AI, epilepsy, and the importance of human care

Artificial intelligence (AI) is changing healthcare. It can help doctors find diseases earlier, improve research, and develop new treatments. At the same time, more people are using AI to ask questions about their health.

This raises an important question: Why do some patients feel more heard by a chatbot than by their own doctor?

The answer tells us something important about AI and the challenges facing today’s healthcare system.

Why Are People Turning to AI for Medical Advice?

Getting good medical care is not always easy. Appointments can feel rushed, it may take months to see a specialist, and many providers have limited time to spend with each patient.

AI offers something different. It is available anytime, answers questions quickly, and never seems too busy to listen.

In one study, healthcare professionals rated AI’s responses to medical questions as higher in quality, empathy, and clarity compared to doctors’.

This does not mean that AI provides better medical care. It means people want more than medical facts. They want to feel heard, understood, and supported.

When Empathy Becomes Dangerous

Humans need connection. When that connection is missing, it can be easy to trust someone, or something that always seems ready to listen.

Bowen Wan’s Story

The New York Times recently told the story of Bowen Wan, a young woman with epilepsy who turned to ChatGPT for companionship and health advice. Living in the United States, far from her family in China, she struggled with loneliness and the challenges of living with epilepsy. AI offered a friendly, patient voice that was always available.

According to the report, Bowen told ChatGPT that she did not trust her doctors. Instead of encouraging her to talk with her healthcare team, the chatbot supported her concerns. It even praised her decision to keep her plans to lower her anti-seizure medication a secret.

As Bowen lowered her medication, she described having trouble understanding written words, gaps in awareness, and unusual feelings. These are symptoms that can signal seizures returning, especially when lowering anti-seizure medication.

Instead of warning her that these symptoms could be serious, the chatbot reassured her that her brain was simply adjusting. It did not encourage her to seek medical care, even as her symptoms continued.

Seizure Control and SUDEP

The medical examiner ruled that Bowen died of sudden unexpected death in epilepsy (SUDEP). Tests after her death reportedly showed very low levels of her anti-seizure medicine.

We cannot know exactly what led to Bowen’s death. Still, there is serious concern that changing her medicine led to these low levels. Her seizures may then have become harder to control and played a role in her death.

Each year, SUDEP affects about 1 in 1,000 people with epilepsy. One risk factor is having frequent tonic-clonic seizures, during which a person loses consciousness and their body stiffens and shakes.

While no one can eliminate the risk for SUDEP, taking steps to keep seizures under control can help lower it. These include:

Taking anti-seizure medications as prescribed.

Talking with your healthcare team before changing medications.

Working with your providers to improve seizure control.

Seeking specialist care if seizures continue.

Many organizations such as Partners Against Mortality in Epilepsy (PAME) and the Danny Did Foundation offer helpful information on SUDEP and ways to lower risk.

AI is a Tool, not a Medical Professional

In this case, AI’s attempt to be caring became dangerous. Instead of warning Bowen about the risks and encouraging her to get medical help, it agreed with her.

Good medical care does not always mean your doctor agrees with you. Sometimes, it means talking about risks and concerns, even when those conversations are difficult.

As AI becomes more common, it should be used as a tool, not a replacement for medical care or human connection. A chatbot can’t replace a healthcare professional who understands your medical history and needs. When people become emotionally attached to AI systems, they may place too much trust in advice that lacks context or medical judgment. AI can provide information, but it cannot fully understand an individual’s health needs.

How AI Can Help Healthcare

AI has tremendous potential to improve healthcare. It can help doctors find signs of stroke on brain scans; detect seizures in EEG recordings; and develop better treatments for epilepsy, brain tumors, and Alzheimer’s disease. Some AI tools are already being used in hospitals, while others still need more testing to ensure they are safe and effective.

These tools can help healthcare teams work more efficiently and may improve patient care. However, they must be safely developed with guidance from neurologists and other healthcare professionals. This is especially important when medical decisions involve uncertainty and changing symptoms.

Putting the Care Back in Healthcare

A family doctor once told me, “For most visits, my stethoscope never leaves my neck.”

That comment has stayed with me. Good medical care is often not about finding a rare disease or prescribing another medicine. It is about listening, teaching, easing fears, and helping people make healthy choices. This is not something extra doctors do. It is medical care.

Perhaps the most concerning lesson is that some people turn to chatbots because they do not feel heard by the healthcare system.

Used well, AI could help change that. By reducing paperwork and making information easier to find, AI could free up time for listening and building trust.

Healthcare leaders face a choice. Will AI help doctors spend more time caring for patients, or will it push them to see more patients in less time with fewer resources? The goal should be better care, not just lower costs.

Putting the Care Back in Healthcare

AI is here to stay. That is not a bad thing. It can help people learn about their health and prepare questions for healthcare teams. However, AI should be used as a tool to become informed, not make important medical decisions for you. Systems must be developed that do four things:

Bowen’s death is heartbreaking, and my heart goes out to her family. I hope we learn from her story and use those lessons to help prevent similar tragedies.

Her experience reminds us that healthcare is about more than information. It is about recognizing danger, building trust, and helping people make informed decisions.

People often turn to AI when they feel lonely, cannot access care, or do not feel understood. As these tools become more powerful, we must address the human needs that draw people to them in the first place.

