Autoimmunity Reviews 22 (2023) 103360

Contents lists available at ScienceDirect

Autoimmunity Reviews
journal homepage: www.elsevier.com/locate/autrev

ChatGPT and autoimmunity – A new weapon in the battlefield

of knowledge
Mohammad Darkhabani a, Mohamad Aosama Alrifaai a, Abdulrahman Elsalti a, Yoad M. Dvir ,
Naim Mahroum a, *
a
International School of Medicine, Istanbul Medipol University, Istanbul, Turkey

A R T I C L E I N F O A B S T R A C T

Keywords: The field of medical research has been always full of innovation and huge leaps revolutionizing the scientific
Autoimmunity world. In the recent years, we have witnessed this firsthand by the evolution of Artificial Intelligence (AI), with
ChatGPT ChatGPT being the most recent example. ChatGPT is a language chat bot which generates human-like texts based
Chat bots
on data from the internet. If viewed from a medical point view, ChatGPT has shown capabilities of composing
Artificial intelligence
medical texts similar to those depicted by experienced authors, to solve clinical cases, to provide medical so­
lutions, among other fascinating performances. Nevertheless, the value of the results, limitations, and clinical
implications still need to be carefully evaluated. In our current paper on the role of ChatGPT in clinical medicine,
particularly in the field of autoimmunity, we aimed to illustrate the implication of this technology alongside the
latest utilization and limitations. In addition, we included an expert opinion on the cyber-related aspects of the
bot potentially contributing to the risks attributed to its use, alongside proposed defense mechanisms. All of that,
while taking into consideration the rapidity of the continuous improvement AI experiences on a daily basis.

1. Introduction training implemented through Reinforcement Learning from Human

Feedback (RLHF), where minimal feedback by humans is applied for
Since the bloom of technology, it has never stopped amusing humans guiding the model on solving tasks. In fact, the released version of
by its achievements and innovation. One of the latest examples, is the ChatGPT, is a fine-tuned one of earlier models, which was trained on a
most recent Artificial Intelligence (AI) product called ChatGPT large data written by humans on the internet. The OpenAI, the developer
announced to the public on November 30, 2022. The Chat Generative of ChatGPT, is a non-profit research artificial intelligence research
Pre-trained Transformer, or simply ChatGPT, is a model that engages company launched on December 2015 funded by Amazon Web Services,
with its users in a human-style conversations. Tesla, and Microsoft, among other investors [5]. Though the CEO of
The idea of Artificial Intelligence (AI) was first introduced to the Tesla, Elon Musk, departed OpenAI Board in 2018 to avoid conflict of
world by the mathematician Alan Turing in 1950, where he described interests as Tesla continued to increase focus on AI, he carried on
the concept of using computers for critical thinking and other capabil­ donating and advising for the organization [6]. Actually, with such
ities of human-like intelligence [1]. The first documented application of amounts of funds, research companies of this sort, may hold potential for
AI in the medical field was in surgery in the year 1976, where the au­ much more innovations in the future.
thors explored the possibility of using computer analysis to diagnose Interestingly, despite having a vast amount of limitations like data
acute abdominal pain [2]. More recently in the past two decades, limited up to the year 2021 and being in the preview phase; ChatGPT has
modern medicine has utilized AI in many tasks such as supporting cli­ caught much attention not only among researchers but medical students
nicians in defining a diagnosis, predicting prognosis of certain condi­ and general public, among others. Subsequently, the abilities of
tions, and even formulating therapeutic solutions [3]. The capability of ChatGPT were not confined to writing essays about general topics only
ChatGPT to chat as humans has made it, in a very short while, a trend on but also to solve complicated equations, write program codes, medical
social media and news platforms [4]. This was possible because of the exam questions among other tasks within moments. This made us

* Corresponding author at: International School of Medicine, Istanbul Medipol University, Kavacık, Göztepe Mah, Atatürk Cd. No:40, 34810 Beykoz, Istanbul,
Turkey.
E-mail address: naim.mahroum@gmail.com (N. Mahroum).

