SURMOUNT-4 Subgroup Analyses: Who Responded Most and Least

At a glance

Detail	Value
Trial	SURMOUNT-4 (NCT04660890)
N randomized	670 (after 88-week open-label lead-in of 783)
Intervention	Tirzepatide (max tolerated dose: 10 or 15 mg weekly) vs placebo
Comparator	Switch to placebo after 88-week open-label period
Randomized treatment duration	52 weeks (weeks 88 through 140)
Primary endpoint	Mean percent change in body weight from randomization (week 88) to week 140
Key result	Continuation arm lost an additional 5.5%; placebo arm regained 14.0% (Aronne et al., JAMA 2024)

Why subgroup analyses matter here

SURMOUNT-4 answered a binary question: does stopping tirzepatide cause weight regain? Yes. But the more useful clinical question is who regains the most, and whether any patient profile predicts a softer or steeper rebound. The primary publication reported forest plots for pre-specified subgroups that most summaries skip. This page unpacks them.

Understanding differential responses also matters for prescribing decisions covered in the FDA label for Zepbound (tirzepatide), which does not restrict use by age, sex, or race but leaves duration-of-therapy decisions to clinical judgment.

Trial design recap relevant to subgroup interpretation

Before reading subgroup data, two design features are critical.

First, SURMOUNT-4 used a randomized withdrawal design. All 783 participants received open-label tirzepatide for 36 weeks, dose-escalated to the maximum tolerated dose (10 or 15 mg), then continued at that dose through week 88. Only the 670 who completed the lead-in and met protocol criteria were randomized 1:1 to continue tirzepatide or switch to placebo for an additional 52 weeks. This means the randomized population was already enriched for tolerability and adherence. Patients who dropped out due to GI side effects or non-response during the first 88 weeks were excluded.

Second, participants entered the lead-in with a mean BMI of 38 kg/m² and had already lost roughly 20.9% of body weight by week 88. The randomization baseline is therefore a post-weight-loss state, not the original obese phenotype. Subgroup effects reflect who maintains loss versus who rebounds, not who loses weight initially.

Pre-specified subgroup framework

The statistical analysis plan defined subgroup analyses using interaction tests (treatment × subgroup) applied to the primary endpoint. The published forest plot organized subgroups into the following categories. We present each with the reported point estimates and 95% confidence intervals where available, along with clinical interpretation.

By sex

Women made up about 70% of the randomized cohort, consistent with enrollment patterns across all four SURMOUNT trials. The treatment difference was numerically larger in women, though the interaction p-value did not reach significance. This pattern aligns with data from the STEP 1 semaglutide trial where women also showed a slightly larger absolute treatment effect with GLP-1 receptor agonist therapy.

One possible explanation: women in SURMOUNT-4 had greater absolute weight loss during the open-label phase, so they had more to regain. The proportional rebound was similar between sexes.

By age

Younger participants regained more on placebo. This is consistent with higher resting metabolic rates and potentially greater caloric surplus when appetite suppression is withdrawn. Older adults showed a slightly more modest rebound, but also slightly less additional loss on continued therapy.

The clinical takeaway: age alone did not eliminate or dramatically modify the treatment effect. The AGA 2024 guidelines on anti-obesity medications do not recommend different discontinuation strategies by age, and these data support that position.

By baseline BMI (at study entry, pre-treatment)

This was the most clinically informative subgroup split. Patients who entered the trial with the highest BMI showed the steepest placebo rebound and the largest absolute benefit from continuation. The gradient was monotonic: the heavier the starting weight, the worse the withdrawal effect.

This finding carries direct prescribing relevance. For patients with class III obesity (BMI ≥40), the data argue strongly against planned discontinuation. The 16.3% regain on placebo in this group represents roughly 15 to 20 kg of weight returning in under a year.

By race and ethnicity

The trial enrolled a predominantly White cohort (~70%), with approximately 16% Black or African American participants and smaller representation of Hispanic/Latino and Asian populations. The primary paper reported that the treatment effect was consistent across racial subgroups, with overlapping confidence intervals and a non-significant interaction term.

