“In response to a comment by ‘Stats is my name‘, (Friday, July 29, 2016 9:36:00 pm): ‘But you can’t compare shedders at 1 month vs shedders at 3 months. You can only compare like with like so you should use the numbers by time period’. This got me thinking about the study design. Even though EBV salivary shedding is intermittent there is no reason why a study design based on a survival analysis, i.e time to an event or in this case time to EBV shedding, won’t work.”
|An example of a survival curve from Wikipedia.|
The following is a correction from when the post went live:
“Survival analysis is a branch of statistics for analysing the expected duration of time until one or more events happen, such as death in biological organisms. It is also called reliability theory or reliability analysis in engineering, duration analysis or duration modelling in economics, and event history analysis in sociology. Survival analysis attempts to answer questions such as: what is the proportion of a population which will survive past a certain time? It is necessary to define ‘lifetime’. In the case of our study survival can be defined as being free of EBV shedding. I therefore propose randomising pwMS to placebo, or FamV, and then do monthly saliva samples and follow them up until we have enough events (shedding events). In the placebo arm we would expect 45% of subjects to shed EBV by 6 months and we predict that this figure will be reduced by 80%, i.e. to 9.2%, in the group of subjects being treated with FamV. With a power of 80% and an alpha of 5% we will need 44 subjects (22 per arm) to have enough events. This study will be followed until for a variable period after the last subject has been randomised.”
“In comparison the proposed power calculations using a proportional analysis: based on the conservative assumption that the antiviral drug will only be 80% effective in suppressing EBV reactivation over 6 months with monthly salivary sampling, a power of 80% and an alpha of 5%, we would need at least 36 subjects or 18 per arm. With a 10% dropout rate this design would need 40 subjects.”
“I am now going to armed with the above information and speak to our statistician to find-out which is the best study design. Once again thank you for your advice and support. It is truly very inspiring to have the community help by suggesting ideas about statistics. In this way the Charcot 2 study is going truly be a study designed and funded by the community.”
5 thoughts on “CrowdSpeak: using survival analysis makes Charcot 2 cheaper”
glad i could teach you professor!what is the 80% effect based on? why not 90?
With 90% power the trial size would be much bigger than 80% power but 80% power is come starting ground for phase II for academic studies
no md. effect size, not power
Professor! What if effect size is 60%http://www.ncbi.nlm.nih.gov/m/pubmed/17538513/
Maybe a silly question: if the person doesn't shed EBV in their saliva, it would be because the study research stage these people were EBV in a latent phase, or in fact they have never been infected? And this sample of almost 50% in EBV shedding investigation of the research is because the virus was in lytic stage? I wonder about this because if EBV is one of the most conragiosos existing viruses, have children who have contact with him, are infected with EBV before the maternal antibodies disappear, then how the other 50% of the sample um the research had not that moment?