SAS_code.sas

* Read in the SDTM lab data (LB) dataset;
data sdtm_lb;
    set SDTM.LB;
run;

* Check the structure of the SDTM LB dataset;
proc contents data=sdtm_lb; 
run;

* Create the ADaM LB (ADLB) dataset;
data work.adlb;
    set sdtm_lb;

    * Retain only the variables we need for analysis;
    keep USUBJID LBCAT LBTEST LBSTRESN LBSTRESU VISITNUM;

    * Create derived variables for baseline and change from baseline;
    by USUBJID LBCAT LBTEST VISITNUM;

    * The baseline is defined as the first non-missing result for each subject, test, and category;
    if first.USUBJID or first.LBTEST or first.LBCAT then do;
        LBBLFL = 'Y';
        LBSTRESN_BL = LBSTRESN;
    end;
    else LBBLFL = 'N';

    * The change from baseline is the result minus the baseline value;
    LBCHG = LBSTRESN - LBSTRESN_BL;

    * Create a flag for significant change from baseline;
    if LBCHG > 5 then LBCHGFL = 'Y';
    else LBCHGFL = 'N';

run;

* Apply ADaM metadata attributes;
* This is a placeholder - actual metadata would be study-specific and would be applied based on the ADaM Implementation Guide;
run;

* Check log for any errors or warnings, and validate outputs;
* This step is crucial for quality control, ensuring the analysis was conducted correctly;

* Documentation is essential for every step to ensure reproducibility and inspection readiness;
* Remember to follow ADaM Implementation Guide and relevant regulatory guidelines;
#############################################################################################################
* Set the input data location;
filename indata '/path/to/clinical_trial_data.csv';

* Import the data into a SAS dataset;
proc import datafile=indata
    out=work.clin_data
    dbms=csv
    replace;
run;

* Check the structure of the data, ensuring it's been read correctly;
proc contents data=work.clin_data; 
run;

* Conduct data cleaning and manipulation;
data work.clean_data;
    set work.clin_data;

    * Recode categorical variables and handle missing values;
    if var1 = 'old category' then var1 = 'new category';
    if var2 = . then var2 = 0;

    * Create a derived variable: the interaction between var1 and var2;
    var1_var2 = var1 * var2;

run;

* Create a macro to automate repetitive tasks: generating summary statistics for a list of variables;
%macro summarise_vars(dataset, vars);
    %let var_count = %sysfunc(countw(&vars));

    %do i = 1 %to &var_count;
        %let var = %scan(&vars, &i);

        proc means data=&dataset mean median min max;
            var &var;
        run;
    %end;

%mend summarise_vars;

* Use the macro to generate summary statistics for var1, var2, and the interaction term;
%summarise_vars(work.clean_data, var1 var2 var1_var2);

* Perform logistic regression;
proc logistic data=work.clean_data descending;
    class var1(param=ref ref='new category');
    model response_var(event='1') = var1 var2 var1_var2;
run;

* Check log for any errors or warnings, and validate outputs;
* This step is crucial for quality control, ensuring the analysis was conducted correctly;

* Documentation is essential for every step to ensure reproducibility and inspection readiness;
* Remember to follow SOPs, WIs, and relevant regulatory guidelines;