(T-114) SAS Macro for Validation of NONMEM Data Items Pertaining to Implicit Expansion of Dose Records in Oral Dose Population Pharmacokinetic Datasets

Matt Maloney, MPH, BS – Data Science Lead, Clinical Pharmacology and Pharmacometrics, Bristol Myers Squibb; Weidong Chen, PhD – Associate Director, Data Science, Clinical Pharmacology and Pharmacometrics, Bristol Myers Squibb

Objectives: Oral dose drugs are often not exclusively administered in clinic, leaving patients to take at home doses and report their dosing history. As a result, complete dosing information is not always available for oral dose studies. In analysis of Population Pharmacokinetic (popPK) datasets, two methods can account for the unrecorded doses. The first is to insert a record for each unrecorded dose event into the dataset. The second is an implicit expansion of dose records with NONMEM standard data items ADDL (additional identical doses given) and II (interdose interval). Oral dose studies have also been shown to contain multiple dose interruptions or reductions due to adverse events, unrecorded dose times, as well as dose titrations or other protocol specified dose regimen complexities that can at times deviate from the actual administered dose frequency. Given these complexities, a validation tool for the data integrity of PopPK dose records was targeted for development to identify anomalies resulting from ADDL/II expansion.

Methods: A SAS macro was created to accompany PopPK dataset generation for validation of oral dose records. The macro is called using the following syntax: “%check_addl(dsin, )” where “dsin” is the dosing records of a PopPK dataset to use as the input, and “obs” optionally specifies how many records to print in the output with default as "max". By utilizing NONMEM, CDISC ADaM PopPK and SDTM rules and guidance, the macro first ensures that certain data variables are present and structured as expected to mitigate user error. The macro works by performing a series of queries on the data where the expected return from the macro is no results. Any return from the macro indicates possible areas of exploration including fixing errors in PopPK data preparation, consultations with data management teams if applicable, or left to the discretion of the pharmacometrician.

Results: The following queries are returned by the macro:
• Check that ADDL=0 when II=0
• Check if records with dose time >=24:00hr (next day) are created
• Check for different dosing schedules on the same day
• Check if doses with EXDOSFRQ > 24hrs overlap with other dose records
• Check for periods of unrecorded dosing
• Check if the actual doses on a specific day does not match EXDOSFRQ
• Check if BID doses have both doses in AM or both doses in PM

Conclusions: Given potential for data abnormalities and the need for imputation of missing dose times, a macro that can be run on a completed dosing dataset provides useful validations on data integrity prior to PopPK analysis steps. The macro efficiently runs queries that check on standard items that can be seen for all PopPK datasets that use ADDL/II dose expansion. The macro has the potential to save both time and needless exclusions that could be avoided if issues are found and dealt with at the time of PopPK dose dataset generation.

Citations: NA