Resource Overlap and Reporting of Independence in Public Prostate Cancer Research
Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Importance
Public prostate cancer data may appear under different repository identifiers or releases while representing overlapping patients, specimens, or sample records. The implications for reported analytic independence require assessment at both resource and publication levels.
Objective
To characterize supported nonindependence among selected public prostate cancer resources and assess reported overlap control in publications using related resources.
Design
Cross-sectional methodological audit. Resource records were reconciled through September 2, 2026; recovery of retained sources and adjudication continued through September 8, 2026. Language-model extraction and targeted original- source checks informed deterministic classification. The authors manually verified all positive findings, including supplementary materials, in September 2026.
Setting
Public prostate cancer repositories and Europe PMC and PubMed retrieval, with a publication cutoff of August 15, 2026.
Participants
A nonrepresentative frame of 189 repository study records and 108 publication/version records contributing 230 publication–resource-pair records.
Exposures
Analytic use of both endpoints of a confirmed resource relation in the same linked claim.
Main Outcomes and Measures
Supported identity levels and resolvable patient-identifier counts; reported overlap control when linked analyses required independent records. Unresolved evidence was retained separately. Published results were not reanalyzed.
Results
The audit identified 101 supported resource pairs: 97 with direct patient, specimen, or sample-record identity, 1 with component containment, and 3 with study/source nonindependence. Among 108 publication/version records, 8 (6 journal articles and 2 preprint versions) used related resources in analyses requiring independence without reporting applicable overlap controls. Another 3 reported applicable controls, 16 retained reporting gaps, and 81 fell outside the qualifying use or relation criteria. One of the 8 had 45 shared sample identifiers in two validation sheets, corroborating sample-record reuse but not donor identity. No global independent-patient total or prevalence estimate was established.
Conclusions and Relevance
Resource nonindependence coexisted with absent claim-specific control reporting, with direct sample-record corroboration in one publication. The findings support provenance-aware use and explicit reporting, without establishing effects on published results.
