Title: Statistics 542 Intro to Clinical Trials Data Monitoring, Monitoring Committee Function
1Statistics 542Intro to Clinical TrialsData
Monitoring, Monitoring Committee Function
Statistical Methods
2Some References
- Texts/Chapters
- 1. Friedman, Furberg DeMets (1998) 3rd
edition, Fundamentals of Clinical Trials,
Springer-Verlag, NY, NY - 2. Pocock (1983) Clinical Trials, Wiley.
- 3. Ellenberg S, Fleming T and DeMets D Data
Monitoring Committees in Clinical Trials A
Practical Perspective. John Wiley Sons, Ltd.,
West Sussex, England, 2002. - 4. Jennison C and Turnbull B (2000) Group
Sequential Methods with Application to Cinical
Trials. Chapman Hall, NY. - 5. DeMets DL (1998) Data and Safety Monitoring
Boards. In Encyclopedia of Biostatistics.
John Wiley and Sons, West Sussex, England, Vol.
2, pp. 1067-71. - 6. DeMets and Lan. The alpha spending function
approach to interim data analysis. In, Recent
Advances in Clinical Trials Design and
Analysis. Kluwer Academic Publishers, Boston,
MA, 1995.
3Some References
- Review Papers
- 1. Greenberg ReportOrganization, review, and
administration of cooperative studies.
Controlled Clinical Trials 9137-148, 1988. - 2. DeMets and Lan (1994) Interim analyses
The alpha spending function approach.
Statistics in Medicine, 13(13/14)1341-52, 1994. - 3. Lan and Wittes. The B-value A tool for
monitoring data. Biometrics 44579-585, 1988. - 4. Task Force of the Working Group on
Arrhythmias of the European Society
of Cardiology The early termination of clinical
trials causes, consequences, and control.
Circulation 89(6)2892-2907, 1994. - 5. Fleming and DeMets Monitoring of clinical
trials issues and recommendations. Controlled
Clin Trials 14183-97, 1993. - 6. DeMets, Ellenberg, Fleming, Childress, et al
The Data and Safety Monitoring Board and AIDS
clinical trials. Controlled Clin Trials
16408-21, 1995. - 7. Armstrong and Furberg Clinical trial data
and safety monitoring boards The search for a
constitution. Circulation 1, Sess6, 1994.
4Data MonitoringRationale
- 1. Ethical
- 2. Scientific
- 3. Economic
5A Brief History
- A 35-year history
- Greenberg Report (1967)
- Coronary Drug Project (1968)
- NIH Experience and Guidelines
- Industry and ICH Guidelines
- Department of Health Human Services Policy
(Shalala, 2000)
6Greenberg Report Recommendations
- Develop a mechanism to terminate early if
- Question already answered
- Trial cant achieve its goals
- Unusual circumstances
- Hypothesis no longer relevant
- Sponsor decision to terminate should be based on
advice of external committee
7Coronary Drug Project (CDP)
- References
- Design (Circulation, 1973)
- Monitoring Experience (CCT, 1981)
- Major Outcome (JAMA 1970, 1972, 1973, 1975)
- Tested several lipid lowering drugs in post MI
patients - Multicenter study
- Mortality as primary outcome
- Began recruitment in 1965
8Coronary Drug Project
- First trial to benefit from Greenberg Report
- Policy Advisory Board
- Senior Investigators, External Experts, NHI
- Initially reviewed interim data
- Data Coordinating and Statistical Center
- Safety Monitoring Committee formed (1968), after
trial was underway
9Early NHLBI CT Model
Funding Agency
Policy Advisory Board
Data and Safety Monitoring Board
Steering Committee
Central Lab(s)
Multiple Clinics
Working Committees
Data Coordinating Center Data Management Statistic
al Analysis
10NHLBI CT Model
Funding Agency
Data Monitoring Committee
Steering Committee
Central Lab(s)
Coordinating Data Center
Clinics
Working Committees
11NIH DMC Activity
- Ref Statistics in Medicine (1993)
- CDP became model for National Heart, Lung, and
Blood Institute (NHLBI) - heart, lung, blood disease trials
- National Eye Institute (NEI) (1972)
- Diabetic Retinopathy Study
- National Institute Diabetes, Digestive and Kidney
(NIDDK) - Diabetes Complication and Control Trial (1980)
- National Cancer Institute (NCI)
- Prevention Trials, Cooperative Group Therapeutic
Trials - National Institute Allergy and Infectious Disease
(NIAID) - AIDS Clinical Trial Group (ACTG) (1986)
12Industry/FDA/ICH
- Industry sponsorship of RCTs expanded
dramatically since 1990 in several disease areas
(e.g. cardiology, cancer, AIDS) - Industry use of DMCs growing as well
- FDA 1989 guidelines very brief mention of data
monitoring and DMCs - International Conference on Harmonization (ICH)
- ICH/E9
- Section 4.5 Interim Analyses
- Section 4.6 Independent DMCs
- ICH/E6
13Independent DMCsWhen are they Needed?
