# Competing Risks and Their Impact on Treatment Efficacy Assessment in Fractionated Stereotactic Radiotherapy for Brain Metastases: A Retrospective Study

## TL;DR

This study compares two statistical methods for analyzing treatment outcomes in brain metastases patients, showing that one method overestimates risks due to not properly accounting for early deaths.

## Contribution

The study demonstrates the bias introduced by using the 1-KM method over CIF in assessing treatment efficacy when competing risks are present.

## Key findings

- The 1-KM method overestimates cumulative incidence of local failure and new brain metastases by 9-14% compared to CIF.
- Survival curves using 1-KM show higher incidence rates due to improper handling of competing risks.
- Non-small cell lung cancer and malignant melanoma were the most common cancer types in the study cohort.

## Abstract

The Kaplan‐Meier (KM) method and competing risk analysis are two statistical approaches for analyzing time‐to‐event data. These methods differ in their treatment of competing events, such as deaths occurring before the event of interest, which can impact the interpretation of treatment efficacy in oncology.

This retrospective study included 73 patients who underwent fractionated stereotactic radiotherapy (six fractions of 5 Gy) for brain metastases at the University Hospital Regensburg between January 2017 and December 2021. The events of interest were the cumulative incidences of local failure within the planning target volume and the development of new brain metastases. Premature deaths occurring before the events of interest were treated as competing events. The complement of the KM method (1‐KM), which censors patients who die prematurely, was compared to the cumulative incidence function (CIF), which accounts for the fact that patients who die without experiencing the event of interest are no longer at risk for that event.

The median follow‐up was 56 months. The most common cancer types were non‐small cell lung cancer (n = 26, 35.6%) and malignant melanoma (n = 25, 34.2%). Using CIF, the cumulative incidences of local failure and new brain metastases were 27% and 55% at 24 months, respectively. The 1‐KM method overestimated the cumulative incidence of local failure by 9% at 24 months and 14% at 36 months, and new brain metastases by 13% at both 24 and 36 months. Survival curves for 1‐KM and CIF showed simultaneous increases for each event, with 1‐KM consistently higher, reflecting differing approaches to competing risks.

This study highlights the impact of statistical method selection on clinical data interpretation and underscores the bias inherent in studies that fail to account for competing risks. Recognizing these differences is crucial for accurately assessing treatment outcomes.