Metabolic and Bariatric Surgery and Change in the AI-ECG-Derived Biological-Age Gap: A Longitudinal Cohort Study.
Key claim: Metabolic and bariatric surgery was associated with a modest late decline in AI-ECG-derived biological-age gap, while chronic adiposity—not within-person weight change—tracked with higher δage.
Abstract
BACKGROUND: Obesity is associated with accelerated aging, yet whether metabolic and bariatric surgery (MBS) can reverse or attenuate this process, as measured by the artificial intelligence-electrocardiogram (AI-ECG)-derived age gap, remains uncertain. METHODS: We retrospectively studied 2,380 adults who underwent primary MBS from 2008 to 2023 and contributed 6,149 ECG observations. The outcome, δage, was AI-ECG-predicted age minus chronological age. Linear mixed-effects models used continuous time with a clinically prespecified 1-year knot and decomposed weight into between-person and within-person components. Secondary analyses evaluated alternative adiposity measures, observed ECG-age normalization, exploratory mediation, and mortality. RESULTS: Median preoperative δage was 3.86 years (IQR, - 1.30 to 9.03), and 69.6% had δage > 0. δage changed by + 0.275 years/year during the first postoperative year (95% CI, - 0.187 to 0.738; P=.244) and by - 0.192 years/year thereafter (95% CI, - 0.303 to - 0.081; P=.001). The piecewise model was not superior to a linear model (P=.093). Higher between-person mean weight was associated with higher δage (+ 0.548 years per 10 kg; P<.001), whereas within-person weight change was not (+ 0.117 years per 10 kg; P=.084). Baseline, observed postoperative nadir, and maximum BMI were positively associated with δage; %EWL and observed weight regain were not. The exploratory indirect effect through weight change was near zero (ACME, + 0.0008 years; P=.814). Among 1,657 patients with baseline δage > 0, 335 (20.2%) had an observed δage ≤ 0 during follow-up. CONCLUSIONS: MBS was associated with a modest late decline in AI-ECG-derived δage. Chronic adiposity, but not within-person weight change, was associated with δage.
Why this matters
This is one of the first longitudinal tests of whether a major metabolic intervention moves an AI-ECG biological-age estimate, and the answer is: barely, and only after a delay. The dissociation between within-person weight loss and δage — while chronic adiposity does track — is a concrete example of the area’s recurring trap, where a clock reflects a stable phenotype more than it responds to change, so ‘moving the clock’ cannot be assumed from moving the exposure.