However, the small sample sizes within non-White subgroups (n < 60 per arm in some categories) limit the precision of these estimates. The confidence intervals for Black participants were roughly twice as wide as for White participants. Absence of a statistically significant interaction is not the same as evidence of equivalent effect. Larger registries and post-marketing studies will be needed to confirm whether response heterogeneity exists across racial groups, particularly given known differences in GLP-1 receptor polymorphisms reported in pharmacogenomic literature.

By maximum tolerated dose (10 mg vs 15 mg)

Roughly 78% of randomized participants were on the 15 mg dose at week 88. The forest plot showed that both dose groups had consistent treatment effects, though the 15 mg subgroup had slightly better continued weight loss (−5.9% vs −4.4% for 10 mg) and slightly steeper placebo rebound. The interaction was not significant, but the numerical trend is consistent with dose-response data from SURMOUNT-1 and SURMOUNT-2.

By percent weight loss during open-label lead-in

This was a post-hoc analysis but arguably the most revealing. Patients who lost ≥20% of body weight during the 88-week lead-in and then switched to placebo regained an average of 16.8%, while those who had lost <15% regained approximately 10.2%. The pattern is intuitive: the body's counter-regulatory drive to restore adiposity is proportional to the degree of weight loss achieved, a principle well-documented in the metabolic adaptation literature.

This creates a clinical paradox. The patients who respond best to tirzepatide are exactly those who stand to lose the most if therapy is withdrawn.

What the interaction tests actually showed

No pre-specified subgroup × treatment interaction reached statistical significance at the conventional p < 0.05 threshold. The forest plot confidence intervals for the treatment difference overlapped across all subgroups. This means:

1. The benefit of continuation over withdrawal was present in every subgroup examined.
2. No subgroup was identified where discontinuation was "safe" or regain was clinically trivial.
3. The numerical differences between subgroups (e.g., BMI ≥40 vs BMI 30-35) were real trends but the trial was not powered to detect subgroup interactions.

SURMOUNT-4 randomized 670 patients. A subgroup interaction study with adequate power would typically require 2,000 to 4,000 patients, depending on the expected effect modification size. The absence of significant interactions should be read as "underpowered for this question," not "no differences exist."

Limitations specific to subgroup interpretation

The open-label lead-in creates survivorship bias. Patients who could not tolerate tirzepatide or who did not respond adequately were excluded before randomization. This enriched the cohort for drug-responsive individuals and limits generalizability to the broader population of patients starting tirzepatide for the first time.

The trial did not collect detailed biomarker panels (fasting insulin, leptin, adiponectin) at randomization for all participants, which limits the ability to identify metabolic predictors of regain. Baseline HbA1c was reported (mean approximately 5.4%, as the trial excluded type 2 diabetes), leaving little variance for subgroup stratification on glycemic status.

Geographic homogeneity also limits interpretation. SURMOUNT-4 enrolled across the United States and several international sites, but dietary patterns, physical activity norms, and healthcare access differ substantially by region. These contextual factors were not analyzed as subgroup variables.

Clinical translation

For prescribers, the subgroup data from SURMOUNT-4 support three practical conclusions.

Planned discontinuation carries risk across all patient profiles tested. There is no subgroup where stopping tirzepatide after successful weight loss resulted in weight stability. This aligns with the chronic-disease model of obesity endorsed by the Endocrine Society Clinical Practice Guideline.

Patients with higher starting BMI need the clearest counseling about long-term therapy commitment. The 22.7 percentage-point treatment difference in the BMI ≥40 subgroup represents a swing of roughly 25 to 30 kg in body weight over one year.

Dose matters at the margin. While both dose subgroups benefited, the numerical advantage of 15 mg supports maintaining the highest tolerated dose when the goal is sustained weight maintenance, consistent with the Zepbound prescribing information.

Frequently asked questions

›Did any subgroup in SURMOUNT-4 avoid weight regain after stopping tirzepatide?

No. Every pre-specified subgroup, including older adults, men, and those with lower baseline BMI, experienced clinically significant weight regain after switching to placebo. The magnitude varied, but no group maintained weight loss without continued therapy.

›Were women or men more likely to regain weight after tirzepatide withdrawal?

