Chapter 20 Randomization for Clinical Trials with R
There are a number of packages for doing key functions for clinical trials in R. You can find many of these on the CRAN Task View for Clinical Trials, at https://cran.r-project.org/web/views/ClinicalTrials.html.
This is a curated list of packages that anyone might find useful in designing, monitoring, or analyzing clinical trials, and is often a good place to start in looking for packages that might be relevant for clinical trials.
If you use Ctrl-F to search the web page for “rand”, several packages address randomization, including
blockrand
randomizeR
pwr
experiment
clusterPower
CRTSize
cosa
PowerupR
Several of these are specifically for more complex designs, including cluster and multilevel randomization (clusterPower, cosa, CRTSize). For today, we will focus on the straightforward randomization packages including {blockrand} and {randomizer}. The {blockrand} package creates randomizations for clinical trials with can include stratified enrollment and permuted block randomization, and can produce a PDF file of randomization cards.
Let’s start with an example in {blockrand}.
Details on the package can be found at https://cran.r-project.org/web/packages/blockrand/blockrand.pdf
or by running help(blockrand) in your Console.
You want to randomize 180 inpatients with severe ulcerative colitis to one of 3 arms: corticosteroids alone (control), corticosteroids + tofacitinib, or corticosteroids + upadacitinib. You want to stratify the participants by (1) prior biologic failure and (2) Albumin level above or below 3.0. To be prepared for dropouts and imbalanced enrollment, you want to have a randomization list with at least 60 assignments available for each arm and stratum. To avoid a recognizable pattern in the randomization, you want to have a permuted block design with blocks of sizes 3, 6, and 9.
Below, you will see how to do this for the biologic failure - low albumin stratum.
bfla <- blockrand(n = 60,
num.levels = 3, # three treatments
levels = c("CS", "CS/Tofa", "CS/Upa"), # arm names
stratum = "Bfail.LowAlb", # stratum name
id.prefix = "BfLA", # stratum abbrev
block.sizes = c(1,2,3), # times arms = 3,6,9
block.prefix = "BfLA") # stratum abbrev
bfla## id stratum block.id block.size treatment
