About

XeraTTE

A public interface to a living systematic review of target trial emulations that report a randomized benchmark. It holds 129 studies and 474 extracted comparisons, of which 279 enter the pooled agreement models.

Where the data comes from

Every number on this site is produced by the review’s reproducible pipeline and loaded here unchanged. The site performs no analysis of its own: pooled estimates, coverage, concordance and the sensitivity analyses are all fit in R and Stan and imported as results.

One value is computed at import time, and only because it is arithmetic rather than inference: whether two stored confidence intervals intersect, so comparisons can be filtered on it without recomputing bounds in SQL.

Pipeline

last synced 2026-09-03
Search
PubMed and OpenAlex, deduplicated into one corpus.
Screening
Title-and-abstract screening, then full-text eligibility assessment, with decisions and reasons recorded per record.
Extraction
One canonical workbook, with every manual fix recorded in a citable correction ledger applied before validation.
Benchmark gate
Each randomized benchmark checked against the trial’s primary report before the comparison enters the models.
Analysis
Bayesian multilevel models in Stan, with comparisons nested in studies; frequentist REML fit alongside as a cross-check.
Validation
Two gates fail the pipeline rather than emit a half-updated state: one checks every numeric claim in the manuscript against its generating table, the other checks that counts agree across artifacts.

What this site is not

It is not a verdict on individual studies. A comparison where the emulation and trial disagree is not evidence that the emulation was done badly: trials and emulations can answer subtly different questions, populations differ, and follow-up horizons rarely match exactly. That is why every comparison carries its PICO alignment grade rather than a pass or fail.

It is not a quality ranking. The transparency indicator counts which disclosures are present. A study with five out of five may still be poorly designed, and a study with two may be excellent.

It is not complete. The review is living; comparisons are added as new emulations are published and as benchmarks are verified.

Citing this database

Cite the underlying review for the findings, and this site for the data interface.

Xera Research. XeraTTE: a database of target trial emulations
benchmarked against randomized trials. Xera.ac OpenScience, 2026.
https://openscience.xera.ac/tte

Data are available under CC BY 4.0. Bulk downloads are on every browse page and through the API.

Xera OpenScience

XeraTTE is one of the tools in the Xera.ac OpenScience platform, which shares a design system and a technical stack across projects.