interim analysis of clinical trial liying xu ccter, cuhk

45
Interim Analysis of Clinical Trial Liying XU CCTER, CUHK

Post on 20-Dec-2015

224 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: Interim Analysis of Clinical Trial Liying XU CCTER, CUHK

Interim Analysis of Clinical Trial

Liying XUCCTER, CUHK

Page 2: Interim Analysis of Clinical Trial Liying XU CCTER, CUHK

Clinical Trial Protocol

• Data safety and monitoring

• Safety Analysis

• Monitoring of the trial Data and Safety Monitoring Committee External Advisory Committee

Page 3: Interim Analysis of Clinical Trial Liying XU CCTER, CUHK

Data and Safety Monitoring Committee (DSMC)

• independent with the investigators, participants or sponsors.

• Including experts: clinicians, epidemiologist, and bio-statisticians (and lay representatives).

• To review accumulating data related to treatment effects, adverse events and trial performance.

• To protect the integrity of the clinical trial from adverse impact resulting from access to trial information with pre-defined SOPs.

• DSMC is a separate entity from an IRB or IEC.

Page 4: Interim Analysis of Clinical Trial Liying XU CCTER, CUHK

The Evaluation of Subcutaneous Proleukin(interleukin-2) in a Randomized International Trial (ESPRIT)

• Patients with HIV diseases,CD4>=300cells/mm3

• Primary objective: to determine whether the addition of IL-2 to combination antiretroviral therapy improves morbidity and mortality

• Sample size:4000

• Follow-up: 5 years

• 27 sites and 23 countries

Page 5: Interim Analysis of Clinical Trial Liying XU CCTER, CUHK

Statistical considerations

• Time to event methods including stratified proportional hazard models, log-rank tests, and Kaplan-Meier cumulative event curves will use used to summarize the major outcomes of HIV-disease progression including death and survival.

• These analysis will be stratified by centre.

Page 6: Interim Analysis of Clinical Trial Liying XU CCTER, CUHK

Data monitoring

• The independent DSMC will meet twice each year to review interim analyses.

• O’Brien-Fleming boundaries and the Lan-DeMets spending function will be used as monitoring guidelines for for the primary endpoint comparisons.

Page 7: Interim Analysis of Clinical Trial Liying XU CCTER, CUHK

Other reason for early termination

• The DSMB will also consider results from other studies and recommend early termination or modification of ESPRIT only when there is clear and substantial evidence of benefit or harm.

Page 8: Interim Analysis of Clinical Trial Liying XU CCTER, CUHK

Other reason for early termination

• The EC or DSMB may consider early termination of trial for reasons of poor accrual, less than anticipated CD4 cell count differences between treatment groups, or excessive loss to follow-up.

Page 9: Interim Analysis of Clinical Trial Liying XU CCTER, CUHK

Data monitoring?

• Asking the same question several times, with the only difference being the amount of data available to answer it.

Page 10: Interim Analysis of Clinical Trial Liying XU CCTER, CUHK

The Prehospital Treatment of Status Epilepticus (PHTSE) study:

• 1) to determine whether administration of bezodiazepines by paramedics is an effective and safe means of treating status epilepticus in the prehospital setting and whether this therapy influences longer-term patient outcome

• 2) to determine whether lorazepam is superior to diazepam for the treatment of status

Page 11: Interim Analysis of Clinical Trial Liying XU CCTER, CUHK

Safety analysis

• Interim safety analysis are performed after the enrollment of 25, 50,100 and 150 unique subjects.

Page 12: Interim Analysis of Clinical Trial Liying XU CCTER, CUHK

Interim analysis

• Detect: early dramatic benefits potential harmful effects.

• Done by independent person.

• Statistic technique(s) is not the sole basis for the decision to stop or continue the trial.

Page 13: Interim Analysis of Clinical Trial Liying XU CCTER, CUHK

Rational for interim analysis

• Ethical

• Scientific

• Economic

Page 14: Interim Analysis of Clinical Trial Liying XU CCTER, CUHK

Decision making process

• Statistical results of interim analysis.

• The merits of the treatment.

• The availability and usefulness of alternative treatments.

• The seriousness of the conditions being treated.

• The acceptability of the treatment to patients,

• Other findings

Page 15: Interim Analysis of Clinical Trial Liying XU CCTER, CUHK

Decision to extend a trial

• To maintain the power

To increase the sample size. To extend the length of follow-up.

Page 16: Interim Analysis of Clinical Trial Liying XU CCTER, CUHK

Frequency of interim analysis

• Long term clinical trials, 4 to 6 month intervals.

• The time lag between entry and response evaluation.

• Special meeting for unexpected toxicity of one intervention.

• Rate of patients accrual.

• 10%, 25% 75% and 100% of the primary outcomes have been observed.

