---
title: Evidence Hierarchy (Evidence-Based Medicine)
type: concept
tags: [philosophy-of-science, evidence-based-medicine, rct, methodology]
sources: ["[[sources/2023-anjum-rocca-phi403-causation-in-science]]", "[[sources/2023-anjum-rocca-phi403-lecture-01-metascience]]", "[[sources/2023-anjum-rocca-phi403-lecture-12-explanatory-power-of-mechanisms]]"]
created: 2026-04-20
updated: 2026-04-20
---

# Evidence Hierarchy (Evidence-Based Medicine)

The explicit **ranking of evidence types** used in evidence-based medicine (EBM) and — by extension — evidence-based policy and evidence-based management. Higher-ranked evidence trumps lower-ranked, regardless of the causal question.

## The canonical pyramid (from highest to lowest)

1. **Meta-analyses / systematic reviews of RCTs**
2. **Randomised Controlled Trials (RCTs)** — the *gold standard*
3. **Comparative / observation studies** (cohort, case-control)
4. **Statistical correlation studies**
5. **Experimental methods** (bench science)
6. **Mechanistic reasoning / expert opinion**

The course presents this hierarchy in [[sources/2023-anjum-rocca-phi403-lecture-01-metascience]] (figure on the back of the handout) and again in [[sources/2023-anjum-rocca-phi403-lecture-12-explanatory-power-of-mechanisms]].

## The philosophical bias

The hierarchy's ranking **encodes a philosophical commitment** to a Humean / difference-making theory of causation: correlation-based evidence is treated as more secure than mechanistic evidence. This biases intervention against causal claims that are hard to operationalise via RCTs (complex interventions, rare effects, individual-level questions).

Andersen, Anjum & Rocca 2019 (*"Philosophical bias is the one bias that science cannot avoid"*, eLife) is the course's reference paper on this — see [[concepts/philosophical-bias]].

## The course's critique
- Mechanisms hold **more explanatory power** than regularities ([[sources/2023-anjum-rocca-phi403-lecture-12-explanatory-power-of-mechanisms]]).
- RCTs systematically exclude causally relevant information ([[concepts/rct-limitations]]).
- The **Russo-Williamson thesis** advocates pairing statistical and mechanistic evidence rather than ranking one above the other.
- Methodological pluralism ([[concepts/methodological-pluralism]]) rejects a fixed hierarchy in favour of context-sensitive evidence weighting.

## Related
[[concepts/rct-limitations]] · [[concepts/methodological-pluralism]] · [[concepts/causal-pluralism]] · [[concepts/mechanisms-in-science]] · [[concepts/philosophical-bias]]