## 1 BfLA01 Bfail.LowAlb BfLA1 6 CS/Upa
## 2 BfLA02 Bfail.LowAlb BfLA1 6 CS
## 3 BfLA03 Bfail.LowAlb BfLA1 6 CS/Upa
## 4 BfLA04 Bfail.LowAlb BfLA1 6 CS
## 5 BfLA05 Bfail.LowAlb BfLA1 6 CS/Tofa
## 6 BfLA06 Bfail.LowAlb BfLA1 6 CS/Tofa
## 7 BfLA07 Bfail.LowAlb BfLA2 9 CS
## 8 BfLA08 Bfail.LowAlb BfLA2 9 CS/Upa
## 9 BfLA09 Bfail.LowAlb BfLA2 9 CS
## 10 BfLA10 Bfail.LowAlb BfLA2 9 CS/Upa
## 11 BfLA11 Bfail.LowAlb BfLA2 9 CS/Upa
## 12 BfLA12 Bfail.LowAlb BfLA2 9 CS/Tofa
## 13 BfLA13 Bfail.LowAlb BfLA2 9 CS/Tofa
## 14 BfLA14 Bfail.LowAlb BfLA2 9 CS
## 15 BfLA15 Bfail.LowAlb BfLA2 9 CS/Tofa
## 16 BfLA16 Bfail.LowAlb BfLA3 6 CS/Tofa
## 17 BfLA17 Bfail.LowAlb BfLA3 6 CS/Upa
## 18 BfLA18 Bfail.LowAlb BfLA3 6 CS
## 19 BfLA19 Bfail.LowAlb BfLA3 6 CS/Upa
## 20 BfLA20 Bfail.LowAlb BfLA3 6 CS
## 21 BfLA21 Bfail.LowAlb BfLA3 6 CS/Tofa
## 22 BfLA22 Bfail.LowAlb BfLA4 6 CS/Tofa
## 23 BfLA23 Bfail.LowAlb BfLA4 6 CS/Upa
## 24 BfLA24 Bfail.LowAlb BfLA4 6 CS
## 25 BfLA25 Bfail.LowAlb BfLA4 6 CS/Tofa
## 26 BfLA26 Bfail.LowAlb BfLA4 6 CS
## 27 BfLA27 Bfail.LowAlb BfLA4 6 CS/Upa
## 28 BfLA28 Bfail.LowAlb BfLA5 6 CS
## 29 BfLA29 Bfail.LowAlb BfLA5 6 CS/Tofa
## 30 BfLA30 Bfail.LowAlb BfLA5 6 CS/Upa
## 31 BfLA31 Bfail.LowAlb BfLA5 6 CS/Upa
## 32 BfLA32 Bfail.LowAlb BfLA5 6 CS/Tofa
## 33 BfLA33 Bfail.LowAlb BfLA5 6 CS
## 34 BfLA34 Bfail.LowAlb BfLA6 6 CS/Upa
## 35 BfLA35 Bfail.LowAlb BfLA6 6 CS/Tofa
## 36 BfLA36 Bfail.LowAlb BfLA6 6 CS
## 37 BfLA37 Bfail.LowAlb BfLA6 6 CS/Upa
## 38 BfLA38 Bfail.LowAlb BfLA6 6 CS
## 39 BfLA39 Bfail.LowAlb BfLA6 6 CS/Tofa
## 40 BfLA40 Bfail.LowAlb BfLA7 9 CS
## 41 BfLA41 Bfail.LowAlb BfLA7 9 CS/Upa
## 42 BfLA42 Bfail.LowAlb BfLA7 9 CS/Upa
## 43 BfLA43 Bfail.LowAlb BfLA7 9 CS/Tofa
## 44 BfLA44 Bfail.LowAlb BfLA7 9 CS
## 45 BfLA45 Bfail.LowAlb BfLA7 9 CS/Tofa
## 46 BfLA46 Bfail.LowAlb BfLA7 9 CS
## 47 BfLA47 Bfail.LowAlb BfLA7 9 CS/Upa
## 48 BfLA48 Bfail.LowAlb BfLA7 9 CS/Tofa
## 49 BfLA49 Bfail.LowAlb BfLA8 9 CS/Tofa
## 50 BfLA50 Bfail.LowAlb BfLA8 9 CS
## 51 BfLA51 Bfail.LowAlb BfLA8 9 CS/Upa
## 52 BfLA52 Bfail.LowAlb BfLA8 9 CS/Tofa
## 53 BfLA53 Bfail.LowAlb BfLA8 9 CS/Tofa
## 54 BfLA54 Bfail.LowAlb BfLA8 9 CS
## 55 BfLA55 Bfail.LowAlb BfLA8 9 CS/Upa
## 56 BfLA56 Bfail.LowAlb BfLA8 9 CS/Upa
## 57 BfLA57 Bfail.LowAlb BfLA8 9 CS
## 58 BfLA58 Bfail.LowAlb BfLA9 3 CS/Upa
## 59 BfLA59 Bfail.LowAlb BfLA9 3 CS
## 60 BfLA60 Bfail.LowAlb BfLA9 3 CS/TofaYou can see the id for each participant, their stratum, the block.id for their permuted block, the block.size, and their assigned treatment. You can imagine this as a randomization list, or as assignments that you could print out on cards and seal in security envelopes for the time of randomization. Of course, this is only one of our four strata. We should do the same for the 3 other strata.