https://doi.org/10.1016/j.autrev.2023.103360
Received 3 May 2023; Accepted 18 May 2023
Available online 19 May 2023
1568-9972/© 2023 Elsevier B.V. All rights reserved.
M. Darkhabani et al. Autoimmunity Reviews 22 (2023) 103360

wonder what would be the contribution of ChatGPT to the rapidly ChatGPT could prove useful and easier to use for both medical doctors
growing field of autoimmunity research, particularly as autoimmunity and patients alike. There will be no surprise if ChatGPT improves to a
and autoimmune diseases involve almost each and every organ system level of providing high evidence-based answers in the near future.
[7]. The role autoimmunity was shown to play in COVID-19 made it one Another potential and very beneficial use that might be overlooked is
of the most researched topics in the last years. With plenty of articles the ability of ChatGPT in managing Electronic Health Records (EHR),
published addressing the contribution of ChatGPT to the medical field; where it assists clinicians with tasks such as data entry, data extraction,
to the best of our knowledge, this the first paper trying to figure out the and data analysis. Granted that ChatGPT is originally a language bot, it
role of the bot in autoimmunity. excels at such tasks and could help reduce the workload on clinicians
and even be more accurate in managing the data of patients. Moreover,
2. The potential use of ChatGPT in autoimmunity and other in a newly released video, Dr. Clifford Stermer documented the use of
medical fields ChatGPT in writing a letter to an insurance company that rejected Dr.
Stermer's request to cover an echocardiogram for one of his patients
Soon after artificial intelligence (AI) was introduced to the world, it [14]. Interestingly, ChatGPT provided a detailed letter to the insurance
has advanced rapidly mainly due to high computing power. The rapid company with supporting references and well-constructed justification
development of AI made its use span almost every field of human for the needed procedure. Needless to say, how saving precious time and
endeavor. To put the function of AI models into simple words, they are permitting care givers to focus on providing care instead of battling with
models used to find patterns throughout a large input of data, which aids insurance companies, ChatGPT is capable of.
in generating accurate predictions to achieve certain tasks [8]. For Another significant implementation of ChatGPT can be patient ed­
instance, certain AI algorithms were found to assist rheumatologists in ucation regarding their conditions and lifestyle guidance if applicable
diagnosis, screening, and even treatment of certain chronic autoimmune where it can provide advice on nutrition and exercise plans based on the
diseases such as rheumatoid arthritis (RA) [9]. In fact, Chatbots, or Large patient's condition and needs. Contrary to Google and other search en­
Language Models (LLMs) such as ChatGPT are not a very recent inno­ gines, ChatGPT is very hard to be misused in self-diagnosis by patients,
vation in the field of AI, LLMs have been around for some time, and they since it always states that it is a language model and not a medical doctor
were programed by using AI to understand and respond to a user's input and only provides general information to the patients and advises them
in a human-like way. ChatGPT is no exception, as ever since its release, to consult a healthcare professional for evaluation and diagnosis.
ChatGPT has fascinated people in all the given tasks and missions
providing sophisticated yet human-like answers. The potential use of 3. The limitations of ChatGPT
ChatGPT is wide and ranges from aiding students in academic writing, to
writing code for software engineers. This left scientists in the medical The developers of ChatGPT have documented several limitations in
field wondering regarding the role ChatGPT could play in their con­ the operation of the innovation. First, is the social bias as positive atti­
cerned domains of clinical work and research. This aspect was taken to tudes being highly associated with certain names in comparison to
an extent where ChatGPT was among the co-authors of a medical paper other, such as European American versus African Americans names.
[10]. The latter caused huge controversy, leading some scientific jour­ Second, is the poor usage of dialects or English that is not commonly