- Department of Health and Human Services Policy
- Shalala (NEJM, 2000) All NIH FDA trials must
have a monitoring plan, for some a DMC may be
required - NIH policy (1998)
- all sponsored trials must have a monitoring
system - safety, efficacy and validity
- DMC for Phase III trials
- FDA guidelines (Nov 2001)
14Need for Independent DMCs
- Phase I Trials (dose)
- Monitoring usually at local level
- Phase II Trials (activity)
- Most monitoring at local level
- Some randomized, blinded, multicenter Phase II
trials may need IDMC - Phase III IV (effectiveness, risk, benefit)
- Most frequent user of IDMC
- Structure of monitoring depends on risk (e.g.
Phase I-IV)
15Data Monitoring Committee
- FDA suggests a need for an
- Independent DSMB for
- Pivotal Phase IIIs
- Mortality or irreversible
- morbidity outcome
16Industry-Modified NIH Model
Pharmaceutical Industry Sponsor
Steering Committee
Regulatory Agencies
Independent Data Monitoring Committee (IDMC)
Central Units (Labs, )
Data Management Center (Sponsor or CRO)
Statistical Analysis Center
Clinical Centers
Institutional Review Board
Patients
17DMC Relationshipsand Responsibilities
- Patients
- Study Investigators
- Sponsor
- Local IRBs
- Regulatory Agencies
18Early Administrative AnalysisDMC and Executive
Committee
- 1. Recruitment/Entry Criteria
- 2. Baseline Comparisons
- 3. Design Assumptions
- a. Control only
- b. Combined groups
19Design Modifications
- 1. Entry Criteria
- 2. Treatment Dose
- 3. Sample Size Adjustment
- 4. Frequency of Measurements
20DMC Data ReviewInterim Analysis
- 1. Recruitment
- 2. Baseline Variables
- -Eligibility
- -Comparability
- 3. Outcome Measures
- -Primary
- -Secondary
- 4. Toxicity/Adverse Effects
- 5. Compliance
- 6. Specified Subgroups
21DMC Recommendations
- 1. Continue Trial / Protocol Unmodified
- 2. Modify Protocol
- 3. Terminate Trial
22Reasons for Early Termination
- 1. Serious toxicity
- 2. Established benefit
- 3. Futility or no trend of interest
- 4. Design, logistical issues too serious to fix
23DMC Decision Making Process Complex (1)
- Recruitment Goals
- Baseline risk and comparability
- Compliance
- Primary and secondary outcomes
- Safety
24DMC Decision Making Process Complex (2)
- Internal consistency
- External consistency
- Benefit/Risk
- Current vs future patients
- Clinical/Public impact
- Statistical issues
25DMC Decision Making Role
- DMC makes recommendations, not final decisions
- Independent review provides basis for DMC
recommendations - DMC makes recommendations to
- Executive Committee who recommends to sponsor,
or - Sponsor
- DMC may, if requested, debrief Executive
Committee and/or sponsor - Rarely are DMC recommendations rejected
26DMC Meeting Format
- Open Session
- Progress, blinded data
- Sponsor, Executive Committee, DMC, SAC
- Closed Session
- Unblinded data
- DMC, SAC
- Sponsor Rep? (Not recommended)
- Executive Session
- DMC only
- Debriefing Session
- DMC Chair, Sponsor Rep, Executive Committee Rep
27DMC Relationships
- Regulatory Agencies (e.g. FDA)
- Could perhaps brief DMC about specific concerns
at Open Session - Should not participate in DMC Closed Sessions
- Should be briefed about DMC recommendations/decisi
ons ASAP following Executive Committee
28DMC Membership
- Monitoring is complex decision process and
requires a variety of expertise - Needed expertise
- Clinical
- Basic science
- Clinical trial methodology
- Biostatistics
- Epidemiology
- Medical ethics
- Helpful expertise
- Regulatory
- Some experience essential
29DMC Confidentiality
- In general, interim data must remain confidential
- DMC may rarely release specific/limited interim
data (e.g. safety issue) - Members must not share interim data with anyone
outside DMC - Leaks can affect
- Patient Recruitment
- Protocol Compliance
- Outcome Assessment
- Trial Support
30DMC Liability
- Recent events (eg Cox-IIs, Vioxx) have raised the
potential for litigation - Members have been gotten a subpoena
- DMC Charters for industry trials now often cover
indemnification clauses - No indemnification yet for NIH trials
31DMC Needs On-LineData Management and Analysis
- DMC reluctant to make decisions on old data
- Minimize data delay and event verification (e.g.
