The EDS diagnosis problem solves itself

Hypermobile Ehlers-Danlos remains the only EDS subtype in the 2017 international classification without an identified causal gene. The condition has been recognized clinically since the early twentieth century. Candidate genes have been investigated. The diagnostic criteria have been refined twice. The fundamental scientific problem of identifying the gene or genes responsible for hEDS has not been solved.

The political problem of who counts as having hEDS has been argued continuously for two decades. The 2017 criteria narrowed the diagnostic boundary relative to the previous Villefranche framework, with the explicit goal of producing a more clinically homogeneous research population. Many people previously diagnosed with EDS no longer met the revised criteria. Hypermobility Spectrum Disorder was introduced as a catch-all category. Many clinicians and many affected individuals consider HSD a demotion rather than a diagnosis. The 2026 revision under discussion at the time of this writing proposes additional adjustments. None of the adjustments will produce a confirmatory genetic test, because the underlying gene or genes have not been found.

Why the gene-hunt has been hard

The conventional approach to identifying a gene for a condition is to recruit clinically diagnosed cases, sequence their genomes, look for variants that segregate with the condition more often than chance would predict, and follow up the candidates with functional studies. The approach has worked for the other thirteen EDS subtypes and for hundreds of other rare diseases.

The approach has not worked for hEDS for two reasons. The first is that the clinical diagnosis is not stable. The 2017 criteria identify a population that is not the same population the Villefranche criteria identified, which is not the same population a clinical diagnosis from the 1990s identified. The genetic heterogeneity within the population may be partly an artifact of the changing diagnostic boundary. Pooling cohorts across diagnostic eras to gain statistical power introduces noise that masks any signal that might exist.

The second is the hypothesis that hEDS is genetically heterogeneous. If the clinical category captures multiple underlying conditions, each with its own genetic basis, the gene-hunt looking for a single shared cause cannot succeed because there is no single shared cause. Detecting multiple weak signals within a heterogeneous population requires either much larger sample sizes than current cohorts or a different analytic approach that stratifies the population before the gene-hunt begins.

The phenotype-cloud approach is the second of those alternatives. Instead of starting with the clinical category and looking for shared genetics, the approach starts with structured longitudinal phenotypic data and clusters patients by symptom trajectories, comorbidity patterns, and biomarker signatures. The clusters that emerge are biological subgroups that may correspond to distinct underlying conditions. Each subgroup, being more genetically homogeneous than the broader hEDS category, supports gene discovery with much more statistical power than the unstratified population.

What the data-driven approach requires

The phenotype cloud requires structured longitudinal data from a large cohort. The cohort needs to span the diagnostic boundaries that the political fights have been about. People with hEDS, with HSD, with chronic pain plus hypermobility who never received a formal diagnosis, with autonomic features clustered around connective tissue laxity, with the broader phenotypic territory the 2017 criteria narrowed away from. Limiting the cohort to people who meet 2017 hEDS criteria would replicate the diagnostic-homogeneity problem at the data layer.

The structured fields required are the same fields a careful clinical research protocol would specify. Joint hypermobility scores at each visit. A standardized symptom inventory. Comorbidity status for the recognized associated conditions (POTS, MCAS, gastroparesis, chronic pain, the rest). Treatment response data for the medications and interventions the patient receives. Wearable device data where available. Patient-reported outcomes on standardized instruments. The list is long and the data per visit is voluminous. The infrastructure to capture it is the infrastructure question that has not been solved.

The infrastructure is achievable with current technology. Ehlers-Danlos Society registries, condition-specific natural history studies, and patient-led data collection efforts each capture pieces of what would be needed. None covers the entire phenotypic territory at the depth and longitudinal duration the cloud approach requires. Building the missing infrastructure is the project. The technology to do it is mature.

What the data could resolve

The hypothesis that hEDS captures multiple underlying conditions is a hypothesis that data of sufficient scale and structure can test. The test does not require any judgment from a criteria committee. The test is a clustering analysis of the phenotypic data, followed by gene-discovery work within each emergent cluster. If the clusters correspond to distinct genetic causes, the field has identified the causes of hEDS. If the clusters do not correspond to distinct genetic causes, the field has learned that the heterogeneity is not genetic at the level the analysis can detect, and the next round of investigation has a clearer baseline.

Either outcome shifts the locus of authority. Currently the locus is the criteria committee that meets every several years to revise the boundaries. The committee deliberates from the same data the field has had for decades, augmented incrementally by case series and small cohort studies. Any committee revising criteria from that data is making a structural judgment, not a data-driven one, because the data of the kind needed to make a data-driven judgment does not exist at the required scale.

The data-driven shift would put the criteria committee out of the central role and into a derivative role. The committee would interpret what the data has shown rather than deciding what the data has been allowed to show. The political fights become moot because the question changes from "where do we draw the line" to "what does the clustering tell us about the biology." The argument that has been corrosive to the EDS community for years gets answered by data the community itself contributes.

The data does not exist yet. Building it is the project that closes the diagnostic question that committees have not closed.