Skip to content

Bariatric and metabolic surgery

TL;DR — Metabolic surgery produces one of the largest histological effects reported for MASH and has the strongest dedicated observational analysis of major adverse liver outcomes. In BRAVES — the only randomised trial comparing surgery with lifestyle plus best medical care — NASH resolution without fibrosis worsening at one year occurred in 56% after Roux-en-Y gastric bypass and 57% after sleeve gastrectomy versus 16% with lifestyle (p<0.0001; per-protocol 70%, 70%, 19%), with resolution probability 3.60-fold (95% CI 2.19–5.92) and 3.67-fold (2.23–6.02) greater than lifestyle (Verrastro 2023, PMID 37088093, NCT03524365). Effects deepen with time rather than fading: in a prospective 5-year sequential-biopsy cohort, NASH had resolved without worsening fibrosis in 84% at 5 years, fibrosis had decreased in 70.2% and disappeared entirely in 56%, including 45.5% of those with baseline bridging fibrosis (Lassailly 2020, PMID 32553765). The SPLENDOR analysis links this to outcomes: 10-year cumulative incidence of major adverse liver outcomes 2.3% versus 9.6% (adjusted HR 0.12, 95% CI 0.02–0.63, p=0.01) and of MACE 8.5% versus 15.7% (adjusted HR 0.30, 0.12–0.72, p=0.007) (Aminian 2021, PMID 34762106). The costs are real: 0.6% died from surgical complications within the first year in SPLENDOR, and surgery has been linked to increased risk of alcohol use disorder (Lefere 2021, PMID 34002452). Cirrhosis remains the frontier — safety data in cirrhosis are thin and incidental liver failure after surgery has been reported.

The randomised evidence: BRAVES

Multicentre, open-label, three-arm randomised trial at three hospitals in Rome, April 2019 – June 2021 (Verrastro 2023, PMID 37088093):

Element Detail
Eligibility age 25–70, BMI 30–55 kg/m², with or without T2D, histologically confirmed NASH
Screening yield 431 biopsy-screened; 103 (24%) did not have histological NASH; 40 (9%) declined
Randomised 288, 1:1:1 to lifestyle + best medical care (96), Roux-en-Y gastric bypass (96), sleeve gastrectomy (96)
Primary endpoint histological NASH resolution without worsening of fibrosis at 1 year
Arm ITT primary endpoint Per-protocol (n=236 completers) Probability vs lifestyle
Lifestyle + best medical care 16% (15/96) 19% (15/80) reference
Roux-en-Y gastric bypass 56% (54/96) 70% (54/77) 3.60× (2.19–5.92), p<0.0001
Sleeve gastrectomy 57% (55/96) 70% (55/79) 3.67× (2.23–6.02), p<0.0001

No deaths or life-threatening complications; severe adverse events in 10 (6%) surgical participants, none requiring re-operation, all resolved medically or endoscopically. The two procedures were indistinguishable on the primary endpoint.

Two features are worth noting. The 24% biopsy-screen failure rate is a reminder of how imprecise clinical NASH diagnosis is without histology (histology and biopsy). And the lifestyle arm achieved 16–19% resolution, close to the pooled meta-analytic estimate of 0.12 for lifestyle-based MASH resolution (lifestyle and weight loss).

The long-term histological trajectory

Prospective single-centre French cohort of 180 severely obese patients with biopsy-proven NASH, with sequential liver samples at surgery and at 1 and 5 years; samples obtained from 125 of 169 (76%) reaching 1 year and 64 of 94 (68%) reaching 5 years (Lassailly 2020, PMID 32553765):

Endpoint at 5 years Result (95% CI)
NASH resolved without worsening fibrosis 84% (73.1–92.2)
Fibrosis decreased vs baseline 70.2% (56.6–81.6)
Fibrosis disappeared entirely 56% (42.4–69.3)
Fibrosis disappeared, among those with baseline bridging fibrosis 45.5%

NASH resolution was already 84% at one year with no significant recurrence between years 1 and 5 (p=0.17), while fibrosis regression was progressive, beginning by year 1 and continuing to year 5 (p<0.001). Persistence of NASH was associated with no fibrosis reduction and less weight loss — BMI reduction 6.3 ± 4.1 kg/m² in those with persistent NASH versus 13.4 ± 7.4 kg/m² in those with resolution (p=0.017), reproducing the weight-loss dose–response seen with lifestyle.

The 68% five-year biopsy retention is the main limitation: patients who do worse are plausibly less likely to return for a research biopsy, which would bias the estimates optimistically.

