LLM-guided diagnosis and self-treatment of familial Mediterranean fever in one patient

Source article: LLM-supported self-diagnosis and self-treatment of familial Mediterranean fever: a case report

Large language models (LLMs) are increasingly being used by patients to seek explanations for persistent or unexplained symptoms. However, their impact on real-world diagnostic pathways and patient behavior remains poorly understood. We report the case of a 34-year-old male of Egyptian origin with recurrent episodes of fever, severe migratory abdominal pain, arthralgia, myalgia, and intermittent erythematous skin lesions since early childhood. Despite almost 3 decades of extensive investigations, no unifying dia…

By the Truvace deskOctober 15, 20262 min read1 sourcePeer-reviewedTRV-2026-1505· cite & verify

LLM-supported self-diagnosis and self-treatment of familial Mediterranean fever: a case report — "Chatbots linked" by Ashhstarr is licensed under CC BY-SA 4.0. To view a copy of this license, visit https://creativecommons.org/licenses/by-sa/4.0/.

Trace impact readingContested

P 69

Both sides are scored from claims and sources, not community votes.

G 69

In brief

A 34-year-old man had suffered recurrent fevers, severe migratory abdominal pain, joint and muscle pain and skin lesions since early childhood, with no unifying diagnosis despite nearly three decades of investigation. In June 2025 he discussed his symptoms with the LLM DeepSeek, which steered the differential toward familial Mediterranean fever and supplied details on genetic testing and colchicine.

He arranged testing himself and obtained colchicine in Egypt. His attacks stopped, but liver enzymes rose and treatment was paused and the dose reduced. The case shows that patient-facing chatbots can help surface rare diagnoses while also nudging people toward unsupervised care. Because it is a single case, how often such outcomes occur remains unknown.

Main points

The patient was a 34-year-old man of Egyptian origin with symptoms since early childhood and no diagnosis after almost 3 decades.
The model repeatedly recommended specialist assessment yet also gave specific information on genetic testing and colchicine.
He started colchicine at 0.5 mg twice daily without prior specialist assessment, and his inflammatory attacks ceased completely.
Raised liver enzymes led to temporary discontinuation and dose reduction of the drug.
A specialist confirmed FMF from phenotype, family history, genotype and colchicine response.

The gain

After about three decades without a unifying diagnosis, a patient's iterative dialogue with DeepSeek narrowed the differential to familial Mediterranean fever, later confirmed by a specialist, and colchicine stopped his inflammatory attacks.

The problem

The same chatbot exchange, with confident diagnostic language and actionable treatment information, was followed by independent genetic testing and colchicine self-treatment without specialist review, and liver enzymes rose.

The rundown

The report describes an iterative conversation in which the chatbot gradually refined the differential diagnosis. Although the model repeatedly advised seeing a specialist, it also provided concrete guidance on genetic testing and colchicine treatment.

Acting on this, the patient arranged genetic testing independently and obtained colchicine in Egypt. Treatment began at 0.5 mg twice daily with no prior specialist assessment, and his attacks stopped entirely.

Liver enzymes then increased, so treatment was temporarily discontinued and the dose reduced. A specialist later confirmed FMF using the characteristic clinical picture, family history, genotype and response to colchicine. The authors call for clear safety boundaries and specialist confirmation for patient-facing LLMs.