bfha <- blockrand(n = 60,
num.levels = 3, # three treatments
levels = c("CS", "CS/Tofa", "CS/Upa"), # arm names
stratum = "Bfail.HiAlb", # stratum name
id.prefix = "BfHA", # stratum abbrev
block.sizes = c(1,2,3), # times arms = 3,6,9
block.prefix = "BfHA") # stratum abbrev
bfha## id stratum block.id block.size treatment
## 1 BfHA01 Bfail.HiAlb BfHA01 3 CS
## 2 BfHA02 Bfail.HiAlb BfHA01 3 CS/Upa
## 3 BfHA03 Bfail.HiAlb BfHA01 3 CS/Tofa
## 4 BfHA04 Bfail.HiAlb BfHA02 6 CS/Upa
## 5 BfHA05 Bfail.HiAlb BfHA02 6 CS
## 6 BfHA06 Bfail.HiAlb BfHA02 6 CS/Upa
## 7 BfHA07 Bfail.HiAlb BfHA02 6 CS/Tofa
## 8 BfHA08 Bfail.HiAlb BfHA02 6 CS
## 9 BfHA09 Bfail.HiAlb BfHA02 6 CS/Tofa
## 10 BfHA10 Bfail.HiAlb BfHA03 6 CS
## 11 BfHA11 Bfail.HiAlb BfHA03 6 CS/Upa
## 12 BfHA12 Bfail.HiAlb BfHA03 6 CS/Tofa
## 13 BfHA13 Bfail.HiAlb BfHA03 6 CS/Upa
## 14 BfHA14 Bfail.HiAlb BfHA03 6 CS/Tofa
## 15 BfHA15 Bfail.HiAlb BfHA03 6 CS
## 16 BfHA16 Bfail.HiAlb BfHA04 9 CS/Upa
## 17 BfHA17 Bfail.HiAlb BfHA04 9 CS
## 18 BfHA18 Bfail.HiAlb BfHA04 9 CS
## 19 BfHA19 Bfail.HiAlb BfHA04 9 CS/Tofa
## 20 BfHA20 Bfail.HiAlb BfHA04 9 CS/Upa
## 21 BfHA21 Bfail.HiAlb BfHA04 9 CS/Tofa
## 22 BfHA22 Bfail.HiAlb BfHA04 9 CS
## 23 BfHA23 Bfail.HiAlb BfHA04 9 CS/Upa
## 24 BfHA24 Bfail.HiAlb BfHA04 9 CS/Tofa
## 25 BfHA25 Bfail.HiAlb BfHA05 9 CS/Tofa
## 26 BfHA26 Bfail.HiAlb BfHA05 9 CS/Tofa
## 27 BfHA27 Bfail.HiAlb BfHA05 9 CS/Upa
## 28 BfHA28 Bfail.HiAlb BfHA05 9 CS/Upa
## 29 BfHA29 Bfail.HiAlb BfHA05 9 CS/Upa
## 30 BfHA30 Bfail.HiAlb BfHA05 9 CS
## 31 BfHA31 Bfail.HiAlb BfHA05 9 CS/Tofa
## 32 BfHA32 Bfail.HiAlb BfHA05 9 CS
## 33 BfHA33 Bfail.HiAlb BfHA05 9 CS
## 34 BfHA34 Bfail.HiAlb BfHA06 6 CS
## 35 BfHA35 Bfail.HiAlb BfHA06 6 CS/Tofa
## 36 BfHA36 Bfail.HiAlb BfHA06 6 CS/Upa
## 37 BfHA37 Bfail.HiAlb BfHA06 6 CS/Upa
## 38 BfHA38 Bfail.HiAlb BfHA06 6 CS/Tofa
## 39 BfHA39 Bfail.HiAlb BfHA06 6 CS
## 40 BfHA40 Bfail.HiAlb BfHA07 6 CS/Upa
## 41 BfHA41 Bfail.HiAlb BfHA07 6 CS/Upa
## 42 BfHA42 Bfail.HiAlb BfHA07 6 CS/Tofa
## 43 BfHA43 Bfail.HiAlb BfHA07 6 CS
## 44 BfHA44 Bfail.HiAlb BfHA07 6 CS/Tofa
## 45 BfHA45 Bfail.HiAlb BfHA07 6 CS
## 46 BfHA46 Bfail.HiAlb