Top Clinical SAS Interview Questions for Beginners & Experienced

In clinical trials, SDTM (Study Data Tabulation Model) datasets provide raw, standardized data, while ADaM (Analysis Data Model) datasets prepare analysis-ready data derived from SDTM. Inconsistencies between these datasets can impact data integrity, analysis outcomes, and regulatory compliance.

Here’s a step-by-step approach to identifying, investigating, and resolving inconsistencies between SDTM and ADaM datasets.

In one of my previous projects, I was tasked with generating Tables, Figures, and Listings (TFLs) for multiple clinical trial datasets. The study required creating the same summary tables and adverse event listings for different treatment groups and time points. Initially, I noticed that I was repeating similar code blocks multiple times, which was time-consuming and prone to manual errors.

Debugging errors in SAS is a crucial skill to ensure the code runs successfully and produces accurate results. When errors or unexpected outputs occur, a systematic approach can help identify and resolve them efficiently. Below are the steps and techniques I follow to debug errors in my SAS programs:

The SAS Log is the primary tool for identifying issues in SAS programs.

• Errors: Highlighted in red. These prevent the program from running successfully.
• Warnings: Highlighted in green. These indicate potential problems but do not always stop the program.
• Notes: Provide information about the program execution (e.g., variable creation, data read/write counts).
• • read/write counts).

  Steps to Debug:

  • Look for the line number where the error occurred (ERROR:).
  • Identify the problematic statement or syntax issue.
  • Use the messages to determine what caused the error.

In one of my previous clinical SAS projects, I was tasked with analyzing and summarizing large adverse event (AE) datasets that contained millions of records across multiple domains. Initially, the SAS program I inherited ran for several hours due to inefficient code logic, which delayed project timelines and impacted reporting deadlines. My role was to identify the bottlenecks and optimize the program for better performance.

Adverse event (AE) analysis is a crucial part of clinical trial data analysis to evaluate the safety profile of a drug or treatment. SAS provides powerful tools to analyze adverse event data, ensuring accurate reporting for regulatory submissions and decision-making.

Here is a step-by-step approach to performing adverse event analysis using SAS.

Adverse event data is typically stored in the SDTM AE dataset or derived into an ADaM ADAE dataset. Key variables include:

• USUBJID: Unique Subject Identifier.
• AETERM: Adverse Event Term.
• AESEV: Severity (e.g., mild, moderate, severe).
• AEDECOD: Coded Adverse Event Term.
• AEREL: Relationship to treatment.
• TRT01P: Planned treatment group (from ADSL dataset).

Duplicate records in a demographics (DM) dataset can cause inconsistencies and inaccuracies in clinical trial analyses. Identifying and resolving duplicates ensures data integrity and compliance with regulatory standards. Here’s how I would approach this problem.

Step 1: Understand Key Identifiers

• In the DM dataset, each record should represent one unique subject.
• The USUBJID (Unique Subject Identifier) variable is typically used to identify records.
• Additional variables like SITEID, SUBJID, or VISIT may also need to be considered.

Step 2: Use PROC SORT with DUPOUT Option

Sort the dataset using PROC SORT to find duplicate records based on the key identifier (USUBJID).

sas

Copy code

proc sort data=dm out=dm_sorted no dup key dupout=dm_duplicates; 

 by usubjid;  

run;

proc print data=dm_duplicates;  

    title “Duplicate Records in DM Dataset”;  

run;

• Output: The DUPOUT dataset (dm_duplicates) contains all duplicate records.

Step 3: Use PROC SQL to Identify Duplicates

Another way to identify duplicates is by grouping data with PROC SQL.

sas

Copy code

proc sql;  

    create table dm_duplicates as  

    select usubjid, count(*) as record_count  

    from dm  

    group by usubjid  

    having count(*) > 1;  

quit;

proc print data=dm_duplicates;  

    title “Duplicate Records in DM Dataset with PROC SQL”;  

run;

• Output: This table lists duplicate USUBJID values along with their counts.

