This systematic review and meta-analysis looked at how well diagnostic prediction models work for cancer-related fatigue. The researchers combined data from 13 studies involving 444,447 cancer patients. They found that these models showed moderate-to-good discrimination, with a pooled area under the curve of 0.83.
However, the results were not consistent across all studies. There was substantial heterogeneity, meaning the models performed differently depending on the specific study. Most of the included studies also had a high risk of bias related to how the statistics were analyzed and reported.
Because of these issues and the wide variation in results, the pooled estimate should be interpreted with caution. The analysis did not report safety concerns because it focused on model accuracy rather than patient outcomes. Readers should understand that while these models show promise, they are not yet ready for routine clinical use without further validation.
This evidence provides a starting point for future research but does not change current practice. Clinicians should continue to use standard assessment tools until more reliable models are developed and tested in real-world settings.