DeepSeek-R1 proposes FMF; MEFV testing confirms the diagnosis

AI system

System: DeepSeek-R1
Model: DeepSeek-R1-0528.
Release: DeepSeek-R1-0528.
Application developer: DeepSeek
Role: The model suggested a possible diagnosis and further investigations during a conversation.
Configuration: Not assessed
Inputs and outputs: Not assessed
Autonomy: Not assessed
Workflow: A patient with a long-standing, characteristic but undiagnosed phenotype describing it in detail to a reasoning LLM over multiple turns. Access to genetic testing for the suspected condition.
Human oversight: Not assessed
Relation to AI: The large language model (LLM) generated the FMF hypothesis and the suggested work-up over a multi-turn dialogue. Confirmation came from MEFV genotyping and clinical response to colchicine. The authors stress that the model repeatedly advised specialist supervision. The self-directed genetic testing and self-medication were the patient's own actions.
AI contribution: The large language model (LLM) generated the FMF hypothesis and the suggested work-up over a multi-turn dialogue. Confirmation came from MEFV genotyping and clinical response to colchicine. The authors stress that the model repeatedly advised specialist supervision. The self-directed genetic testing and self-medication were the patient's own actions.

Benefits

Outcome identifiers: Not assessed
Benefit: colchicine stopped attacks after ~30 years of symptoms
Category: Improved Physical Health
Status: Not assessed
Beneficiary groups: People.
Beneficiaries: Not assessed
Reach: Not assessed
Distribution and access: Not assessed
Duration: Not assessed
Expected?: Yes
Costs and harms: Not assessed

Measurement

Studies: Not assessed
Study design: Not assessed
Measurement: Not assessed
Setting and sample: Not assessed
Measures: colchicine stopped attacks after ~30 years of symptoms
Comparison: almost three decades of conventional investigations
Results: Reference: Despite almost 3 decades of symptoms, extensive investigations, including contrast-enhanced abdominal CT and MRI, gastroscopy, colonoscopy, and urological assessment, failed to establish a diagnosis. Result: the diagnosis of FMF was confirmed.
Additional value from AI: Not assessed
Limitations: Not assessed
Denominators: Not assessed
Materials for replication: Not assessed

Money

Kind of value: Not assessed
Amount: Not assessed
Currency and prices: Not assessed
Period and unit: Not assessed
Realization: Not assessed
Recipients and payers: Not assessed
Gross or net: Not assessed
Included costs: Not assessed
Comparison: Not assessed
Method: Not assessed
Return and payback: Not assessed

People and places

Participants: Developer: DeepSeek.
Jurisdictions: Benefit: Hamburg, Germany. Developer: China.
Ownership and control: Not assessed
Dependencies: Not assessed
Cross-border connections: Not assessed
Geopolitical implications: Not assessed

Follow-up

Conditions for replication: Genetic or clinical confirmation by a specialist before the diagnosis is accepted. Treatment initiated under medical supervision (the case documents hepatotoxicity from self-dosing).
Replication attempts: Not assessed
Later findings: 29 September 2026. After a colchicine pause for raised liver enzymes and resumption at a lower dose, the patient remained attack-free under clinic follow-up as of publication.
Availability: Not assessed
Institutional decision: Not assessed
Grounds: Not assessed
Action and result: Not assessed
Success criterion: Not assessed

Evidence

Source numbers: AIBD-0061-source-0001–0002
Citations: Numbered references below.
Language: Not assessed
Source relationships: Independent
Original evidence: Not assessed
Access and versions: Not assessed
Supported findings: Not assessed
Evidence links: Not assessed
Evidence strength: Not assessed
Missing evidence: Not assessed

Grades

Assessed benefits: People.
Depth: People: 3.
Reasons: People: Inflammatory attacks that had recurred since early childhood ceased on colchicine after a genetically confirmed diagnosis. The benefit is life-changing, tempered by a documented transaminase rise from self-dosing before medical supervision.
Observed or intended: Not assessed
Rubric: Not assessed
Assessor and date: Not assessed
Alternative assessments: Not assessed
Fill: Independently studied

Entry details

Kind

Not assessed

Related entries

Same task, a consumer chatbot assessing a patient's own symptoms in dialogue: ChatGPT-4o attributed Scott Winters's dizziness and pain to benign causes and discouraged hospital care before a massive pulmonary embolism, the inverse of DeepSeek-R1's correct rare diagnosis.

Field

Healthcare - rheumatology / autoinflammatory disease. Patient-initiated use of a consumer LLM.

Purpose

DeepSeek-R1 is a general-purpose reasoning chatbot. Answering a user's medical questions with a differential diagnosis falls within its general question-answering function. It is not a validated medical device.

Dates

Date basis unassessed: 2025 (precision unassessed). Updated: 21 October 2026.

Stage