# Data — DMCausalTriangulate-Kor

This directory holds curated, **publicly sourced** effect estimates and
two small ontologies. No private patient data, no live API calls.

## Files

- `drug_ontology.csv` — ATC class · drug · canonical target · mechanism (manually curated).
- `outcome_ontology.csv` — outcome · MeSH descriptor · category · desirable direction.
- `effects_sample.csv` — 5-design effect estimates for representative T2DM
  drug × outcome pairs. Each row is one design × pair. 52 rows total.

## Schema — `effects_sample.csv`

| column | description |
| --- | --- |
| `drug_class` | one of SGLT2i, GLP-1RA, tirzepatide, DPP-4i, metformin, ... |
| `drug` | specific molecule, or `class` if the estimate is class-level |
| `atc_code` | ATC code (drug- or class-level) |
| `outcome` | human-readable outcome name |
| `outcome_mesh` | MeSH descriptor ID |
| `outcome_category` | macro-CV / microvascular / safety / oncology / ... |
| `design` | RCT, observational, target-MR, ex vivo, within-subject |
| `effect_estimate` | hazard ratio / odds ratio / risk ratio (point estimate) |
| `ci_low`, `ci_high` | 95% confidence interval |
| `sample_size` | n (or `NA`) |
| `follow_up_years` | median follow-up |
| `population` | study population string |
| `source_citation` | author year journal short form |
| `source_url` | persistent URL (DOI / NEJM / journal site / ClinicalTrials.gov) |

## Sources

The sample CSV draws on publicly available estimates from:

- **RCTs** — CVOT / DKD / HF / MASH publications indexed via PubMed and the
  ClinicalTrials.gov results database (EMPA-REG OUTCOME, CANVAS, DECLARE,
  CREDENCE, DAPA-CKD, LEADER, SUSTAIN-6, REWIND, SAVOR-TIMI 53, EXAMINE,
  TECOS, SURMOUNT-1, STEP 1, SYNERGY-NASH, etc.)
- **Observational meta-analyses** — published in Diabetes Care, BMJ, NEJM,
  Nature Medicine (EMPRISE, FAERS, NHIS-NSC Korea, TriNetX, CPRD).
- **Mendelian randomization** — drug-target MR studies using cis-pQTL of
  SLC5A2 / GLP1R / DPP4 (Schmidt 2017 *Lancet Diab Endo*; Scott 2016 *Sci
  Transl Med*; Katzmann 2021 *Diabetes Care*; Walker 2023 *Diabetes Care*).
- **Ex vivo / animal** — selected systematic reviews via SyRF / CAMARADES
  and individual mechanism papers (Uthman 2018 *Cardiovasc Res*; Holscher
  2018 *Alzheimers Dement*; Bjerre Knudsen 2010 *Endocrinology*).

## Disclaimers

- Numbers are **summary statistics extracted from publicly published
  estimates** and may be rounded or simplified for illustration.
- The CSV is intended as a **starter set** — researchers should curate
  their own evidence table for any production use.
- This data does **not** constitute clinical advice.

## Adding rows

To extend, append rows in the same schema. The tool auto-detects designs
and rebuilds the grid; no schema migration needed.