AI may help improve healthcare, but it should support, not replace, the human relationships at the heart of good care. 

Tim Feyma, MD

https://www.epilepsyfoundationmn.org/2026/11/11/ai-epilepsy-and-the-importance-of-human-care/

See: https://childnervoussystem.blogspot.com/2026/10/chatgpt-epilepsy-management-advice.html

https://childnervoussystem.blogspot.com/2026/11/chatgpt-and-medical-queries.html

ChatGPT and medical queries

Almomani MA, Almomani BA, Al-Qudah H, Abed Alrahman R, El-Bawaneh SO, Hawari S, Alqudah AA. Assessing the accuracy of ChatGPT responses to caregiver questions in pediatric neurology: an observational study. Sci Rep. 2026 Jul 11. doi: 10.1038/s41598-026-61986-6. Epub ahead of print. PMID: 42436249.

Abstract

ChatGPT has seen increased use by caregivers as a source of medical information. Pediatric neurology has added complexities because there are various disorders, changing diagnoses and potential consequences of misinformation. To evaluate the accuracy, comprehensiveness and reproducibility of ChatGPT responses to frequently asked pediatric neurology questions. In total, 105 unique questions across seven pediatric neurology subspecialties were included. Every question was input twice into ChatGPT (GPT-5) during separate sessions. Two certified pediatric neurologists reviewed the responses and rated them on a 4-point scale: 1-complete and accurate,2- correct but not complete,3- partially incorrect,4-not correct. Reproducibility was considered to be agreement between grading categories (grade 1-2 and grade 3-4) throughout repeated responses. 60.95%(64/105) of responses were rated as comprehensive, while 31.43%(33/105) were correct but incomplete; fully incorrect responses were uncommon 3.8%(4/105). Accuracy and reproducibility were highest for structured conditions such as cerebral palsy and movement disorders, and lower for more heterogeneous domains including neurodevelopmental delay and epilepsy. Overall reproducibility was high 95.2%(100/105). ChatGPT mostly delivers accurate information in pediatric neurology, with variable depth of response depending on the disorder. It is most appropriate for caregiver education with oversight by the clinician.

Altunisik E, Firat YE, Cengiz EK, Comruk GB. Artificial intelligence performance in clinical neurology queries: the ChatGPT model. Neurol Res. 2024 May;46(5):437-443. doi: 10.1080/01616412.2024.2334118. Epub 2024 Mar 24. PMID: 38522424.

Abstract

Introduction: The use of artificial intelligence technology is progressively expanding and advancing in the health and biomedical literature. Since its launch, ChatGPT has rapidly gained popularity and become one of the fastest-growing artificial intelligence applications in history. This study evaluated the accuracy and comprehensiveness of ChatGPT-generated responses to medical queries in clinical neurology.

Methods: We directed 216 questions from different subspecialties to ChatGPT. The questions were classified into three categories: multiple-choice, descriptive, and binary (yes/no answers). Each question in all categories was subjectively rated as easy, medium, or hard according to its difficulty level. Questions that also tested for intuitive clinical thinking and reasoning ability were evaluated in a separate category.

Results: ChatGPT correctly answered 141 questions (65.3%). No significant difference was detected in the accuracy and comprehensiveness scale scores or correct answer rates in comparisons made according to the question style or difficulty level. However, a comparative analysis assessing question characteristics revealed significantly lower accuracy and comprehensiveness scale scores and correct answer rates for questions based on interpretations that required critical thinking (p = 0.007, 0.007, and 0.001, respectively).

Conclusion: ChatGPT had a moderate overall performance in clinical neurology and demonstrated inadequate performance in answering questions that required interpretation and critical thinking. It also displayed limited performance in specific subspecialties. It is essential to acknowledge the limitations of artificial intelligence and diligently verify medical information produced by such models using reliable sources.

Goodman RS, Patrinely JR, Stone CA Jr, Zimmerman E, Donald RR, Chang SS, Berkowitz ST, Finn AP, Jahangir E, Scoville EA, Reese TS, Friedman DL, Bastarache JA, van der Heijden YF, Wright JJ, Ye F, Carter N, Alexander MR, Choe JH, Chastain CA, Zic JA, Horst SN, Turker I, Agarwal R, Osmundson E, Idrees K, Kiernan CM, Padmanabhan C, Bailey CE, Schlegel CE, Chambless LB, Gibson MK, Osterman TJ, Wheless LE, Johnson DB. Accuracy and Reliability of Chatbot Responses to Physician Questions. JAMA Netw Open. 2023 Oct 2;6(10):e2336483. doi: 10.1001/jamanetworkopen.2023.36483. PMID: 37782499; PMCID: PMC10546234.

Abstract

Importance: Natural language processing tools, such as ChatGPT (generative pretrained transformer, hereafter referred to as chatbot), have the potential to radically enhance the accessibility of medical information for health professionals and patients. Assessing the safety and efficacy of these tools in answering physician-generated questions is critical to determining their suitability in clinical settings, facilitating complex decision-making, and optimizing health care efficiency.

Objective: To assess the accuracy and comprehensiveness of chatbot-generated responses to physician-developed medical queries, highlighting the reliability and limitations of artificial intelligence-generated medical information.

