1 / 37

Interim Monitoring in Randomized Trials

Interim Monitoring in Randomized Trials. Which trials require monitoring? Why alter/stop a clinical trial early? Who should decide? What should be monitored? How often should you monitor? What statistical methods to use? How can a trial be altered? Fascinating examples.

Télécharger la présentation

Interim Monitoring in Randomized Trials

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Interim Monitoring inRandomized Trials • Which trials require monitoring? • Why alter/stop a clinical trial early? • Who should decide? • What should be monitored? • How often should you monitor? • What statistical methods to use? • How can a trial be altered? • Fascinating examples...

  2. Which Trials to Monitor • Trials in which there is a possibility • of unexpected harm from the intervention • of unexpected benefit from the intervention • that the question will not be answered • Increasingly viewed by the NIH as an independent advisory board • May not be feasible in small, short trials

  3. Why Stop a Trial Early? • Harm clearly demonstrated • Benefit clearly demonstrated • Not possible to demonstrate benefit • trial design flawed • low enrollment, high noncompliance, poor data, high drop-out • no difference between groups • Research question answered by another study

  4. Who Should Decide? • Sponsor • Investigators • Independent monitoring board, without conflict of interest • experts -? investigators • ethicists - ? representative of Sponsor • statisticians -? representative of NIH - ? lay persons

  5. What Should You Monitor? • Issues early in the trial • trial design • recruitment • compliance • loss to follow-up • data quality and timeliness • information from other studies

  6. What Should You Monitor? • Issues later in the trial • primary and secondary outcomes • clear benefit • clear harm • no conditional power • adverse events • side effects • subgroups

  7. How Often to Monitor? • Initial meeting several months before start-up • review protocol, classification of outcomes and adverse events, adjudication, etc • review and accept DSMB guidelines, statistical methods and meeting schedule • Subsequent meetings • often enough to achieve goals • not so often that there is no new data • depends on the duration of the trial and perceived risk of the intervention

  8. Confidentiality • DSMB usually does not share interim results with sponsors, investigators, ppts • Open session • DSMB, sponsor, investigators, NIH, FDA • recruitment, retention, data quality • overall findings • Closed session • DSMB members, + statistician • between group findings • discussion regarding modifications

  9. Statistical Methods for Interim Monitoring • Perform tests of significance and • stop the trial if any p<.05 • Simple, but wrong total testsoverall alpha 1 .05 2 .08 5 .14 10 .20 20 .35

  10. Interim Analyses in the CDP +2 +1 0 -1 -2 Z Value 10 20 30 40 50 60 70 80 90 100 Month of Follow-up

  11. Statistical Methods • Perform tests of significance and adjust the test-wise alpha • Bonferroni • Classical sequential methods • Group sequential methods • Pocock • Haybittle and Peto • O’Brien and Flemming • Lan and DeMets

  12. Group Sequential Methods • O’Brien - small i for early tests Flemming gradually increasing i N=5 interim tests;  = .05 initial 1=.00001; f=.041 • Lan-  spending function DeMets defined by N previous “looks” proportion of data/time between

  13. Alpha Spending LookZ-valueP-value 1 4.56 .00001 2 3.23 .0001 3 2.63 .009 4 2.28 .023 5 2.04 .041

  14. Symmetric Stopping Boundaries 6 4.56 3.23 2.63 4 2.28 2 2.04 Z Stop for Harm 0 Stop for Benefit 1st Look 2nd Look 3rd Look 4th Look 5th Look -2 -2.04 -2.28 -4 -2.63 -3.23 -4.56 -6

  15. Conditional Power Compute p(reject Ho given data so far) Deterministic Curtailed Sampling • assume all future outcomes in treated • assume all future outcomes in placebo Stochastic Curtailed Sampling • assume Ho true • assume Ha true

  16. How Can a Trial Be Altered? • Early alterations • change entry criteria • increase sample size • change outcome • adjust dose • Goals of these changes • make the trial successful (definitive result) • change original protocol as little as possible • Timely, high quality data crucial

  17. How Can a Trial Be Altered? • Later alterations • increase duration of the trial • modify the trial protocol • stop one arm of the intervention • terminate high risk groups • add safety measures • others • terminate the trial early

  18. Interim Monitoring • NOT simply a statistical issue • Must weigh: • Possible baseline differences in groups • Possible bias in assessment of outcome • Impact of missing data • Differential co-intervention or noncompliance • Internal consistency of findings • Impact of early termination on medical practice and public health

  19. Nuts and Bolts • Board chosen early • Data Monitoring Plan • board members • variables and analyses • frequency of monitoring • statistical methods • guidelines for decisions • Timely, accurate and complete data

