EUGM 2011 | DAY | Experiences with interim analyses and dm cs
0 likes211 views
This document discusses experiences with interim analyses and data monitoring committees in clinical trials. It describes several examples of trials that conducted interim analyses where decisions about continuing or stopping the trial early were considered. The key lessons are that plans and decisions based on interim data need to be cautious, as interim results may not be statistically precise enough. Pre-specifying analysis plans and allowing data monitoring committees to make independent recommendations can help avoid suboptimal decisions based on interim trial data.
![9
Interim
analysis,
N=722
• Mortality
rates:
placebo
38.9%
Afacogin
29.1%
P
=
0.006
• Tifacogin
significantly
a:enuated
prothrombin
fragment
and
thrombin:anA‐thrombin
complex
levels
[secondary
endpoints]
in
paAents
with
high
and
low
INR
[pre‐specified
subgroups]](https://image.slidesharecdn.com/eugm2011-day-experienceswithinterimanalysesanddmcs-150701043223-lva1-app6891/85/EUGM-2011-DAY-Experiences-with-interim-analyses-and-dm-cs-9-320.jpg)
![24
Interim
analysis,
N=722
• Mortality
rates:
placebo
38.9%
Afacogin
29.1%
P
=
0.006
• Tifacogin
significantly
a:enuated
prothrombin
fragment
and
thrombin:anA‐thrombin
complex
levels
[secondary
endpoints]
in
paAents
with
high
and
low
INR
[pre‐specified
subgroups]](https://image.slidesharecdn.com/eugm2011-day-experienceswithinterimanalysesanddmcs-150701043223-lva1-app6891/85/EUGM-2011-DAY-Experiences-with-interim-analyses-and-dm-cs-24-320.jpg)
EUGM 2011 | DAY | Experiences with interim analyses and dm cs
- 2. Experiences with Interim Analyses and Data Monitoring Commi:ees Or: I d never be so stupid as to do that, would I?
- 3. 3 plans or decisions based on staAsAcally imprecise interim data may oKen be subopAmal Section 6.5 Sponsor access to interim data for planning purposes
- 5. 5 Design and objecAves • Double blind, placebo‐controlled, mulAcentre, phase 3 randomised controlled clinical trial • Primary outcome – all cause 28 day mortality • Target sample size 1550 (Increased by 400 at an interim analysis)
- 8. 8 Interim analysis, N=722 • Mortality rates: placebo 38.9% Afacogin 29.1% P = 0.006
- 9. 9 Interim analysis, N=722 • Mortality rates: placebo 38.9% Afacogin 29.1% P = 0.006 • Tifacogin significantly a:enuated prothrombin fragment and thrombin:anA‐thrombin complex levels [secondary endpoints] in paAents with high and low INR [pre‐specified subgroups]
- 10. 10 What should happen? • Pre‐specified interim analysis at ¼, ½ and ¾ of paAents compleAng 28 days • Purpose: safety, fuAlity and efficacy • Safety and efficacy – kind of the same thing • Tifacogin is effecAve – more paAents are dying on placebo
- 11. 11 What should happen? • If this trial con.nues, pa.ents randomised to the control arm will needlessly die • Another 1100 paAents due to be randomised • We should order 160 coffins for paAents who will be given .facogin • We should order 215 coffins for paAents who will be given placebo
- 12. 12 Phew, I got the real stuff!
- 14. 14 The trial • 5 courses of therapy versus 4 courses • Primary endpoint – survival • 1078 paAents • Yearly interim analyses
- 16. 16 Interim analysis (3) • Hazard raAo 0.47 (in favour of 5 courses) • 95% confidence interval 0.29 – 0.77, P=0.003 • Hazard raAo 0.55 (in favour of 5 courses) • 95% confidence interval 0.38 – 0.80, P=0.002
- 17. 17 That ll be five courses for me, please, waiter
- 20. 20 New (changed) endpoint • All‐cause mortality • At an interim analysis, changed to: – All‐cause mortality (P<0.005) Or – All‐cause mortality or cardiovascular hospitalisaAon (P<0.045)
- 21. 21 Results Relative risk 0.77 (0.60 – 0.98) P=0.031 Relative risk 0.92 (0.80 – 1.07) P=0.296 All-cause mortality All-cause mortality or Cardiovascular hospitalisation
- 22. 22 Results
- 24. 24 Interim analysis, N=722 • Mortality rates: placebo 38.9% Afacogin 29.1% P = 0.006 • Tifacogin significantly a:enuated prothrombin fragment and thrombin:anA‐thrombin complex levels [secondary endpoints] in paAents with high and low INR [pre‐specified subgroups]
- 25. 25 What did happen? • They kept the trial going (Yes, and people died!) • How many died? High INR Low INR placebo 296 (33.9%) 118 (23%) Afacogin 301 (34.2%) 83 (12%) P = 0.?
- 26. 26 What did happen? • They kept the trial going (Yes, and people died!) • How many died? High INR Low INR placebo 296 (33.9%) 118 (23%) Afacogin 301 (34.2%) 83 (12%) P = 0.051 Pearson χ2
- 27. 27 What did happen? • They kept the trial going (Yes, and people died!) • How many died? High INR Low INR placebo 296 (33.9%) 118 (23%) Afacogin 301 (34.2%) 83 (12%) P = 0.03 logisAc regression
- 28. 28 Conclusion plans or decisions based on statistically imprecise interim data may often be suboptimal
- 30. 30 Interim analysis (3) • Hazard raAo 0.47 (in favour of 5 courses) • 95% confidence interval 0.29 – 0.77, P=0.003 • Hazard raAo 0.55 (in favour of 5 courses) • 95% confidence interval 0.38 – 0.80, P=0.002
- 32. 32 The sequence of events plans or decisions based on statistically imprecise interim data may often be suboptimal
- 33. 33 So the quesAons to ask yourselves 1. That trial we stopped early… what would have happened if we had kept going? 2. That fixed sample size trial that we did… what would have happened if it recruited more paAents?