Abstract 92P
Background
Immune checkpoint inhibitors (ICIs) are used for many non-small cell lung cancer (NSCLC) patients. Immune-related adverse events (irAEs) pose a risk, impacting efficacy and quality of life. Retrieving retrospective information on irAEs from electronic medical records (EMR) is challenging, requiring time intensive efforts. We introduced an innovative method based on objective measurements for efficient irAE identification.
Methods
The analysis included adult cancer patients who received ICIs between September 2015 to December 2023. We used objective parameters that can be pulled from the EMR. We created 4 conditions to independently identify irAEs: 1) abnormal laboratory results with use of prednisone/methylprednisolone; 2) use of infliximab; 3) discontinuation of treatment with abnormal laboratory results; and 4) discontinuation of treatment with use of prednisone/methylprednisolone. The conditions were applied to our cancer patient database using electronic algorithm code written in R software.
Results
The cancer cohort has 21,568 patients, 1200 received ICIs. Among them, 457 had NSCLC and 109 (23.8%) were identified as meeting ≥1 condition. Eight (7.3%) were identified as having irAE based on abnormal laboratory results and use of prednisone or methylprednisolone. None were treated with infliximab. Eighteen (16.5%) discontinued treatment for >7 weeks, exhibiting abnormal laboratory results, while 62 (56.9%) discontinued treatment for >7 weeks and used prednisone or methylprednisolone, leading to identification as experiencing an irAE. Twenty-one (19.3%) were identified >1 condition. Manual evaluation of 25 cases produced by the algorithm and found no false positives (0%). Regarding false negatives, we manually reviewed 20 cases of NSCLC patients known to have irAE, and 19 (95%) were correctly identified by the code.
Conclusions
Our research introduces an innovative and efficient approach to address irAEs in NSCLC patients receiving ICI. Objective measurements within the EMR and an R software-based code has the potential to improve our understanding and management of irAEs. We plan to validate this approach with an AI algorithm on a larger cancer patient dataset.
Legal entity responsible for the study
The authors.
Funding
Has not received any funding.
Disclosure
A. Merchant: Financial Interests, Institutional, Research Grant: Pfizer; Financial Interests, Institutional, Other: Amgen; Financial Interests, Personal, Other: Oncovalent. N. Tatonetti: Financial Interests, Personal, Other: CARI Health; Financial Interests, Institutional, Funding: Pfizer, Amgen. K. Reckamp: Financial Interests, Personal, Other, consultant: Amgen, AstraZeneca, Blueprint, Daiichi Sankyo, Genentech, GSK, Janssen, Lilly, Mirati, Takeda; Financial Interests, Personal, Advisory Board: EMD Serono, Merck KGA; Financial Interests, Personal, Invited Speaker: Seattle Genetics,; Financial Interests, Institutional, Invited Speaker: Genentech, Blueprint, Calithera, Daiichi Sankyo, Elevation Oncology, Janssen. All other authors have declared no conflicts of interest.