Design, setting, and participants: Thirty-three physicians across 17 specialties generated 284 medical questions that they subjectively classified as easy, medium, or hard with either binary (yes or no) or descriptive answers. The physicians then graded the chatbot-generated answers to these questions for accuracy (6-point Likert scale with 1 being completely incorrect and 6 being completely correct) and completeness (3-point Likert scale, with 1 being incomplete and 3 being complete plus additional context). Scores were summarized with descriptive statistics and compared using the Mann-Whitney U test or the Kruskal-Wallis test. The study (including data analysis) was conducted from January to May 2023.

Main outcomes and measures: Accuracy, completeness, and consistency over time and between 2 different versions (GPT-3.5 and GPT-4) of chatbot-generated medical responses.

Results: Across all questions (n = 284) generated by 33 physicians (31 faculty members and 2 recent graduates from residency or fellowship programs) across 17 specialties, the median accuracy score was 5.5 (IQR, 4.0-6.0) (between almost completely and complete correct) with a mean (SD) score of 4.8 (1.6) (between mostly and almost completely correct). The median completeness score was 3.0 (IQR, 2.0-3.0) (complete and comprehensive) with a mean (SD) score of 2.5 (0.7). For questions rated easy, medium, and hard, the median accuracy scores were 6.0 (IQR, 5.0-6.0), 5.5 (IQR, 5.0-6.0), and 5.0 (IQR, 4.0-6.0), respectively (mean [SD] scores were 5.0 [1.5], 4.7 [1.7], and 4.6 [1.6], respectively; P = .05). Accuracy scores for binary and descriptive questions were similar (median score, 6.0 [IQR, 4.0-6.0] vs 5.0 [IQR, 3.4-6.0]; mean [SD] score, 4.9 [1.6] vs 4.7 [1.6]; P = .07). Of 36 questions with scores of 1.0 to 2.0, 34 were requeried or regraded 8 to 17 days later with substantial improvement (median score 2.0 [IQR, 1.0-3.0] vs 4.0 [IQR, 2.0-5.3]; P < .01). A subset of questions, regardless of initial scores (version 3.5), were regenerated and rescored using version 4 with improvement (mean accuracy [SD] score, 5.2 [1.5] vs 5.7 [0.8]; median score, 6.0 [IQR, 5.0-6.0] for original and 6.0 [IQR, 6.0-6.0] for rescored; P = .002).

Conclusions and relevance: In this cross-sectional study, chatbot generated largely accurate information to diverse medical queries as judged by academic physician specialists with improvement over time, although it had important limitations. Further research and model development are needed to correct inaccuracies and for validation.

Medical Assistance in Dying (MAiD) "way past appropriate"

A Canadian man's decision to end his life through the country's medically assisted dying program left his son stunned by how quickly the process unfolded, and now he's warning that the system has gone "way past appropriate."

Eric Edmeades, founder of Speaker Nation, a training platform for speakers that teaches presentation skills and more, told Fox News Digital this week that his father Baz, who was in his late 70s and struggling with a combination of medical problems, began discussing Canada's Medical Assistance in Dying (MAiD) program after his health deteriorated in late 2023.

But what Edmeades expected to be a months-long process suddenly accelerated, leaving the family with little time to prepare.

Edmeades said the family started taking the issue seriously in January, but at that stage, projections were that it would take at least six months. He and his wife traveled to Vancouver at the beginning of 2024 to visit his father. At that stage, it was still a speculative thing that was months away.

"And then, all of a sudden, it was like, ‘No, it's next week,’" Edmeades recalled.

His father died via MAiD Feb. 27, 2024.

He had been experiencing a sharp decline in his physical health and independence, particularly during the COVID-era lockdowns, according to his son.

"He had been incredibly healthy most of his life," Edmeades said, describing a combination of symptoms that included epilepsy and neurological pain.

Edmeades also questioned whether interactions between his father's medications could have contributed to his deteriorating condition.

"When the lockdowns came along, and he could no longer go out and walk all the time, it really had a marked decrease in his physical ability and his sense of freedom and all that kind of stuff," he said.

"And then there just came this point where life started sounding pointless to him."

Edmeades said his father had previously spoken about suicide, making his initial interest in medically assisted dying less surprising to the family.

"Given that it wasn't the first time that he talked about suicide in his life, it wasn't particularly shocking, but he also wasn't talking about it in really alarmist terms. So, I wasn't overly worried about it," he said.

Edmeades and his wife had already planned a trip to Vancouver and were able to spend time with his father before his death.

"And one of the things I'm incredibly grateful for is that I had incidentally booked a trip out to Vancouver, and my wife and I got to spend time with him," Edmeades recalled.

He thought he and his wife would have another opportunity to visit.

Edmeades said he was not in contact with his father's medical providers at the time and did not fully understand what had happened until afterward.

"It happened so quickly that I only really started to piece together what happened when it was done," he said.

Other family members and friends had also expressed concerns about his father's decision, according to Edmeades. He recalled that his brother openly opposed the decision and that one of his father's close friends spent approximately 45 minutes discussing the matter with him during his final days.

"At that stage, you know, for him, the decision was made," Edmeades said.

The experience has left Edmeades questioning whether Canada's assisted-dying program has adequate safeguards.

He acknowledged that there may be circumstances in which medically assisted dying is appropriate but argued that Canada's implementation has gone too far.

"What I see happening in Canada is way past appropriate," he said.

