AI analysis of Reddit reveals public interest in GLP-1 drugs for weight loss and mental health benefits

A recent study published in the journal Communication Medicine, US researchers used large-scale language models (LLMs) to analyse over 391,000 unique discussions about glucagon-like peptide-1 (GLP-1) receptor agonists (GLP-1 RAs) on the social media platform Reddit. The study found that interest in GLP-1 RAs was high, with discussions focusing on weight loss experiences, side effects, access issues and positive psychological benefits, with mostly neutral to positive sentiment.

study: Using large-scale language models to assess public perceptions of glucagon-like peptide-1 receptor agonists in social mediaImage credit: Caroline Ruda/ Shutterstock

background

More than 38% of the world's population is overweight or obese, and this is projected to reach 51% by 2035. Obesity is known to significantly increase the risk of cardiometabolic disease and all-cause mortality. GLP-1 RAs are drugs that mimic the function of natural GLP-1, a gut hormone that controls glucose metabolism and satiety. This class of drugs was initially approved for the treatment of type 2 diabetes, but has recently attracted worldwide attention for cardiovascular risk reduction and weight loss in obese patients with and without diabetes. However, public opinion towards GLP-1 RAs, which is essential for treatment acceptance and adherence, has not been thoroughly investigated.

Social media platforms like Reddit provide anonymous, public conversations about health topics, revealing real-world experiences that are often missed in clinical settings and trials. Manually analyzing this large amount of data requires a lot of resources, but artificial intelligence systems like LLM can speed up the analysis. Therefore, researchers in this study used LLM to analyze over 390,000 Reddit discussions about GLP-1 RAs, identifying topics such as weight loss, side effects, and concerns. They aimed to monitor side effects, gauge public sentiment, and use the findings to guide future research and public health initiatives.

About the Research

Reddit hosts user-generated discussions in the form of posts and comments, which are organized into public, topic-specific communities called “subreddits.” GLP-1 RA-related discussions were curated by indexing Reddit content based on generic and brand names of GLP-1 RA drugs, including semaglutide. The dataset contained 391,461 unique discussions (mostly from 2021 onwards) by 116,216 authors, 71,982 posts, and 319,479 comments.

The analysis employed the aforementioned “topic modeling” approach, using various tools and algorithms. Discussions were converted into numerical representations and clustered to identify topics. Each topic was labeled and grouped based on the similarity of discussion content. This approach aimed to extract key themes and insights from the extensive discussions on GLP-1 RA available on Reddit. Furthermore, the study employed a model called “RoBERTa” (short for Robustly Optimized Bidirectional Encoder Representations from Transformers pre-training approach) to classify sentiment. The model used three probabilities (ranging from 0 to 1) to classify the nature of sentiment in the text as “negative, neutral, or positive sentiment.”

Results and discussion

Approximately 97.1% of the discussions were focused on GLP-1 RA drugs prescribed for weight loss, such as semaglutide, tirzepatide, and liraglutide, with Ozempic being the most discussed (41.4%), despite not being approved by the US Food and Drug Administration (FDA) as a weight loss drug. Only 2.9% of the discussions were about GLP-1 RAs approved only for diabetes. The amount of discussions increased significantly since 2022, following the approval of Wegovy by the FDA.

The model identified 168 discussion topics of high public interest that focused on experiences with weight loss drugs. Topics included: efficacyTopics discussed included comparison with other treatments, effects on appetite, and side effects. The most frequent side effect was nausea, followed by vomiting, injection site problems, constipation, pancreatitis, and gastroparesis. Additionally, access issues, market shortages, insurance coverage, and the ethics of off-label use were discussed. Positive effects on motivation and mental health, as well as the value of avoiding bariatric surgery, were also discussed. Topics were categorized into 33 groups reflecting themes such as comparison with other treatments, side effects, access concerns, and psychological benefits. Sentiment analysis revealed that 31.8% of discussions were negative, 50.1% were neutral, and 17.4% were positive. Notably, two topics were excluded due to illegal content related to obtaining illegal drugs.

Scatter plot showing 2D projection of all discussion embeddings. Each point represents a discussion. The color on top represents the relevant group of that discussion based on topic modeling. The x-axis and y-axis represent the two axes (feature 1, feature 2) along which the embeddings were dimensionally reduced using uniform manifold approximation and projection for visualization.

This study, powered by large-scale AI-based analysis of social media discussions, illuminates public perceptions and experiences with medicines, providing insights beyond traditional clinical research. However, the study has limitations including possible mislabeling due to spelling errors, inability to verify reported side effects, limited generalizability, and suboptimal general task benchmarking of LLMs.

Conclusion

In conclusion, this study used LLM to analyze a large-scale GLP-1 RA-related discussion on social media. Findings revealed discussions centered around weight loss experiences, comparative side effects, access issues, and positive psychological benefits. This indicates high public interest in GLP-1 RAs and highlights priorities for the clinical and policy communities, including monitoring for side effects, addressing access barriers, and recognizing both the physical and psychological benefits of these medications.