Complex Design Randomization

Creating a randomization list for a complex design using Create randomization list (complex design)

Consider an example in which there will be 3 treatment groups, placebo, new drug, standard drug. The ratio of sample sizes will be 1:2:2. That is, one-fifth of patients will be assigned to the placebo group, and two-fifths to each of the two active drugs. Patients will be entered at three sites, Center 1, Center 2, and Center 3, and patients will be stratified using two stratifying factors: gender (male or female) and smoking (smoker or non-smoker). It is expected that the three Centers will enter about the same number of patients, and that about 30% of patients will be in the “smoker” category. The total sample size is planned to be 500.

To create a randomization list for this example:

Select Create Randomization List (Complex designs) from the Randomize menu. In the Randomization dialog box, enter the study title, Example. The screen shows the basic information about the number of treatments, centers, strata, and sample size already filled in.

To enter the rest of the information about the treatments we specify that there are 3 treatments, and click on the Specify button.

The Specify dialog box will appear with the Ratio column filled in for the default situation in which all treatments would have equal sample sizes.

Type in the three treatment group names, Placebo, New drug, and Standard. Leave the default treatment code numbers, or type in other numbers or letters such as P, N, and S. They will appear as a column in the randomization list.

Edit the Ratio column for New drug and Standard to specify sample size ratios in these groups of twice the planned sample size for the Placebo group.

Enter 1, 2, and 2 in the Ratio column.

Click OK to return to the Randomization List dialog box

You must now specify the number of centers. In this example there are 3 centers, so enter the number 3 and click on the Specify button. The Specify centers dialog box will appear.

Enter the center names and edit the proportions column if necessary. For this example, we expect the same sample size in each center, so the filled-in box will retain the default proportions.

Click OK to return to the Randomization List dialog box.

There are two stratifying factors, factor 1 is gender, and factor 2 is smoking status; each has two strata. Enter the number 2 for Factor 1, and click on the Specify 1 button to specify the strata for Factor 1.

The Specify factor 1 dialog box will appear.

Enter the stratum names and the proportions of patients expected in each stratum. Here we expect an equal number of males and females, so we enter the names male and female and leave the default proportions.

Click on OK to return to the Randomization List dialog box.

Specify two strata for Factor 2.

Now the Specify 2 button is activated, so click on this to specify the strata for Factor 2. We enter smoker and non-smoker and change the proportions to 0.3 and 0.7.

Click OK to return to the Randomization List dialog box.

If you expect that the proportions in the four combined strata are what would be expected if the stratifying factors were independent, continue on and specify the total sample size. If you want to check the proportions for the combined strata or to edit them, click on Edit joint proportions for strata.

For this example, we assume that fewer women smoke and change the proportions to 0.2 and 0.3 for males and 0.1 and 0.4 for females to maintain a 50-50 ratio of males to females.

Click OK and the Randomization dialog box will appear again.

Using Mixed Block Sizes

When you click on OK this automatically updates the proportions in the Specify Factor 1 and Specify Factor 2 dialog boxes if necessary.

We specify a total sample size of 500. Or we can type in a somewhat larger sample size to allow for extra patients in one or more of the centers or strata.

Specify whether you want all blocks to be size 5 (the minimum allowable with a 1:2:2 randomization), size 10, size 15, or size 20. Or you can use the nQuery mixed blocks randomization. (If you choose the mixed block randomization, about 60% of the sample size within each center/stratum will be assigned using blocks of size 5 and the rest using blocks of size 10.)

nQuery Advisor starts the randomization with a random seed to initiate the pseudo-random number generator used for the simulations. If no random seed is specified, nQuery Advisor uses the clock on your computer to create the random seed.

If you want to specify a random seed for the simulation, check the box labeled Select this box to specify seed (Optional) and enter an integer value greater than 0 into the box labeled Random seed. Here we specify a random seed of 123. To see the randomization list, click on View List.

The Randomization List screen will appear with the study title, date, and a viewable spreadsheet with subject ID numbers, center/stratum labels, and the treatment or group assignment for each subject. Use the slider bar to move down in the list to see the treatment assignments for the other center/stratum combinations.

You can save the list to a file, you can copy it and paste it into a document or spreadsheet, or you can print it. If you wish to change the specifications, click on re-specify and re-specify the desired information. If you want to have Codes for the Center/stratum combinations appear in the List, click on Edit IDs to specify.

nQuery Advisor automatically creates RandList ID numbers for the subjects by using the sequence numbers (in this case the numbers 1 to 525 since more than 500 subjects were required to satisfy the treatment, center, and stratum specifications) and padding them with leading zeros. If you wish to edit these ID numbers, perhaps to number separately within each center or stratum, or to change the minimum width of the ID number, click on the button labeled Edit IDs.
The Specify starting ID numbers for each center/stratum dialog box will appear. This box shows the center and stratum names, the starting IDs, and the sample sizes for each center/stratum combination. For this example, we edit the starting IDs to begin the numbering for each center with a unique digit but allow sequential numbering within each center. That is, we change the starting IDs for the four Center 1 strata to 1001, 1036, 1086, 1106; we change the starting IDs for the four Center 2 strata to 2001, 2036, 2086, 2106; and we change the starting IDs for the four Center 3 strata to 3001, 3036, 3086, 3106.

When you click OK you will see the randomization list with the changes in the codes and the ID numbers.
To save the randomization list to a file:

Click on the Save or Save as button. The file will be saved as an nQuery Advisor file with the nQuery Advisor randomization list file extension .nqr. You can open this file again at a later time. You also have the option to save the randomization list in a .csv or .txt format to use in other applications.

To copy the file to be included in a document:

Click on the Copy button. Then you can paste the list directly into a spreadsheet or into a document where it will appear as a table. Information about the study title, treatment names, etc will appear in the pasted table. You can then edit the headers and/or subject ID numbers as desired and save or print from these programs. The following listing shows the header information and selected portions of the randomization list as pasted into Microsoft Word from Microsoft Excel.