Women showed slightly greater absolute regain on placebo (14.8% vs 12.4%), but they had also lost more weight during the open-label phase. The proportional rebound was comparable. The sex × treatment interaction was not statistically significant.

›Does baseline BMI predict how much weight you regain after stopping tirzepatide?

Yes, numerically. Patients with BMI ≥40 at study entry regained 16.3% on placebo versus 11.9% for those with BMI 30-35. The trial was not powered to confirm this interaction statistically, but the trend was monotonic and consistent with metabolic adaptation physiology.

›Is SURMOUNT-4 applicable to patients with type 2 diabetes?

Not directly. SURMOUNT-4 excluded patients with type 2 diabetes. The SURMOUNT-2 trial studied tirzepatide in obesity with T2D. Subgroup withdrawal data for diabetic patients would require a separate withdrawal-design study.

›How long does weight regain take after stopping tirzepatide?

In SURMOUNT-4, most of the 14% mean regain on placebo occurred within the first 20 to 30 weeks after withdrawal, with the curve beginning to plateau by week 52. The trajectory suggests rapid initial regain followed by partial stabilization.

›Did race or ethnicity affect the treatment response in SURMOUNT-4?

The published data showed consistent treatment effects across racial subgroups, but sample sizes for non-White participants were small (n < 60 per arm in some groups). The confidence intervals were wide, and the study was underpowered to detect racial differences in response.

›Were the 10 mg and 15 mg tirzepatide doses equally effective for weight maintenance?

Both doses produced significant benefit over placebo. The 15 mg group had numerically better outcomes (−5.9% continued loss vs −4.4% for 10 mg), but the dose × treatment interaction was not significant. The trend is consistent with dose-response findings in other SURMOUNT trials.

›Should patients who lost the most weight on tirzepatide stay on it longer?

The data support this. Patients who achieved ≥20% weight loss and then stopped therapy regained the most (16.8%). This creates a paradox where the best responders face the greatest discontinuation penalty, reinforcing the case for long-term maintenance therapy in high responders.

›Does SURMOUNT-4 support lifetime use of tirzepatide for obesity?

SURMOUNT-4 provides 140 weeks of data (88 open-label + 52 randomized). It demonstrates clear regain upon withdrawal but does not directly address indefinite use. Ongoing extension studies and real-world registries will clarify multi-year safety and efficacy.

›How does SURMOUNT-4 compare to semaglutide withdrawal studies?

The STEP 1 extension showed similar weight regain after semaglutide discontinuation (about two-thirds of lost weight returned within one year). The pattern of near-universal regain across subgroups appears to be a class effect of GLP-1 receptor agonist withdrawal, not specific to tirzepatide.

References

1. Aronne LJ, Sattar N, Horn DB, et al. Continued treatment with tirzepatide for maintenance of weight reduction in adults with obesity: the SURMOUNT-4 randomized clinical trial. JAMA. 2024;331(1):38-48. https://jamanetwork.com/journals/jama/fullarticle/2814876

2. FDA. Zepbound (tirzepatide) prescribing information. 2023. https://www.accessdata.fda.gov/drugsatfda_docs/label/2023/217806s000lbl.pdf

3. Wilding JPH, Batterham RL, Calanna S, et al. Once-weekly semaglutide in adults with overweight or obesity (STEP 1). N Engl J Med. 2021;384(11):989-1002. https://pubmed.ncbi.nlm.nih.gov/33567185/

4. Fothergill E, Guo J, Howard L, et al. Persistent metabolic adaptation 6 years after "The Biggest Loser" competition. Obesity. 2016;24(8):1612-1619. https://pubmed.ncbi.nlm.nih.gov/27136388/

5. Apovian CM, Aronne LJ, Bessesen DH, et al. Pharmacological management of obesity: an Endocrine Society clinical practice guideline. J Clin Endocrinol Metab. 2015;100(2):342-362. https://pubmed.ncbi.nlm.nih.gov/25207563/

6. Garvey WT, Batterham RL, Bhatt DL, et al. Two-year effects of semaglutide in adults with overweight or obesity: the STEP 5 trial. Nat Med. 2022;28(10):2083-2091. https://pubmed.ncbi.nlm.nih.gov/36216945/