coreSR.AIcoreSR
Usage Guide

Dose-Response Meta-Analysis

How dose-response works

1

Create a project and outcome

  1. 1Click "New project" and name the analysis
  2. 2Create an outcome: name the endpoint, pick binary or continuous, and choose the effect measure
  3. 3Set the reference dose the curve is expressed against
Tip: A project groups outcomes that share a treatment family. Each outcome carries its own dataset, model, and GRADE assessment.
2

Prepare your data

  1. 1Always required: a study identifier, and a numeric dose for each arm
  2. 2Binary outcomes: events and total per arm, or effect and standard error per contrast
  3. 3Continuous outcomes: mean, SD and n per arm, or effect and standard error per contrast
  4. 4Optional but recommended: a per-study risk-of-bias column, which feeds the risk-of-bias GRADE row
3

Upload and map columns

  1. 1Upload a CSV or Excel file and pick the data format — wide, long, or dose-response
  2. 2Map the study, dose, and outcome columns; AI detection pre-fills the suggestions
  3. 3Optionally point the PICO indirectness attributes at their columns so the GRADE indirectness engine can read them
  4. 4Check the data preview, then continue
Tip: Columns holding numeric values stored as text are flagged before they can break the fit — clean them in place from the mapping dialog.
4

Fit the models

  1. 1The app fits both a linear and a restricted cubic spline model via dosresmeta and overlays them on one curve with confidence intervals
  2. 2Enter a dose in "Look up dose" to read the estimated effect at that dose with its 95% CI
  3. 3Model Details gives the slope, standard error, p-value, AIC, and the heterogeneity statistics, alongside the assumptions each model makes
  4. 4Model Comparison reports both AICs and a Wald non-linearity test, and recommends which fit to report
Tip: The recommendation is keyed on the Wald non-linearity test; ΔAIC is used only as a tiebreaker.
5

Grade and export

  1. 1The GRADE tab pools by risk-of-bias level, tests whether the estimates differ across strata, and shows the weighted evidence distribution behind the certainty rating
  2. 2Review each domain and revise any rating before saving
  3. 3Outputs renders the GRADE decisions table or a Summary of Findings, in your choice of style, and exports it as PDF, PNG, HTML, or DOCX
Tip: The submission estimand is the persisted per-unit linear trend versus the reference dose. Dose-specific curve predictions on the Results tab are exploratory and are not submission SoF estimands.
coreSR.AICoreSR

AI-powered evidence synthesis