NOTT, ACTG 019) - Be prepared from start (e.g., CAST)
- Focus on key variables, not complete case reports
(delays can be problematic)
32Levels of Independence
- Totally Inhouse Coordinating Center
- Internal DM, Internal SAC, External DMC
- Internal DM, External SAC, External DMC
- External DM(e.g. CRO), External SAC, External DMC
33Industry-Modified NIH Model
Pharmaceutical Industry Sponsor
Steering Committee
Regulatory Agencies
Independent Data Monitoring Committee (IDMC)
Central Units (Labs, )
Data Management Center (Sponsor or CRO)
Statistical Analysis Center
Clinical Centers
Institutional Review Board
Patients
34DMC Summary
- NIH Clinical Trial Model - long history of
success - Adaptation for industry can be made
- SC, DMC, SAC or DM are critical components
- Independence of DMC essential
- Best way to achieve this goal is for external SAC
and external DMC
35Data Monitoring Process
- 1. DMC and the decision process
- 2. A brief introduction to statistical
monitoring methods - a. Group Sequential
- b. Stochastic Curtailment
- 3. Examples
- Ref BHAT, DeMets et al. Controlled
Clin Trials,1984
36Decision Factors
- 1. Comparability
- 2. Bias
- 3. Compliance
- 4. Main effect vs. Potential side effects
- 5. Internal Consistency
- a. Outcome measures
- b. Subgroups
- c. Centers
- 6. External Consistency
- 7. Impact
- 8. Statistical Issues/Repeated Testing
37Beta-blocker Heart Attack Trial (BHAT)
- Preliminary Report. JAMA 2462073-2074, 1981
- Final Report. JAMA 2471707-1714, 1982
- Design Features
- Mortality Outcome 3,837 patients
- Randomized Men and women
- Double-blind 30-69 years of age
- Placebo-controlled 5-21 days post-M.I.
- Extended follow-up Propranolol-180 or 240 mg/day
38BHATAccumulating Survival Data
- Date Data Monitoring
- Committee Meeting Propranolol Placebo Z(log
rank) - May 1979 22/860 34/848 1.68
- Oct 1979 29/1080 48/1080 2.24
- March 1980 50/1490 76/1486 2.37
- Oct 1980 74/1846 103/1841 2.30
- April 1981 106/1916 141/1921 2.34
- Oct 1981 135/1916 183/1921 2.82
- June 1982
- Data Monitoring Committee recommended
termination
39Beta-Blocker Heart Attack Trial October 1,
1981LIFE-TABLE CUMULATIVE MORALITY CURVES
40Beta-Blocker Heart Attack TrialBaseline
Comparisons
- Propranolol Placebo
- (N1,916) (N1,921)
- Average Age (yrs.) 55.2 55.4
- Male () 83.8 85.2
- White () 89.3 88.4
- Systolic B.P. 112.3 111.7
- Diastolic B.P. 72.6 72.3
- Heart rate 76.2 75.7
- Cholesterol 212.7 213.6
- Current smoker () 57.3 56.8
41Beta-Blocker Heart Attack TrialTotal
Mortality(Average 24-Month Follow-Up)
- Propranolol Placebo
- Age 30-59 5.9 7.1
- 60-69 9.6 14.4
- Sex Male 7.0 9.3
- Female 7.1 10.9
- Race White 6.7 9.0
- Black 11.0 15.2
42Beta-Blocker Heart Attack TrialTotal
Mortality(Average 24-Month Follow-Up)
- Propranolol Placebo
- Risk Group I 13.5 16.9
- Risk Group II 7.8 11.4
- Risk Group III 5.2 7.1
43DMC Interim Analysis
- Ethical, scientific and financial reasons
- Repeated analysis of accumulating data causes a
statistical problem
44Coronary Drug Project
Cumulative morality rate
Month of Follow-up
Life-table cumulative mortality rates, Coronary
Drug Research Project Group
45Coronary Drug Project Research Group
Placebo Superior
Clofibrate Superior
z values for clofibrate-placebo differences in
proportion of deaths by calendar month since
beginning of study (Month 0 March 1966, Month
100 July 1974)
46Repeated Significance Testing
- Repeated testing increases Type I error AMR
(JRSS, 1969) - Example
- Critical Value 1.96
- Reject H0 if Z gt 1.96
- No. Of Looks (Planned) Type I Error
- 1 0.05
- 2 0.08
- ...