Hard outcomes: SPLENDOR

From 25,828 liver biopsies at a US health system 2004–2016, 1,158 adults with obesity and biopsy-confirmed NASH with fibrosis stage 1–3 were identified: 650 who had bariatric surgery (Roux-en-Y gastric bypass or sleeve gastrectomy) with simultaneous liver biopsy, and 508 non-surgical controls, balanced by overlap weighting on baseline characteristics including histological severity. Median follow-up 7 years (IQR 4–10); 63.9% women, median age 49.8 years, median BMI 44.1 (Aminian 2021, PMID 34762106):

Outcome at 10 years (overlap-weighted cumulative incidence) Surgery Non-surgical Adjusted difference Adjusted HR
Major adverse liver outcomes (cirrhosis, HCC, transplant, liver-related death) 2.3% (0–4.6) 9.6% (6.1–12.9) 12.4% (5.7–19.7) 0.12 (0.02–0.63), p=0.01
MACE (coronary, cerebrovascular, heart failure, cardiovascular death) 8.5% (5.5–11.4) 15.7% (11.3–19.8) 13.9% (5.9–21.9) 0.30 (0.12–0.72), p=0.007

Unweighted event counts were 5 versus 40 major adverse liver outcomes and 39 versus 60 MACE — small numbers behind wide intervals, and the confidence interval on the liver hazard ratio spans a 30-fold range. Against that, 4 patients (0.6%) died of surgical complications within the first year (gastrointestinal leak n=2, respiratory failure n=2).

This is the only MASH intervention with published data on major adverse liver outcomes. It is observational with overlap weighting, not randomised; the selection into surgery cannot be fully adjusted away.

Cirrhosis: the frontier has moved

The statement that surgery is untested in cirrhosis is now out of date for compensated cirrhosis. SPECCIAL (Surgical Procedures Eliminate Compensated Cirrhosis In Advancing Long-term) retrospectively classified 62 patients who underwent metabolic surgery and 106 non-surgical controls as having obesity and histologically proven compensated MASH cirrhosis, using doubly robust estimation to balance baseline characteristics; mean follow-up was 10.0 ± 4.5 years (Aminian 2025, PMID 39870816):

Outcome, 15-year cumulative incidence Surgery (n=62) Non-surgical (n=106) Adjusted HR
Major adverse liver outcomes 20.9% (2.5–35.9) 46.4% (25.6–61.3) 0.28 (0.12–0.64), p=0.003
Decompensated cirrhosis 15.6% (0–31.3) 30.7% (12.9–44.8) 0.20 (0.06–0.68), p=0.01

This is the only intervention of any kind with published data on hepatic outcomes in MASH cirrhosis — no drug has been shown to reduce major adverse liver outcomes at this stage, which is precisely the population excluded from the resmetirom label (resmetirom and thyromimetics) and the population where semaglutide's cirrhosis programme did not meet its endpoint (GLP-1 and incretin therapy). The caveats are the same as SPLENDOR's and larger: 168 patients, observational, single health system, the surgical group selected as fit enough to operate on with portal hypertension, and lower bounds of the cumulative-incidence intervals touching zero. What it establishes is not that surgery should be done in cirrhosis but that the question is now answerable and the answer is not obviously "no".

Mechanism

Surgery acts through reduced food intake, altered gut hormone secretion, changes in metabolic signalling and improvement in adipose tissue dysfunction; reduction in hepatic fat and improvement in hepatic insulin resistance are among the earliest effects, before most weight loss has occurred (Lefere 2021, PMID 34002452). The incretin overlap with pharmacological therapy is not coincidental — see GLP-1 and incretin therapy — and the shared adipose-tissue mechanism is set out in pathogenesis.

Risks specific to the liver

Risk Evidence
Perioperative mortality 0.6% within 1 year in SPLENDOR (PMID 34762106)
Severe adverse events 6% in BRAVES, none requiring re-operation (PMID 37088093)
Safety in cirrhosis uncertain; a specific open question raised as the scope of surgery expanded (PMID 34002452)
Incidental liver failure after surgery reported; not quantified (PMID 34002452)
Liver transplant outcomes after prior bariatric surgery an open question raised in the same review (PMID 34002452)
Alcohol use disorder studies in humans and rodents link bariatric surgery to increased risk — a major risk factor for liver disease in its own right (PMID 34002452)

The alcohol signal deserves emphasis in this condition specifically: a patient with metabolic steatosis who develops an alcohol use disorder after surgery moves along the SLD spectrum toward MetALD or ALD, where outcomes are worse (nomenclature and definitions). It is a plausible mechanism by which surgical benefit could be partly offset in a subgroup, and it is not measured in the histological trials.