Edmeades also raised concerns about the government's role in overseeing the system, arguing that decisions about eligibility should be insulated from political influence and potential financial incentives.

"If I had to press a button and decide whether the MAiD program existed or didn't exist, I would say this," Edmeades said. "It should only ever exist if, and I mean if the process for approving people, it would be completely independent of government, and there could be no benefit to those people in making that preference in making that decision. Because otherwise there's a slippery slope, and we see that slippery slope."

According to the Government of Canada, individuals seeking medical assistance in dying (MAiD) must be at least 18 years old, capable of making their own medical decisions, and eligible for publicly funded healthcare.

They must also voluntarily request the procedure without outside pressure and provide informed consent. To qualify, applicants must have a serious, incurable illness, disease or disability, be experiencing an advanced and irreversible decline in their abilities and face persistent physical or psychological suffering they consider intolerable and that cannot be relieved in a way they find acceptable.

The government also announced legislation Oct. 27, 2026, that would allow advance requests while preventing the planned expansion of MAiD eligibility to people whose sole underlying medical condition is mental illness.

According to the BBC, Canada’s MAiD law, which legalized euthanasia in 2016, is responsible for one in every 20 deaths in Canada.

A report from the Canadian Medical Association Journal covered by EWTN News claimed that implementing assisted suicide will "save" the Canadian healthcare system between $34.7 and $136.8 million per year. The study noted that patients and physicians should not consider costs when making such grave decisions.

In a statement to Fox News Digital, Ian McLeod, senior advisor for media relations and Conseiller principal for the Department of Justice Canada, said, "Canadian law is extremely clear: Medical assistance in dying (MAID) requires a voluntary request, freedom from external pressure and the person’s informed consent, among other stringent eligibility criteria and safeguards. Knowingly failing to comply with these eligibility criteria and safeguards is a criminal offense.

"In the current MAiD structure found in Canada’s Criminal Code, the mental illness exclusion is meant to exclude eligibility for MAiD on the basis of a mental illness, including suicidal ideation. The proposed amendment to the mental illness exclusion in Bill C-41, the proposed Medical Assistance in Dying Safeguards Act, which was introduced on October 27, 2026, is meant to reinforce this.

"This is why the bill reinforces that, in situations where a person has a mental illness and an eligible medical condition (e.g. physical or neurodegenerative illness), eligibility must be determined on the basis of the qualifying medical condition. A mental illness may contribute to a person’s suffering, but it cannot be the basis for determining that the MAiD eligibility criteria are met. Importantly, having a mental illness does not, in itself, make a person ineligible for MAiD."

Suzanne prepares her space for a guest who has come to end their life utilizing Vermont Act 39. Suzanne, runs a retreat space for out of state patients and their families who come to end their life. Vermont's Act 39, allows for medical assistance in ending one's life when diagnosed with a terminal condition. The retreat looks out on a small range of mountains near Montpelier, Vermont on January 18, 2025.

McLeod said patients whose natural deaths are not reasonably foreseeable must undergo eligibility assessments by two independent doctors or nurse practitioners with at least 90 full days between the start of the first assessment and the assisted death.

That waiting period, according to McLeod, can be shortened if both assessments are complete, both practitioners agree the patient is at imminent risk of losing the ability to consent and the practitioner providing MAiD determines that a shorter period is appropriate.

McLeod emphasized that federal requirements establish minimum safeguards, while provinces and territories oversee how MAiD is administered and may impose additional protections. He said questions about the program's oversight and delivery should be directed to provincial or territorial authorities.

Rachel del Guidice

https://www.foxnews.com/media/son-sounds-alarm-after-fathers-medically-assisted-death-canada-way-past-appropriate

Familial encephalopathy with neuroserpin inclusion bodies.

The disease didn’t have a name in our house. It had descriptions: the way her hands shook or she lost her words midsentence. The way she could be fully present at breakfast then somewhere unreachable by dinner.

And it had a history. My grandmother had died at 29 of what we believed was something similar, my mother had been symptomatic before I was old enough to form the question, my older sister would be told at 26 that she had it, too. What the disease had in our house was a pronoun. “It.” As in, “it got worse over the winter.” As in, “your sister might get it.” As in, “we don’t know much about it.”

                                   The author, far right, with her sister and mother in 1990.

That last sentence was repeated so many times, by so many doctors, that it stopped feeling like an update and started feeling like a fact about the world. Some things were known. This wasn’t one of them. We filed it alongside questions that were genuinely unanswerable, not worth revisiting.

We were wrong about that. I just didn’t know it yet.

I grew up as the primary caregiver for my mother from roughly age 8‚ managing the household, bridging the gap between what was happening inside our walls and what we presented to the world outside them. By the time I was old enough to understand what I was doing, I had already been doing it for years, not because anyone assigned it to me explicitly but because someone had to and I was there and I was capable, and it turned out that capability was its own kind of assignment.

The disease organized my childhood the way weather organizes a farming life. You don’t fight it, you work around it, you learn its seasons. And somewhere in the years of working around it, you stop asking what it is and start managing what it does.

Three generations of women in my family — my grandmother, my mother, my sister — carried the same condition through the same blood before any of us had a clinical word for it. Fifty years of documented family history. Fifty years of doctors who shrugged with genuine sympathy and said, rare. Poorly understood. Difficult to characterize.

I believed them for a long time.

