# Reproduce the study

This is an aggregate-data evidence synthesis, using factual extractions from the linked primary trial reports. No patient-level records are included. The code does not download or execute material from the cited authors.

The reference journal harness recognizes `env/requirements.txt` and runs `sh code/run` in Python 3.12. No network is needed while computing. It recreates the main JSON, CSV and PNG files under `results/`. The manuscript binds displayed values to those JSON results.

For a manual Docker reproduction from the extracted bundle directory:

```sh
docker build -t stable-coronary-study -f env/reproduce.Dockerfile .
docker run --rm --network none --read-only --cap-drop ALL --security-opt no-new-privileges --memory 2g --cpus 2 --pids-limit 256 --tmpfs /tmp:rw,size=128m -v "$PWD:/study:ro" -v "$PWD/results:/study/results:rw" stable-coronary-study
```

Building the environment requires package downloads; the calculation itself runs without network access. Package versions are pinned in `env/requirements.txt`.

`data/endpoints.csv` holds reported event counts with their original denominator types. It does not contain estimated individual outcomes. `data/hr_effects.csv` preserves published hazard ratios separately. Screening decisions, exact searches, source locators and extraction caveats are in `data/`. The analysis plan discloses the prior outcome information read before analysis and the extraction-stage amendment.

Do not interpret a ratio of the reported fractions as a censoring-adjusted fixed-horizon risk. In particular, HREVS uses complete cases and FAME-3 uses randomized subgroup denominators with incomplete follow-up. The optional R validation in `env/validation.md` checks the numerical implementation, not the validity of all clinical inference. Its archived CSVs and validation JSON are records of the additional check; the default Python run regenerates the main scientific results and figures.

The manuscript is a secondary analysis, not a newly conducted clinical trial. Data and code are provided for inspection, correction and independent reanalysis. Source papers retain their own rights; full-text copies and source-table images are excluded from this bundle.
