Most obesity trials run a lifestyle program alongside the drug. SURMOUNT-3 ran one before the drug, and then used the result as an entry ticket. Twelve weeks of intensive counseling on a 1,200 to 1,500 kcal diet came first, and only the participants who had lost at least 5% of their body weight by the end of it were randomized to anything. That design is usually described as giving tirzepatide a head start. It is more accurately described as a screening instrument, and the trial’s own authors say so in their limitations: the people the lifestyle program failed never reached the randomization that produced the headline. The dose-response row this trial sits in belongs to the tirzepatide weight article. What follows is the machinery.
The funnel, in three numbers
A total of 972 people were assessed for eligibility. 806 entered the 12-week intensive lifestyle lead-in, which delivered eight in-person counseling sessions from a dietitian or equivalent professional, instructed women to eat roughly 1,200 kcal/day and men roughly 1,500, and permitted up to two meal replacements a day. Of those 806, 579 (71.8%) reached the 5% threshold and were otherwise eligible, and were randomized 1:1 to tirzepatide at the maximum tolerated dose (n = 287) or placebo (n = 292) for 72 weeks.[1]
Across the lead-in, mean body weight in those 579 fell from 109.5 kg at screening to 101.9 kg at randomization, a mean reduction of 6.9% among successful completers. The randomized cohort had a mean age of 45.6 years, was 62.9% female and 86.0% white, had lived with obesity for a mean of 15.1 years, and 66.1% carried at least one obesity-related complication.[1]
What the two arms did next
The coprimary endpoints were the additional percentage weight change from randomization to week 72 and the share achieving an additional reduction of 5% or more. Under the treatment-regimen estimand, which counts everybody regardless of adherence, the tirzepatide group changed by −18.4% (SE 0.7) and the placebo group by +2.5% (SE 1.0) — an estimated treatment difference of −20.8 percentage points (95% CI, −23.2 to −18.5; P < 0.001). An additional 5% or more was reached by 87.5% against 16.5% (odds ratio 34.6; 95% CI, 19.2 to 62.6).[1]
A second estimand sits beside it in the same tables and must never be stacked on the first. The efficacy estimand, which asks what happens on treatment, returned −21.1% against +3.3%, a difference of −24.5 percentage points (95% CI, −26.1 to −22.8).[1] Three points of apparent extra effect come from changing the question, not from changing the drug.
There is a third denominator as well. Measured from the start of the lead-in rather than from randomization, the sequence produced a total weight change of −24.3% against −4.5%, a difference of 19.9 percentage points (95% CI, −23.5 to −16.2).[1] Note which way that moves: the total figure is larger than the randomized one, and the gap between the arms is smaller, because the lifestyle program contributed to both columns.
The filter, in the authors’ own words
The limitations section is unusually direct. It records that 17.5% of participants did not lose at least 5% of baseline weight during the lifestyle intervention and were therefore not randomized to medication, and states that — to the extent that response to lifestyle intervention predicts response to medication — excluding them may have produced a higher mean weight loss with tirzepatide than would have been seen otherwise. It goes on to call for trials of drug response in people who are unsuccessful with lifestyle intervention, noting that a failed lifestyle attempt has commonly been a prerequisite for starting drug therapy at all.[1]
That is the shape of the finding. A telehealth service typically prescribes because diet and exercise did not work, which is close to the opposite of SURMOUNT-3’s inclusion rule. The trial does not say the drug fails in lifestyle nonresponders; it says nobody has measured it, and its authors are the ones saying so.
The placebo arm and the maintenance endpoint
Under the treatment-regimen estimand, 94.0% (270) of the tirzepatide group held at least 80% of the weight the lead-in had removed, against 43.8% (128) on placebo (odds ratio 19.7; 95% CI, 10.3 to 37.6).[1] Read the placebo column as a standalone statement: more than half of the people who had just succeeded on a supervised program, and who stayed on lifestyle counseling for another seventy-two weeks, could not keep four-fifths of what they had lost.
That is not a fixed law of lead-in trials, and one built the same way disagrees. SCALE Maintenance randomized 422 adults who had lost at least 5% during a low-calorie-diet run-in — a mean of 6.0% (SD 0.9) — to liraglutide 3.0 mg or placebo for 56 weeks. Its drug arm lost a further 6.2% (SD 7.3) and its placebo arm lost 0.2% (SD 7.0), a difference of 6.1 percentage points (95% CI, −7.5 to −4.6), with 81.4% against 48.9% maintaining the 5% or more they had lost during the run-in (odds ratio 4.8; 95% CI, 3.0 to 7.7).[2] One placebo arm gained 2.5% and the other held flat, over different durations and after different lead-ins. Regain after a diet is a tendency, not a constant.