- 5 0.14
- ...
- 10 0.20
- Must adjust interpretation of z to be
conservative.
47Three Procedures for Conservative Interim
Monitoring
- 1. Group Sequential
- A modification of classical sequential
- 2. Curtailed Sampling/Conditional Power
- 3. Bayesian (not discussed here)
- FIRST TWO METHODS HAVE PRIMITIVE VERSIONS IN
CORONARY DRUG PROJECT
48Group Sequential BoundariesBasic Idea
- Compute summary statistic at each
interim analyses, based on additional group of
new subjects (events) - Compare statistic to a conservative critical
value - 2? 0.05 overall
- Various Methods
- Haybittle-Peto (1971, 1976)
- Pocock (1977)
- OBrien-Fleming (1979)
- Slud and Wei (1982)
- Lan and DeMets (1983)
49Group Sequential Boundaries
50Haybittle-Peto Group Sequential Boundaries
- Simple, Ad Hoc
- For interim analyses, use very conservative
critical value - e.g., 3.0
- For final analysis, no adjustment really needed
- e.g., 1.96
- Significance level approximate
- e.g., 0.05
- References Haybittle. British Journal of
Radiology, 1971 - Peto et al. British Journal of Cancer, 1976
- Refinement
- Fleming, Harrington, OBrien. Controlled
Clinical Trials, 1984
51Classical Group Sequential Boundaries
- Proposal
- Analysis after groups of 2n subjects
- Maximum of K groups, independent
- Same critical value at each analysis
- Basic Model
- Zj statistics for jth group
- j 1, 2, ..., K (K groups, 2n each)
- Zj N(?,1), independent
- H0 ? 0
- References
- Armitage, McPherson, and Rowe, J Royal
Statistical Society, 1969 - McPherson and Armitage, Journal of the Royal
Statistical Society, 1971 - Pocock. Biometrika, 1977
52Summary Statistic
- Add up statistic for each sequential group
- That is
- This is often the usual test statistic
53Summary Statistic
- - Let
-
-
- - H0 ? 0
- - Decision
- 1. Continue if -Zp ltZilt Zp for i lt K
- 2. Otherwise, stop and reject H0 for i lt K
- - Pick Zp to give overall ?
- 3. If i K
- Reject H0 if Zi gt Zp
- Accept H0 if Zi lt Zp
54- the summary statistics for all data in
the 1st i groups - Binomial a.
- Normal b.
- Survival c.
-
-
- Process assumes independent increments
55Pocock (1977) Group Sequential Boundaries
- Boundaries
- Zp
- Example
- - " 0.05, K 5, Zp 2.413
- - 1 - ? 0.90 ? ? 1.59
- Compare Two Sample Means where H0µC µT
56Pocock (1977) Group Sequential Boundaries
- Design
- - e.g., HA µT - µC 0.5 ?