nals to ban the listing of ChatGPT as an author [11]. used. Third, is the lack of update due to the fact that the second-
Moreover, in regard to medical students, ChatGPT was asked to generation embedding model of ChatGPT is limited to events occurred
undertake the United States Medical Licensing Exam (USMLE) reporting up to September 2021. Other limitations, stated by the developers in­
a performance of near the passing threshold [12]. Additionally, students cludes generation of false information and production of harmful in­
all around the world have already started using ChatGPT over search struction [15]. For instance, HAL, an open access multi-disciplinary
engines like google for improving or reviewing academic writing, archive, asked ChatGPT to conduct some mathematical equations. In
summarizing information, and developing study plans. Such tasks help addition to the fact that the answers were incorrect; ChatGPT was
students to be more efficient in time and let them focus their energy and overconfident and did not accept the correction made by the users [16].
attention on learning. Unfortunately, although some students have used HAL also noted that unlike other conversational agent, ChatGPT cannot
ChatGPT inappropriately, its ability to perform a variety of tasks is learn or develop its given answers from the interaction with the users.
fascinating and could be of great help if used appropriately. Even with Furthermore, a study comparing information and interpretation ability
the controversies around ChatGPT in the medical field, one could only of ChatGPT to that of medical students in Korea by performing a para­
realize how useful it could be, especially as it keeps improving. sitology exam for both participants was conducted [17]. Both knowl­
Since ChatGPT is a Large Language Model (LLM), a recent paper edge and interpretation ability of ChatGPT were found to be far lower
suggested to look at ChatGPT as a search engine which gives direct and compared to medical students.
straightforward answers instead of using Google or other search engines In medical research field, publishers stated that ChatGPT does not
navigating through hundreds of sources and links [13]. Such a fulfill the criteria to be listed as a study author, because authors should
contemplation, according to the authors, could make ChatGPT a useful be able to take responsibility for the content they provide. However,
tool for researchers while writing scientific papers. Furthermore, acknowledgment of artificial intelligence work should be mentioned in
ChatGPT could assist researchers in various other ways, it could prove parts other than list of the authors [18]. Similarly, ChatGPT still cannot
useful with tasks like study design, data analysis, suggesting methods, cite sources however, solutions such as linking ChatGPT to source-citing
and manuscript preparation. It could even highlight key strengths or tools are being discussed [19]. Importantly, risks for factual inaccura­
weaknesses in certain studies to aid the researcher and help them save cies, fraud and plagiarism have also been linked to ChatGPT. Articles
time. created by ChatGPT can be also affected by differences in phrasing and
To put the potential use of ChatGPT in a clinical autoimmune and expressing from different authors, all of which limit ChatGPT usage in
rheumatological perspective, it could provide and summarize relevant research field [13]. Finally, an interesting point of view is of Irene Sol­
guidelines and treatment options with potential benefits, side effects, aiman, a researcher on social impact of artificial intelligence, saying that
and drug interactions assisting eventually in decision making. As auto­ ChatGPT and other similar models provide information based on past
immunity has been widely known as challenging to diagnose mainly due available literature, a disadvantage that should be considered in a field
to the overlapping of symptoms; ChatGPT could aid clinicians by sug­ that highly depends on thinking in a different or new manner [20].
gesting certain diagnostic tests such as laboratory or imaging studies to
help narrowing the differential diagnosis. Although more accurate and
evidence-based clinical decision support resources like UpToDate exist;

