What the chart shows
The diet switch chart on the home page draws, for each diet you tap, the mean body weight change that a published study reported, scaled to a 96 kg start. The light jitter on the line is cosmetic, under 0.3 kg, so it reads like a weigh in series rather than a plot.
The studies are randomised trials where one exists, with a meta-analysis or a second trial as a cross check. The table below names each one, with links to the publisher.
How a study becomes a line
Each study reports weight at a few time points. We take each as a percentage of that trial’s own starting weight, apply it to 96 kg and draw straight lines between the points.
Nothing is extended past a study’s last measurement. Where a trial stopped before week 30, the line holds flat and the chart says so.
What an average is not
An average is the middle of a wide spread. In STEP 1 the mean was 14.9% lower at 68 weeks, yet about one in seven people on the drug lost under 5% and about one in three lost 20% or more.
The people in these trials also had counselling, regular weigh ins and a team watching, and in the drug trials a placebo group lost weight too. Their average is not a forecast for anyone reading this page.
Undiet describes what the papers report; it does not advise. Before a diet, a fasting protocol or a medication, talk to your doctor or a dietitian.
Undiet 90 has no study
Undiet 90is the rotation program built into the app, and no trial has tested it. Its line is labelled as a model on the chart: the mean of two balanced diet arms, the PREDIMED-Plus Mediterranean group and the low fat group of Bazzano et al. 2014.
When a study of the program exists, this page and the chart will use it instead.
What happens when you switch
The second period on the chart follows its own study from that study’s first week, starting at the weight the first period ended on. No study tested one diet after another, so this is the honest shape, not a measured one.
Three findings about leaving a program are added on top. After semaglutide stopped in the STEP 1 extension, people regained two thirds of what they had lost within a year; after behavioural programmes end, each kilogram lost comes back at about 0.13 to 0.19 kg a year; and the 1.6 kg of water lost in the first two weeks of a ketogenic diet is modelled as returning when a low carb period ends.
Reading the table
Each row names the study, who was in it, how long it ran, the figure as the paper reports it, how it was turned into the line, and the caveat we think you should know. Links go to the publisher’s page.
Sample sizes are the number randomised unless the row says otherwise. Where a trial reported only a percentage, the kilograms are that percentage of 96 kg.
The studies, one by one
GLP-1
STEP 1
randomised, placebo controlled · n 1,961 · 68 weeks
Reported:semaglutide 2.4 mg once weekly: body weight 14.9% lower at 68 weeks vs 2.4% lower with placebo (15.3 kg vs 2.6 kg from a 105.3 kg baseline).
On the chart:Weekly means read off the published Figure 1A (observed in trial data, weeks 0 to 68, digitised from the article PDF), applied to 96 kg.
Caveat:With placebo controlled injections, dose escalation over 16 weeks, counselling every 4 weeks, a 500 kcal daily deficit and 150 minutes of activity a week. The placebo arm lost 2.4%.
GLP-1, cross check
SURMOUNT-1
n 2,539 · 72 weeks
Reported:tirzepatide 5, 10 and 15 mg: 15.0%, 19.5% and 20.9% lower at 72 weeks vs 3.1% with placebo.
On the chart:Not drawn; listed so the line can be checked against a second source.
Keto
Foster et al. 2010
randomised, low carbohydrate (20 g a day induction) vs low fat · n 307 · 2 years
Reported:low carbohydrate arm: 9.5 kg lower at 3 months, 12.2 kg at 6 months, 10.9 kg at 12 months, 6.3 kg at 24 months (103.3 kg baseline); the low fat arm lost the same.
On the chart:Table 2 kg values divided by the 103.3 kg baseline, applied to 96 kg.
Caveat:With weekly group behavioural sessions for 20 weeks and prescribed walking. The low fat arm lost the same amount, so most of this line is the programme, not the carbohydrate rule.
Keto, cross check
Bueno et al. 2013 meta-analysis
13 trials · n 1,577 · 12 months or longer
Reported:very low carbohydrate ketogenic vs low fat diets at 12 months or longer: 0.91 kg more lost on keto (95% CI 0.17 to 1.65 kg).
On the chart:Not drawn; listed so the line can be checked against a second source.
Fasting 16:8
TREAT
randomised, 16:8 (noon to 8 pm) vs three structured meals · n 116 · 12 weeks
Reported:0.94 kg lower (1.17%) at 12 weeks on 16:8 vs 0.68 kg lower (0.75%) in the control group; the difference was not significant.
On the chart:The reported 1.17% of the 99.3 kg baseline, applied to 96 kg; straight line to week 12, then held flat because the trial stopped there.
Caveat:No calorie target, no counselling, daily weigh ins on a connected scale. 141 were randomised and 116 analysed.
Fasting 16:8, cross check
Chen et al. 2023 meta-analysis
9 trials · n 665 · 8 to 14 weeks
Reported:time restricted eating vs control in adults with overweight or obesity, mostly 8 to 14 week trials: 1.28 kg more lost (95% CI 0.52 to 2.05 kg).