- -
- 20
- - 2nK 2(20)5
- 200 subjects
57OBrien-Fleming (1979)Group Sequential Boundary
- Basic Model
- Continue if
-
- for i lt K
- Critical value decreases as i increases
- For i K, Reject H0 if ZKgtZOBF
- Boundary Example
-
- Values
- i1 2.04 v(5/1) 4.88
- i2 2.04 v(5/2) 3.36
- i3 2.04 v(5/3) 2.68
- i4 2.04 v(5/4) 2.29
- i5 2.04 v(5/5) 2.04
- Reference OBF, Biometrics, 1979
58OBrien-Fleming (1979)Group Sequential Boundary
- Design Example
-
-
-
-
- Reference OBF, Biometrics, 1979
59Fixed Sample SizeComparison of Means
- Sample size 2? .05 Power .90 ?/? .5
N 84 2N 168
60Group Sequential Boundaries
- 2? .05
- Power .90, .80
- OBrien-Fleming Pocock
- ?OBF ?P
- K ZOBF 90 80 ZP 90 80
- 1 1.96 3.24 2.80 1.96 3.24 2.80
- 2 1.98 2.30 1.99 2.18 2.40 1.98
- 3 1.99 1.88 1.63 2.29 2.01 1.63
- 4 2.03 1.64 1.42 2.36 1.76 1.53
- 5 2.04 1.46 1.27 2.41 1.59 1.38
61Group Sequential Boundaries
- 2? .01
- Power .90, .80
- OBrien-Fleming Pocock
- ?OBF ?P
- K ZOBF 90 80 ZP 90 80
- 1 2.58 3.86 3.42 2.58 3.86 3.42
- 2 2.58 2.73 2.42 2.77 2.84 2.55
- 3 2.59 2.23 1.98 2.87 2.36 2.12
- 4 2.60 1.94 1.72 2.94 2.07 1.86
- 5 2.62 1.74 1.54 2.99 1.87 1.67
62Group Sequential Boundaries
63BHAT GSB
64Lan-DeMets (1983)Group Sequential Boundaries
- Criticism of classical GSB
- Number of analyses specified in advance
- Equal increments in information
- Lan and DeMets specify ?(t)
- ?(t) defines rate at which Type I error is used
where t is proportion of total information
accumulated by calendar time tC - So 0 lt t lt 1
- Thus ?(t) increasing
- ? (0) 0
- ?(1) ?
65- Information and Calendar Time
- t proportion of information accumulated by tc
- Examples
- A. Immediate Response
- x1, x2, . . . . , xn, . . . ., xN,
- y1, y2, . . . . , yn, . . . ., yN,
- tc
- t 2n/2N n/N
- B. Failure Time (e.g. logrank)
-
66Group Sequential Methods
K 5, ? .05
67Boundary Crossing Probability
- E.g., K 5, ? 0.025
- Upper Boundary
- C1 C2 C3 C4 C5
- Pocock (2.41, 2.41, 2.41, 2.41, 2.41)
- OBF (4.56, 3.23, 2.63, 2.28, 2.04)
- Pocock OBF
- 1. P Z1 gt C1 0.0079 (0.000)
-
- 2. P Z1 gt C1 or Z2 gt C2
- 0.0079 0.0059 0.0138 (0.0006)
-
- 3. P Z1 gt C1 or Z2 gt C2 or Z3 gt C3
- 0.0138 0.0045 0.0183 (0.0045)
-
- 4. P Z1 gt C1, ..., Z4 gt C4
- 0.0183 0.0036 0.0219 (0.0128)
-
- 5. P Z1 gt C1, ..., Z5 gt C5
- 0.0219 0.0031 0.0250 (0.0250)
68(No Transcript)
69 (t2) - ? (t1)
70Examples of ?(t)
- Approximates
- 1. OBF
- 2. ?2 (t) ? ln 1 (e - 1)t Pocock
- 3. ?3 (t) ?t
- Comparison of Boundaries (? .025, N 5)
- Values C1 C2 C3 C4 C5
- 1. OBF 4.56 3.23 2.63 2.28 2.04
- ?1(t) 4.90 3.35 2.68 2.29 2.03
- 2. Pocock 2.41 2.41 2.41 2.41 2.41
- ?2(t) 2.44 2.43 2.41 2.40 2.38
- 3. ?3(t) 2.58 2.49 2.41 2.34 2.28
71Cardiac Arrhythmia Suppression Trial (CAST)
- Ref NEJM 321(6)406-12, 1989
- Cardiac arrhythmias associated with increased
risk of sudden death - New class of drugs (eg, encainide, flecanide)
suppressed arrhythmias - CAST designed to test effect on sudden death
72CAST GSB
- ? spending function approach
- ?(t) ½ ? t t lt 1
- ? t 1
- for benefit ? 0.025
- Used symmetric ? 0.025 boundary for harm
73CAST Interim DataSudden Death
- Time Placebo Drug LogRank ZL ZU
- 9/01/88 5/576 22/571 -2.82 -3.18 3.01
- 3/30/89 9/725 33/730 -3.22 -3.04 2.71
- Initially expected 100 events/arm
74CAST Sequential Boundaries
75Estimation of Treatment Effect
- If terminated early
- Need adjusted estimates
- Naive estimate typically too low
- Naive confidence interval (CI) does not have
correct coverage - Need an adjusted CI
- Some References
- Classical GSB Tsiatis, Rosner, Mehta.
Biometrics, 1984 - Fan DeMets, 2000.
