Who Really Drives the Weight Loss Drug Conversation on X (Twitter)?

Weight loss drugs have become one of the most talked about subjects on social media. But when a prescription medication turns into a cultural phenomenon, who is actually shaping what the public believes about it?

A new peer-reviewed study in the Saudi Pharmaceutical Journal put that question to the test, and the answer should give every health communicator pause. Researchers mapped the conversation about Mounjaro (tirzepatide) on X and found that the most influential voices were overwhelmingly not medical professionals. They were public figures, political personalities, and ordinary people sharing personal weight-loss stories.

The study was conducted using NodeXL.

The study at a glance

“Influence without medical expertise: a social network analysis of Mounjaro discussions on Twitter (X)” was authored by Saja AlHazmi, Hayam A. AlRasheed, and Basma Gomaa, of Nova Southeastern University and Princess Nourah Bint Abdulrahman University. It was published open access in 2026.

Mounjaro is a medication approved for type 2 diabetes, but it is increasingly used off-label for weight loss, driven in large part by social media attention. The researchers set out to answer two questions: who are the most influential users in Mounjaro conversations on X, and what is the overall sentiment of those conversations?

Their dataset: 5,566 tweets involving 5,641 unique users, collected between January 9 and February 24, 2025, connected by 6,810 interactions including replies, mentions, and retweets.

How they did it

This is a textbook example of combining network structure with content analysis, and it is worth walking through because each step maps to a specific NodeXL capability.

• Betweenness centrality identified the bridges: users who sit on the shortest paths between otherwise separate communities. These are the people who move information across the network, and they are frequently not the loudest or most followed accounts.
• In-degree centrality identified the primary information sources, measured by how many incoming connections each account received.

• Cluster analysis using the Clauset-Newman-Moore algorithm detected sub-communities inside the network, revealing the distinct topics people were actually discussing.

• Sentiment analysis using NodeXL’s Words and Word Pairs function classified the tone of the conversation and surfaced the specific words driving positive and negative attitudes.

That combination is the point. Content analysis alone tells you what is being said. Network analysis alone tells you how information moves. Together they tell you who shapes the conversation, how their message travels, and what the crowd feels about it.

Why this matters

The authors are careful about identifying the implications of their work. Positive sentiment is not inherently a problem, and it reflects genuine public interest in new treatments. But when optimism is amplified mainly by people without clinical training, and when the loudest cluster is broadcasting rather than discussing, the conditions are right for unrealistic expectations, off-label misuse, and misinformation to spread faster than correction.

Their recommendation is practical rather than alarmist: clinicians, pharmacists, and public health organizations need a stronger presence in these digital spaces, and campaigns should work with the existing optimism while adding balanced, evidence-based context about risks and limitations.

The study also makes a methodological point worth underlining. Isolated users and cohesive sub-communities call for different communication strategies. You cannot design an effective intervention if you cannot see the shape of the crowd you are trying to reach.

The limits, stated plainly

The authors are transparent about the limits of this work. The data covered only public, English-language tweets within a narrow time window. Automated sentiment analysis struggles with sarcasm and nuance. X users skew younger and more politically engaged than the general population. And findings from one platform do not necessarily transfer to TikTok, Instagram, Reddit, or Facebook, which differ in norms and demographics.

Given these caveats, the authors point toward useful future work: longer timeframes, multilingual analysis, and additional network measures such as density and modularity to probe echo chambers and expert-to-layperson interaction more deeply.

Seeing the pattern beneath the noise

What makes this study a good example of network analysis is that its central finding could not have been reached any other way. You can read a thousand tweets about Mounjaro and come away knowing that people are excited about it. Only a network map tells you that the excitement is being routed through people with no medical training, that the largest group is talking past itself rather than to itself, and that the concerns about cost and side effects live in small, tightly connected pockets off to the side.

That is the difference between reading a conversation and understanding it as a structural whole.

We were glad to support this work. The Social Media Research Foundation’s Marc Smith assisted with the NodeXL graphs and provided guidance throughout the data analysis.

Read the study

AlHazmi, S., AlRasheed, H. A., and Gomaa, B. (2026). Influence without medical expertise: a social network analysis of Mounjaro discussions on twitter (X). Saudi Pharmaceutical Journal, 34:41. https://doi.org/10.1007/s44446-026-00097-9

Open access under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.

Run this kind of analysis yourself

Every measure in this study, betweenness centrality, in-degree centrality, Clauset-Newman-Moore clustering, and sentiment analysis, is built into NodeXL and runs inside the Excel spreadsheet you already know. No programming required.

Start mapping your own networks.