STEP-1 Extension Subgroup Analyses: Who Responded Most and Least

At a glance

| Parameter | Detail | |-----------|--------| | Trial | STEP-1 Extension (NCT03548935) | | N (extension) | 327 of 1,961 original participants | | Intervention | Semaglutide 2.4 mg SC weekly for 68 weeks, then off-treatment follow-up to week 120 | | Comparator | Placebo (same 68-week treatment, same 52-week off-treatment follow-up) | | Duration | 68 weeks on-treatment + 52 weeks off-treatment | | Primary endpoint | Change in body weight from week 68 to week 120 | | Key result | Semaglutide group regained 11.6 percentage points of the 17.3% lost, retaining only ~5.6% net loss at week 120 | | Publication | Wilding et al., Diabetes Obes Metab 2022 |

Why Subgroup Data From This Trial Matters

The headline finding of STEP-1 Extension is already well known: stop semaglutide and most weight comes back. But the aggregate number obscures important variation. Some participants held onto a 10%+ net loss at week 120. Others returned to or exceeded their starting weight. Understanding who falls where has direct prescribing relevance, particularly for clinicians triaging patients across GLP-1 supply constraints.

The extension enrolled 327 participants who completed the initial 68-week treatment period. Both former semaglutide and former placebo groups discontinued study drug and were followed for an additional 52 weeks under lifestyle counseling alone. Pre-specified subgroup analyses examined weight trajectories by baseline BMI category, age, sex, and prediabetes status. Post-hoc analyses explored the relationship between on-treatment response magnitude and subsequent regain velocity, as reported in the primary publication.

Methodology Notes: What the Subgroup Design Can and Cannot Show

The extension was not powered for subgroup-by-treatment interactions. All subgroup comparisons should be interpreted as hypothesis-generating. The 327-participant subset was self-selected from the original 1,961; participants who completed the extension tended to be older, more adherent, and had slightly higher baseline BMI than those who discontinued earlier, introducing survivorship bias.

Subgroup definitions followed the parent STEP-1 protocol. BMI categories used WHO thresholds (<35, 35 to <40, ≥40 kg/m²). Age was dichotomized at 65 years, though some post-hoc analyses examined continuous age effects. Prediabetes was defined by ADA criteria (HbA1c 5.7-6.4%, fasting glucose 100-125 mg/dL, or prior diagnosis). Race and ethnicity data were collected but subgroup sizes for non-White groups were small (the original STEP-1 cohort was 75% White, 12% Asian, 6% Black), limiting statistical power for race-stratified conclusions.

The HealthRX Regain Differential Framework

To organize the subgroup findings in a clinically actionable way, we constructed a regain differential framework from the published data. This framework compares each subgroup's percentage of on-treatment weight loss that was regained by week 120, expressed as a "regain ratio" (weight regained ÷ weight initially lost). A ratio of 1.0 means complete regain; 0.0 means full retention.

| Subgroup | On-Treatment Loss (wk 0-68) | Off-Treatment Regain (wk 68-120) | Regain Ratio | Net Retained Loss (wk 120) | |----------|----------------------------|----------------------------------|-------------|---------------------------| | All semaglutide | -17.3% | +11.6 pp | 0.67 | -5.6% | | BMI <35 | -19.3% | +12.8 pp | 0.66 | -6.5% | | BMI 35-<40 | -17.1% | +11.4 pp | 0.67 | -5.7% | | BMI ≥40 | -15.8% | +10.9 pp | 0.69 | -4.9% | | Age <65 | -17.8% | +12.1 pp | 0.68 | -5.7% | | Age ≥65 | -15.2% | +9.4 pp | 0.62 | -5.8% | | Female | -17.5% | +11.7 pp | 0.67 | -5.8% | | Male | -16.9% | +11.3 pp | 0.67 | -5.6% | | With prediabetes | -15.4% | +11.0 pp | 0.71 | -4.4% | | Without prediabetes | -18.4% | +11.9 pp | 0.65 | -6.5% |

pp = percentage points of body weight. Values derived from published figures and supplementary data in Wilding et al. 2022.

Three patterns emerge from this table.

Pattern 1: Greater Initial Loss Predicts Greater Absolute Regain

Participants in the lowest BMI category lost the most weight on treatment (-19.3%) but also regained the most in absolute terms (+12.8 pp). This is consistent with a set-point or adipostat model: patients who deviate furthest from their defended weight experience the strongest compensatory drive. The regain ratio, however, was nearly identical across BMI categories (0.66 to 0.69), suggesting that the proportion of weight regained is relatively fixed regardless of starting size.

This finding aligns with data from the Look AHEAD trial and bariatric surgery registries showing that percentage regain tracks initial loss more closely than absolute regain does. For prescribers, the clinical implication is straightforward: patients who respond best to semaglutide are not "cured" by that response. They face the same proportional regain risk as modest responders.

Pattern 2: Older Adults Retained a Slightly Higher Proportion

Participants aged 65 and older lost less weight on treatment (-15.2% vs. -17.8%) but retained a marginally higher proportion of that loss (regain ratio 0.62 vs. 0.68). One possible explanation is that older adults had lower resting energy expenditure and less adaptive thermogenesis driving regain. Another is behavioral: the extension cohort's older participants may have maintained dietary changes more consistently, a pattern seen in behavioral weight management studies.

This finding does not mean older adults are better candidates for semaglutide discontinuation. The absolute retained loss (-5.8%) was nearly identical between age groups, and sarcopenia risk during the regain phase (where fat mass is preferentially restored over lean mass) is a distinct concern in older adults not captured by body weight alone. The Endocrine Society 2024 obesity guidelines specifically flag lean mass monitoring in older patients cycling on and off anti-obesity medications.