BfHA08 6 CS/Tofa
## 47 BfHA47 Bfail.HiAlb BfHA08 6 CS
## 48 BfHA48 Bfail.HiAlb BfHA08 6 CS/Upa
## 49 BfHA49 Bfail.HiAlb BfHA08 6 CS/Upa
## 50 BfHA50 Bfail.HiAlb BfHA08 6 CS
## 51 BfHA51 Bfail.HiAlb BfHA08 6 CS/Tofa
## 52 BfHA52 Bfail.HiAlb BfHA09 6 CS/Upa
## 53 BfHA53 Bfail.HiAlb BfHA09 6 CS/Upa
## 54 BfHA54 Bfail.HiAlb BfHA09 6 CS/Tofa
## 55 BfHA55 Bfail.HiAlb BfHA09 6 CS
## 56 BfHA56 Bfail.HiAlb BfHA09 6 CS
## 57 BfHA57 Bfail.HiAlb BfHA09 6 CS/Tofa
## 58 BfHA58 Bfail.HiAlb BfHA10 3 CS
## 59 BfHA59 Bfail.HiAlb BfHA10 3 CS/Upa
## 60 BfHA60 Bfail.HiAlb BfHA10 3 CS/Tofabnha <- blockrand(n = 60,
num.levels = 3,
levels = c("CS", "CS/Tofa", "CS/Upa"),
stratum = "Bnaive.HiAlb",
id.prefix = "BnHA",
block.sizes = c(1,2,3, 4),
block.prefix = "BnHA")
bnha## id stratum block.id block.size treatment
## 1 BnHA01 Bnaive.HiAlb BnHA1 9 CS/Upa
## 2 BnHA02 Bnaive.HiAlb BnHA1 9 CS/Tofa
## 3 BnHA03 Bnaive.HiAlb BnHA1 9 CS/Tofa
## 4 BnHA04 Bnaive.HiAlb BnHA1 9 CS/Tofa
## 5 BnHA05 Bnaive.HiAlb BnHA1 9 CS/Upa
## 6 BnHA06 Bnaive.HiAlb BnHA1 9 CS/Upa
## 7 BnHA07 Bnaive.HiAlb BnHA1 9 CS
## 8 BnHA08 Bnaive.HiAlb BnHA1 9 CS
## 9 BnHA09 Bnaive.HiAlb BnHA1 9 CS
## 10 BnHA10 Bnaive.HiAlb BnHA2 12 CS
## 11 BnHA11 Bnaive.HiAlb BnHA2 12 CS/Upa
## 12 BnHA12 Bnaive.HiAlb BnHA2 12 CS/Tofa
## 13 BnHA13 Bnaive.HiAlb BnHA2 12 CS/Upa
## 14 BnHA14 Bnaive.HiAlb BnHA2 12 CS
## 15 BnHA15 Bnaive.HiAlb BnHA2 12 CS
## 16 BnHA16 Bnaive.HiAlb BnHA2 12 CS/Upa
## 17 BnHA17 Bnaive.HiAlb BnHA2 12 CS/Tofa
## 18 BnHA18 Bnaive.HiAlb BnHA2 12 CS
## 19 BnHA19 Bnaive.HiAlb BnHA2 12 CS/Tofa
## 20 BnHA20 Bnaive.HiAlb BnHA2 12 CS/Upa
## 21 BnHA21 Bnaive.HiAlb BnHA2 12 CS/Tofa
## 22 BnHA22 Bnaive.HiAlb BnHA3 9 CS/Tofa
## 23 BnHA23 Bnaive.HiAlb BnHA3 9 CS/Upa
## 24 BnHA24 Bnaive.HiAlb BnHA3 9 CS/Tofa
## 25 BnHA25 Bnaive.HiAlb BnHA3 9 CS
## 26 BnHA26 Bnaive.HiAlb BnHA3 9 CS/Tofa
## 27 BnHA27 Bnaive.HiAlb BnHA3 9 CS
## 28 BnHA28 Bnaive.HiAlb BnHA3 9 CS/Upa
## 29 BnHA29 Bnaive.HiAlb BnHA3 9 CS/Upa
## 30 BnHA30 Bnaive.HiAlb BnHA3 9 CS
## 31 BnHA31 Bnaive.HiAlb BnHA4 12 CS
## 32 BnHA32 Bnaive.HiAlb BnHA4 12 CS/Tofa
## 33 BnHA33 Bnaive.HiAlb BnHA4 12 CS
## 34 BnHA34 