The first constraint on any analytical project in this field is sample size. Classic CJD affects about 1 to 2 people per million per year in many countries, including the United States. Median survival from symptom onset is about 4 to 5 months. A hospital or registry sees few cases, and each case has a short window in which data can be collected.
What has the field already tried to measure?
Seven completed drug studies are summarised by cureffi.org. They enrolled 360 people in total, between 7 and 121 per study, and only three were randomised. Six of the seven used survival as the main endpoint. The exception was a German flupirtine trial of 28 people that used a cognitive scale, and it did not lead to a standard treatment. Apart from that trial, the results were negative or were positive results that the investigators attributed to confounders. The summary covers symptomatic people only, mostly with sporadic CJD.
Survival is a blunt endpoint in a disease where it is short and varies between people. Better ways to measure progression in small groups are an open problem, covered from the drug developer side on the pharma biomarkers and endpoints page. See also open questions.
Which problems have a checkable answer?
Diagnosis is the clearest. The CDC describes a definite diagnosis as requiring brain tissue, obtained by biopsy or at autopsy. Probable diagnosis relies on clinical features together with tests such as EEG, MRI, a CSF 14-3-3 assay or RT-QuIC. A project that classifies cases can be checked against neuropathology, which makes confirmed cases a useful reference standard. The number of confirmed cases available to any one group is the limit.
Other candidate problems are earlier recognition of rapidly progressive dementia in routine records, and helping clinicians reach the right test faster. Whether a given method works in practice is not established by the sources used here, and this page does not claim it.
What is not known?
- How well a model built on one country’s cases transfers to another country’s, because surveillance rules and test availability differ.
- Whether clinical records in routine care contain enough signal for early detection. No source opened for this page reports a validated tool.
- How many confirmed cases with matched imaging, fluid test and genetic data can be shared across institutions. See data availability.
Treatment remains supportive care only, and no therapy has been shown to stop progression. Any AI product that addresses people affected by prion disease should say so and should not interpret a person’s symptoms, test results or genetic status. See responsible use, the AI in prion research page, and the AI partners overview.