Survival Data Quality in Retrospective Oncology Publications

Survival endpoints are the cornerstone of cancer research. Their accuracy depends on the quality of the data sources, and many sources of survival data, such as electronic medical records (EMRs), provide incomplete data. It is unknown how often retrospective studies in oncology journals report their survival data sources or use high-quality sources.

Retrospective studies that included survival as an endpoint and were published in nine oncology journals in 2001, 2011, and 2021 were included. For each paper, endpoint(s) and outcome data source(s) were extracted. Data sources were categorised as high quality, uncertain quality, or unknown. The primary outcome was the proportion of studies utilising a high-quality survival data source.

Of 514 included papers, the most commonly reported endpoint was overall survival (OS, 80%), followed by progression-free survival (PFS) and disease-free survival; 14% did not report survival data source. All journals more often used uncertain-quality than high-quality data sources, but papers published in 2021 used more high-quality sources than those in 2001/2011 (26% vs 13%). In multivariable regression, radiation oncology papers were less likely to use high-quality data sources (p < 0.001), but there was no association of data quality with impact factor.

Most studies used sources of uncertain quality, and higher impact-factor journals did not publish a higher proportion of studies with high-quality data sources, though there was improvement over time. This suggests a need to improve compliance with reporting guidelines and encourage researchers to use high-quality survival data sources.