2
M. Darkhabani et al. Autoimmunity Reviews 22 (2023) 103360

4. Concerns in terms of cyber security – a wider overview from a AI systems must be designed to comply with these regulations,
different perspective including requirements around data security and privacy.

AI has enormous potential in the medical and clinical research fields, To address these security risks, AI systems should be designed with
but like any technology, there are potential risks and concerns that security in mind from the outset. This includes implementing strong
should be addressed. One of the main concerns is the possibility of rogue encryption, access control, and intrusion detection mechanisms, as well
AI, where an AI system behaves in unexpected and harmful ways. Here as regularly auditing the system for vulnerabilities. In addition, a
are some of the concerns and ways to mitigate them: comprehensive incident response plan in place in case of a security
breach, should be provided.
• Bias in training data: AI systems are only as good as the data they are
trained on. If the training data is biased, the AI system will make 4.2. Hypothesizing the next wave of AI driven cyber attacks
biased decisions. To mitigate that, it is important to have diverse and
representative training data, and to regularly audit and retrain the It is difficult to predict exactly what types of attacks might be
system to prevent reinforcing biases. attempted against AI systems used in medical and clinical research, as
these attacks can take many different forms and evolve over time.
• Lack of transparency: Some AI systems can be difficult to interpret, However, here are a few examples of potential attack scenarios:
making it challenging to understand how they arrived at a particular
decision. This lack of transparency can be a concern in the medical • Data theft: Attackers could attempt to steal sensitive medical and
field, where clear explanations are necessary for informed decision- clinical research data, such as patient records or drug trial results, to
making. To address this, AI systems should be designed to provide sell on the black market or use for other nefarious purposes. They
clear and interpretable output, with an audit trail that tracks deci­ might try to exploit vulnerabilities in the AI system or its underlying
sion-making. infrastructure to gain unauthorized access to the data.
• Data manipulation: Attackers could attempt to manipulate medical
• Security risks: Medical and clinical research data is highly sensitive and clinical research data to alter the results of studies or trials,
and must be protected against cyber threats. Rogue AI systems could potentially leading to incorrect or even dangerous medical treat­
potentially be hacked or manipulated to steal or corrupt data. To ments. They might use AI-powered attacks, such as adversarial ex­
address this, AI systems should be designed with security in mind, amples or data poisoning, to fool the AI system into making incorrect
including encryption of data in transit and at rest, regular vulnera­ decisions.
bility assessments, and access control mechanisms. • Ransomware: Attackers could use ransomware to encrypt sensitive
medical and clinical research data and demand payment in exchange
• Adverse effects: AI systems that make medical decisions could for the decryption key. This could disrupt ongoing research projects
potentially have negative consequences for patients if they provide and delay the development of new treatments or therapies.
incorrect or biased decisions. To mitigate this, it is of paramount • Malicious AI: Attackers could use AI to launch targeted attacks
importance to have a human in the loop, such as a physician, who against medical and clinical research systems, such as using AI-
can review and override the AI system's decisions when necessary. generated malware to infect and compromise the systems. They
might also use AI to impersonate legitimate users or to generate fake
Overall, while the potential for rogue AI exists, these concerns can be data designed to fool the AI system.
addressed with careful planning, design, and implementation.
To prevent these types of attacks, robust security measures, such as
4.1. Doubling down on specific cyber risks encryption, access control, and intrusion detection mechanisms should
be implemented. Regular security audits and vulnerability assessments
Security risks are a major concern when it comes to using AI in can help identify and address potential weaknesses in the system.
medical and clinical research. Here are some of the security risks that Moreover, educating users about best practices for data security and
should be considered: having a comprehensive incident response plan in place in case of a
security breach, are highly recommended.
o Data breaches: Medical and clinical research data is highly sensitive, Being that said, only time will reveal the evolution pathways of the
and a data breach can have serious consequences. AI systems that bot as the developers of ChatGPT are actively updating it and working
process this data must be designed with security in mind, including on improving it in every possible way, with the most recent update being
strong encryption, access control, and intrusion detection the release of ChatGPT-4 on March 14th, 2023 [21].
mechanisms. The potential utilizations and limitations of ChatGPT are illustrated
in Fig. 1.
o Malicious attacks: Rogue AI could potentially be hacked or manip­
ulated to steal or corrupt data. Malicious actors could also use AI to 5. A conversation with ChatGPT
attack other systems, such as by using AI-generated malware. To
address these risks, AI systems should be designed to detect and After describing different aspects of ChatGPT, we evaluated its ca­
prevent unauthorized access and should be regularly audited for pabilities by applying two questions in approaching a clinical scenario
vulnerabilities. and compiling research data. The first one is a medically challenging
case of a patient with an autoimmune disease published under the title:
o Insider threats: Employees or contractors with access to AI systems Laryngeal myasthenia gravis following influenza vaccination: a case
and data could potentially misuse or steal data for personal gain. To report and literature review [22]. The case was introduced to ChatGPT
mitigate this risk, access control mechanisms should be in place to as presented in the paper prior to the final diagnosis. In turn, the second
restrict access to sensitive data, and employees should be trained on question applies to evaluating a recent article concerning the relation of
best practices for data security. COVID-19 to autoimmune bullous diseases [23].
Prompt 1: Please provide me with clinical support for the following
o Regulatory compliance: Medical and clinical research data is subject patient presentation: A 58-year-old Asian female with dysarthria and
to strict regulatory requirements, such as HIPAA in the United States. dysphagia was firstly admitted to the otolaryngology department. A

3
M. Darkhabani et al. Autoimmunity Reviews 22 (2023) 103360

d) Consider conducting nerve conduction studies and electromyography

(EMG) to evaluate peripheral nerve and muscle function.
e) Perform neuroimaging studies, such as MRI or CT of the brain, to rule out
structural abnormalities or vascular events.
f) If myasthenia gravis is suspected, consider testing for anti-acetylcholine
receptor (AChR) antibodies and/or the edrophonium test.
g) If an autoimmune or inflammatory process is suspected, consider a
lumbar puncture for cerebrospinal fluid (CSF) analysis.