Step 4: Use DATA Step to Flag Duplicates

Flag duplicate records using the FIRST. and LAST. variables in a BY-group processing.

sas

Copy code

proc sort data=dm;  

    by usubjid;  

run;

data dm_flagged;  

    set dm;  

    by usubjid;  

    if first.usubjid and last.usubjid then dup_flag = 0;  

    else dup_flag = 1;  

run;

proc print data=dm_flagged;  

    where dup_flag = 1;  

    title “Flagged Duplicate Records in DM Dataset”;  

run;

• Output: Records with dup_flag = 1 are duplicates.

Step 1: Investigate the Cause of Duplicates

Source Issues: Duplicates may originate from errors during data collection, merging, or dataset creation.

Check Attributes: Compare all variable values for the duplicates to determine differences or redundancies.

Step 2: Retain Only One Record Per Subject

If duplicates are identical, retain only the first occurrence using PROC SORT.

sas

Copy code

proc sort data=dm out=dm_deduplicated nodupkey;  

    by usubjid;  

run;

proc print data=dm_deduplicated;  

    title “DM Dataset After Removing Exact Duplicates”;  

run;

Step 3: Resolve Conflicting Records

If duplicates have conflicting data, create rules to resolve them:

• Rule 1: Retain records with the most complete data.
• Rule 2: Retain records based on a priority variable (e.g., SITEID).
• Rule 3: Retain records with the earliest or latest visit date.

Example:

sas

Copy code

data dm_resolved;  

    set dm;  

    by usubjid;  

    if first.usubjid; /* Retain the first record */  

run;

proc print data=dm_resolved;  

    title “Resolved DM Dataset”;  

run;

• Compare Original and Cleaned Datasets: Use PROC COMPARE to ensure no unintended changes were made.

sas

Copy code

proc compare base=dm compare=dm_resolved;  

    id usubjid;  

    title “Validation of Deduplicated DM Dataset”;  

run;

• Check Record Count: Confirm the number of unique USUBJID values matches the deduplicated dataset.

sas

Copy code

proc sql;  

    select count(distinct usubjid) as unique_subjects  

    from dm_resolved;  

quit;

• Audit Trail: Maintain a record of steps taken to identify and resolve duplicates, including criteria used for resolving conflicts.
• Update Metadata: Reflect changes in the DEFINE.XML file or dataset documentation.

Generating a summary table of baseline characteristics by treatment group is a common task in clinical SAS programming. It involves summarizing variables like age, gender, weight, and other demographic or baseline measurements across treatment groups.

Here’s a step-by-step approach to accomplish this in SAS.

1. Understand the Dataset

• The ADSL (Subject-Level Analysis Dataset) is typically used for baseline characteristics.
• Key variables:
  • TRT01P: Planned treatment group.
  • AGE, WEIGHT, HEIGHT: Numeric demographic variables.
  • SEX, RACE: Categorical demographic variables.

Use PROC MEANS to calculate descriptive statistics (e.g., N, Mean, SD, Min, Max) for numeric variables.

Example Code:

sas

Copy code

proc means data=adam.adsl n mean std min max;  

    class trt01p;  

    var age weight height;  

    title “Summary of Baseline Characteristics – Numeric Variables”;  

run;

Output:

Use PROC FREQ to calculate the distribution of categorical variables like gender and race.

Example Code:

sas

Copy code

proc freq data=adam.adsl;  

    tables trt01p*sex / nocol nopercent;  

    title “Summary of Baseline Characteristics – Categorical Variables (Gender)”;  

run;

Output:

To combine numeric and categorical summaries into a single table, you can use PROC REPORT for customization.

Example Code:

sas

Copy code

proc report data=adam.adsl nowd;  

    columns trt01p sex age weight;  

    define trt01p / group “Treatment Group”;  

    define sex / across “Gender”;  

    define age / analysis mean “Mean Age”;  

    define weight / analysis mean “Mean Weight”;  

    title “Baseline Characteristics Summary Table”;  

run;

For large datasets with many variables, automate the process using SAS macros to reduce repetitive code.

