“A Father Trusted AI Over Doctors—Now He’s Battling Stage 4 Esophageal Cancer

Warren Tierney, a 37-year-old father of two from the UK, developed a sore throat and ignored the initial symptoms since he was busy taking care of his family including his wife and two young kids. Early visits to the hospital was also not fruitful and he was sent home with reflux tablets. So, he decided to check his symptoms using ChatGPT. The artificial intelligence chatbot suggested a mild infection and reassured him it was not serious. Trusting this advice, he delayed seeing a doctor again. Months later, persistent symptoms led him to seek emergency medical care, where specialists diagnosed him with stage 4 adenocarcinoma of the oesophagus. Stage 4 cancer have only palliative care in Ireland. The family has started a GoFundMe page to raise money for treatment outside. He is currently undergoing treatment, including chemotherapy and radiotherapy.

The overall prognosis in stage IV esophageal adenocarcinoma remains poor. The estimated 5-year mortality for stage IV disease exceeds 85% to 90%. Following diagnosis, many patients suffer significant comorbidities and require interventions such as esophageal stenting and feeding tube placement. A significant factor in improving survival after chemoradiation involves local disease control according to a study published in Pubmed .^{[1] [2]}

The family is devastated by the news and advocates not to rely on ChatGPT for diagnosing medical conditions. His case has sparked debate on the dangers of relying on AI tools for medical decision-making without professional consultation.

Medical experts emphasize that while AI chatbots can provide general health information, they are not substitutes for clinical evaluation. Throat cancer symptoms often overlap with benign conditions such as infections, making accurate diagnosis challenging without physical examination, imaging, or biopsy. Early detection is crucial because delayed diagnosis can allow cancers to progress to more advanced stages, reducing survival rates and complicating treatment.

The issue extends beyond individual cases. A recent study published in JAMA Pediatrics evaluated the reliability of large language models for medical questions and found that while responses were generally empathetic and easy to understand, they were not consistently accurate or comprehensive. There was 80% error in diagnosis. ^[3]

The convenience of online symptom checkers and AI tools has led to growing use worldwide, but health authorities stress the importance of viewing them only as supplementary resources. Chatbots cannot replace the expertise of licensed healthcare professionals, who can assess a patient’s history, perform examinations, and order diagnostic tests.

Doctors advise patients to use AI and online tools cautiously, as guides for general awareness rather than diagnostic certainty. The safest course remains consulting a physician for accurate diagnosis and timely treatment.

References:

1. Minashi, Kei, Shigetoyo Saji, Hirofumi Isohata, Toshiro Ito, Tsutomu Imai, Hiroshi Okuno, Norihisa Nakamura, et al. "Endoscopic Management of Esophageal Cancer after Definitive Chemoradiotherapy." Endoscopy 45, no. 3 (2013): 200–207. https://pubmed.ncbi.nlm.nih.gov/23325163/.

2. Ilson, David H. "Esophageal Cancer Chemotherapy: Recent Advances." Gastrointestinal Cancer Research 2, no. 2 (2008): 85–92. https://pmc.ncbi.nlm.nih.gov/articles/PMC2630822/.

3. Bressler, Nicholas M., Michael F. Chiang, Diana W. Bianchi, John D. Birkmeyer, David S. Jones, Tracy A. Lieu, and Rita F. Redberg. "Diagnostic Accuracy of a Large Language Model in Pediatric Case Studies." JAMA Pediatrics 178, no. 2 (2024): 130–36. https://jamanetwork.com/journals/jamapediatrics/fullarticle/2813283.

(Rh/Eth/TL/MSM)