Determinants of Non-Receipt of Guideline-Concordant Treatment in Pancreatic Cancer: A Classification Tree Analysis
Manish Tripathi1,2, Ioannis Liapis1,3, Nritya Nair1,3
1Institute of Cancer Outcomes and Survivorship, University of Alabama at Birmingham, Birmingham, AL, USA.
Background:
Non-receipt of guideline-concordant treatment (GCT) in pancreatic ductal adenocarcinoma (PDAC) is associated with worse outcomes. Studies have identified associated factors, but less is known about how these factors co-occur within patient subgroups at the time of presentation to oncology care.
Methods:
We conducted a retrospective cohort study of patients with PDAC treated at two academic centers. Candidate variables available at presentation were selected a priori across demographic, clinical, socioeconomic, and healthcare utilization domains. Classification and regression tree (CART) models were used to identify conditionally defined subgroups associated with non-receipt of GCT. Models were developed using an 80/20 training/test split with repeated cross-validation.
Results:
Of 1040 patients, 337 (32.4%) did not receive GCT. In both unpruned and pruned models, Area Deprivation Index (ADI) defined the initial split. Among patients with ADI ≥7, non-receipt of GCT was more common among those entering care through the emergency department, with further stratification by race, social support, and timeliness of care. Among patients with ADI <7, stratification was based on age, social support, and performance status. In sensitivity analyses excluding age, ADI remained the initial split, with downstream stratification based on cancer screening, insurance, social support, and albumin. The model demonstrated an area under the receiver operating characteristic curve (AUROC) of 0.849 on the training set and 0.662 on the independent test set.
Conclusion:
Non-receipt of GCT in PDAC is associated with distinct clinically identifiable patterns involving socioeconomic context, access to care, and patient vulnerability at presentation, highlighting potential targets for early intervention.