Bnaive.HiAlb BnHA4 12 CS/Tofa
## 35 BnHA35 Bnaive.HiAlb BnHA4 12 CS
## 36 BnHA36 Bnaive.HiAlb BnHA4 12 CS/Upa
## 37 BnHA37 Bnaive.HiAlb BnHA4 12 CS/Upa
## 38 BnHA38 Bnaive.HiAlb BnHA4 12 CS/Upa
## 39 BnHA39 Bnaive.HiAlb BnHA4 12 CS/Upa
## 40 BnHA40 Bnaive.HiAlb BnHA4 12 CS/Tofa
## 41 BnHA41 Bnaive.HiAlb BnHA4 12 CS
## 42 BnHA42 Bnaive.HiAlb BnHA4 12 CS/Tofa
## 43 BnHA43 Bnaive.HiAlb BnHA5 12 CS
## 44 BnHA44 Bnaive.HiAlb BnHA5 12 CS/Upa
## 45 BnHA45 Bnaive.HiAlb BnHA5 12 CS/Tofa
## 46 BnHA46 Bnaive.HiAlb BnHA5 12 CS
## 47 BnHA47 Bnaive.HiAlb BnHA5 12 CS/Upa
## 48 BnHA48 Bnaive.HiAlb BnHA5 12 CS
## 49 BnHA49 Bnaive.HiAlb BnHA5 12 CS/Tofa
## 50 BnHA50 Bnaive.HiAlb BnHA5 12 CS/Tofa
## 51 BnHA51 Bnaive.HiAlb BnHA5 12 CS
## 52 BnHA52 Bnaive.HiAlb BnHA5 12 CS/Tofa
## 53 BnHA53 Bnaive.HiAlb BnHA5 12 CS/Upa
## 54 BnHA54 Bnaive.HiAlb BnHA5 12 CS/Upa
## 55 BnHA55 Bnaive.HiAlb BnHA6 6 CS/Upa
## 56 BnHA56 Bnaive.HiAlb BnHA6 6 CS/Upa
## 57 BnHA57 Bnaive.HiAlb BnHA6 6 CS
## 58 BnHA58 Bnaive.HiAlb BnHA6 6 CS/Tofa
## 59 BnHA59 Bnaive.HiAlb BnHA6 6 CS/Tofa
## 60 BnHA60 Bnaive.HiAlb BnHA6 6 CSbnla <- blockrand(n = 60,
num.levels = 3,
levels = c("CS", "CS/Tofa", "CS/Upa"),
stratum = "Bnaive.LoAlb",
id.prefix = "BnLA",
block.sizes = c(1,2,3),
block.prefix = "BnLA")
bnla## id stratum block.id block.size treatment
## 1 BnLA01 Bnaive.LoAlb BnLA1 3 CS/Tofa
## 2 BnLA02 Bnaive.LoAlb BnLA1 3 CS
## 3 BnLA03 Bnaive.LoAlb BnLA1 3 CS/Upa
## 4 BnLA04 Bnaive.LoAlb BnLA2 9 CS/Upa
## 5 BnLA05 Bnaive.LoAlb BnLA2 9 CS
## 6 BnLA06 Bnaive.LoAlb BnLA2 9 CS/Tofa
## 7 BnLA07 Bnaive.LoAlb BnLA2 9 CS
## 8 BnLA08 Bnaive.LoAlb BnLA2 9 CS/Tofa
## 9 BnLA09 Bnaive.LoAlb BnLA2 9 CS/Upa
## 10 BnLA10 Bnaive.LoAlb BnLA2 9 CS
## 11 BnLA11 Bnaive.LoAlb BnLA2 9 CS/Tofa
## 12 BnLA12 Bnaive.LoAlb BnLA2 9 CS/Upa
## 13 BnLA13 Bnaive.LoAlb BnLA3 6 CS
## 14 BnLA14 Bnaive.LoAlb BnLA3 6 CS/Tofa
## 15 BnLA15 Bnaive.LoAlb BnLA3 6 CS/Upa
## 16 BnLA16 Bnaive.LoAlb BnLA3 6 CS/Tofa
## 17 BnLA17 Bnaive.LoAlb BnLA3 6 CS/Upa
## 18 BnLA18 Bnaive.LoAlb BnLA3 6 CS
## 19 BnLA19 Bnaive.LoAlb BnLA4 9 CS
## 20 BnLA20 Bnaive.LoAlb BnLA4 9 CS/Tofa
## 21 BnLA21 Bnaive.LoAlb