Clinical and Psychometric Predictors of a Positive Alcohol Use Disorder Screen after Roux-en-Y Gastric Bypass: A...

Alcohol use disorder (AUD) after bariatric surgery is increasingly recognized but remains under-detected because routine screening relies on self-report. We aimed to identify clinical and psychometric predictors of a positive alcohol screen and to build a simple risk-stratification model; as a secondary objective, we critically evaluated whether a routinely available hepatic enzyme, temporally anchored gamma-glutamyl transferase (GGT), adds information.

This was a single-center cross-sectional analysis of 130 adults who had undergone Roux-en-Y gastric bypass (RYGB) and completed the Cut-down/Annoyed/Guilty/Eye-opener (CAGE) and AUDIT questionnaires during routine follow-up at a private clinic in Brazil. Questionnaire responses were linked to an institutional clinical-laboratory database. The primary outcome was a positive AUDIT screen (score ≥ 8). Multivariable logistic regression with 5,000 bootstrap resamples identified independent predictors. As a secondary, exploratory analysis, the GGT measurement closest to the questionnaire date was selected within ± 6-, ± 12-, and ± 24-month windows and compared between groups, with Benjamini-Hochberg correction across a four-marker hepatic-enzyme set; a linear mixed model examined longitudinal GGT trajectories.

A positive AUDIT screen was identified in 32 patients (24.6%). Median age was 47 years (interquartile range [IQR] 37-58; mean 47.4 ± 12.6), and 80% were women. Male sex (OR 5.3, 95% CI 1.9-14.3) and paternal alcohol use (OR 5.5, 95% CI 2.2-13.7) were independent predictors; a two-variable model served as a candidate risk-stratification approach with selection-aware optimism-corrected internal performance (c-index 0.71; calibration slope 0.97), stratifying predicted risk from 9% (neither factor) to 75% (both, small stratum), and would require external validation before clinical use. In the AUDIT-anchored ± 6-month subgroup (n = 58), GGT was higher among screen-positive patients (median 31 vs. 15.5 U/L; AUC 0.72) but did not survive false-discovery-rate (FDR) correction (q = 0.055), and both medians remained within the reference range. A secondary, exploratory database analysis (Supplement) was directionally consistent at the group level but rested on an imperfect, construct-distinct label and is not confirmatory. The mixed model did not detect a divergent longitudinal GGT trajectory (β = 0.02 log-units/year, 95% CI - 0.02 to 0.06).

After bariatric surgery, male sex and paternal alcohol use are strong, easily ascertained predictors of a positive alcohol screen in this cohort and can be translated into a simple, bedside risk-stratification approach (a candidate requiring external validation). GGT shows a weak, group-level association with alcohol use - specific within the hepatic-enzyme panel tested, yet within the reference range and uninformative at the individual level - so it may complement, but cannot replace, psychometric screening. These findings favor psychometric screening focused on identifiable risk strata, with routine GGT as at most a supportive signal.