76Repeated Confidence Intervals
- Use RCIs
- Calculate RCI using GSB boundary
- Specify as least medically significant
difference -
- Otherwise continue
- References
- Jennison and Turnbull. Controlled Clinical
Trials, 1984 - Jennison and Turnbull. Biometrika, 1985
zL(i)lower sequential boundary zU(i)upper
sequential boundary
77Repeated Confidence Intervals
d
1.0
78Repeated Confidence Intervals
d
o
o
o
1.0
information fraction
79Symmetric or Asymmetric GSBs
- For two existing or alternative treatments
- Symmetric GSBs
- Level of evidence same in either direction
- For a new treatment vs. standard of care
- Asymmetric GSBs
- Level of evidence less for negative (harmful)
trends than for positive (beneficial) trends - If new treatment not proven but already in
practice, may require more symmetric GSBs than
asymmetric GSBs - Need convincing evidence to discourage or stop a
practice - Reference
- DeMets, Pocock Julian, Lancet, 1999
80Futility
- Generally defined as a trial not being able to
meet its primary objective - Benefit
- Effect in either direction
- Non-inferiority
- May also involve other secondary objectives
81Types of Decision Errors
- Type I making a false claim of an intervention
effect - Beneficial
- Harmful
- Type II making a false claim of no intervention
effect
82Negative Trends
- Trends are variable, especially early
- Different trend patterns
- Settle down on no difference
- Reverse to become beneficial (positive)
- Drift downwards to demonstrate harm
- Plunge towards harm
83Negative Trend Examples
- Inotropic drugs for chronic heart failure
- Milrinone/PROMISE
- Vesnarinone/VEST
- Anti-arrhythmic drugs for cardiac arrhythmia
patients/CAST - Hormone replacement therapy (HRT) / Womens Health
Initiative (WHI) HERS - Coumadin vs Aspirin/ CARS
- Atenolol vs Placebo in MI / ISIS-I
84Methods for Assessing Negative Trends
- Harm
- Symmetric Boundaries such as group sequential
boundaries - Asymmetric Group Sequential Boundaries
- Futility
- Beta Spending Functions
- SPRT/Triangular Boundaries
- Conditional Power
- Predictive Power (Bayesian)
85Group Sequential Boundaries 1? 0.25
86Prospective Randomized Milrinone Survival
Evaluation (PROMISE)New England Journal of
Medicine, 1991
- Test Milrinone to reduce mortality in class III
and - IV congestive heart failure patients
- Milrinone known to increase cardiac contractility
- Randomized, double-blind
- 119 clinical centers
- Outcome
- Primary All cause mortality
- Secondary Cardiovascular mortality
Hospitalizations
87PROMISE
- 1088 patients randomized (561 M, 527 P)
- Median follow-up 6 months (0, 20 months)
- 90 compliance
- None lost to follow-up
- DSMB stopped PROMISE 5 months early
- 168 deaths on Milrinone (30)
- 127 deaths on placebo (24)
- 95 deaths cardiovascular
88Survival Curves for the PROMISE Trial
89Group Sequential Boundaries for the PROMISE Trial
90PROMISE MORTALITY (milrinone vs. placebo)
Group Sequential Bounds (1a .025)
91PROMISE DMC
- Agonizing negative trend
- Issue of symmetric vs. asymmetric GSB
- PROMISE crossed zone of futility
- Milrinone approved for IF use for high risk
patients - Milrinone improved cardiac function
- Needed to rule out neutrality
- from harm on mortality
92Conditional Power
- One method to assess likelihood of achieving a
statistically significant result at the trial
conclusion. - Compute probability of rejecting Ho at the trial
conclusion, given current trend and assumed
effect for remaining portion of the trial
93Stochastic Curtailed Sampling
- Also called Conditional Power
- Likelihood of a Trend Reversal
- Lan, Simon Halperin.
- Communications in Statistics-C 1207-219, 1982
- Lan and Wittes, Biometrics, 1988.
- Bound Type I Type II Error
94- In General
- Let Z(T) test statistic at end of trial
- Z(t) current value at time t
- R rejection region
- R acceptance region
- P Z(T) ? R H0 ?
- P Z(T) ? R HA ?
- or P Z(T) ? R HA 1 - ?
95- Lan, Simon, Halperin (1982) compute
- Decision Compute When
- reject P Z(T) ? R H0, Z(t) ?0 positive
trend -
- accept P Z(T) ? R HA, Z(t) ?A negative
trend - Then
- P Type I error ? ?/?0 ??
