Iron Status and Cardiac Conduction in Humans: Aggregate Data and Analysis Code for a Systematic Review
Description
This dataset accompanies a systematic review of iron status and cardiac conduction in humans. It provides aggregate data extracted from published reports and Python code for reproducing individual-study estimates, sensitivity analyses, exploratory syntheses, and figures. The review examines iron deficiency, iron-deficiency anemia, systemic iron overload, and myocardial iron burden in relation to PR interval, QRS duration, atrioventricular block, and supporting electrophysiological measures. Evidence is organized by population, iron exposure, clinical setting, outcome, and comparison type to distinguish disease-status associations, within-disease iron comparisons, and treatment-related changes. The package includes: • A bibliography and clinical characteristics of 44 quantitative-candidate reports. • Extracted statistical inputs and 52 display results from 15 reports, comprising 47 results with usable comparative confidence intervals, three conditional results, and two descriptive results. • Six paired-correlation sensitivity records, 13 supplementary evidence records, and four explicitly post hoc exploratory pooling scenarios. • Preliminary outcome-specific risk-of-bias and evidence-certainty assessments. • Offline Python scripts, dependency specifications, data documentation, numerical verification outputs, and editable SVG figures. Analyses include mean differences, exact odds ratios, correlation estimates, and exploratory random-effects models. Conditional assumptions, shared cohorts, overlapping comparisons, and source locations are retained to support interpretation and reuse. Report and result counts should not be interpreted as counts of independent studies. Exploratory pooled estimates are distinguished from the protocol-defined primary synthesis. Data are supplied in JSON and CSV formats. The scripts reproduce the numerical results and generate 52 individual-result SVGs and four exploratory SVGs. These figures support numerical reproducibility and are separate from the manuscript’s composite figures. English and Chinese documentation explains the file structure, methods, and local execution steps. Version 0.1.0 is a manuscript-development release. Final study eligibility and human confirmation of AI-assisted evidence assessments remain in progress. The package contains aggregate published-study data; no individual participant data or publisher article full texts are included. Related protocol registration: https://doi.org/10.17605/OSF.IO/RPBU8
Files
Steps to reproduce
1. Download Iron_Conduction_Data_Code_v0.1.0_UPLOAD.zip and extract it to a local folder. 2. Install Python 3.12. Open a terminal in the extracted folder containing requirements.txt. 3. Create a virtual environment and install the required packages. Windows PowerShell: py -3.12 -m venv .venv .\.venv\Scripts\python.exe -m pip install -r requirements.txt macOS/Linux: python3.12 -m venv .venv .venv/bin/python -m pip install -r requirements.txt 4. Run the analysis. Windows PowerShell: .\.venv\Scripts\python.exe scripts/rerun_analysis.py macOS/Linux: .venv/bin/python scripts/rerun_analysis.py 5. Inspect the regenerated files in outputs/. These include single-study effect estimates, paired-correlation sensitivity analyses, supplementary calculations, four post-hoc exploratory syntheses, numerical validation results, and editable SVG figures. 6. Review outputs/validation/ to check agreement between recalculated estimates and the supplied reference results. Consult README.md and docs/METHODS.md for analytical assumptions and interpretation. The plotting workflow generates 52 individual-result SVG figures and four exploratory-synthesis SVG figures. Figures use an 80 mm width and 7-point text, with Arial requested. Install Arial locally to reproduce the intended typography. These figures are separate from the manuscript’s assembled multipanel figures. An internet connection is required to install dependencies; subsequent analysis runs use the supplied local data. No individual participant data or publisher full-text files are required.