A maximum tolerated dose is not a dose
Participants were not assigned to 10 mg or to 15 mg. They were assigned to whichever of the two they could tolerate, with de-escalation and reescalation permitted during titration.[1] That choice reflects prescribing practice and it removes something: SURMOUNT-3 cannot report a dose-response, cannot say what 15 mg does on its own after a lead-in, and cannot be lined up against a fixed-dose arm anywhere else without noting that its exposure was individualized. The fixed-dose version of the same drug is described in the SURMOUNT-1 article, and the mechanics of climbing to either dose in the titration article.
Going lifestyle-first cost tolerability
Adverse events ended treatment in 30 participants (10.5%) on tirzepatide against 6 (2.1%) on placebo. Nausea was reported by 39.7% against 14.0%, diarrhea by 31.0% against 9.2%, constipation by 23.0% against 6.8% and vomiting by 18.1% against 1.4%. Serious adverse events ran 5.9% against 4.8%, with one death in each arm.[1]
Set the 10.5% beside the same molecule without a lead-in. In SURMOUNT-1, adverse events caused treatment discontinuation in 4.3%, 7.1% and 6.2% of the 5 mg, 10 mg and 15 mg arms, against 2.6% on placebo.[3] SURMOUNT-3’s figure is higher than any of them, and its authors say so, offering the hypothesis that caloric restriction reduces endogenous glucagon-like peptide-1 and other gastrointestinal satiety hormones, and that this may worsen early tolerability of an incretin drug.[1]
The supporting example they reach for does not hold on this particular measure, and a page repeating their argument should say so. They cite STEP 3 against STEP 1 for semaglutide. STEP 3, with thirty counseling visits and an eight-week low-calorie diet, discontinued treatment for gastrointestinal events in 3.4% of its semaglutide group;[4] STEP 1, with a far lighter lifestyle program, did so in 4.5%.[5] That runs the other way. The SURMOUNT-3 observation stands on its own comparison with SURMOUNT-1; the semaglutide pair does not reinforce it. The trial that ran that semaglutide design is covered in the STEP 3 article.
Who finished, and who left
Of the 579 randomized, 479 (82.7%) completed the study — 87.8% on tirzepatide and 77.7% on placebo — and 429 (74.1%) completed it on treatment, 78.7% and 69.5% respectively. The leading reason for stopping treatment on tirzepatide was an adverse event at 10.5%, followed by withdrawal by the participant at 6.3%. On placebo, the leading reasons were withdrawal by the participant at 14.4% and loss to follow-up at 6.5%.[1]
The two arms therefore emptied for different reasons: one because of how the drug felt, the other because people stopped turning up. A trial with a gaining placebo arm gives its comparator group little to stay for, and that asymmetry is worth holding onto when reading the 20.8-point gap.
What SURMOUNT-3 does not answer
It had no arm that received tirzepatide without the lead-in, so it cannot say from inside itself whether the sequence beat the drug alone. It excluded diabetes, enrolled only in North and South America, and ran a cohort 86.0% white, which its authors name as a limit on generalizability.[1] It stopped at week 72 and followed nobody afterward — what happens then is the subject of the stopping article.
And it did not test the order. Lifestyle first, then drug, is one sequence; starting both at once is another, and nothing here randomized anyone between them. The discussion’s suggestion that going lifestyle-first could maximize weight reduction is a reading of two separate trials, not a result.[1]
What was in the syringe
The figures above describe branded, FDA-approved tirzepatide at 10 or 15 mg, titrated under supervision with permission to step down, following twelve weeks of professional dietary counseling that no subscription includes. Sellers on the tirzepatide board largely dispense compounded preparations, which are not FDA-approved and are not reviewed by the FDA for safety, efficacy or quality before they are dispensed — the distinction drawn in the compounded-versus-brand article.
The subtraction a buyer should make here is unusual, because it runs in both directions. The 24.3% total belongs to a program plus a drug, and only one of those is for sale. But the 10.5% discontinuation belongs to that same sequence too, and a person starting a compounded vial with no preceding diet is not obviously buying the harder version.