- P Type II error ? ?/?A ??
96- Example (? 0.05)
- ?0 ??
- 0.70 0.071
- 0.80 0.063
- 0.85 0.059
- Note
- If ?0 1 or ?A 1
- ? Curtailment
97B-ValueA Method for Computing Conditional
PowerLan Wittes (1988) Biometrics
- Let t n/N (or d/D)
- Z(t) current standardized statistic
-
- Now Z(1) B(1) and
- ( observed remaining)
-
98Visual Aid
H0 ? 0 HA e.g. ?
B(1)
B(t)
99Conditional Power
100Conditional Power
1. Survival D total events 2. Binomial N
total sample size
101Conditional Power (2)
3. Means N total sample size
102 Conditional Power Table
- Typically, compute conditional power for a series
of assumed intervention effects (for observed
negative trends) - Observed effect to date
- Assumed effect
- Null effect
- Effects in between
- For a specified conditional power (e.g. 10, 20
30), boundaries can be generated
103Conditional Power Boundaries
104Example BHAT
- Expected Deaths D 398
- Observed Deaths 183 Placebo d 318
- 135 Propranolol D d 80 398
- Observed logrank Zd 2.82 t 318/398 .80
- Compute Conditional Power under H0
- 1 - ? - 1.25
- 0.89
105BHAT Stochastic Curtailed Sampling
106Conditional Power
- One method to assess likelihood of achieving a
statistically significant result at the trial
conclusion. - Compute probability of rejecting Ho at the trial
conclusion, given current trend and assumed
effect for remaining portion of the trial
107Response Adaptive Designs
- Adjust/increase sample size if treatment effect
assumed was too large - Traditionally, this approach discouraged
- Recent methodology suggests possible approaches
108Proschan Hunsberger Method
- Simple method may make Type I error substantially
less than 0.05 - Developed method to obtain exact Type I error as
a function of Z1 and n2, using a conditional
power type calculation
109Proschan Hunsberger
Conditional Power and p value required in stage 2
as a function of R n2/n1 for the NHLBI Type II
study example
110Proschan Hunsberger
- Allows for sample size adjustment based on
observed treatment effect - Requires increasing final critical value
111Conditional Power Method
- Lan, Simon Halperin (1982, Com in Stat)
- Lan Wittes (1988, Biometrics)
- Lan Zucker (1995, Stat in Med)
112Conditional Power Method
- Cui, Hung Wang (1997, ASA Biopharm Proceedings)
- CP(?) gt 1.2 CP(?) Decrease N
- CP(?) lt 0.8 CP(?) Increase N
- Properties not well characterized
113Conditional Power Method
- Chen, DeMets Lan (2004, Stat Med)
- Can extend to a second stage if CP gt 50 for
current trend - Lan Trost (Proceeding of Biopharm, ASA, 1997)
114Lan Trost
- If CP (n) lt cl , then terminate for futility
accept null (REQUIRED) - If CP(n) gt cu , then continue with no
change - If cl lt CP(n) lt cu , then increase sample size
from N0 to N to get conditional power to desired
level - Type I error is controlled at nominal level
115Chen, DeMets Lan
- Increase N0 if interim result promising
- Conditional power gt 50 for current trend
- Increase in N0 not greater than 1.75 times
- Under these conditions, Type I error is not
increased and no practical loss in power
116Why does it work?