BnLA4 9 CS
## 22 BnLA22 Bnaive.LoAlb BnLA4 9 CS
## 23 BnLA23 Bnaive.LoAlb BnLA4 9 CS/Tofa
## 24 BnLA24 Bnaive.LoAlb BnLA4 9 CS/Upa
## 25 BnLA25 Bnaive.LoAlb BnLA4 9 CS/Tofa
## 26 BnLA26 Bnaive.LoAlb BnLA4 9 CS/Upa
## 27 BnLA27 Bnaive.LoAlb BnLA4 9 CS/Upa
## 28 BnLA28 Bnaive.LoAlb BnLA5 9 CS/Upa
## 29 BnLA29 Bnaive.LoAlb BnLA5 9 CS
## 30 BnLA30 Bnaive.LoAlb BnLA5 9 CS/Upa
## 31 BnLA31 Bnaive.LoAlb BnLA5 9 CS
## 32 BnLA32 Bnaive.LoAlb BnLA5 9 CS/Tofa
## 33 BnLA33 Bnaive.LoAlb BnLA5 9 CS/Tofa
## 34 BnLA34 Bnaive.LoAlb BnLA5 9 CS
## 35 BnLA35 Bnaive.LoAlb BnLA5 9 CS/Upa
## 36 BnLA36 Bnaive.LoAlb BnLA5 9 CS/Tofa
## 37 BnLA37 Bnaive.LoAlb BnLA6 9 CS/Upa
## 38 BnLA38 Bnaive.LoAlb BnLA6 9 CS
## 39 BnLA39 Bnaive.LoAlb BnLA6 9 CS
## 40 BnLA40 Bnaive.LoAlb BnLA6 9 CS/Tofa
## 41 BnLA41 Bnaive.LoAlb BnLA6 9 CS/Tofa
## 42 BnLA42 Bnaive.LoAlb BnLA6 9 CS/Tofa
## 43 BnLA43 Bnaive.LoAlb BnLA6 9 CS
## 44 BnLA44 Bnaive.LoAlb BnLA6 9 CS/Upa
## 45 BnLA45 Bnaive.LoAlb BnLA6 9 CS/Upa
## 46 BnLA46 Bnaive.LoAlb BnLA7 6 CS/Tofa
## 47 BnLA47 Bnaive.LoAlb BnLA7 6 CS/Upa
## 48 BnLA48 Bnaive.LoAlb BnLA7 6 CS
## 49 BnLA49 Bnaive.LoAlb BnLA7 6 CS
## 50 BnLA50 Bnaive.LoAlb BnLA7 6 CS/Tofa
## 51 BnLA51 Bnaive.LoAlb BnLA7 6 CS/Upa
## 52 BnLA52 Bnaive.LoAlb BnLA8 3 CS/Tofa
## 53 BnLA53 Bnaive.LoAlb BnLA8 3 CS
## 54 BnLA54 Bnaive.LoAlb BnLA8 3 CS/Upa
## 55 BnLA55 Bnaive.LoAlb BnLA9 6 CS
## 56 BnLA56 Bnaive.LoAlb BnLA9 6 CS/Tofa
## 57 BnLA57 Bnaive.LoAlb BnLA9 6 CS
## 58 BnLA58 Bnaive.LoAlb BnLA9 6 CS/Upa
## 59 BnLA59 Bnaive.LoAlb BnLA9 6 CS/Tofa
## 60 BnLA60 Bnaive.LoAlb BnLA9 6 CS/Upa20.1 Printing these on Cards
Ideally, you will print out each randomization on a card, and seal it in a security envelope, with the outside of the envelope labeled with the id. You can do this with the plotblockrand() function. This function creates a pdf file of randomization cards for printing. These are designed so that the middle text will show in a standard letter sized envelope with a window, and the top text (the assignment) can be folded over to increase security (against anyone trying to peek through the security envelope to guess the assignment).