How large is it? Three independent cohorts now quantify what the Lefere review described qualitatively, and they agree on direction and roughly on magnitude while disagreeing on absolute importance:

Cohort Design Comparison Effect
Swedish Obese Subjects, 2,007 surgical vs 2,040 matched controls, median 25.2 y (up to 35 y) (Sjöholm 2025, PMID 41066138) prospective controlled gastric bypass (n=266) vs usual obesity care AUD adjusted HR 5.07 (3.11–8.25)
Same vertical banded gastroplasty (n=1,365) AUD HR 2.28 (1.56–3.34)
Same gastric banding (n=376) AUD HR 2.34 (1.37–4.01)
Same — alcohol-related mortality gastric bypass sub-HR 6.18 (2.48–15.40)
Same vertical banded gastroplasty sub-HR 3.56 (1.79–7.08)
Same gastric banding sub-HR 2.52 (0.89–7.15), ns
Danish registers, 13,430 surgical (95% bypass) vs 21,021 obese controls, median 6.9 y (Bramming 2021, PMID 33085738) register cohort, four analytic approaches incl. IPTW and self-controlled before/after surgery vs no surgery AUD HR 7.29 (5.06–9.48); already HR 2.77 (1.39–5.53) in year 1

Three points follow. First, the risk is procedure-dependent — bypass carries roughly twice the hazard of restrictive procedures in SOS — which matters because BRAVES found bypass and sleeve indistinguishable for the liver at one year (PMID 37088093); if the hepatic benefit is equal and the alcohol hazard is not, the procedures are not interchangeable over a lifetime. Second, excess diagnoses appeared within the first postoperative year (PMID 33085738), a timing compatible with altered alcohol pharmacokinetics but not sufficient to exclude psychosocial explanations. Third, the Danish authors are explicit that despite the large relative risk, few individuals developed AUD in absolute terms, and that severe obesity should not be a contraindication to surgery on these grounds (PMID 33085738). Both relative and absolute scales belong in the decision, and neither cohort measured liver endpoints.

A possible mitigation. In 15,382 post-bariatric patients who subsequently received an anti-obesity medication (2020–2024, US EHR network, 1:1 propensity-matched to 3,990 per group), incretin-based therapy versus non-incretin anti-obesity drugs was associated with lower new-onset AUD (2.4 vs 5.2 per 1,000 person-years; HR 0.45, 0.25–0.81, p=0.006) and lower initiation of AUD medications (15.2 vs 25.6 per 1,000 person-years; HR 0.59, 0.46–0.75, p<0.001) (Fakhoury 2025, PMID 41632137). This is retrospective and confounded by indication, and the authors note that long-term liver outcomes were not examined — but it raises the possibility that the same drug class that treats MASH also attenuates surgery's principal liver-relevant harm.

Comparison with drug therapy

Intervention MASH resolution Fibrosis improvement Follow-up Design
Sleeve gastrectomy / RYGB (BRAVES) 57% / 56% vs 16% lifestyle — 1 year randomised
Bariatric surgery (Lassailly) 84% 70.2% decreased 5 years prospective cohort
Semaglutide 2.4 mg (ESSENCE) 62.9% vs 34.3% 36.8% vs 22.4% 72 weeks randomised, placebo
Resmetirom 100 mg (MAESTRO-NASH) 29.9% vs 9.7% 25.9% vs 14.2% 52 weeks randomised, placebo
Lifestyle, ≥10% weight loss 90% (subgroup) 45% regression 52 weeks prospective, non-randomised subgroup

These are not directly comparable — placebo/comparator arms differ by a factor of three across trials, follow-up differs, and BRAVES and Lassailly used different comparators or none. What can be said is that surgery has randomised evidence against an active lifestyle/medical comparator, five-year paired histology, and a dedicated analysis of major adverse liver outcomes. The hard-outcome result remains observational rather than randomised.

