This cross-sectional methodological audit examined public prostate cancer data and the reporting of analytic independence when related resources were used. The objective was to characterize supported nonindependence among selected public prostate cancer resources and to assess whether publications using related resources reported applicable overlap controls when analyses required independent records.
The audit reconciled resource records through September 2, 2026, with recovery of retained sources and adjudication continuing through September 8, 2026. Publication retrieval used Europe PMC and PubMed, with a publication cutoff date of August 15, 2026. The authors applied an approach combining language-model extraction and targeted original-source checks to inform deterministic classification. All positive findings and relevant supplementary materials were manually verified in September 2026.
The nonrepresentative frame included 189 repository study records and 108 publication/version records, which together contributed 230 publication-resource-pair records for evaluation. Exact resource identifiers and source records are provided in the supplementary tables accompanying the audit.
Deterministic classification was informed by automated extraction and manual source checks. The audit distinguished types of supported resource relations, including direct identity at the patient, specimen, or sample-record level; component containment; and study/source nonindependence. When evidence was unresolved, it was retained separately rather than forced into a supported category.
The audit identified 101 supported resource pairs among the evaluated records. Of these: 97 pairs showed direct patient, specimen, or sample-record identity; 1 pair demonstrated component containment; and 3 pairs reflected study/source nonindependence. These supported relations indicate that distinct repository identifiers or releases can represent overlapping records drawn from the same donors, specimens, or samples.
Among the 108 publication/version records assessed, the audit classified publications by whether they used related resources in analyses that required independent records and whether they reported applicable overlap controls. Results were:
The audit did not reanalyze published results to measure any impact of resource nonindependence on study conclusions.
In one of the 8 publications flagged for nonreporting, the authors identified 45 shared sample identifiers appearing in two validation sheets. This finding corroborated sample-record reuse across resources in that specific case; however, the audit notes that this corroboration did not by itself establish donor identity for those samples.
The audit retained unresolved evidence separately rather than making unsupported assertions. It did not establish a global independent-patient total or estimate the prevalence of overlap across all public prostate cancer resources. Published results from the assessed articles were not reanalyzed for effects attributable to resource overlap. The frame was nonrepresentative and limited to specified repositories and publications up to the stated cutoffs.
The authors report that resource nonindependence coexisted with absent claim-specific reporting of overlap controls in a subset of publications using related resources. The findings support use practices that are provenance-aware and recommend explicit reporting of overlap-control methods when analyses require independent records. The audit does not assert effects on published results but highlights gaps in reporting that could affect interpretations of analytic independence.
The supplementary workbook contains the derived resource tables and publication-level audit summaries. The authors permit public sharing of the derived audit data and analysis code; complete versioned files and SHA-256 manifests are retained locally pending public repository deposition. No controlled-access individual-level dbGaP files were used. The authors declared no competing interests and stated that the audit of public metadata and publications did not require institutional review board approval.