We’re planning our next deep-dive survey and want your input! After the success of our GLP-1s survey, we're looking for another intervention with active community discussion, but limited structured data.
If you have a strong preference, or a treatment you think is underrepresented in the conversation, we’d love to hear from you.
GLP-1 agonists are the most actively discussed experimental treatment in Long COVID and ME communities — and the most polarizing. Our survey collected 120 free-text anecdotes from patients who had tried Tirzepatide, Semaglutide, or other GLP-1 drugs, making it the largest structured patient dataset on the subject.
The analysis covers overall outcome distributions, onset timing, which symptoms improved most, side effect profiles, and subgroup breakdowns by baseline severity, illness duration, diagnosis, and drug. Every section is paired with verbatim patient quotes.
The problem
Trying treatments for Long COVID and ME is like a game of minesweeper: patients don't know if the next one they try is going to make them better, worse, or just waste their time and money to no effect.
Make no mistake: this is an information asymmetry problem. Despite there being hundreds of potential treatments, detailed information is sparse and patients end up blindly choosing which one they're going to add next to their stack.
Projects like TREATME, Longhaulwiki, and Stuffthatworks are helpful for rough prioritization: they ran large surveys to rank treatments by how much they helped, but once you zoom in on any single treatment and the data thins out: usually just a rough outcome score and little else.
Clinical trials offer the promise of more rigor and better data, but they take years to complete, and even the best-run trials rarely offer sufficient subtyping: too often a trial fails its primary endpoint, while a subset of patients actually improved!
Reddit is the only remaining option to get on-the-ground data, but redditors aren't incentivized to leave detailed notes. Searching for treatment anecdotes returns dozens of posts and comments that might say "it helped" or "made me worse", but nothing else.
Luckily, there's another way.
Patient questions that keep going unanswered
- Am I in a group likely to respond - or likely to get worse?
- What biomarkers correlate with response or harm?
- What prior treatments might correlate with better or worse outcomes?
- What dosing strategy gives the best outcomes?
- What side effects should I watch for?
- How long until I see an effect, and how long will it last?
- What is the full distribution of outcomes - no effect, partial, remission, worsening?
- Which treatment combinations improve results, and which blunt them?
How we're solving it
One treatment at a time, surveyed in depth
Rather than broad multi-treatment surveys, we publish one focused survey per round, examining one popular treatment that lacks structured trials and organized anecdata.
Surveys are short and easy to fill out, with a minimal number of structured questions - accessible even for patients dealing with significant cognitive symptoms.
Finally, instead of constraining patients to endless banal multiple-choice questions, each survey hinges on a freeform anecdote section - a large open text field where you describe your experience in your own words.
AI-assisted analysis
We use AI to simplify and speed up analysis, looking for correlations across responses - surfacing patterns that constrained checkboxes would miss entirely. Freeform anecdotes are translated into structured data, so your choice of words becomes analyzable signal without forcing you into boxes upfront.
Treatments are stratified by, among others, suspected subtype, baseline condition severity, and other successful treatments tried, and plotted against outcome and side effect profiles. Analyses are published here as each survey closes - not just did it work, but who it worked for, and why.
Stay in the loop
Get notified when new surveys open and analyses are published.
About
This project is organized by Highly Agentic LC/ME, a group of patients from tech and research backgrounds volunteering their time to give back to the community.