Does it read scanned PDFs?

No; a text layer is required (scans trigger a warning).

What happens when the two readings disagree?

The cell is left blank on purpose and listed in Conflicts for you to settle on the page.

How many PDFs per run?

Up to 20 PDFs and 40 outcomes (15 on Deep); failed articles are not charged.

Extraction Data

The double data extraction Cochrane asks for — in minutes per PDF

200 hours of spreadsheet work; and single extraction lets sign errors and SD-that-is-SE slip through.

Run this tool See all tools

You give

The study PDFs (up to the page limit of the tier) and the list of outcomes, one per line.

You get

One workbook with 7 sheets: Reconciled, Arm A, Arm B, Conflicts, Validation, Studies and Legend — with source, confidence and quote for every value.

How it works

# Extraction
$ scopetools run dual-extraction
Model-assisted, with deterministic checks.
Two arms: reading A starts from the text, reading B from tables and figures. Values that agree are accepted; disagreements are left blank on purpose and listed in a Conflicts sheet for you to settle on the page. Executable checks flag SD that looks like SE, events greater than n, unit mismatches and conversions (Wan, Hozo) with the formula used.

Two independent arms (A reads the text first, B the tables first); what agrees is accepted, what diverges goes to a Conflicts sheet for you to settle on the page. Executable checks flag SD≈SE, events > n, unit mismatches and Wan/Hozo conversions with the formula used. Seven sheets plus a raw .json that can become Supplementary Data. Only extracted articles are charged.

What it costs

Standard tier: 15 credits per article — 8 PDFs ≈ 120 credits (US$6). Light 5 · Deep 35 (the tier selector shows the total before you run).

Charged only when it delivers — a failed run costs nothing.

Credits are shared across every MetaScope tool — buy once, spend anywhere.

Limits

Scanned PDFs without a text layer cannot be read. Values marked low confidence must be checked on the page.

Frequently asked questions

Does it read scanned PDFs?: No; a text layer is required (scans trigger a warning).
What happens when the two readings disagree?: The cell is left blank on purpose and listed in Conflicts for you to settle on the page.
How many PDFs per run?: Up to 20 PDFs and 40 outcomes (15 on Deep); failed articles are not charged.

Extracted? Run the Data check for meta-analysis before pooling — 3 credits.