# Risk Factors and Prediction Model for Early‐Onset Immune‐Related Adverse Events in Pan‐Cancer Patients Undergoing Anti‐PD‐(L)1 Therapy: A Retrospective Study in a Tertiary‐Level Hospital

## TL;DR

This study develops a prediction model to identify patients at high risk of immune-related adverse events during anti-PD-(L)1 cancer therapy.

## Contribution

A novel risk assessment model for early-onset immune-related adverse events in cancer patients undergoing anti-PD-(L)1 therapy is introduced.

## Key findings

- Endocrinal toxicities were the most common immune-related adverse events reported.
- A prediction model with a C-index of 0.727 was developed using clinical and laboratory parameters.
- The model can help screen and monitor high-risk patients to improve prognosis.

## Abstract

Anti‐programmed death 1 (PD‐1) and anti‐programmed death ligand 1 (PD‐L1) immune checkpoint inhibitors (ICIs) have changed the treatment landscape of many advanced malignancies. However, immune‐related adverse events (irAEs) bring great challenges to clinical benefits. The prediction of irAEs is urgently demanded for early detection and intervention.

Patients in our center who received anti‐PD‐(L)1 immunotherapy between January 2019 and May 2023 were collected. Logistic least absolute shrinkage and selection operator (LASSO) regression analysis with 10‐fold cross‐validation was performed to identify the most relevant variables associated with irAEs. Multivariate logistic regression analysis was used to build a prediction model by introducing features selected in LASSO regression analysis.

Overall, 680 eligible patients were included, of whom 330 patients were included in the irAEs group. In the irAEs group, 455 different irAEs were reported, of which 52 events were grade 3 or higher in severity. Endocrinal toxicities (174/680, 25.59%) were the most commonly reported irAEs. Through LASSO and logistic regression analysis, we developed a risk assessment model to predict the risk of irAEs based on basophil percentage (BASO%), hemoglobin (Hb), absolute lymphocyte count (ALC), platelet‐to‐lymphocyte ratio (PLR), lymphocyte‐to‐monocyte ratio (LMR), blood urea nitrogen level (BUN), the Charlson comorbidity index (CCI) score, Eastern Cooperative Oncology Group Performance Status (ECOG PS), and hepatitis B/hepatitis B surface antigen carriers. The model had a C‐index of 0.727, with good discrimination and calibration capabilities.

The prediction model developed in our study can screen and monitor patients with high risk of developing irAEs. It may improve prognosis for pan‐cancer patients receiving anti‐PD‐(L)1 immunotherapy.