−==
ZZn
formulaeapproximatForZZn
ForLiuChowandLiu
RT
RT
![Page 46: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/46.jpg)
Adjustment of CovariatesCovariates are factors that affect the primary efficacy endpoints
prognostic, risk, or confounding factorsage, gender, race, disease severity, etc.
Patient-specific covariatesCovariates measured before randomization
Baseline FEV1, FVC, etc.Time-dependent covariates Covariates measured after randomization
May be affected by the treatmentsCd4 level during the trials
![Page 47: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/47.jpg)
Adjustment of CovariatesStratification based on known covariates before randomization and conduct of the trials (pre-randomization adjustment).Adjustment of covariates in the analysis improvement of the precision of the estimated treatment effects (post-randomization adjustment).Adjustment of covariates reduces the variability associated with the estimated treatment effect. The estimated treatment effect is unbiased without adjustment of covariates as long as assignment of treatments is random (That is the rewards when you pay the price of randomization).
![Page 48: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/48.jpg)
Adjustment of CovariatesNeed to check the treatment-by-covariate interaction.If the treatment-by-covariate interaction exists, generalizability of the results is limited. Avoid to adjust the primary endpoints for the covariates measured after randomization.Specify the covariates in the protocol.
![Page 49: Selection of Endpoints and Sample Size Estimation in ...ntur.lib.ntu.edu.tw/bitstream/246246/20060927123048179199/1/sample size...Selection of Endpoints and Sample Size Estimation](https://reader031.fdocuments.in/reader031/viewer/2022040123/5e0c0062229b046864456a2e/html5/thumbnails/49.jpg)
SummaryEfficacy endpoints should be clinically meaningful.Minimal sample size should be chosen to provide sufficient power to detect a clinically significant difference.A under-powered trial is unethical.Adjustment of covariate can reduce variability of the estimated treatment effect.For a randomized trial, unadjusted treatment effect is still unbiased.