Monitoring of Trial Results Flashcards
What is the point of interim analysis of data?
To allow decisions to be made as to whether the trial should be stopped or modified in some way
Why do many trials require some form of monitoring to be ethically acceptable?
To detect treatment differences that require action
To safeguard the interests of participants already in the trial
To safeguard future participants who may be enrolled or have the disease under scrutiny
Adherence solely to individual ethics (judgment based on what is best for the next patient to be enrolled in the trial) means that decisions about treatment may interfere in what tenants of trial design?
Randomisation
Blinding and placebo
Large-scale and unbiased comparison of treatments
The need for a Data Monitoring Committee should be based off what factors?
Size of a trial Population of trial participants (vulnerable or not) How serious the disease is under study Trial endpoints Type of intervention being studied (i.e. novel or high risk) Length of study Impact on public health Trial design Commercial interest
What is the DCM’s most important roles?
Evaluate accumulating trial data and focussing on safety and efficacy
Make recommendations to trial leadership regarding continuation, modification or termination of a trial
Review trial conduct for ethical, safety and integrity
What is a statistical stopping rule?
Pre-set conditions for the statistical comparisons which, if met, trigger consideration for stopping the trial early
Outline the fixed nominal statistical stopping rule
alpha (interim) is the same for all analyses and which will maintain the overall type 1 error rate. The calculations are difficult and not necessary to understand. Pocock et all 1977
What is the main flaw of the fixed nominal rule?
It makes it too easy to stop a trial early when numbers are smaller and the evidence is not convincing enough. Even if at the final analysis p-value <0.05 this would be rejected. Making it more difficult to detect true treatment differences and reducing the power of the study
Outline the O’Brien and Fleming rule of statistical stopping
Similar to fixed nominal rule in that it has a fixed number of analyses where the overall type 1 error rate maintains agreed level
How does the O’Brien and Fleming rule differ from the fixed nominal rule?
O/F has a fixed number of analyses but crucially starts with low p-values which get progressively larger and still add up to overall pre-determined type 1 error rate (e.g. 0.05)
What is a flaw of the O’Brien and Fleming rule of statistical stopping?
The last interim analysis the significance level to stop becomes too lenient (more lenient than even the fixed nominal rule)
Outline the Peto-Haybittle rule of statistical stopping
The rule states that in order to consider stopping the trial early, the p-value needs to be very small (e.g. <0.001) at interim analysis.
What is the difference between Peto-Haybittle rule and fixed nominal/O’Brien and Fleming rules?
The number of interim analyses do not need to be pre-determined and additional interim analyses have virtually no effect on overall type 1 error rate
What is the Lan and DeMets alpha-spending function
A spending function of alpha is devised and p-values can be calculated based of past and planned analyses.
What are the benefits of the Lan and DeMets alpha-spending function?
Allows different strategies for weighting analyses at any point of the trial. And can emulate other stopping rules such as fixed nominal / O’Brien and Fleming
