Background

• In clinical trials, the participants are randomly allocated to either test drug or comparator group
• We will have different groups of subjects based on the number of treatments or treatment combinations being studied in the trial
• In parallel arm study, a subject will only receive the treatment to which he/she is randomized to 
• To assess if the subject characteristics are similar across different treatment groups, we create summary reports
• Subject characteristics can be of numeric type (age, baseline BMI, etc) or categorical type(Sex, Race etc)
• For numeric or continuous variables, data is summarized in terms of sample size, number of records with missing data, mean, standard deviation, quartiles (q1, median(q2), q3), minimum and maximum etc
• These descriptive statistics are obtained for each group so that a comparison can be made between the groups
• Below is a table summarizing the age value and the results are presented in a tabular format with different treatments side by side for easy comparison

Points to note in this kind of summary table

• The variable being analyzed is called an analysis variable. 
• In the above example, age is the analysis variable
• The variable which contains the treatment group information is called a 'grouping' or 'class' or 'by' variable
• In the above example, we are creating a summary to compare different 'dose levels' - the variable that contains the 'dose level' of a subject will be called a 'class' variable
• If we follow ADaM standard, planned or actual treatment variables (TRT01P, TRT01A etc.) will be used as 'class' variables

• We need to check the different descriptive statistics that are being presented
• In this example, n, missing, mean, standard deviation, q1, median, q3, minimum and maximum are being presented
• The number of decimal values presented for different statistics are to be noted.
• Generally,
  • mean, median, q1, q3 will be presented to 1 additional decimal place than in data
  • minimum and maximum will be presented to the same number of decimals as in data
  • standard deviation will be presented to 2 additional decimal places than in data
• Some of the statistics are presented side by side 
  • n (missing)
  • Mean (SD)
  • Q1, Q3
  • Min, Max
• Treatment groups are presented as columns in the table
• Statistics are presented as rows in the table with a specific order

How do we program this in SAS?

• There are multiple ways in which descriptive statistics for numeric variables can be obtained in SAS
• Most commonly used procedures for this purpose are proc means or proc summary 
• Both proc means and proc summary can produce output to the output window or to an output dataset
• We need extensive transformations on the obtained numbers reported by these procedures
  • control the number of decimals to be presented for different statistics
  • concatenate two different statistics into a single row
  • use descriptive labels for the statistics etc
• We need to use a combination of data steps, proc steps (like sort, transpose, report etc) to produce the final output