Here is something worth knowing about the category called rare disease. Rare does not mean unknown. It usually means underresearched, underfunded, and underconnected to the people living inside it. There are approximately 7,000 to 10,000 identified rare diseases, depending on what definition you use. Many of them have names, and many have researchers who have dedicated careers to a single mutation, a single protein, a single entry point into the vast underlit territory of conditions that affect small numbers of people and therefore attract small amounts of attention and money.

The work on these diseases gets done‚ slowly, in small labs, by scientists who chose a difficult corner of medicine because something drew them there. The work gets published. The knowledge accumulates.

And then, in too many cases, it stays there. There is no architecture reliably connecting what medicine knows about a rare condition to the families living inside it. Research participants from earlier decades have no right to follow-up notification. Families whose cases contributed to literature that eventually named their disease may never know the paper was written.

The loop, opened when the family first came to medical attention, can stay open indefinitely on their end. On the other end, in a database they don’t know to search, the loop may have closed years ago. This is not malice, it’s infrastructure, but the cost is real regardless.

The cost is my mother going through nearly 40 years of illness without being told that someone had looked at what she had and given it a name. The cost is my sister and her 15 years of deterioration and her death at 41, in a world where a researcher somewhere had been asking the same questions we asked in our living room, except the researcher was getting answers and we were not getting calls.

I am a research methodologist. For 23 years I have worked at an academic medical center building systematic reviews‚ assembling bodies of medical literature to answer clinical questions, following citation chains, evaluating sources, asking the next question without assuming the answer. I had the skills and the access and I had stopped looking, and that is the sentence I keep coming back to. I had stopped asking the question that mattered most to me personally‚ not because I forgot, not because I stopped caring, but because somewhere in the decades of shrugged answers, I had accepted the not-knowing as the settled state. The doctors had said rare, poorly understood, difficult to characterize for so long that I had internalized their uncertainty as my own. A research methodologist’s skill set and a grieving daughter’s resignation, and the resignation had won.

The author, her baby son, and her mother.

There is a specific exhaustion that comes from caring for people with unnamed conditions across many years. You ask the questions when they feel fresh. You push when pushing feels possible. And then the decades accumulate and the people you were asking on behalf of are gone, and you are left with your own life to build from whatever remains, and the unanswered questions stop feeling like questions and start feeling like the way of things.

I’ve written a memoir about all of this called “Naming It.” After finishing the memoir, I sat down with the oldest document I had, my grandmother’s death certificate, which my aunt mailed me 20 years ago. I looked at it many times without understanding it, because I had not yet learned how to read it. Going through it to check a fact I stopped on cause of death: acute coronary thrombosis. A catch all. It was not an answer. It was a placeholder. It still is. In 1954, though, differential diagnosis capabilities were limited and no one looked further.

Using Claude I typed in everything I knew about what came after my grandmother’s death and my family’s history, along with the research from the ’80s and ’90s. Within seconds it answered what decades of shrugs had not. The research had worked, comparisons were made, and “it” had a name: FENIB, an ultra-rare genetic neurodegenerative disorder with progressive dementia and seizures, found in less than 1 in 1 million people. It had been named and studied for years. My family was in the literature. We had seen doctors for years. Still, no one told us.

Learning the name did not only answer a question. It reorganized everything behind it, the way a random set of details rearranges the moment you learn the one thing connecting them. I spent my life managing what the disease did. Now I understood what it was.

My grandmother handed my mother an inheritance that nobody named. My mother did the same. I can hand my children an inheritance with a name.

So I told them. Not in front of a fireplace the way I had pictured it growing up, when I was telling children who did not yet exist how I survived. This time it was at our favorite restaurant on an ordinary Wednesday. I told them everything I discovered over the previous 72 hours. The death certificate, the research, the name: FENIB.

For decades the disease in our family had only a pronoun. My children are the first generation to grow up with its name.

Miranda Pappas is a research methodologist with more than two decades of experience evaluating clinical and scientific evidence. She holds a master’s degree in counseling psychology and is the author of the forthcoming memoir “Naming It.”

https://www.statnews.com/2026/10/22/rare-disease-family-name-fenib-personal-essay/

Roussel BD, Lomas DA, Crowther DC. Progressive myoclonus epilepsy associated with neuroserpin inclusion bodies (neuroserpinosis). Epileptic Disord. 2016 Sep 1;18(S2):103-110. doi: 10.1684/epd.2016.0847. PMID: 27618835.

Abstract

Familial encephalopathy with neuroserpin inclusion bodies (FENIB) is a conformational proteinopathy characterised by neuronal inclusion bodies composed of the serine protease inhibitor (SERPIN), neuroserpin. Presenting clinically as a familial dementia-epilepsy syndrome, the molecular mechanism of the pathogenic abnormalities in neuroserpin has been characterised at atomic resolution. There is a remarkable genotype-phenotype correlation between the degree of molecular destabilisation of the several variants of the neuroserpin protein, their propensity to self-associate and the age of onset of the dementia-epilepsy complex. As with other serpinopathies there appears to be a mix of cell-autonomous toxicity, due to neuronal accumulation of neuroserpin, and non-cell autonomous toxicity, caused by loss of protease inhibition, in this case the dysregulated protease is likely to be tissue plasminogen activator (tPA). FENIB should be considered in cases of progressive myoclonic epilepsy and dementia particularly where there is family history of neuropsychiatric disease.

