Key takeaways
- A trial curve is an average of many people. It is not a prediction for one person.
- Papers report change as a percentage of starting weight, at set weeks.
- Two kinds of analysis can give different averages from the same trial.
- Spread between people is as important as the average, and is often shown separately.
- This page holds no numbers. Every figure is a placeholder to fill from the paper, with a citation.
Why this guide has no numbers on it
Tools that estimate "how long until my goal" usually draw a trial's average curve and rescale it to your weight. The idea is easy to misread. This guide explains how the curves are made, so you can read the originals yourself. We do not compare results with other people’s data, and we do not copy figures from memory. Each figure below is marked to be filled from the paper.
Where to look
Two well-known places to start are the main papers for the large trials of tirzepatide and semaglutide in people with obesity. Find them by name:
- SURMOUNT-1, tirzepatide. [citation: authors, journal, year, DOI]
- STEP 1, semaglutide 2.4 mg. [citation: authors, journal, year, DOI]
Each has a table or figure of weight change by week and by dose group. Open the full text and look for the figure that plots the mean percentage change in body weight over time.
How the figures are built
Percentage of starting weight
The usual unit is percent change from baseline. Baseline is the person’s weight at the start. For one person, change in % = (weight now − baseline) ÷ baseline × 100. The paper then averages across everyone in the group.
Percentages let people of different sizes be averaged together. They also mean that a given percentage is a different number of pounds for different people.
Time points
Papers report at scheduled visits, for example every few weeks. The curve between visits is a line joining the dots. It is not a measurement.
| Week | Mean change, group A | Mean change, group B | Placebo |
|---|---|---|---|
| [week] | [% from paper] | [% from paper] | [% from paper] |
| [week] | [% from paper] | [% from paper] | [% from paper] |
| [week] | [% from paper] | [% from paper] | [% from paper] |
Which people are in the average
Trial reports often give two versions of the same result:
- Treatment-regimen (intention-to-treat) estimand. Everyone who was randomised counts, including people who stopped the medicine or changed dose.
- Efficacy (on-treatment) estimand. Only people still taking the medicine as planned count, or the result is modelled as if all did.
The second average is usually larger. Neither is wrong. They answer different questions. A tool that does not say which it uses can mislead. [state which estimand each figure comes from, with the paper’s table]
Dose groups
Trials often have several dose arms, including a placebo arm. A curve for one arm applies to people who took that dose. Placebo arms also lose or gain weight on average, because trial participants usually receive lifestyle support.
Average versus spread
Averages hide a range. Papers show this in a few ways: the share of people who reached given thresholds of weight loss, confidence intervals, or a spread of individual responses. [share of participants reaching each threshold, from paper] A person at the low end of the spread and a person at the high end took the same medicine.
What a curve does not tell you
- How you will do. It describes a group.
- What happens after stopping. Look for a separate report or extension study. [citation]
- Safety or tolerability. Those are reported in other sections of the paper. [citation]
- Whether a goal is right for you. That is for you and your clinician.
A worked example of the arithmetic
Example. These numbers are made up to show the steps and do not come from any trial.
A person starts at 200 lb. A made-up curve reads −10% at some week.
- Change = 200 × 0.10 = 20 lb.
- Weight at that week = 200 − 20 = 180 lb.
Running it in reverse: for a goal of 170 lb from 200 lb, the needed change is (200 − 170) ÷ 200 = 15%. A tool would look for the week where the curve reaches −15%. If the curve never reaches it, there is no answer. That is a fact about the average, not about you.
Frequently asked questions
Can I use a trial curve to set a date?
It can give a rough shape. It cannot give a date. Individual results spread widely around the average.
Why do two sources quote different averages?
They may use different estimands, different dose groups or different weeks.
Why is the placebo group in the table?
Because it shows what happened without the medicine, under the same support.
Where do I find the curve?
In the main results paper, usually a figure of mean percent change in body weight by week. [figure number in each paper]
Does Reko predict my weight?
No. Reko records your weights and doses and draws your own line.
How Reko does this in the app
Reko does not forecast. It draws your own weight over time, in lb, st or kg, next to your logged doses, side effects, and the calendar. You set the units. The record is yours to show a clinician. Doses export as a spreadsheet and everything as a file.
Reko records what you and your prescriber decide. It never suggests a dose, and this page is not medical advice.
- SURMOUNT-1 main results paper. [full citation and DOI]
- STEP 1 main results paper. [full citation and DOI]
- Statistical note on estimands in clinical trials: ICH E9(R1). [link to guideline]