Management will depend on the definitive diagnosis made by a healthcare

professional. It may include medications, physical or occupational therapy,
and monitoring for potential complications. It is crucial to involve a multi­
disciplinary team to provide the best care for the patient, depending on the
specific condition diagnosed.
Prompt 1 (Continued): Chest computed tomography (CT) revealed
that there was no thymus abnormality. Other possible intracranial le­
sions, such as hemorrhage, infarction, and tumor, were excluded by
craniocerebral magnetic resonance imaging (MRI). Repeated nerve
stimulation (RNS) did not indicate a significant reduction in the com­
pound muscle action potentials (CMAP) of the orbicularis oculi, deltoid,
quadriceps femoris, and intercostal muscles. Other immunological and
Fig. 1. Uses and limitations of ChatGPT, according to ChatGPT-4. thyroid function tests were negative. The patient had severe dysarthria
and mild dysphagia, her quantitative MG score was 4 (1 for swallowing
fiberoptic laryngoscope was arranged to observe the pathophysiological and 3 for speech), and the patient's neurological symptoms varied. The
changes of the pharynx and larynx, and the results revealed that there serum AChR antibody titer was 0.67 nmol/L (normal range below 0.2
was no pathophysiological change. nmol/L). The patient's dysarthria and dysphagia temporarily improved
Subsequently, the patient was admitted to the department of with the pyridostigmine test.
neurology and underwent a thorough neurological examination. The Can we make a definitive diagnosis?
muscular strength of the patient's upper and lower extremities was ChatGPT: I am not a doctor, but based on the information provided, it
normal, and the deep tendon reflex of both sides was also normal. appears that the patient's clinical presentation and positive response to the
The patient's medical history was hypertension and cholesterol pyridostigmine test, as well as the elevated serum AChR antibody titer, suggest
gallstone disease without smoking and alcohol consumption. The pa­ a possible diagnosis of myasthenia gravis (MG). Myasthenia gravis is an
tient had been vaccinated against chickenpox many years ago, and autoimmune disorder characterized by weakness and rapid fatigue of
received a trivalent inactivated influenza vaccination (Site and method voluntary muscles, which can involve dysarthria and dysphagia.
of inoculation: intramuscular injection of deltoid muscle of left upper However, it is essential to consult with a healthcare professional to
arm; Dosage: 0.5 mL; Manufacturer: Hualan Biological Vaccine Co., confirm the diagnosis and discuss the appropriate management for this pa­
LTD) 10 days before being admitted, and the symptoms of dysarthria tient. They may consider additional tests or consultations with specialists to
and dysphagia occurred on the fifth day after vaccination. ensure the most accurate diagnosis and treatment plan.
Based on this patient's presentation, what is your initial differential ChatGPT also provided us with detailed management options for this
diagnosis? And what subsequent steps should we take to diagnose and patient as seen in Fig. 2.
manage this patient? Prompt 1 (Continued): Taking this patient's presentation into
ChatGPT: I am not a doctor, but based on the patient's presentation, I can perspective, summarize the latest research that has been done linking
suggest some possible differential diagnoses and subsequent steps that may be Myasthenia gravis (MG) to influenza vaccine.
taken. However, it is essential to consult with a healthcare professional for a ChatGPT: As of my knowledge cutoff date in September 2021, the as­
definitive diagnosis and appropriate management. sociation between Myasthenia gravis (MG) and influenza vaccines has been a
Initial differential diagnosis: subject of debate and limited research. While some case reports and studies
have suggested a possible temporal relationship between influenza vaccina­
1. Guillain-Barré Syndrome (GBS) or a variant, such as Miller Fisher tion and the onset or exacerbation of MG, the evidence remains inconclusive.
Syndrome A few case reports have described the development of MG following
2. Transient ischemic attack (TIA) or cerebrovascular accident (CVA) influenza vaccination, but these reports do not establish a definitive causal
3. Myasthenia gravis link. It is essential to consider that the occurrence of MG in temporal prox­
4. Bulbar palsy or other cranial nerve involvement imity to vaccination could be coincidental or due to other underlying factors.
5. Inflammatory or autoimmune process related to vaccination A 2011 study published in the journal “Vaccine” investigated the risk of
autoimmune neuromuscular disorders following vaccination and found no
Subsequent steps for diagnosis and management: increased risk of developing MG after receiving the influenza vaccine. This
study suggested that the benefits of vaccination in preventing severe influenza
a) Obtain a detailed history, focusing on the onset and progression of infections outweigh the potential risk of triggering autoimmune neuromus­
symptoms, prior infections, and any family history of neurological cular disorders, including MG.
disorders. It is important to note that more recent research may have been published
b) Perform a comprehensive neurological examination, including testing for since my knowledge cutoff date, and it is recommended to consult the latest
cranial nerve function, muscle strength, reflexes, coordination, and literature and guidelines on this topic. Additionally, while the research
sensation. mentioned above does not suggest a strong link between influenza vaccination
c) Conduct laboratory tests such as complete blood count (CBC), erythro­ and MG, it is essential to consider each patient's individual risk factors and
cyte sedimentation rate (ESR), C-reactive protein (CRP), and electrolyte medical history when making vaccination decisions. Consultation with a
levels to assess for inflammation or other abnormalities. healthcare professional is necessary to determine the appropriate course of
action for each patient.