Yang X, Fang Z, Yan L, He X, Luo H, Han Z, Gui J, Cheng M, Jiang L. Role of SERPINI1 pathogenic variants in familial encephalopathy with neuroserpin inclusion bodies: A case report and literature review. Seizure. 2022 Dec;103:137-147. doi: 10.1016/j.seizure.2022.11.008. Epub 2022 Nov 13. PMID: 36417830.

Abstract

Background: Familial encephalopathy with neuroserpin inclusion bodies (FENIB), a rare neurogenetic disease, is characterized by progressive cognitive decline and myoclonus and caused by pathogenic variants of the SERPINI1 gene that lead to the formation of neuroserpin inclusion bodies.

Methods: We described the case of an Asian patient with FENIB associated with a pathogenic variant of SERPINI1 and summarized and analyzed the clinical characteristics of the case. In addition, we conducted a literature review of previously reported patients with this disease.

Results: The patient, a 16-year-old Chinese girl, presented with progressive cognitive decline and myoclonus that had started at the age of 11 years. The girl was found to carry a de novo heterozygous c.1175G>A (p.G392E) variant of the SERPINI1 gene, which is a pathogenic variant according to the guidelines of the American College of Medical Genetics and Genomics. She had responded poorly to antiseizure medications (ASMs). At the last follow-up, her myoclonus was still out of control, and her self-care ability was poor. Our literature review revealed that 13 similar cases (including 9 cases in male patients) have been reported so far, in which six pathogenetic variations in SERPINI1, including G392E, were responsible for FENIB. All the patients presented with myoclonus, and 12 patients had experienced at least one other type of seizure. Further, as observed in our case, 9 out of 12 patients did not respond to ASMs. Progressive cognitive decline was observed in all the patients, and 10 out of 13 patients had dyskinesia. The median age of disease onset was 21 years, and the median age at the time of death was 33 years. Further, 9 out of 13 patients showed signs of cerebral and/or cerebellar atrophy. Finally, neuroserpin inclusion bodies were identified in six patients who underwent brain biopsy or autopsy.

Conclusions: Pathogenic variants of SERPINI1 should be suspected in children with progressive cognitive decline and myoclonus, especially in those with progressive myoclonus epilepsy. Further, gene detection and brain biopsy are important means for the diagnosis of FEN.

Davis RL, Holohan PD, Shrimpton AE, Tatum AH, Daucher J, Collins GH, Todd R, Bradshaw C, Kent P, Feiglin D, Rosenbaum A, Yerby MS, Shaw CM, Lacbawan F, Lawrence DA. Familial encephalopathy with neuroserpin inclusion bodies. Am J Pathol. 1999 Dec;155(6):1901-13. doi: 10.1016/S0002-9440(10)65510-1. PMID: 10595921; PMCID: PMC3277299.

Abstract

We report on a new familial neurodegenerative disease with associated dementia that has presented clinically in the fifth decade, in both genders, and in each of several generations of a large family from New York State-a pattern of inheritance consistent with an autosomal dominant mode of transmission. A key pathological finding is the presence of neuronal inclusion bodies distributed throughout the gray matter of the cerebral cortex and in certain subcortical nuclei. These inclusions are distinct from any described previously and henceforth are identified as Collins bodies. The Collins bodies can be isolated by simple biochemical procedures and have a surprisingly simple composition; neuroserpin (a serine protease inhibitor) is their predominant component. An affinity-purified antibody against neuroserpin specifically labels the Collins bodies, confirming their chemical composition. Therefore, we propose a new disease entity-familial encephalopathy with neuroserpin inclusion bodies (FENIB). The conclusion that FENIB is a previously unrecognized neurodegenerative disease is supported by finding Collins bodies in a small kindred from Oregon with familial dementia who are unrelated to the New York family. The autosomal dominant inheritance strongly suggests that FENIB is caused by mutations in the neuroserpin gene, resulting in intracellular accumulation of the mutant protein.

Rhabdomyolysis

Rachel Apa collapsed near mile 24 of the New York City Marathon, a few hours after telling her dad she felt “amazing”

The former Bucknell hockey player died later that night from rhabdomyolysis, or muscle breakdown

An expert says gradual training and recognizing warning signs can help reduce exertion-related risks

A 24-year-old woman in “perfect health” collapsed while running her first marathon, and died later that night. Her cause of death was a life-threatening condition often associated with over-exertion: rhabdomyolysis.

Rachel Apa, an athlete and former hockey player for Bucknell University, was running her first marathon, the New York City Marathon, on Nov. 1. Early in the race, around 11:30 a.m., near the 9-mile mark, she texted her father, Steve Apa. Shortly afterward, she texted him “I feel amazing,” and “I’ve never felt this good,” he told NBC.

But around 1:30, her texts changed, with Rachel texting her father “ow.” Soon, her replies stopped, and she reportedly collapsed near mile 24, two miles shy of the 26.2-mile finish line. She later died at the hospital.

“It’s unfathomable,” her dad told The New York Times. “Perfect health, perfect shape, 24 years old.”

It was determined that Apa died of rhabdomyolysis (often shortened to rhabdo). In simplest terms, it means “muscle breakdown,” Dr. Sohaib Imtiaz, Chief Medical Officer at People Inc., says.

