Case Definition Heterogeneity
The single most pervasive confound in ME/CFS research is not measurement error or sample size — it is that researchers have been studying different diseases under the same name. The choice of diagnostic criteria determines who enters a study, and each set of criteria selects a population with different biological characteristics, symptom profiles, and illness severity. Findings that replicate in CCC/ICC cohorts frequently fail to replicate in Fukuda cohorts — and vice versa — because the study populations are biologically different.
1 Diagnostic Criteria Comparison
The four major case definition systems select fundamentally different patient populations:
| Criterion | Fukuda (1994) | CCC (2003) | ICC (2011) | IOM/SEID (2015) |
|---|---|---|---|---|
| PEM required | No | Yes | Yes | Yes |
| Orthostatic intolerance | No | Yes | Yes | Yes |
| Cognitive impairment | Optional | Required | Required | Yes |
| Minimum symptoms | 4 of 8 | 7 of 10+ | Meet multiple clusters | Mild-moderate-severe |
| Fatigue duration | ≥6 months | ≥6 months | ≥6 months | ≥6 months + substantial reduction |
| Functional impairment | Substantial | Substantial | ≥50% pre-illness | Substantial |
The practical consequence: Fukuda criteria capture 2–2.5× more patients than CCC criteria when applied to the same population (Jason et al. 2015). When applied to a large US registry (n=2,143), prevalence rates varied from 0.21% (ICC) to 0.84% (Fukuda) — a 4-fold difference (Jason et al. 2020). When applied to a CDC survey (n=2,762), Fukuda yielded 16% “CFS-like” while the CDC empirical definition yielded 2.5% — a 6-fold difference (Brimmer et al. 2016).
2 Biological Divergence by Criteria
The UK ME/CFS Biobank provided the first systematic evidence that criteria-selected groups are biologically distinct, not just clinically distinct. CCC-selected patients had more severe symptoms, lower quality of life, and different immune and metabolic profiles than Fukuda-selected patients from the same biobank (Nacul et al. 2019) (Nacul et al. 2017).
Most strikingly, 25% of patients meeting Fukuda criteria do not meet CCC criteria (Strand et al. 2019). These “Fukuda-only” patients have idiopathic chronic fatigue under CCC — not ME/CFS. Their biology is more similar to healthy controls than to CCC-positive patients (Nacul et al. 2017). This means roughly one quarter of findings from Fukuda-based studies may reflect non-ME/CFS biology.
The DecodeME GWAS (n \(>\) 15,000) (DecodeME Consortium 2025) is the first large-scale study to systematically compare genetic architecture across diagnostic criteria. Preliminary results indicate that genetic signals differ when analyzed under different criteria — providing molecular-level validation that criteria choice determines which disease you are studying.
Argument for standardization: The IOM/SEID criteria (2015) represent the closest the field has come to institutional consensus, were developed through a rigorous systematic review process, and require PEM — the cardinal feature of ME/CFS. Adopting SEID as the universal standard would eliminate criteria heterogeneity as a confound and make cross-study comparisons possible.
Argument against standardization: SEID criteria risk the same problem as CCC/ICC — restricting to more severely ill patients may exclude mild cases and racial/SES minority patients who present differently (Haney, Dimmock, and Jason 2024). A standardized but biased entry criterion trades one confound for another.
Consequence: Until the field agrees on a standard case definition, readers must check which criteria each study used — and should treat findings from Oxford/Fukuda studies as preliminary until replicated in a PEM-required cohort. Research that stratifies results by criteria (showing what holds across definitions vs what is criteria-specific) is more valuable than research that reports results from a single criteria-defined cohort.
3 Recommendations for Readers
When evaluating published ME/CFS research, the single most important metadata field is the diagnostic criteria used. A study reporting “elevated cytokine X in ME/CFS” requires checking whether the cohort was Fukuda-defined (likely ~25% idiopathic chronic fatigue), Oxford-defined (likely depression/deconditioning), or CCC/ICC-defined (more likely genuine ME/CFS). The evidentiary weight of a finding scales with criteria specificity.
Consequence: The “replication crisis” in ME/CFS is largely a criteria problem — not a failure of science but a predictable consequence of studying different diseases under the same name. Standardizing on PEM-required criteria would resolve more replication failures than any improvement in study design alone.
4 A Formal Test of the Central Claim: Criteria-Dose-Response Meta-Analysis
The chapter’s core qualitative argument — that diagnostic criteria stringency determines which biology is studied — has never been quantified across the full ME/CFS literature. This can be tested: (a) score every set of diagnostic criteria by stringency (PEM required, orthostatic intolerance required, cognitive impairment required, minimum symptom count, functional impairment threshold), producing a “criteria stringency index” (CSI) from ~1 (Oxford) to ~8 (ICC), (b) meta-analyze every published finding separately by criteria group, computing the effect size for each CSI tier, (c) regress effect size against CSI.
For disease-specific biomarkers (NK cell function, lactate, Day-2 CPET decline), the slope should be positive — stricter criteria → larger, more specific effect sizes. For non-specific markers (fatigue, depression, deconditioning), the slope should be zero or negative. A significant CSI×outcome interaction would confirm that criteria stringency predicts what you find, converting the chapter’s qualitative narrative into a single quantifiable meta-regression parameter.
Origin: brainstorm.