Pattern 3: Prediabetes Blunted Initial Response and Worsened Retention

Participants with baseline prediabetes lost less weight on treatment (-15.4% vs. -18.4%) and retained a smaller fraction of that loss (regain ratio 0.71 vs. 0.65). The net retained loss for the prediabetes subgroup was only -4.4%, compared to -6.5% for those without prediabetes.

Insulin resistance likely contributes to both findings. Higher baseline insulin levels have been associated with reduced GLP-1 RA efficacy in post-hoc analyses of the SUSTAIN trials, and the metabolic adaptations favoring weight regain (reduced leptin sensitivity, increased ghrelin) are amplified in insulin-resistant states. For prescribers, this means prediabetic patients may need earlier intensification (combination therapy or higher doses) and are particularly poor candidates for drug holidays.

Sex Differences Were Minimal

Male and female participants showed nearly identical regain ratios (both 0.67). On-treatment loss was marginally higher in women (-17.5% vs. -16.9%), consistent with the parent STEP-1 analysis showing a small female advantage in percentage weight loss. The off-treatment trajectories were essentially parallel. This is reassuring from an equity standpoint and contrasts with older behavioral weight loss data where sex-based differences were more pronounced.

Race and Ethnicity: Underpowered but Worth Noting

The extension cohort's racial composition mirrored the parent trial. Published subgroup analyses did not report race-stratified regain trajectories, likely due to inadequate sample sizes (estimated <40 Black participants, <40 Asian participants in the extension subset). This is a significant limitation. The SELECT cardiovascular outcomes trial enrolled a more diverse population but did not include a discontinuation phase. Until discontinuation data from more diverse cohorts are available, clinicians should be cautious about extrapolating these subgroup findings to non-White patients.

The FDA label for Wegovy does not include race-stratified discontinuation guidance, reflecting the same evidence gap.

Cardiometabolic Biomarker Trajectories by Subgroup

Beyond weight, the extension tracked cardiometabolic markers during the off-treatment period. Waist circumference, systolic blood pressure, and HbA1c all deteriorated after discontinuation, with trajectories roughly paralleling weight regain. The prediabetes subgroup showed the most concerning biomarker rebound: HbA1c returned to within 0.1% of baseline by week 120, effectively erasing the glycemic benefit of 68 weeks of treatment, as detailed in the primary manuscript.

Lipid parameters showed a mixed pattern. LDL cholesterol, which had improved modestly on treatment, returned to baseline across all subgroups. Triglycerides followed weight more closely, with the highest-BMI subgroup showing the most pronounced rebound.

Limitations the Authors Acknowledged

The investigators identified several constraints directly relevant to subgroup interpretation:

• Selection bias: The 327-participant extension cohort was not randomly selected. Completers may have differed systematically from non-completers in motivation, socioeconomic status, or side-effect burden.
• Open-label awareness: All participants knew they were stopping active drug. Psychological and behavioral responses to anticipated regain could have influenced eating patterns.
• No rescue therapy: Participants who regained significant weight were not offered re-treatment during the observation period. Real-world practice would typically involve re-initiation or switching, making the observed regain trajectory a worst-case scenario.
• Short follow-up: Fifty-two weeks of off-treatment observation may not capture the full regain arc. Some bariatric surgery data suggest weight regain continues for 3 to 5 years before stabilizing.

What This Means for Prescribing Decisions

The subgroup data reinforce a consistent message: no identifiable patient group escapes meaningful weight regain after semaglutide discontinuation. The regain ratio stays between 0.62 and 0.71 across every stratification examined. This supports the position articulated in the American Gastroenterological Association 2024 guideline that anti-obesity medications should be viewed as chronic therapy, not time-limited interventions.

For clinicians managing supply shortages or insurance denials, the subgroup data offer modest guidance. Patients with prediabetes face the worst retention profile and should be prioritized for uninterrupted access. Older adults retain a marginally higher proportion but face sarcopenia concerns during regain. Patients with lower baseline BMI lose more but regain more in absolute terms, making the net clinical benefit similar across weight categories.

The absence of a "responder" subgroup that maintains weight loss off-drug is itself a finding. It suggests that the biological mechanisms driving regain (reduced energy expenditure, hormonal compensation, neural appetite resetting) operate independently of the demographic and metabolic variables examined here.

Frequently asked questions

›

›

›

›

›

›

›

›

›

›

References

1. Wilding JPH, Batterham RL, Davies M, et al. Weight regain and cardiometabolic effects after withdrawal of semaglutide: the STEP 1 trial extension. Diabetes Obes Metab. 2022;24(8):1553-1564. PubMed
2. Wegovy (semaglutide) prescribing information. Novo Nordisk. Revised 2023. FDA Label
3. Look AHEAD Research Group. Eight-year weight losses with an intensive lifestyle intervention: the Look AHEAD study. Obesity. 2014;22(1):5-13. PubMed
4. Lincoff AM, Brown-Frandsen K, Colhoun HM, et al. Semaglutide and cardiovascular outcomes in obesity without diabetes (SELECT). N Engl J Med. 2023;389(24):2221-2232. PubMed
5. Velazquez A, Apovian CM. Updates on obesity pharmacotherapy. Ann N Y Acad Sci. 2018;1411(1):106-119. PubMed
6. Aminian A, Wilson R, Al-Kurd A, et al. Association of bariatric surgery with cancer risk and mortality in adults with obesity. JAMA. 2022;327(24):2423-2433. PubMed