Open questions

  • Can surgery be done safely in MASH cirrhosis? No completed randomised trial was identified. A live ClinicalTrials.gov search on 2026-09-02 found one recruiting randomised study of TIPS plus sleeve gastrectomy versus medical/lifestyle care in 70 patients with cirrhosis, portal hypertension and severe obesity (NCT07058155); its primary endpoint is 6-month SF-36 physical-component score, not hepatic events. A completed observational intragastric-balloon study enrolled 56 people with compensated NASH cirrhosis (NCT03753438). The observational answer now exists: SPECCIAL reports a 15-year major-adverse-liver-outcome incidence of 20.9% versus 46.4% (adjusted HR 0.28, 0.12–0.64) in 168 patients with compensated MASH cirrhosis (PMID 39870816). What remains open is decompensated cirrhosis, and whether the effect survives randomisation.
  • Which procedure, once alcohol risk is priced in? BRAVES found bypass and sleeve equivalent for hepatic histology at one year (PMID 37088093), while SOS found bypass carries roughly twice the AUD hazard of restrictive procedures (HR 5.07 vs 2.28–2.34) and a six-fold alcohol-related mortality sub-hazard (PMID 41066138). No study has combined hepatic benefit and alcohol harm into a single decision model for a MASH population.
  • Does incretin therapy after surgery protect the liver by protecting against alcohol? Post-bariatric incretin use was associated with 55% lower new-onset AUD (PMID 41632137), but liver-related outcomes were not examined — the authors say so explicitly. This is a designable study in existing claims data.
  • Is the SPLENDOR outcome benefit causal? Overlap weighting balanced measured covariates including histological severity, but surgical candidacy encodes unmeasured fitness and adherence. Only a randomised trial powered for liver outcomes would settle it, and BRAVES was powered for histology at one year.
  • How much of the 5-year fibrosis regression is real versus survivor bias? 68% of eligible patients provided a 5-year biopsy (PMID 32553765).
  • Does surgery beat incretin therapy? No completed head-to-head trial exists. A recruiting randomised study compares metabolic surgery with incretin therapy in 120 participants and uses two-year biopsy fibrosis change as an endpoint (NCT06374875); until it reports, comparative efficacy remains unknown.
  • How large is the post-surgical alcohol-use-disorder effect on liver outcomes? The association is now quantified in general bariatric populations (adjusted HRs 2.28–7.29 across cohorts; alcohol-related mortality sub-HR up to 6.18 after bypass — PMIDs: 41066138, 33085738) but still has not been quantified within a MASH surgical cohort with liver endpoints. SOS and the Danish registers captured AUD diagnoses and alcohol-related death, not hepatic histology or decompensation.
  • Does the choice of procedure matter for the liver? BRAVES found gastric bypass and sleeve gastrectomy indistinguishable at one year (PMID 37088093). A 2026 meta-analysis of 16 studies found no overall difference in NAS; a greater effect beyond one year was subgroup evidence, while bypass produced more weight loss (PMID 42420150). Five-year randomised hepatic comparisons remain unavailable.

References

  1. Verrastro O, Panunzi S, Castagneto-Gissey L, et al. Bariatric-metabolic surgery versus lifestyle intervention plus best medical care in non-alcoholic steatohepatitis (BRAVES): a multicentre, open-label, randomised trial. Lancet. 2023;401(10390):1786-1797. PMID 37088093
  2. Lassailly G, Caiazzo R, Ntandja-Wandji LC, et al. Bariatric Surgery Provides Long-term Resolution of Nonalcoholic Steatohepatitis and Regression of Fibrosis. Gastroenterology. 2020;159(4):1290-1301.e5. PMID 32553765
  3. Aminian A, Al-Kurd A, Wilson R, et al. Association of Bariatric Surgery With Major Adverse Liver and Cardiovascular Outcomes in Patients With Biopsy-Proven Nonalcoholic Steatohepatitis. JAMA. 2021;326(20):2031-2042. PMID 34762106
  4. Lefere S, Onghena L, Vanlander A, et al. Bariatric surgery and the liver-Mechanisms, benefits, and risks. Obes Rev. 2021;22(9):e13294. PMID 34002452
  5. Sanyal AJ, Newsome PN, Kliers I, et al. Phase 3 Trial of Semaglutide in Metabolic Dysfunction-Associated Steatohepatitis. N Engl J Med. 2025;392(21):2089-2099. PMID 40305708
  6. Harrison SA, Bedossa P, Guy CD, et al. A Phase 3, Randomized, Controlled Trial of Resmetirom in NASH with Liver Fibrosis. N Engl J Med. 2024;390(6):497-509. PMID 38324483
  7. Vilar-Gomez E, Martinez-Perez Y, Calzadilla-Bertot L, et al. Weight Loss Through Lifestyle Modification Significantly Reduces Features of Nonalcoholic Steatohepatitis. Gastroenterology. 2015;149(2):367-78.e5. PMID 25865049
  8. Liu C, et al. Sleeve gastrectomy versus Roux-en-Y gastric bypass for nonalcoholic fatty liver disease: a systematic review and meta-analysis. Surg Obes Relat Dis. 2026. PMID 42420150
  9. Aminian A, Aljabri A, Wang S, et al. Long-term liver outcomes after metabolic surgery in compensated cirrhosis due to metabolic dysfunction-associated steatohepatitis. Nat Med. 2025;31(3):988-995. PMID 39870816
  10. Sjöholm K, Peltonen M, Jacobson P, et al. Alcohol use disorder and alcohol-related mortality after metabolic bariatric surgery: prospective controlled cohort study. Br J Surg. 2025;112(10):znaf211. PMID 41066138
  11. Bramming M, Becker U, Jørgensen MB, et al. Bariatric surgery and risk of alcohol use disorder: a register-based cohort study. Int J Epidemiol. 2021;49(6):1826-1835. PMID 33085738
  12. Fakhoury B, Sierra L, Rama K, et al. Incretin-Based Therapies and Post-Bariatric Surgery Alcohol Use Disorder. JAMA Netw Open. 2025;8(12):e2549086. PMID 41632137