Page 17: Interim Analysis of Clinical Trial Liying XU CCTER, CUHK

Examples

One full interim analysis will be undertaken when half the participants have followed for 3 years

Three planned analysis when the subjects are 18 months, 3 years and 5 years of age

Page 18: Interim Analysis of Clinical Trial Liying XU CCTER, CUHK

Statistical stopping rules

• A simple rule to ensure that the overall probability of type I error is controlled.

• In one anticipates no more than 10 interim analyses and there is one main response variable, one can adopt P<0.01 as the criterion for stopping the trial, since the overall type I error will not exceed 0.05.

Page 19: Interim Analysis of Clinical Trial Liying XU CCTER, CUHK

Group sequential methods (Pocock 1977)

• What is the maximum number of interim analyses (or groups)?

• How many patients should be evaluated between successive analysis, i.e. what should be the size of each group?

Page 20: Interim Analysis of Clinical Trial Liying XU CCTER, CUHK

Risk of false positive

• If one carries out 10 interim analyses the chance of at least one analysis showing a treatment difference significant at the 5% level increases to 0.19 even if the treatments are truly equally effective.

Page 21: Interim Analysis of Clinical Trial Liying XU CCTER, CUHK

Repeated significance tests on accumulating data

No. of repeated tests at the 5% level Overall significant level

1

2

3

4

5

10

20

50

100

1000

0.05

0.08

0.11

0.13

0.14

0.19

0.25

0.32

0.37

0.53

1.0

For two treatments, a normal response with known variance and equally spaced analyses

Though broadly similar results for other type of data

Page 22: Interim Analysis of Clinical Trial Liying XU CCTER, CUHK

Nominal significance levels

• A more stringent nominal significant level for each repeated test, to keep overall significant level at some reasonable value, 0.05 or 0.01.

Page 23: Interim Analysis of Clinical Trial Liying XU CCTER, CUHK

Nominal significance level required for repeated two-sided significance testing with overall significance level (0.01 or 0.05) and various N

N (Max. tests) = 0.05 = 0.01

2

3

4

5

10

15

20

0.029

0.022

0.028

0.016

0.0106

0.0086

0.0075

0.0056

0.0041

0.0033

0.0028

0.0018

0.0015

0.0013

Page 24: Interim Analysis of Clinical Trial Liying XU CCTER, CUHK

Fig. 1 Three group sequential stopping boundaries for the standard normal statistic (Zi) for up to five sequential groups with two sided significance level of 0.05

Page 25: Interim Analysis of Clinical Trial Liying XU CCTER, CUHK

Table 1. Nominal P values for overall type I error of (=0.05)

k Pocock O’Brien—Fleming

1

2

1

2

3

1

2

3

4

1

2

3

4

5

0.0294

0.0294

0.0221

0.0221

0.0221

0.0182

0.0182

0.0182

0.0182

0.0158

0.0158

0.0158

0.0158

0.0158

0.0051

0.0415

0.001

0.0151

0.0471

0.001

0.0039

0.0184

0.0412

0.001

0.0013

0.0085

0.0228

0.0417

Page 26: Interim Analysis of Clinical Trial Liying XU CCTER, CUHK

Group sequential tests boundaries (see Fig.1)

• Pocock (1977) Divides equally the overall significance levels

• Peto (1976) Interim tests are run with a very low level of

significance (0.001), which has little impact on the level of significance of the final test.

• O’Brien and Fleming (1979) It is a intermediate between the previous two methods,

have slightly increase in the significance level on each following test.

Page 27: Interim Analysis of Clinical Trial Liying XU CCTER, CUHK

Pocock’s method

• A trial in non-Hodgkins lymphoma compared two drug combinations CP (cytoxan-prednisone) and CVP (cytoxan-vincristine-prednisone)

• Outcome measure: tumor shrinkage.

• Patient accrual lasted over two years

• 120-130 patients were entered.

• Five interim analysis were planned: one after about every 25 patients were entered.

Page 28: Interim Analysis of Clinical Trial Liying XU CCTER, CUHK

Interim analyses for a trial innon-Hodgkins lymphoma

Response Rate

CP CVP X2 (without continuity correction )

Analysis 1

Analysis 2

Analysis 3

Analysis 4

Analysis 5

3/14

11/27

18/40

18/54

23/67

5/11

13/24

17/36

24/48

31/59

1.63

0.92

0.04

3.25

4.15

0.05<P<0.1

0.016<P<0.05

Page 29: Interim Analysis of Clinical Trial Liying XU CCTER, CUHK

Conclusion

• The superiority of CVP is interesting but inconclusive.

• However, further data on response duration and survival eventually clarified that CVP did appear to be a better therapy.

Page 30: Interim Analysis of Clinical Trial Liying XU CCTER, CUHK

How many interim analysis

• The theoretical results indicate that there is little statistical advantage in having a large number of repeated significant tests. As a general rule, it would seem sensible to plan on a maximum of five interim analysis.

