Randomisation

Question 1

essence of randomization

Answer 1

based purely on chance

Question 2

2 elements

Answer 2

1. creating a randomization scheme (before the start of a study)
2. allocating patients to intervention or control group (once patients start to participate in the study)

Question 3

simple randomization

Answer 3

like throwing a dice
issues:
- uneven groups/ imbalance (not 50/50)

Question 4

how to solve imbalance with simple randomization

Answer 4

use block randomization:

essence: after every block, balance is 50/50
- you can choose the size of the block (after e.g. 6/8 etc. the groups are even)

Question 5

problem with block size of 2

Answer 5

predictable, after the first randomization outcome, you know what the second one will be

Question 6

problems with large blocks

Answer 6

if you stop recruiting earlier than expected, you have uneven groups again (e.g., block of 200 –> 100 in intervention, 50 in control)

Question 7

best block size

Answer 7

vary block size randomly, e.g., between 2 and 4 (then even when you’re in a block size of 2 atm, you don’t know that, so you don’t know the outcome of the next person)

Question 8

creating a randomization scheme

Answer 8

• use block randomization

- with randomly varying block size

Question 9

problem of unequal distribution of variables

Answer 9

-problematic if they are prognostic factors (e.g., if gender is associated with the effect)

Question 10

when stratification?

Answer 10

when you expect a variable to be strongly related to effect (i.e. prognostic factor) –> to ensure equal distribution of this variable

• not really necessary in large trials, because you expect variables to be distributed equally (known and unknown prognostic factors)
• use sparsely, mainly with strong prognostic factors

Question 11

how stratification? procedure

Answer 11

1. determine the variables you want to use
2. determine the number of groups (stratum/ strata) for each variable
   - not continuous! groups! e.g. age: 18-40, 41-60, 60+ (3 strata)
3. calculate total number go groups/ strata, e.g. 2 variables with 2 strata –> 4 total number of strata
4. create a randomization scheme for each of the groups/ strata: e.g. 4 randomization schemes

Question 12

problem with stratification

Answer 12

you quickly have too many strata: 4 variables with 2 groups results into 16 strata!

• -> too complicated
• -> you need a huge sample

Question 13

minimisation

Answer 13

another way of randomization
- the chance on intervention/ control depends on characteristics of those already assigned, e.g. all women are assigned to intervention group –> chance of next women to draw control group is increased (best used when you have more factors to randomize on)

Question 14

allocating patients: what is that?

Answer 14

• after eligibility screen –> allocation
  (patients enter RCT gradually!)
• allocation needs to be blinded! might create bias, if person screening for eligibility knows the next allocation outcome

Question 15

allocating patients steps

Answer 15

1. independent person stores randomization scheme
2. researcher includes patient
3. researcher asks independent person to reveal randomization outcome (=allocates the next patient)

Question 16

randomization and group therapies

Answer 16

• patients have to wait until they are enough people in the group to start with the treatment (not a problem with quick inclusion rate)
  solutions:
• start with people who are not in your study
• collaborate with different centers
• smaller group sizes
  –> measure people again before they start treatment, so you can see what happened while they waited