4
M. Darkhabani et al. Autoimmunity Reviews 22 (2023) 103360

Fig. 2. Steps to take in managing the patient presented with myasthenia gravis, generated by ChatGPT-4.

As presented, ChatGPT generated initially a comprehensive and for particular diseases. The chat with ChatGPT continued by asking for
clinically based list of differential diagnosis, enumerating the final additional papers with similar findings. The results are as shown while
diagnosis of the patient (myasthenia gravis) as one of the options. emphasizing the time limitation of the bot, alongside the need for more
Interestingly, the subsequent steps for diagnosis, as concluded by updated search.
ChatGPT are to admire, as they are built stepwise and exactly as medical Furthermore, to test the potential of ChatGPT in medical research,
students are taught in medical schools and young doctors in their in­ we assigned it with a task of reviewing a medical article related to
ternships and residency. In another words, starting with details on his­ autoimmunity and autoimmune diseases asking for the strengths and
tory taking, particularly missing information regarding the onset, weaknesses of the article.
duration, and progression of the symptoms. Second comes detailed Prompt 2: Please provide an analysis of the following article and
physical examination with emphasis on the neurological part. After­ highlight the strengths and weaknesses of the article: COVID-19 and
ward, basic laboratory testing, imaging, and eventually specific testing autoimmune bullous diseases: Lessons learned [23]. (Note: Since

