The Impact of HbA1c on Post-hysterectomy Complications as a Continuous Variable vs. Dichotomous Using Traditional (HbA1c <8) and Novel Cut-Points

INTRODUCTION: Diabetes impacts 11.6% of U.S. adults, a continually increasing number. 600,000 hysterectomies are performed annually in the U.S. Previous studies linking diabetes and surgical complications show increased surgical site infections, wound dehiscences, abscesses, postoperative fevers, systemic infections, and postoperative transfusions. However, there is discord about the ideal HbA1c cutoff for minimizing post-op complications. OBJECTIVE: To investigate the correlation between HbA1c levels and complications after hysterectomies. METHODS: We utilized Cerner Health Facts database with 519 million patient encounters from 750 hospitals January 2010 to November 2018. Our retrospective cohort identified patients undergoing abdominal, laparoscopic +/ − robotic-assisted, or vaginal hysterectomy. We included patients with perioperative HbA1c values 3 months prior to 6 months after hysterectomy, regardless of diabetes diagnosis, including both those diagnosed with diabetes or undergoing screening. We performed univariable and multivariable logistic regressions, adjusting for HbA1c, demographic factors and comorbidities. The outcome variable was a composite of any post-op complication. HbA1c was both a continuous variable and categorical, using HbA1c<8 and HbA1c<8.85 as cutpoints. HbA1c<8.85 was chosen after ROC curves demonstrated this maximized sensitivity/specificity. RESULTS: Of 86,420 hysterectomies, 2,727 (3.2%) had HbA1c. Excluding 34 obviously erroneous HbA1c values left 2,693 for final analysis (Figure 1), including 514 (19.1%) with complications and 2,179 (80.9%) without. Median [interquartile range] days from surgery to complication was 18 [5–42]. 75.2% were White, and 83.2% were in urban hospitals. Route was 43.9% abdominal, 30.5% laparoscopic, and 25.7% vaginal, with 69.7% adnexa removed. Complications were older (60.5±15.9 vs 58.4±14.2 years, p=.004), more likely from the Southern U.S. (24.5% vs 18.5%, p=0.001), used more tobacco (26.7% vs 19.0%, p<0.001), and more likely obese (32.3% vs 23.3%, p<0.001). Higher HbA1c had more complications: continuous (p=0.034), dichotomous HbA1c<8.0 (p=0.014), or dichotomous HbA1c<8.85 (p=0.001). HbA1c<8.85 optimized test characteristics (sensitivity=0.14, specificity=0.91, PPV=0.27, NPV=0.82). However, NPV and overall performance were not significantly different using HbA1c<8.85 vs HbA1c<8.0 (p=0.648). Likewise, both cut-points performed marginally better than chance (AUC=0.52–0.53). Predictive value improved with including confounders (multivariable model AUC=0.71–0.72). Multivariable analysis demonstrated HbA1c>8.85 increased complications (odds ratio [OR]=1.72, 95% CI=1.26–2.33) along with covariates (Table 1). Notably, obesity increased complications independently from HbA1c (OR=1.34, 95% CI=1.06–1.70). Contrary to other studies, age and presence of cancer did not impact complications. Urogynecological diagnoses, including urinary frequency, urinary tract infection, and dysuria, increased complications for all hysterectomies, including those not for urogynecological indications (OR=2.33, 3.60, 3.99). CONCLUSIONS: Post-hysterectomy complications are a complex and nuanced outcome currently imperfectly predicted using HbA1c. Since HbA1c as a continuous variable impacted complications, any improvement in blood sugar control and resultant HbA1c reductions will decrease complications. For surgeons deciding whether to proceed with hysterectomy based on HbA1c, the best cut-point was 8.85, but this was not statistically different from current standard of 8. Therefore, either are acceptable thresholds. However, HbA1c as a single variable was marginally better than chance at predicting complications so further research is needed to create prediction models incorporating both HbA1c and other comorbidities to choose appropriate surgical candidates.