On the chart:Not drawn; listed so the line can be checked against a second source.
Vegan
Kahleova et al. 2020
randomised, low fat vegan diet vs no change · n 244 · 16 weeks
Reported:6.4 kg lower at 16 weeks on the low fat vegan diet (93.6 kg baseline) vs 0.5 kg lower in the control group.
On the chart:6.4 kg divided by the 93.6 kg baseline (6.8%), applied to 96 kg; held flat after week 16 because the trial stopped there.
Caveat:Ad libitum, with weekly classes. Weight was measured at weeks 0 and 16 only, so the line between them is straight. Run by the Physicians Committee for Responsible Medicine.
Low-carb
Bazzano et al. 2014
randomised, under 40 g digestible carbohydrate a day vs low fat · n 148 · 12 months
Reported:low carbohydrate arm: 5.7 kg lower at 3 months, 5.6 kg at 6 months, 5.3 kg at 12 months (96.3 kg baseline); low fat arm 1.8 kg lower at 12 months.
On the chart:Table 3 kg values divided by the 96.3 kg baseline, applied to 96 kg.
Caveat:With dietitian sessions, weekly at first then monthly, and one meal replacement a day in both arms.
Low-carb, cross check
DIETFITS
n 609 · 12 months
Reported:healthy low carbohydrate 6.0 kg lower vs healthy low fat 5.3 kg lower at 12 months; the difference was not significant.
On the chart:Not drawn; listed so the line can be checked against a second source.
Mediterranean
PREDIMED-Plus, one year results
randomised, energy restricted Mediterranean diet with activity and support vs usual Mediterranean advice · n 626 · 12 months
Reported:2.4 kg lower (2.7%) at 6 months and 3.2 kg lower (3.7%) at 12 months in the intervention group vs 0.4 kg and 0.7 kg in the control group.
On the chart:The reported percentages (2.7% at 26 weeks, 3.7% at 52 weeks), applied to 96 kg.
Caveat:Energy restricted, with activity promotion and behavioural support, in adults aged 55 to 75 with metabolic syndrome. Measured at 6 and 12 months, so the line to month 6 is straight.
Undiet 90
No study. Modelled as the mean of two balanced diet arms
see the two arms
Reported:No study of Undiet 90 exists. The line is the mean of two balanced diet arms: the PREDIMED-Plus Mediterranean group and the low fat group of Bazzano et al. 2014 (2.6 kg lower at 3 months, 2.3 kg at 6 months, 1.8 kg at 12 months from 97.9 kg)..
On the chart:Mean of the two arms, each interpolated weekly from its reported time points, applied to 96 kg.
Caveat:Not a study of this program. A labelled model, the mean of two balanced diet arms, shown so the chip has an honest line rather than an invented one.
PREDIMED-Plus, one year resultsBazzano et al. 2014, low fat arm
No diet
Control groups of TREAT and Kahleova et al. 2020
the no change control arms of two randomised trials (57 and 122 people) · n 179 · 12 to 16 weeks
Reported:control arms asked to change nothing: 0.68 kg lower (0.75%) over 12 weeks in TREAT and 0.5 kg lower over 16 weeks in Kahleova et al..
On the chart:Mean of the two control arms (0.75% and 0.54%) placed at their mean time point, week 14, applied to 96 kg; held flat after that.
Caveat:People in a trial weigh in and know they are watched, which is why even control arms drift down a little. Outside trials, adults gained about 1.5 kg per four years (Mozaffarian et al. 2011, 120,877 people).
No diet, cross check
Mozaffarian et al. 2011
n 120,877 · up to 20 years
Reported:three cohorts followed up to 20 years: an average gain of 3.35 lb (1.5 kg) per four year period.
On the chart:Not drawn; listed so the line can be checked against a second source.
The switch
STEP 1 extension
n 327
Reported:one year after semaglutide and the lifestyle programme stopped, participants had regained 11.6 of the 17.3 percentage points lost, two thirds of the prior weight loss.
On the chart:After a GLP-1 period ends, two thirds of what was lost is added back evenly over the following 52 weeks.
The switch
Hartmann-Boyce et al. 2021
249 trials · n 59,081
Reported:after behavioural weight loss programmes end, each kilogram lost was associated with faster regain at 0.13 to 0.19 kg a year; the difference from controls lasted at least five years.
On the chart:After any diet period ends, 0.16 kg a year for each kilogram lost is added back, evenly.
The switch
Hall et al. 2016
n 17
Reported:in the first 15 days of an isocaloric ketogenic diet body weight fell 1.6 kg, which the authors attribute mainly to water, with fat mass down only 0.2 kg.
On the chart:When a keto or low carb period ends and the next one is not low carb, that 1.6 kg of water is modelled as returning over 15 days. A metabolic ward study of 17 men; the return is a model, not a measured figure.