Beyond HbA1c: Multivariate Clustering and Propensity-Weighted Analysis Identify High-Risk Patients with Diabetes...

Preoperative glycemic control is a key concern in bariatric surgery. However, the extent to which suboptimal HbA1c levels impact short-term postoperative morbidity remains uncertain. This study evaluated the association between preoperative HbA1c categories and 30-day postoperative morbidity in a large Middle Eastern bariatric surgery cohort, using propensity score weighting and unsupervised clustering to refine risk stratification.

This retrospective cohort analysis included 1778 bariatric surgery patients operated between 2019 and 2024 at a tertiary referral center in the United Arab Emirates. Patients were stratified into two groups: controlled HbA1c (≤6.5%, n = 1557) and elevated HbA1c (>6.5%, n = 221). Baseline demographics, perioperative parameters, and 30-day outcomes were extracted from a prospectively maintained registry. Statistical analysis included inverse probability weighting using propensity scores to adjust for age, BMI, and sex. Principal component analysis (PCA) followed by K-means clustering was performed on age and HbA1c to identify multivariate patient subgroups. Both unweighted and weighted analyses were performed.

Elevated HbA1c patients were older than those with controlled HbA1c (43.2 ± 12.1 versus 34.6 ± 10.8 years, P < .001 unweighted; 43.2 ± 12.1 versus 42.1 ± 12.0 years, P = .364 weighted). BMI was modestly higher in the elevated group but not significant after weighting. Thirty-day complication rates were low across all groups. The highest risk was observed in patients >60 years with diabetes (8.1%), while patients ≤20 years had negligible complications. In adults 21-60 years, complication rates increased modestly with HbA1c (normal 3.9%-5.3%, prediabetes 3.9%-6.3%, diabetes 6.5%-8.6%). No deaths occurred. Other complications, including deep/organ-space SSI, pneumonia, pulmonary embolism, renal failure, myocardial infarction, transfusion, gastrointestinal bleeding, readmission, and reoperation, were rare and showed no significant differences across HbA1c strata (all P > .05).

Elevated HbA1c was not an independent predictor of short-term morbidity after bariatric surgery. PCA-cluster analysis identified a distinct high-risk subgroup: older diabetic patients, who demonstrated the greatest vulnerability to short-term complications. These findings suggest that perioperative risk assessment should move beyond HbA1c cutoffs alone and incorporate multivariate patient profiling to guide individualized care.