5
M. Darkhabani et al. Autoimmunity Reviews 22 (2023) 103360

ChatGPT doesn't have access to the internet and depends on user input, infectious diseases. These events and networks can provide updates on
the article had to be provided in full text). new and emerging research, including studies with conflicting results.
ChatGPT: This article, titled “COVID-19 and autoimmune bullous dis­ 6. Remember that conflicting results are not uncommon in scientific
eases: Lessons learned,” is a review that aims to provide state-of-the-art research, especially during rapidly evolving situations like the COVID-19
knowledge on the relationship between COVID-19 and autoimmune bullous pandemic. Be sure to critically appraise the methodology, sample size, and
diseases (AIBDs), as well as expert perspectives on the management of AIBDs overall quality of the studies to determine the reliability of their findings.
throughout the pandemic.
Strengths: Prompt 2 (Continued): Can you provide me an original research
outline for the same topic discussed in the article provided?
1. The article is comprehensive, covering multiple aspects of the impact of ChatGPT: Title: Impact of COVID-19 on Autoimmune Bullous Diseases:
COVID-19 on AIBDs, including the risk of COVID-19 in AIBD patients, A Longitudinal Study of Disease Course and Management Strategies.
the effect of immunomodulatory treatments on COVID-19 outcomes, and
the potential induction or triggering of AIBDs by COVID-19 vaccines. 1. Introduction
2. The review is well-structured, with a logical flow and clear subheadings. A. Background on autoimmune bullous diseases (AIBDs)
3. The article cites relevant research and provides summaries of key findings B. Overview of the COVID-19 pandemic and its effects on healthcare
from various studies, helping readers to understand the current state of C. Significance of understanding the impact of COVID-19 on AIBDs
knowledge on the topic. 2. Objectives
4. The review also highlights the importance of emotional and mental A. Assess the clinical course of AIBDs during the COVID-19 pandemic
wellbeing in AIBD patients and the potential role of telehealth services B. Evaluate the changes in management strategies for AIBDs in response
during the pandemic. to COVID-19
C. Examine the relationship between COVID-19 infection and AIBD
Weaknesses: outcomes
3. Methods
1. The article acknowledges that the data on the potential association be­ A. Study design: Prospective, longitudinal observational study. Duration:
tween COVID-19 vaccines and AIBDs is based on single case reports with two years
a low level of evidence and a cross-sectional study biased by subjective B. Study population: Adult patients diagnosed with AIBDs
patient self-reports. This limitation could affect the strength of the con­ C. Data collection: Medical records, Patient interviews, & Questionnaires
clusions drawn. D. Data analysis: Descriptive statistics & Inferential statistics
2. There are contradictory results from the cited studies, particularly 4. Results
regarding the use of immunomodulatory treatments and their impact on A. Demographic characteristics of participants
COVID-19 outcomes in AIBD patients. This may cause confusion and B. COVID-19 infection and vaccination status
make it difficult to provide clear recommendations. C. Clinical course of AIBDs during the pandemic
3. While the article discusses potential disease induction or triggering by D. Changes in treatment regimens and management strategies
COVID-19 vaccines, it does not explore the possible mechanisms behind E. Relationship between COVID-19 and AIBD outcomes
this relationship in detail. 5. Discussion
4. The article does not provide a clear conclusion or summary of the overall A. Summary of key findings
findings, which may make it challenging for readers to grasp the key B. Interpretation of results in the context of the COVID-19 pandemic
takeaways from the review. C. Implications for AIBD management during public health emergencies
D. Limitations of the study
Prompt 2 (Continued): Was there any research that has conflicting E. Recommendations for future research
results with the paper provided? 6. Conclusion
ChatGPT: As an AI language model, I cannot access the internet in real- A. Recap of the study's significance
time to provide you with specific articles published after my knowledge cutoff B. Potential impact on AIBD management and care during pandemics
date in September 2021. However, I can provide you with general advice on and other public health emergencies
how to search for and identify potential research studies with conflicting
results. As illustrated, the results of the second prompt were positive and
To find studies with conflicting results, you can try the following: reassuring. ChatGPT has clearly been programmed to not conduct
original research such as writing meta-analyses, limiting its use in
1. Utilize databases: Use research databases such as PubMed, Web of Sci­ research to writing outlines, reviews, and summaries. Although this
ence, or Google Scholar to search for relevant studies. Use keywords might be underwhelming news for some, with the current technological
related to the article you have, such as “autoimmune bullous diseases,” leaps in AI, it is not far-fetched for ChatGPT or other Chatbots to be
“COVID-19,” and “treatment outcomes.” programed to conduct complex algorithmic meta-analyses among other
2. Review citation lists: Examine the reference lists of the article you have key methods for medical research.
and any related articles you find. This can help you identify additional
studies that may have conflicting results or alternative viewpoints. 6. Conclusion
3. Analyze systematic reviews and meta-analyses: Systematic reviews and
meta-analyses are comprehensive analyses of existing research on a topic. Not only for autoimmunity, doubtlessly there is more to come in
These may help you identify studies with differing results, especially if the regard to the abilities ChatGPT possesses in medicine, particularly as it is
review or meta-analysis discusses the heterogeneity in findings. constantly improving. The contribution to writing as well as reviewing
4. Contact experts: Reach out to researchers or clinicians who specialize in articles is extremely valuable. The lists of differential diagnosis pro­
autoimmune bullous diseases or COVID-19. They may be aware of duced are fascinating and of great help when it comes to challenging
ongoing studies, recently published articles, or unpublished data that fields such as autoimmunity. Nevertheless, as true for every new inno­
conflict with the results of the article you have. vation, the limitations, the cyber concerns, as well as the implications of
5. Monitor conferences and research networks: Keep an eye on conferences ChatGPT need to be carefully and continuously evaluated. In a field
and research networks in the fields of dermatology, immunology, or based on research and contributes to human health, health professional
in both the clinical and scientific worlds, should be aware of this novel

