How we produce our data

Why this page exists

We cite our own data on several pages, so the method should be as inspectable as the journals we cite. If we ask you to weigh a number we produced, you should be able to see how it was produced and where it stops being reliable.

The 25,000-post analysis

Before building GLPme we analysed more than 25,000 public GLP-1 discussions to find out what people actually struggle with, rather than what a feature list assumes they struggle with.

The output is a topic distribution: what share of the conversation each theme accounts for. A single post can touch several themes, so the shares sum above 100%.

What this data can and cannot tell you

It measures what people talk about publicly. That is not the same as what people experience. Public discussion over-represents the unusual, the frightening and the frustrating, and under-represents the person whose treatment is going fine and who therefore posts nothing.

So we use it for one thing only: deciding what to build and what to write about. We never present it as clinical prevalence. When this site says 27.3% of discussions concerned side effects, that is a statement about a conversation, not about a population. Every clinical rate on this site comes from a trial or a label, and is cited as such.

Trial curve benchmarking

GLPme compares a user's weight trajectory to published trial curves so a slow month reads as a slow month rather than as failure.

The reference curves come from the published STEP programme for semaglutide and SURMOUNT for tirzepatide. We plot the trial's mean trajectory against elapsed time on treatment.

The limits are worth stating plainly. Trial participants are not a random sample of everyone taking these drugs: they had structured lifestyle support, monitored adherence, and eligibility criteria you may not match. A trial mean also hides enormous individual spread. The comparison is context, not a target and not a prediction, and we say so in the product as well as here. The public version of the same maths is the weight loss projection tool.

How the calculators are verified

Every calculator on this site is checked against an independent implementation before it ships.

The dose conversion, unit conversion, supply and projection tools each have a verification script that recomputes the same results from the same inputs using separately written code, and compares the outputs. A mismatch blocks the build. The tools also run entirely in your browser: nothing you type into them is transmitted, stored or logged anywhere.

What the verification does not do is check clinical appropriateness. A unit conversion can be arithmetically perfect and still be the wrong dose for you. That is why every dose tool states, on the page, that it converts a number your prescriber gave you and does not suggest one.

Dates, versions and drift

Regulatory positions change. NAION was not on the semaglutide label before 2025. The FDA's proposed exclusion of semaglutide from the 503B Bulks List is dated April 2026. Anything we write about the compounding market has a shorter shelf life than anything we write about trial results.

Each page carries the date it was last substantively updated. Where a claim is time-sensitive, we date the claim in the sentence rather than relying on the page date. Sources are listed at the bottom of every guide so you can check whether the underlying document has moved.

Keep reading