117Chen, DeMets Lan
118Adaptive Design Remarks
- A need exists for adaptive designs (even FDA
statisticians agree) - Technical advances have been made through several
new methods - Adaptive designs are still not widely accepted
subject to (strong) criticism - May be useful for non pivotal trials
- Practice precedes theory, perhaps in time
119Some Data Monitoring Examples
120Vesnarinone in Heart Failure
- Two Trials
- First Trial
- (New Engl J of Med 329149-155, 1993)
- Second Trial
- (New Engl J of Med 339 1810-16, 1998)
121Vesnarinone in Heart Failure Trial I
- First Trial (New Engl J of Med 329149-155, 1993)
- Patients with Class II-III heart failure
- A vasodilator and inotrope
- Randomized double blind
- Vesnarinone (120 mg, 60 mg) vs. placebo
- Mortality outcome
122VESNARINONE- Trial I
- A 60 mg dose had no effect on exercise tolerance
or symptoms - A 120 mg dose increases exercise tolerance and
reduces symptoms - 120 mg arm stopped early with increased mortality
(6 vs. 16, plt.01 - 60 mg arm continued observed a 60 reduction in
mortality
123VESNARINONETrial I
- Plbo 60mg P
- Mortality 33/238 13/239 .002
- Mortality and
- Morbidity 50/238 26/239 .003
124Vesnarinone in Heart Failure Trial II - VEST
- Second Trial (New Engl J of Med 3391810-16,
1998) - Two doses (30 60 mg) vs. placebo
- NYHA Class III/IV patients, LVEF LT 30
- Randomize double blind
- Mortality outcome
125VEST(NEJM, 1998)
- 3833 patients randomized
- Primary Outcome observed an increase in
mortality on vesnarinone HR 1.2, unadjusted
p 0.02 - Secondary Outcome
- Worsening Mortality plus CHF hospitalization
- Improved Quality of Life
126VESTSurvival in the Three Groups
127Acumulating Results for VEST
- Information Fraction Logrank Z value
- (high dose)
- .43 0.99
- .19 - 0.25
- .34 - 0.23
- .50 - 2.04
- .60 - 2.32
- .67 - 2.50
- .84 - 2.22
- .90 - 2.43
- .95 - 2.71
- 1.0 - 2.41
128VEST MORTALITY (high dose vs. placebo)
129Conditional Power for VEST
- RR Information Fraction
- .50 .67 .84
-
- .50 .46 .01 lt.01
- .70 .03 lt.01 lt.01
- 1.0 lt.01 lt.01 lt.01
- 1.3 lt.01 lt.01 lt.01
- 1.5 lt.01 lt.01 lt.01
130MERIT-HF Study Design
- Chronic heart failure patients
- Randomized placebo controlled
- Metoprolol (a beta-blocker) vs. placebo
- Two-week placebo run in (compliaance)
- Entered 3991 patients
- Terminated early
- Mean followup approximately one year
- The International Steering Committee on Behalf of
the MERIT-HF Study Group, - Am J Cardiol 1997 80(9B)54J-58J. The MERIT-HF
Study Group, ACC, March 1999.
131MERIT-HF Entry Characteristics
Meto CR/XL Placebo Mean age (years) 64 64 Male
sex () 77 78 NYHA class II () 41 41
III () 56 55 IV () 3.5 3.8 Ejection
fraction 0.28 0.28
Data unblinded by ISaC The MERIT-HF Study Group,
ACC, March 1999
132Total Mortality
133MERIT-HF Monitoring Bounds for Total Mortality
X3.807 corresponds to a marginal p-value of
approx. p0.00015
DeMets, Julian, Chatterjee Guidelines for
Interim Analyses in MERIT-HF
134Relative Risk and 95 Confidence Interval
135Relative Risk
136Diabetic Retinopathy Study (DRS)
- Randomized placebo controlled
- Randomized eyes
- photocoagultion vs. no therapy
- Outcome visual loss
- Patients (1700) with proliferative retinopathy
- Planned 5 yr. follow-up
- Protocol change at 2 yrs.
- Unanticipated dramatic early benefit (RR .4, Z
5.5) - No long term track record for photocoagulation
- Decision
- Treat all untreated eyes at high risk
- Publish interim results
- Follow all patients for long term "adverse" and
beneficial effects
137Diabetic Retinopathy Study (DRS)Ref DRS Group
(1980) Ophthalmology
138ISIS-I Trial
- Ref ISIS-I. Lancet, 1986
- A RCT evaluating the early use of atenolol vs.
placebo in post MI patients
139(No Transcript)
140Data Monitoring Summary
- DMCs useful to assess risk/benefit
- Treatment effect trends emerge with variability
over time - Statistical methods available to minimize false
positive claims - No single method adequate
141Downtown Madison Monona Terrace
142More Details on Conditional Power
143B-ValueA Method for Computing Conditional
PowerLan Wittes (1988) Biometrics
- Let t n/N (or d/D) n tN
-
-
- Now ZN B(1)
- and
- ( obs remaining)
144Conditional Power
- B(1) B(t) B(1) - B(t)
- Properties of B(t) B(1) - B(t)
- (i) B(t) B(1) - B(t) Normal Independent
- (ii) E B(t) ?t
- E B(1) - B(t) ? - ?t ? (1-t)
- (iii) V B(t) t
- V B(1) - B(t) 1 - t
145Conditional Power
- Conditional Distribution
- EB(1) B(t) B(t) ? (1-t)
- V B(1) - B(t) 1 - t
- PZN B(1) ? Z?B(t), ?)
-