Systematic Review of Prostate Cancer Prediction Models with Meta-Analysis Results

A systematic review and meta-analysis by Emmanuel Nsedu Israel, registered in PROSPERO (CRD42025611480), examined artificial intelligence models for predicting prostate cancer outcomes. The analysis, last updated in 2026, pooled data from 85 studies to evaluate model performance across five clinical endpoints. The results show pooled area under the curve (AUC) metrics ranging from 0.792 to 0.845.

Use Cases

• Benchmarking new prostate cancer prediction models based on reported AUC metrics for overall survival.
• Assessing methodological quality of clinical prediction models based on the Prediction Model Quality Score (PMQS).
• Analyzing publication bias in medical AI research based on funnel plot and Egger's regression test results.
• Comparing model performance across clinical endpoints like progression-free survival and treatment response.
• Informing clinical translation efforts based on findings regarding the need for standardized reporting and external validation.

Strengths

• Includes 85 studies in the meta-analysis, providing a substantial evidence base.
• Reports pooled AUC results for five distinct clinical prediction endpoints (OS, PFS, RDM, TR, TQL).
• Follows PRISMA criteria and has a registered protocol, indicating a structured review methodology.
• Assesses publication bias using multiple statistical methods (funnel plots, Egger's test, trim-and-fill).

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The dataset is very small (29.3 KB), suggesting it contains summary statistics rather than raw model data.

Provenance

Source

Emmanuel Nsedu Israel via figshare.

Collection Method

Systematic literature review of PubMed and Scopus, followed by random-effects meta-analysis.

Freshness

Last updated 2026-05-01 05:28:15; freshness should be verified.