Evidence Level Classification
| Evidence Level | Description | ME/CFS Status | Reliability |
|---|---|---|---|
| Established | Replicated across multiple independent cohorts with consistent methodology | Documented in ME/CFS (multiple studies, consistent findings) | Highest confidence |
| Probable | Replicated in at least two cohorts with some methodological differences | Strong suggestive evidence in ME/CFS but requires replication | Moderate-high confidence |
| Preliminary | Single study or limited replication with consistent findings | Emerging evidence in ME/CFS, requires validation | Moderate confidence |
| Theoretical | Based on mechanistic inference from related conditions | Hypothesized but not yet tested in ME/CFS | Low confidence |
| Speculative | Limited mechanistic basis, no direct testing in ME/CFS | Highly uncertain | Lowest confidence |
This chapter synthesizes multi-system pathophysiology documented in Chapters 6–12 into comprehensive models of ME/CFS. The core challenge in understanding ME/CFS is that individual system abnormalities—mitochondrial dysfunction, immune dysregulation, neurological impairment, cardiovascular deficits, endocrine disruption, gut dysbiosis—co-occur in patterns that suggest coordinated, self-reinforcing mechanisms rather than independent parallel pathology.
The paper that follows — across this chapter and the preceding chapters — contains a large number of speculative hypotheses, open questions, and mechanistic proposals (~70+ constructs). Many are built on inferential chains where multiple uncertain links must hold simultaneously for the conclusion to follow. Some explain the same observation through multiple competing mechanisms without resolution. Certainty values are assigned per individual construct, which can obscure the multiplicative uncertainty of dependency chains where speculation B depends on speculation A — the composite probability may be substantially lower than either individual certainty. This is a deliberate feature of a hypothesis-generating work: the intent is to map the space of plausible mechanisms and expose them to falsification, not to assert that all are simultaneously correct. However, the density of speculation should not be mistaken for density of evidence. Each claim carries its own certainty value; readers should treat these as conditional probabilities (the claim if the upstream premises hold), not as marginal probabilities (the claim unconditionally). Where multiple independent mechanisms are proposed for the same observation, the most parsimonious candidate is not identified — this is a limitation the reader should be aware of. The empirically grounded content (the 10-step energy chain framework, the research-gap tables G01–G39, replicated clinical findings with replication status noted) provides the evidential anchor; the speculative layer is scaffolding for hypothesis generation, not established fact.
A related caveat concerns the evidential standards applied to different types of claims throughout this paper. Subjective outcome measures (patient-reported fatigue, symptom severity, quality of life) are used as valid evidence when they support biological hypotheses but are critiqued as unreliable when they support interventions the paper argues against (notably in the PACE trial). This asymmetry is a genuine limitation of the paper’s methodology, not a principled position: unblinding bias affects all unblinded studies regardless of whether the intervention is behavioural or biological, and the paper’s standard should be consistently applied. The reader should note that the same self-report instruments cited as evidence for biological mechanisms elsewhere carry the same epistemological limitations when used in treatment trials. The chapter’s historical analysis (Section Mechanistic Convergence vs. Divergence Across Post-Infectious Conditions) attempts to address this asymmetry by presenting the strongest available form of the positions being critiqued before explaining why the critique survives the steelman, but residual asymmetry may remain and should be weighed by the reader.