Pesquisadores from Universidade from Pensilvânia used artificial intelligence languages to analyze more than 400 thousand posts from Reddit and identify symptoms in users of weight loss medications that are not included in official leaflets. The study, published in Nature Health, revealed that some adverse effects reported by patients were not detected in conventional clinical trials.
The research covered posts from approximately 70 thousand users over more than 5 years. The drugs analyzed include semaglutide and tirzepatide, which are widely used for diabetes and weight loss. Grandes language models, such as GPT and Gemini, processed the posts and classified the described symptoms.
AI reveals undocumented symptoms
Ciclos irregular menstrual periods, intermenstrual bleeding, chills, hot flashes, fever-like sensations and fatigue were among the symptoms recorded by users on social media. Esses side effects do not appear in documentation provided by manufacturers or in traditional clinical trial reports.
The previous difficulty in analyzing this information resided in the fact that patients describe the same symptoms in different ways. Artificial intelligence has managed to standardize and identify patterns that conventional methods do not capture.
Segundo Lyle Ungar, Sistemas Informação professor and study co-author, clinical trials often identify only the most dangerous side effects. “But they may not be able to identify which symptoms worry patients most. Embora social networks are not necessarily representative, a large number of posts may reflect additional concerns,” he explained in a press release.
Vantagem speed detection
The method proposed by the researchers offers a faster alternative to traditional clinical trials. Essa speed is crucial especially when a drug moves from niche use to the mainstream market almost overnight, as occurred with semaglutide and tirzepatide.
Learn more: Artificial intelligence models detect unprecedented adverse reactions in weight loss pens
Sharath Chandra Guntuku, research associate professor at Ciência at Computação and Informação at Penn Engineering and senior author of the study, highlighted: “This solution does not replace clinical trials, but it can be much faster.”
The analysis allowed researchers to extract valuable information from online communities without the need for lengthy bureaucratic processes. Reddit served as a rich source of real reports from patients who share experiences in an organic way.
Próximos steps in research
The researchers plan to expand the analysis beyond Reddit and English-speaking communities. The goal is to compare results across different languages and regions to see if there are similar patterns in reported side effects.
Essa geographic and linguistic expansion may reveal variations in reported symptoms between different populations. Dados collected in Portuguese, Spanish, French, and other language-speaking communities may provide additional information about the safety of these medications.
More on this story: Tadalafil use by women raises questions about effectiveness and risks to female health
The findings will be shared with healthcare professionals to warn patients about side effects not reported by traditional science. Médicos will be able to use this information to better guide their patients about possible adverse experiences.
Impacto in medication monitoring
The study demonstrates how artificial intelligence can complement traditional pharmaceutical surveillance systems. Agências Regulators in several countries monitor drugs after their approval, but the current methodology may be slow to capture all adverse reactions.
On the same topic: Artificial intelligence maps unprecedented symptoms in patients who use pens to lose weight
The innovative approach paves the way for an early warning system based on social media data. Quando many users report a specific symptom, algorithms can flag potential side effects before they become a significant public health problem.
Large language models also reduce analysis operational costs. Anteriormente, studying millions of reports would require large teams of researchers to manually read each post. Agora, machines can process data in a fraction of the time.
Confiabilidade from social media as a scientific source
Embora social media data is not representative of the general population, the massive amount of posts offers valuable insights. Usuários of Reddit, for example, often share detailed and honest experiences about medications.
Pacientes who report side effects tend to seek out online communities to validate their experiences and find support. Esses spaces function as natural laboratories where people describe real symptoms experienced on a daily basis.
The voluntary and anonymous nature of social media may encourage more honest reporting compared to formal medical consultations, where patients may withhold information for a variety of reasons.
More on this story: Regulatory body approves first synthetic semaglutide pen to treat diabetes in the country
The research represents a significant advance in modern pharmacovigilance, showing how technology can improve the detection of side effects in approximately real time for medicines widely used by the population.