RCT Class #8 Flashcards
most basic type of randomization and define it
simple randomization
assign patients to one of two groups, individuals are chosen at random and has the same probality of being chosen but they are not randomly assigned to an intervention.
What does it mean to say individuals are chosen at random but are not randomly assigned to an intervention.
They are chosen from the population to be a candiate for the study so therefore not randomly chosen for an intervention
website he likes for rct
randomization.com
What is blocked randomization?
randomize with short blocks, to account for changing baseline characteristics over time
typical block size
4, 6, or 8
PRos of block randomization
- avoids imbalances in allocation
- if trial stops early, balance is nealy s\assured
- Protects against participatnchanges over time
what is a simple random sample
for a population selected each individual has the same probability of being chosen.
What type of randomization do we need for our prop
stratified randomization with randomly permuted blocks (blocks size are random)
How to avoid problems with randomization
after first round have someone validate it.
What is the problem with simple randomization
patients characteristics change over time, this may casuse imbalances and serious bias.
Blocks are set ratio’s
yeah, that way we protect against things going out of control with balance
formula for differing amount with blocking
for block size b, N per group never differes by >b/2
blocked randomization problem
gives incorrect variabces and tends to bias in a cosdsrvative direction increasing power
blockings frequecny of use
ALWAYS USED IN RCT OF ANY APPRECIABLE SIZE
Randomly permuted blockes why?
With a constant block size you can guess the allocation of the last patient in the block and create bias.
Solution to constant block problem of being able to anticipate the last assignment(s)
randomly permuted blocks. basically an additonal layer of randomization
Should proportions of block sizes be equal?
No.. These are the standards for large trials they must be used in our props
How are we able to evaulate if the randomization scheme is working?
It should be by design. Keep it as simple and as possilbe. Standard operating procedures that are tested and subsequently validated. Monitoring and Quality control at regular intervals.
What is the seed
A randomly gnerated number to serve as a starting point in a table of randomly generated treatment assisnments
always need to stratify by what one variable
Site!!
stratified randomization within the site
do you stratify before or after randomization in RCT
before. it is done to ensure there are equal numbers of a group. Male and female Randomization scheme.
overall message on stratificaiton other than site
we usually dont need to stratify by other factors. bc rct supposed to have balance of measured and unmeasured confounders
for every strata you multiply by
each other for total number of strata.
give number of strata
site 20 strata
gender 2 strata
other 2 strata
20 x 2 x2 =80 randomization schemes .
Shit gets complicated so fast man.
problems with overstratification
gets complex and can empty cells.
Standard procedure for randomization of multicenter rCTS
Blocked and stratified.
blocking before or after stratificaiton
after
Allcoation: why is 1:1 optimal
bc it is c/w equipoise of rct, and maximizes power of each enrolled pt.
unequal alloaction may be when
if one rx is very cheap and one is very expensive
4 parts to consort diagram need to include
- N screened
- N not received allocation
- N LTF + why
- N excluded + why (none in ITT)