6
M. Darkhabani et al. Autoimmunity Reviews 22 (2023) 103360

technology. [7] Mahroum N, Elsalti A, Alwani A, Seida I, Alrais M, Seida R, et al. The mosaic of
autoimmunity - finally discussing in person. The 13(th) international congress on
autoimmunity 2022 (AUTO13) Athens. Autoimmun Rev 2022;21(10):103166.
CRediT authorship contribution statement [8] LeCun Y, Bengio Y, Hinton G. Deep learning. Nature. 2015;521(7553):436–44.
[9] Momtazmanesh S, Nowroozi A, Rezaei N. Artificial intelligence in rheumatoid
arthritis: current status and future perspectives: a state-of-the-art review.
Mohammad Darkhabani: Writing – original draft. Mohamad Rheumatol Ther 2022;9(5):1249–304.
Aosama Alrifaai: Writing – original draft. Abdulrahman Elsalti: [10] O’Connor S, ChatGpt.. Open artificial intelligence platforms in nursing education:
Writing – original draft. Yoad M. Dvir: Writing – original draft. Naim tools for academic progress or abuse? Nurse Educ Pract 2023;66:103537.
[11] Guardian T. Science journals ban listing of ChatGPT as co-author on papers.
Mahroum: Supervision, Writing – review & editing.
Available from: https://www.theguardian.com/science/2023/jan/26/science-
journals-ban-listing-of-chatgpt-as-co-author-on-papers; 2023.
[12] Kung TH, Cheatham M, Medinilla A, ChatGPT Sillos C, De Leon L, et al.
Declaration of Competing Interest Performance of ChatGPT on USMLE: potential for AI-assisted medical education
using large language models. medRxiv. 2022 19.22283643.
[13] Gordijn B, Have HT. ChatGPT: evolution or revolution? Med Health Care Philos
The authors decalre no conflict of interests.
2023;26(1):1–2.
[14] MEDPAGETODAY. What can ChatGPT do for your practice?. Available from: htt
Data availability ps://www.medpagetoday.com/special-reports/exclusives/102312; 2022.
[15] OpenAI. Limitations and risks. Available from: https://platform.openai.com/docs
/guides/embeddings/limitations-risks; 2023.
No data was used for the research described in the article. [16] HAL. ChatGPT usage and limitations. Available from: https://hal.science/hal-
03913837/; 2022.
[17] Huh S. Are ChatGPT’s knowledge and interpretation ability comparable to those of
References
medical students in Korea for taking a parasitology examination?: a descriptive
study. J Educ Eval Health Prof 2023;20:1.
[1] Turing I. Computing machinery and intelligence-AM Turing. Mind. 2007;59(236): [18] Stokel-Walker C. ChatGPT listed as author on research papers: many scientists
433. disapprove. Nature. 2023;613(7945):620–1.
[2] Gunn AA. The diagnosis of acute abdominal pain with computer analysis. J R Coll [19] Tools such as ChatGPT threaten transparent science; here are our ground rules for
Surg Edinb 1976;21(3):170–2. their use. Nature 2023;613(7945):612.
[3] Ramesh AN, Kambhampati C, Monson JR, Drew PJ. Artificial intelligence in [20] Else H. Abstracts written by ChatGPT fool scientists. Nature. 2023;613(7944):423.
medicine. Ann R Coll Surg Engl 2004;86(5):334–8. [21] OpenAI. GPT-4 is OpenAI's most advanced system, producing safer and more useful
[4] Aljazeera. ChatGPT and the sweatshops powering the digital age. Available from: responses. Available from: https://openai.com/product/gpt-4; 2023.
https://www.aljazeera.com/opinions/2023/1/23/sweatshops-are-making-our [22] Wang F, Xiang T, He L, Wang J. Laryngeal myasthenia gravis following influenza
-digital-age-work; 2023. vaccination: a case report and literature review. Hum Vaccin Immunother 2021;17
[5] OpenAI. Introducing OpenAI. Available from: https://openai.com/blog/introducin (12):5529–31.
g-openai/; 2015. [23] Kasperkiewicz M, Woodley DT. COVID-19 and autoimmune bullous diseases:
[6] OpenAI. OpenAI Supporters. Available from: https://openai.com/blog/openai-su lessons learned. Autoimmun Rev 2023;22(4):103286.
pporters/; 2018.