uc_study <- bind_rows(bfha, bfla, bnha, bnla) # bind together the four strata into one dataframe
blockrand::plotblockrand(uc_study, # input dataframe
file = "uc_study.pdf", # output pdf
# top hidden text with assignment
top=list(text=c('My Study','Patient: %ID%','Treatment: %TREAT%'),
col=c('black','black','red'),font=c(1,1,4)),
# middle text to show through window of # 10 envelope
middle=list(text=c("My Study","Stratum: %STRAT%","Patient: %ID%"),
col=c('black','blue','orange'),font=c(1,2,3)),
# bottom text- any instructions to study coordinator
bottom="Call 123-4567 to report patient entry",
cut.marks=TRUE) # add cut marks - 4 per pageOpen up the file uc_study.pdf in your Files tab to see the output pdf file, with assorted fonts and colors.
Just for fun, change (and then re-run) the
text “My Study” to something more interesting
change “Patient” to “Participant”
change “Treatment” to “Arm” or “Assignment”
change some of the colors to standard R colors
change some of the fonts (within 1-4)
Sometimes with equal blocks, and clear treatment effects or side effects, nurses or study coordinators can guess the randomization pattern. If you want to get fancy, and make it even harder to guess treatment assignments, you can add one of the unequal blocks options, to make it hard to find patterns in treatment or in side effects. Set uneq.beg = TRUE for an unequal block in the beginning, or uneq.mid = TRUE for an unequal block in the middle.
20.2 Now, try this yourself
You want to randomize 80 outpatients with Crohn’s disease to one of 8 arms, as part of a 2^3 factorial design to increase patient activation. These arms involve using (A,B, C) or not using (a,b,c) 3 intervention components. The 8 arms then become:
abc
abC
aBc
aBC
Abc
AbC
ABc
ABC
Then we want to stratify the participants by baseline PAM score (a measure of patient activation) with levels of low, medium, and high PAM.
To be prepared for dropouts and imbalanced enrollment, you want to have a randomization list with at least 32 assignments available for each arm and stratum. To avoid a recognizable pattern in the randomization, you want to have a permuted block design with blocks of sizes 8 and 16.
You can hover over top right corner of the code chunk below, and a copy icon will appear - click this to copy the code to your clipboard. You can then paste it into your local version of RStudio, edit it, and run it.
In the code block below, fill in the blanks to complete the code to make a dataframe for the low_pam stratum.
low_pam <- blockrand(n = __,
num.levels = __, #eight treatments
levels = c("abc", "abC", "aBc", "aBC",
"Abc", "AbC", "ABc", "ABC"), # arm names
stratum = "__", # stratum name
id.prefix = "lp", # stratum abbrev
block.sizes = c(1,2,3), # times arms
block.prefix = "LP") # stratum abbrev
low_pamNow that you have one stratum sorted, edit the code block below to create the med_pam and high_pam strata.
med_pam <- blockrand(n = __,
num.levels = __, #eight treatments
levels = c("abc", "abC", "aBc", "aBC",
"Abc", "AbC", "ABc", "ABC"), # arm names
stratum = "__", # stratum name
id.prefix = "__", # stratum abbrev
block.sizes = c(__), # times arms
block.prefix = "__") # stratum abbrev
med_pam
high_pam <- blockrand(n = __,
num.levels = __, #eight treatments
levels = c("abc", "abC", "aBc", "aBC",
"Abc", "AbC", "ABc", "ABC"), # arm names
stratum = "__", # stratum name
id.prefix = "__", # stratum abbrev
block.sizes = c(__), # times arms
block.prefix = "__") # stratum abbrev
high_pamGreat!
