About FormBlends Research

FormBlends Research is a reading service. It takes the primary publication of each pivotal GLP-1 and GIP/GLP-1 trial and writes down, in plain English and with the numbers left in, what that paper says.

What the site is

Twenty-four trial digests covering semaglutide (STEP, SELECT, FLOW, SUSTAIN-6, PIONEER 6), tirzepatide (SURMOUNT, SURPASS-2), retatrutide, orforglipron (ATTAIN) and CagriSema (REDEFINE). Each digest has the same shape: what the trial asked, how it was designed, who was enrolled, the primary endpoint and its result with the confidence interval, the key secondary results, adverse events and discontinuation, funding and registration, what the trial does not show, and where every number came from. Around the digests sit three methods guides, a master comparison table, and an interactive explorer that filters the trials and charts each arm.

Who runs it

The site is written and maintained by the FormBlends editorial team, which also publishes formblends.com. FormBlends operates a GLP-1 weight-loss programme that dispenses compounded medication through licensed pharmacies. That is a commercial interest, and it is disclosed here, in the footer, and wherever a page links to a FormBlends product or report. None of the trials digested here studied a compounded product. Compounded drugs are not FDA approved and are not interchangeable with the brand products that were tested.

No page on this site carries a named clinician, a "medically reviewed by" line, a testimonial or a rating. If a page ever appears to, it is not ours.

How a digest is produced

1. Retrieve the abstract from the source. Every PubMed abstract is fetched directly from the NCBI database through its E-utilities interface, by PubMed id, on the date shown at the bottom of the page. Where the paper is open access, the PMC full text is fetched the same way and used for figures the abstract lacks, chiefly discontinuation tables. The digest says which figures came from the full text.
2. Write only what the source states. If the abstract does not give a baseline weight, a confidence interval, a discontinuation rate or a funder, the digest says "not stated in the abstract" rather than filling the gap from a press release, a review, a label or memory. This is why several digests have blank fields. We would rather show the gap than paper over it.
3. Name the estimand. Treatment-policy, treatment-regimen, intention-to-treat, time-to-event, or "not stated". Numbers under different estimands are not swapped or averaged.
4. Keep the interval with the estimate. A point estimate is not published on this site without its 95 percent confidence interval when the abstract gives one, and the digest says when it does not.
5. Separate primary from secondary. The primary endpoint is what the trial was powered to test. Secondary results are reported as secondary.
6. Say what the trial does not show. Every digest ends with the questions the trial was not designed to answer, including the cross-trial comparisons it tempts you to make.
7. Check the data set against the digests. The explorer runs on a typed data file. Automated tests check that every trial has a PubMed id, a date, funding and at least two arms, that every confidence interval brackets its estimate, and that the file contains no em dashes. Those tests run before every deployment.

What we leave out on purpose

• Numbers from press releases, conference abstracts and investor presentations, even when they are widely reported. When a peer-reviewed primary paper appears on PubMed, we digest it.
• Secondary analyses and post hoc subgroup papers, unless a digest names one explicitly as a secondary source.
• Author names in citations. We cite by title, journal, year and PubMed id, which is unambiguous, and avoid the risk of misattributing a paper to the wrong first author.
• Mechanistic explanations for findings the primary papers do not explain. The sibling site The Science does that job.
• Recommendations. A trial digest tells you what happened to a randomised group. What it means for you is a conversation with a prescriber.

Reporting an error

If a digest states a number you cannot find in the cited abstract or full text, or misses one that is there, write to hello@formblends.com with the page URL, the figure, and the passage of the source you are looking at. Errors of fact are corrected as a priority, the page's updated date moves, and the correction is noted at the foot of the page.

What this site is not

It is not medical advice, it is not a substitute for reading the papers, and it is not a comparison between drugs except where a trial was head-to-head. Read the medical disclaimer before relying on anything here.

Sources

1. FormBlends Accessed September 4, 2026.
2. NCBI E-utilities (PubMed efetch), the retrieval route for every abstract cited on this site Accessed September 4, 2026.
3. FDA: Human Drug Compounding Accessed September 4, 2026.