Illustrative AI review sample — surgery

Key Points

1Prospective dual-cohort study from Iran comparing incident T2DM, hypertension, and dyslipidaemia in bariatric surgery (n=612) versus obese controls (n=593) over three years; impressive risk reductions, but non-randomised allocation and baseline metabolic imbalance between cohorts limit causal attribution.

Major Issues

Methods: The surgery and control cohorts originate from different source populations with fundamentally different recruitment strategies (clinic-volunteers vs. community-sampled), introducing systematic selection bias and confounding by indication that cannot be fully controlled by statistical methods alone.

Methods / Statistics: No propensity score matching, inverse probability weighting, or multivariable regression is used to adjust for baseline differences between groups; relative risk reductions are unadjusted despite acknowledged baseline metabolic imbalance.

Results: Per-procedure subgroup outcomes (SG vs. OAGB vs. RYGB) are reported only for T2DM, not for hypertension or dyslipidaemia; this selective reporting may bias interpretation of procedure-specific effects.

Methods: Loss to follow-up rate and missing data handling strategy are not reported; in a 3-year prospective study of severely obese patients, attrition can be substantial and its pattern (informative vs. random) affects result validity.

The effect of bariatric surgery in comparison with the control group on the prevention of comorbidities in people with severe obesity: a prospective cohort study

Key Points

Major Issues