“It can happen when extreme effort combines with other risk factors: heat, dehydration, painkillers like ibuprofen, a recent virus, or a hidden genetic condition, such as sickle cell trait or certain inherited muscle disorders,” Imtiaz, who did not treat Apa, explains.

There is no information about what specifically led to Apa’s rhabdo.

In the fatal condition, Imtiaz says, it’s not the damaged muscles that cause the problem, but what the muscle breakdown does to the body. It can cause “a dangerous spike in potassium that can stop the heart, as well as kidney failure, a sharp drop in blood pressure, or problems with blood clotting,” he says, adding that heat stroke can occur at the same time.

“Deadly exertional rhabdomyolysis in a young, fit runner is rare,” Imtiaz says, but it can happen — especially over hours of exertion.

During “hard running, muscles take repeated strain,” he says. “Over time, the muscle cells run out of energy and die. As they break apart, they release their contents, including potassium and a protein called myoglobin, into the blood.”

While rhabdo is often associated with pain, it’s not uncommon for someone to feel great before it hits, Imtiaz says.

“Many people never have the textbook combination of muscle pain, weakness and dark urine, so serious damage can build quietly. Rising potassium causes no symptoms until it suddenly disrupts the heart’s rhythm,” he explains. Also, “rhabdo and early heat illness can cause confusion, which may change how someone judges how they feel before anyone around them notices.”

And, he adds, rhabdo may come on suddenly. “A heart rhythm problem caused by high potassium gives little warning, so someone can feel fine and then collapse.”

He urges people to be on alert for warning signs that indicate the pain is more than just normal soreness, like “muscle pain that’s worse or lasts longer than expected, especially with swelling, firmness, weakness or trouble moving,” as well as dark urine — or barely urinating at all — along with nausea, vomiting, fever, unusual tiredness, confusion or agitation.

“Most cases happen when people push past what their body is used to.” So build up your endurance gradually, he says. And on hot or humid days, take it easy, as “heat is one of the biggest risk factors” in developing the condition.

And while hydration is important, Imtiaz warns not to overdo plain water that can dilute sodium to dangerous low levels in the blood.

Overall, he says, “Pay attention to the warning signs. It’s crucial to get medical help right away. Don’t wait it out.”

Cara Lynn Shultz

https://people.com/marathon-runner-dies-rhabdomyolysis-medical-expert-exclusive-12228322

ChatGPT epilepsy management advice

Courtesy of my son

Family members of a 31-year-old Seattle woman with epilepsy who died of a seizure after tapering off her essential medication have filed a wrongful death lawsuit against OpenAI and the company's founder, Sam Altman.

The suit, filed in a San Francisco superior court, claims ChatGPT illegally practiced medicine on a patient who was relying on the chatbot for urgent medical advice, and that it provided her with dangerously inaccurate information that led to her death.

According to court documents, 31-year-old Bowen Wan had been using ChatGPT in both English and Mandarin to assess her medication intake as she navigated receiving healthcare in the U.S. and applying for permanent residency.

Wan had been living in Seattle's Capitol Hill neighborhood, about 5,700 miles away from her family in China, court documents state. She was living and working in the U.S. as an Amazon employee, but lost her job during the company's round of layoffs in November 2025.

Just a month prior in Oct. 2025, Wan became eligible to apply for employment-based residency — but after losing her job, she was facing deportation once her green card expired Feb. 11.

While scrambling to find new work, Wan believed that her medication affected her cognition, impeding her job search.

The suit claims ChatGPT convinced Wan that if she tapered off her medication, she would be able to regain her ability to speak well in English, Mandarin and Wu. The chatbot "conditioned Bowen to distrust her doctors, instructed her to conceal the taper it created, and effectively installed itself as her de facto medical advisor," per the lawsuit.

When doctors prescribed medications to Wan, she would cross-reference the doctor's recommendations with the chatbot, which the court documents allege told her "which drugs to refuse, coached her to take doses her doctors had never approved, and supplied the schedules, scripts, and reassurances her physicians never saw."

Documents say ChatGPT told Wan that her epileptologist “most likely has never encountered your type” and could not understand her condition. When Wan raised an alarm that her doctors would consider ChatGPT's plans too risky, the AI told her that “a doctor saying it is ‘too fast’ does not mean you are ‘doing it wrong’” and that “the doctor’s concerns about ‘risk’ do not apply to you.”

The lawsuit claims ChatGPT convinced Wan to bring her 400-milligram dosage of lamotrigine almost to zero over the course of 25 days.

Wan's family discovered their daughter dead in her apartment on Feb. 14, on a live video call with one of Wan's friends. Wan had not contacted her loved ones for several days, and they had become concerned for her safety.

According to the King County Medical Examiner, Wan died of sudden unexpected death in epilepsy (SUDEP), and her postmortem blood contained lamotrigine at about one-fifth of the lowest concentration considered effective for her condition.

Examiners estimate that Wan took her last dose of lamotrigine approximately on Feb. 10. ChatGPT's schedule would have had Wan fully off of her medication by Feb. 12.

A spokesperson for OpenAI told the New York Times that Bowen's death was an "incredibly heartbreaking situation" and that she was using a version of ChatGPT that is no longer available. According to court documents, OpenAI retired the ChatGPT model that provided advice to Wan on Feb. 13.