Now try to
bind these 3 strata into one dataframe
print these as cards to a pdf file
Edit the code chunk below to produce the pdf file
cd_study <- bind_rows(__,__,__) # bind together the 3 strata into one dataframe
blockrand::plotblockrand(__, # input dataframe
file = "cd_study.pdf", # output pdf
# top hidden text with assignment
top=list(text=c('CD Study','Patient: %ID%','Treatment: %__%'),
col=c('orange','blue','red'),font=c(1,1,4)),
# middle text to show through window of # 10 envelope
middle=list(text=c("CD Study","Stratum: %STRAT%","Patient: %__%"),
col=c('black','red','cadetblue'),font=c(1,2,3)),
# bottom text- any instructions to study coordinator
bottom="Call 123-4567 to report patient entry",
cut.marks=TRUE) # add cut marks - 4 per page20.3 Now Freestyle
Your turn. Create randomization tables and a pdf file of cards for a study of 2 microbiome interventions to reduce the formation of colon adenomas.
your 3 study arms will be - placebo, Streptococcus thermophilus, and S.thermo plus lactose (a preferred sugar for S.t, making this arm a synbiotic, while arm 2 is a probiotic) - aka 3 arms called: pbo, probiotic, synbiotic.
Your stratifications will be by
prior polyps being MSI_hi or MSI_lo (for microsatellite instability mutations)
BMI above or below 35. BMI_hi, BMI_low
block sizes of 4,8,12,16
160 per arm
Edit the code block below for the first stratum
mhbh <- blockrand(n = __, # treatment arms
num.levels = __, # of treatments
levels = c("placebo", "probiotic", "synbiotic"), # arm names
stratum = "__,__", # stratum name
id.prefix = "mhbh", # stratum abbrev
block.sizes = c(__,__,__,__), # times arms
block.prefix = "__") # stratum abbrev
mhbhEdit the code block below for the remaining strata
mhbl <- blockrand(n = __, # treatment arms
num.levels = 3, # of treatments
levels = c("placebo", "probiotic", "__"), # arm names
stratum = "msi_hi.bmi_lo", # stratum name
id.prefix = "__", # stratum abbrev
block.sizes = c(1,2,__,__), # times arms
block.prefix = "MHBL") # stratum abbrev
mhbl
mlbl <- blockrand(n = 160, # treatment arms
num.levels = __, # of treatments
levels = c("placebo", "__", "synbiotic"), # arm names
stratum = "__", # stratum name
id.prefix = "mlbl", # stratum abbrev
block.sizes = c(__,__,3,4), # times arms
block.prefix = "MLBL") # stratum abbrev
mlbl
mlbh <- blockrand(n = __, # treatment arms
num.levels = 3, # of treatments
levels = c("__", "probiotic", "synbiotic"), # arm names
stratum = "msi_lo.bmi_hi", # stratum name
id.prefix = "__", # stratum abbrev
block.sizes = c(1,2,3,4), # times arms
block.prefix = "MLBH") # stratum abbrev
mlbhEdit the code block below to bind the strata together and print the cards
adenoma_study <- bind_rows(mlbl, mlbh, mhbh, mhbl) # bind together the strata into one dataframe
blockrand::plotblockrand(__, # input dataframe
file = "adenoma_cards.pdf", # output pdf
# top hidden text with assignment
top=list(text=c('Adenoma Study','Patient: %__%','Treatment: %TREAT%'),
col=c('orange','blue','red'),font=c(1,1,4)),
# middle text to show through window of # 10 envelope
middle=list(text=c("Adenoma Study","Stratum: %__%","Patient: %ID%"),
col=c('black','red','cadetblue'),font=c(1,2,3)),
# bottom text- any instructions to study coordinator
bottom="Call 123-4567 to report patient entry. \nInstruct participant to avoid antibiotics and stop aspirin",
cut.marks=TRUE) # add cut marks - 4 per page