Methodology

The ARCADE consensus process combines RAND/UCLA appropriateness scoring for research priority statements with a parallel Combined Evidence Dependency (CED) classification framework for diagnostic technologies.

Overview

The RAND/UCLA Appropriateness Method combines a systematic literature review with expert panel voting to determine whether clinical recommendations are appropriate for specific indications. ARCADE uses this method for research priority statements, while CED classification items are judged with predefined criteria rather than by the RAND median alone.

Statement Voting (RAND/UCLA)

Panelists rate research priority statements on a 1–9 scale:

1–3InappropriateThe statement is not appropriate for this indication. Harms likely outweigh benefits.
4–6UncertainEvidence is insufficient or equivocal. Expert opinion is divided.
7–9AppropriateThe statement is appropriate. Benefits outweigh harms based on available evidence.

RAND Consensus Classification

For each statement, the final classification is determined as follows:

  • Appropriate: Median score 7–9 without disagreement.
  • Inappropriate: Median score 1–3 without disagreement.
  • Uncertain: Median score 4–6, or disagreement present regardless of median.

Disagreement

Disagreement is present when at least one third of votes fall in the 1–3 range and at least one third fall in the 7–9 range simultaneously, regardless of the median. This follows the standard RAND/UCLA threshold.

CED Classification

CED items are classified separately from the RAND/UCLA statement score. Each technology is assigned a CED status (solved or unsolved) together with an ARCADE designation (CORE, SUPPORTED, or UNESTABLISHED).

CED Status Criteria

CED-solved

Requires all three criteria to be present:

  • Imminent embolic risk: The finding carries a substantial and universally recognised risk of systemic embolism or recurrent stroke.
  • Pathogenic certainty: The finding is definitively causal, rather than a bystander or nonspecific marker.
  • Established therapy: There is universal agreement that the finding warrants a specific intervention.

CED-unsolved

Assigned when any incomplete evidence-chain criterion is present:

  • Variable recurrence risk: The finding is associated with recurrence risk, but the magnitude of that risk is not uniform.
  • Contextual pathogenicity: The biological relevance of the finding shifts according to patient characteristics or competing stroke mechanisms.
  • Unproven treatment benefit: Finding-guided treatment has not yet been shown to reduce recurrence, has an uncertain risk-benefit ratio, remains controversial, or has no clearly established therapy.

ARCADE Designations

The CED status and the clinical designation answer different questions: CED status describes whether the detection-to-treatment-to-outcome evidence chain is complete, while the ARCADE designation describes the current clinical role of the test.

CORERoutinePerformed in most patients with ischaemic stroke or TIA as part of routine cardiac evaluation. Compatible with CED-solved or CED-unsolved.
SUPPORTEDSelectiveAppropriate when standard evaluation is insufficient, when a specific aetiological target is suspected, or when patient phenotype supports use. Compatible with CED-solved or CED-unsolved.
UNESTABLISHEDResearchPromising but insufficiently supported for guideline endorsement or routine implementation outside adequately designed prospective studies.

CED Research Implications

  • CED-solved | CORE: Optimise protocols, access equity, and implementation.
  • CED-unsolved | CORE: Highest priority is a dedicated RCT to establish whether acting on the finding reduces recurrent ischaemic events.
  • CED-solved | SUPPORTED: Address implementation barriers through training, standardisation, and healthcare system investment.
  • CED-unsolved | SUPPORTED: Prospective RCTs should establish clinical outcome benefit, with the aim of upgrading to CORE when warranted.
  • UNESTABLISHED: Diagnostic accuracy, management feasibility, and clinical outcome benefit require prospective validation before reclassification.

Round Structure

  1. Round 1: Research priority statements are rated independently. CED classification items are classified separately using the CED and ARCADE designation rubric.
  2. Review: Statements with uncertain classification or disagreement are reviewed by the steering committee and may be revised.
  3. Round 2: Statements that did not close under the Steering Committee's rule are re-rated, each shown with its Round 1 result and the panel feedback. Closed statements are shown for information only. (Original plan: research priorities and CED classification items are included again so both outputs complete the two-round consensus process.

How Panelists Contribute

  • Vote (required): Rate each statement on the 1–9 appropriateness scale.
  • Classify CED items (required when shown): Assign CED-solved or CED-unsolved and the corresponding ARCADE designation.
  • Highlight text (optional): Mark words/sentences tied to specific concern areas.
  • Thematic disagreement tags (optional): For lower scores, classify why you disagree (e.g., population, interpretation, scope).
  • General feedback (optional): Add narrative comments only when they are important for clarity, feasibility, scope, or interpretation.
  • Suggested revision (optional): Propose replacement wording only when a critical revision would materially improve the statement.

Privacy

Each panelist receives one voting token that is used across both rounds for completion tracking and within-round workflow access. The administrative contact list is kept separately from the analysis dataset, and final exports report aggregate results without names or email addresses.

After voting is complete, panelists may receive a separate results token that opens an individualized comparison page. This results token is for returning each panelist's own responses alongside the group distribution; it is separate from the voting token and is not used in pooled consensus reporting.

Reference

Fitch K, Bernstein SJ, Aguilar MD, et al. The RAND/UCLA Appropriateness Method User's Manual. Santa Monica, CA: RAND Corporation; 2001.