Macro Example:

sas

Copy code

%macro summarize_baseline(var, group);  

    proc means data=adam.adsl n mean std min max;  

        class &group;  

        var &var;  

        title “Baseline Summary for &var by &group”;  

    run;

%mend summarize_baseline;

%summarize_baseline(age, trt01p);  

%summarize_baseline(weight, trt01p);

Export the summary table to a format suitable for clinical reporting (e.g., PDF, RTF, or Excel) using the ODS (Output Delivery System).

Example Code:

sas

Copy code

ods pdf file=”baseline_summary.pdf”;  

proc report data=adam.adsl nowd;  

    columns trt01p age weight;  

    define trt01p / group “Treatment Group”;  

    define age / analysis mean “Mean Age”;  

    define weight / analysis mean “Mean Weight”;  

run;

ods pdf close;

• Ensure the calculations align with the Statistical Analysis Plan (SAP).
• Use PROC COMPARE to validate against source data if required.

Example Validation Code:

sas

Copy code

proc compare base=source_data compare=summary_table;  

    id trt01p;  

    title “Validation of Baseline Summary Table”;  

run;

Merging datasets in SAS is a common operation used to combine data from two or more datasets based on one or more common variables. Here’s how you can merge datasets and the precautions to consider:

Sort the Datasets by the Common Variable(s):
SAS requires datasets to be sorted by the variable(s) you will use to merge them.
sas
Copy code
PROC SORT DATA=dataset1; BY id; RUN;

PROC SORT DATA=dataset2; BY id; RUN;

Use the MERGE Statement in a DATA

 Step:1
Combine datasets using the MERGE statement and specify the common variable(s) with the BY statement.
sas
Copy code
DATA merged_dataset;

  MERGE dataset1 dataset2;

  BY id;

RUN;

Handling One-to-Many or Many-to-Many Relationships:
Ensure you understand the relationship between the datasets to avoid unexpected results.

Ensure Datasets are Sorted by the BY Variable(s):
Merging unsorted datasets may produce an error or incorrect results.

Verify the Existence and Naming of Common Variables:

• If datasets have variables with the same name but different meanings, rename them before merging to avoid conflicts.
• Use the RENAME= option in the DATA step if needed.

Example:
sas
Copy code
DATA dataset1_renamed;

  SET dataset1(RENAME=(var1=var1_dataset1));

RUN;

Check for Duplicate Keys in the BY Variable(s):
Duplicate values in the BY variable can lead to a one-to-many or many-to-many merge, potentially producing unintended results.
Use PROC FREQ or PROC SORT NODUPKEY to check for duplicates.
sas
Copy code
PROC SORT DATA=dataset1 NODUPKEY; BY id; RUN;

Understand Missing Values Handling:

If a key exists in one dataset but not in the other, SAS will merge it with missing values for the unmatched variables.

Use the IN= option to identify and handle unmatched observations.

Example:
sas
Copy code
DATA merged_dataset;

  MERGE dataset1 (IN=in1) dataset2 (IN=in2);

  BY id;

  IF in1 AND in2; /* Keeps only matched observations */

RUN;

Review the Log for Warnings or Errors:
The SAS log can indicate issues like unmatched variables, missing BY statements, or other merge-related problems.

Validate the Output:
Always validate the merged dataset to ensure the merge was performed correctly. Use PROC PRINT, PROC FREQ, or PROC MEANS to review the results.
sas
Copy code
PROC PRINT DATA=merged_dataset (OBS=10); RUN;

INNER JOIN dataset2 AS b

  ON a.id = b.id;For more flexibility, you can use SQL-style joins with PROC SQL.

Example of an inner join.

sas

Copy code

PROC SQL;

  CREATE TABLE merged_dataset AS

  SELECT a.*, b.*

  FROM dataset1 AS a

QUIT;

By following these steps and precautions, you can merge datasets accurately and avoid common pitfalls in SAS programming.

• Using MERGE or PROC SQL.