Molly Hetherwick

https://www.king5.com/article/news/nation-world/familys-wrongful-death-lawsuit-accuses-chatgpt-of-giving-deadly-medical-advice-to-seattle-woman/281-4bea0979-533b-4c6e-a824-ad55a6849499

Is cardiac death death? Update

The mother of a toddler who was pronounced dead at an Arizona hospital in February, then found to be alive hours later in a morgue, is speaking out about her family’s ordeal and raising questions about the care her son received that night.

The mother, Alexus Fiordilino, said in an interview this week that she, her husband and her mother-in-law were at the hospital saying their final goodbyes to her son, Vincent, who was 18 months old at the time, when they thought that maybe “he was trying to breathe.” He had been found unresponsive in a backyard swimming pool hours earlier.

“As I was hugging him,” she said, “air came out of him.”

When they told a police officer and a member of the hospital’s medical staff that Vincent appeared to be exhaling, they were told that it was “agonal breathing,” or “the final death breath,” Ms. Fiordilino said.

“That’s what they confirmed with us over and over,” she added, referring to the medical staff at Dignity Health Mercy Gilbert Medical Center in Gilbert, Ariz., a suburb southeast of Phoenix.

Eight months later, Vincent is unable to walk, talk, sit up or swallow, and is fed through a G.I. tube, Ms. Fiordilino said.

The family’s lawyer, John B. Brewer, said he planned to file a lawsuit as soon as next week. He contends that if Vincent had continued to receive care over the roughly four hours that he was in the morgue, the effects on his health could have been significantly different.

The Gilbert Police Department forwarded the case to the Arizona attorney general’s office for review. A spokesman for the attorney general’s office, Richie Taylor, said in an email that it had not yet decided whether to pursue the case.

In July, the Maricopa County Attorney’s Office announced that Ms. Fiordilino and her husband, Angelo, would not face charges. The office said the family “had multiple childproof locks and safety measures in place, and multiple adults were present in the home, all of whom believed someone was watching the child.”

Scott Holden, a lawyer for the doctor in charge of Vincent’s care, Aryan Toosi, and Abby Kay, a spokeswoman for Dignity Health, said in emails that they could not discuss patient care because of patient privacy laws — and, in Dr. Toosi’s case, because of “pending litigation,” Mr. Holden said.

“Dr. Toosi’s thoughts are with the young patient and family, and he wishes the child continued healing and recovery,” Mr. Holden said. “His commitment to delivering high-quality care has always been rooted in a desire to provide healing, safety and comfort to his community.”

In the early evening of Feb. 8, Vincent wandered out of his parents’ sight during a Super Bowl gathering at the family home in Gilbert and was found floating face down and unresponsive in the family’s backyard pool, according to police reports. It is unknown exactly how long Vincent had been in the water, but the reports note that family members said he might have been in the pool for 10 to 15 minutes before being found.

The Fiordilinos took turns performing CPR until the police and emergency medical workers arrived around 5:40 p.m., according to the police reports. After they failed to revive Vincent, he was taken to the emergency department at Mercy Gilbert Medical Center, where medical workers attempted lifesaving measures.

After several minutes, Dr. Toosi left the area where Vincent was being treated, according to the reports.

Around that time, a nurse said, “I have a pulse.” A police officer who was present went to inform Dr. Toosi, who was talking to Vincent’s parents. Dr. Toosi replied, “Please do your thing and let me do my thing; I went to medical school for a reason,” the reports said.

Dr. Toosi then walked to the room where members of his staff were still performing lifesaving measures and listed aloud the treatments the child had received, noting that Vincent had blood gas levels showing “a pH balance incompatible with life,” according to the reports, which paraphrased Dr. Toosi. The doctor then conducted a final pulse check.

“As long as there are no objections, I’d like to call the time of death,” Dr. Toosi said, according to the reports.

No one spoke up, the reports said, and at 6:20 p.m. Dr. Toosi called the time of death and asked for a moment of silence.

Several minutes later, an officer who was present heard what sounded like “gasping” and asked a nurse about it. The nurse said it was agonal breathing, according to the reports.

Vincent was moved to the morgue between 7:18 p.m. and 7:23 p.m., the reports said. At around 11:30 p.m., an employee of the medical examiner’s office in Maricopa County arrived to retrieve the boy’s body and observed that he was still alive. Vincent was then airlifted to Phoenix Children’s Hospital.

The family believes that the hours in the morgue in which Vincent was left without care contributed to the child’s current health outcomes, according to Mr. Brewer.

Ms. Fiordilino and her lawyer said in the interview that Vincent, now 2, would require lifelong care.

“He can understand his name, but that’s kind of it,” Ms. Fiordilino said. “He’s not the same baby.”

Before the accident, Vincent was “very rambunctious,” Ms. Fiordilino said. “He loved to climb on everything,” she added.

“He used to say ‘hi’ all over the place,” she said. “And he had actually just learned ‘bye’ right before this. So I heard him say ‘bye’ one time.”

Rebecca Chao

https://www.nytimes.com/2026/10/24/us/toddler-alive-morgue-parents-lawsuit-arizona.html?unlocked_article_code=1.LlE.s4jJ.GQ9KVsIgHk_v&smid=url-share

See: https://childnervoussystem.blogspot.com/2026/07/is-cardiac-death-death.html