  20. Coronary Arrhythmia Suppression Trial • 1727 of planned 4400 subjectsafter MI with ventricular ectopy • Flecainide, encainide or moricizine vs. pbo • Mean follow-up 1 year of planned 5 years • Outcomes - mortality from arrhythmia, total mortality

  21. Coronary Arrhythmia Suppression Trial OutcomeF/EPlacebop N randomized 730 725 Arrhythmic death 33 9 .0006 Total death 56 22 .0003

  22. Beta-blocker Heart Attack Trial • Subjects - 3,837 persons 5-21 days after MI • Intervention - propranolol 180-240mg/day vs placebo • Follow-up - 40 of planned 48 months • Outcome - mortality

  23. Beta-blocker Heart Attack Trial AnalysesMonth Deaths Z Critical Value 1 11 56 1.68 5.88 2 16 77 2.24 5.04 3 21 126 2.37 3.79 4 28 177 2.30 3.19 5 34 247 2.34 2.64 6 40 318 2.82 2.30 7 48

  24. Coronary Drug Project • Subjects - 8,341 men post-MI • Interventions - estrogen 2.5 and 5.0 mg QD dextrothyroxine 6 mg QD clofibrate 1.8 gm QD niacin 3.0 gm QD placebo • Follow-up - 1.5 to 2.5 of planned 5 years • Outcomes - death, MI, cancer, VTE

  25. Coronary Drug Project CEE 5 mgPlaceboRR ( n=1,119)(n=2,789) CHD event 11.0% 7.5% 1.5 PE or DVT 3.5% 1.5% 2.3* Total mortality 9.7% 8.2% 1.2 JAMA, 1970

  26. Coronary Drug Project • Low-dose CEE (2.5 mg/d) stopped after 2.5 years for similar findings • D-thyroxin found to increase death rate at 3 years in men with abnormal baseline EKG • D-thyroxin stopped in all men with abnormal baseline EKG

  27. HERS DSMB ReportMonitoring for VTEs . . . 6 . . . . . 4 . Stop for Harm Stop for Benefit 2 . Z 0 6 mo 1.5 yr 2.5yr 3.5 yr End -2 -4 -6

  28. HERS DSMB ReportMonitoring for CHD Death 6 . . . 4 . Stop for Harm . Stop for Benefit . 2 . . . . Z 0 6 mo 1.5 yr 2.5yr 3.5 yr End -2 -4 -6

  29. WHI - Overall Risk and Benefit Harm% Change in Risk • CHD +29% • Stroke +41% • Breast cancer +26% • Pulm. embolus +2.1X Global Index Benefit • Hip fracture -34% • Colorectal cancer -37%

  30. Letrozole for Breast Cancer • In early stage ER+ breast cancer • 5 years of tamoxifen reduces recurrence • longer treatment LESS effective • tamoxifen may have estrogenic activity • Treatment with letrozole after tamoxifen might reduce recurrence

  31. Letrozole after Tamoxifen Trial • Subjects - postmenopausal women treated with 4 to 6 years of adjuvant tamoxifen for breast cancer • Intervention - letrozole 2.5 mg QD • Follow-up - planned 5 years • Outcomes - disease free survival overall survival quality of life safety Goss, NEJM, 2003

  32. DSMB Guidelines • Expect 515 events (recurrent cancer) • Review safety twice yearly • Interim analysis twice, after • 171 (1/3 expected total) events • 342 (2/3 expected total) events • Lan and DeMets -spending function with O’Brien-Flemming boundaries

  33. First Interim Analysis • 5187 women enrolled (planned 4800) • Mean 2.4 yrs of follow-up (planned 5) • 207 events (40% of expected) LetrozolePlacebo RH P-value N events 75 132 .57 .00008 4-year DFS 93% 87% .001 Deaths 31 42 4-year S 96% 94% .76

  34. First Interim Analysis LetrozolePlaceboP-value Flushes 47% 40% <.001 Arthritis 6% 3% <.001 Arthralgia 21% 17% <.001 Osteoporosis 5.8% 4.5% 0.07 Fracture 3.6% 2.9% 0.24 CVD events 4.1% 3.6% 0.40

  35. Issues • How long does benefit last? • Does benefit strengthen over time? • Does treatment reduce mortality? • Do findings support 5 years of treatment with letrozole? • Were side effects underestimated? • How will other trials be affected?

  36. Stopping Trials • Stopping for harm often not adequately planned in advance • harm goes on too long • stopping boundaries unreasonable • Stopping for benefit raises issues because less data obtained on: • long-term benefit • side effects

  37. Summary - Interim Monitoring • Interim monitoring very important • Should be planned in advance • Should be performed well • Any change in trial protocol should be carefully considered, weighing many issues

More Related