Page 31: Interim Analysis of Clinical Trial Liying XU CCTER, CUHK

• It is sensible to consider just two analyses for the

trial currently undertaken without interim analysis.

One half way through and the other at the end.

There can still be a major reduction in the number

of patients exposed to an inferior treatment since

for such a trial with sufficient overall power there

is a reasonable chance of being able to stop

halfway though.

How many interim analysis

Page 32: Interim Analysis of Clinical Trial Liying XU CCTER, CUHK

Alpha-spending functions (Lan and DeMets)Flexible group sequential procedures (1983)

• The limitation of group sequential methods.

To specified the number K of planned interim analyses in advance.

The requirement for equal numbers of either participants or events between each analysis.

The exact time of the interim analysis is specified.

Page 33: Interim Analysis of Clinical Trial Liying XU CCTER, CUHK

Alpha-spending function

• To allow investigators to determine how they want to “spend” the type-I error or alpha during the course of the trial.

• The alpha spending function guarantees that at the end of the trial, the overall type I error will be the pre-specified value of .

• This approach is a generalization of the previous group sequential methods so that Pocock and O’brien Fleming monitoring procedures become special cases.

Page 34: Interim Analysis of Clinical Trial Liying XU CCTER, CUHK

Calendar time and information fraction

• At any particular calendar time t in the study, a certain fraction t* of the total information is observed such as:

expectednumber totalThe

pointat that randomized tsparticipan The/ Nn

expectednumber totalthe

studies) survivalin ( observed events ofnumber the/ Dd

Page 35: Interim Analysis of Clinical Trial Liying XU CCTER, CUHK

• For regression slops: the information fraction is more generally defined in terms of the ratio of the inverse of the variance of the test statistic at the particular interim analysis and the final analysis.

• The alpha spending function (t*), determines how the pre-specified is allocated at each interim analysis as a function of the information fraction.

Calendar time and information fraction

Page 36: Interim Analysis of Clinical Trial Liying XU CCTER, CUHK

Advantages

• Neither the number nor the time of the interim analyses need to be specified in advance.

• Once the spending function is selected, the information fractions t*1, t*2…. Determine the critical or boundary values exactly.

• The frequency of the interim analyses can be changed during the trial and still preserve the pre-specified level.

Page 37: Interim Analysis of Clinical Trial Liying XU CCTER, CUHK

A Example of group sequential methods in the Beta-Blocker Heart Attack Trial

• Specifications of the group sequential boundary:

The O’ Brien-Fleming group sequential procedure.

Seven meetings scheduled to review interim data.

The trial was designed for a two-sided 5%

significant level.

Page 38: Interim Analysis of Clinical Trial Liying XU CCTER, CUHK

Fig.1 Six interim log rank statistics plotted for the time of data monitoring committee meeting with a two –sided O’Brien-Fleming significance level boundary in the Beta-Blocker Heart Attack Trial. Dashed line represents Z=1.96.

Page 39: Interim Analysis of Clinical Trial Liying XU CCTER, CUHK

• From the second analysis on, the conventional significant value of 1.96 was exceeded.

• the trial was continued. • at the six meeting, the O’Brien-Fleming boundary was

crossed • a decision was made to terminate the trial

• Crossing the O’Brien-Fleming boundary was only one of the factors in this decision!

Interim log rank tests in the Beta-Blocker Heart Attack Trial

Page 40: Interim Analysis of Clinical Trial Liying XU CCTER, CUHK

Figure 2 Cumulative mortality curves comparing propranolol and placebo in the Beta-Blocker Heart Attack Trial.

Page 41: Interim Analysis of Clinical Trial Liying XU CCTER, CUHK

What should be in a interim report

• Patients recruitment progress

• Data quality

• Baseline characteristics

• Patients compliance

• Primary and secondary outcomes

• Adverse events

• Other safety measures

Page 42: Interim Analysis of Clinical Trial Liying XU CCTER, CUHK

Who should have access to the report?

• DSMB members only

• Confidential!!

Page 43: Interim Analysis of Clinical Trial Liying XU CCTER, CUHK

Practical Issues of Interim Results

• Consequences Continuation Termination Modification Unblinding of the code (unmasking) Modification of Patient Information Sheet Notification of ethics committees (and /or FDA,

Human Rights Committee) Re-estimating sample size Timing and extent of unblinding of interim results

Page 44: Interim Analysis of Clinical Trial Liying XU CCTER, CUHK

A software: EaSt 2000

• Interactive Software and Consulting Services for the Design and Interim Monitoring of Group-Sequential Clinical Trials

• CYTEL Software Corporation

• E-mail:[email protected]

Page 45: Interim Analysis of Clinical Trial Liying XU CCTER, CUHK

A book

• Group Sequential Methods with Applications to Clinical Trials

By Christopher Jennison and Bruce W. Turnbull. 2000

CHAPMAN & HALL/CRC