T. Ueland1, J.P. Shelley1, J.D. Mosley2, J. Robinson4, E.R. Gamazon5, L. Maguire3, R. Peek6, A.T. Hawkins7 1Vanderbilt University Medical Center, School Of Medicine, Nashville, TN, USA 2Vanderbilt University Medical Center, Department Of Internal Medicine, Nashville, TN, USA 3University Of Pennsylvania, Department Of Surgery, Philadelphia, PA, USA 4Vanderbilt University Medical Center, Department Of Pediatric Surgery, Nashville, TN, USA 5Vanderbilt University Medical Center, Department Of Medicine, Division Of Genetic Medicine, Nashville, TN, USA 6Vanderbilt University Medical Center, Division Of Gastroenterology, Hepatology, And Nutrition, Nashville, TN, USA 7Vanderbilt University Medical Center, Division Of General Surgery, Nashville, TN, USA
Introduction: Although personalized risk stratification tools for severe presentations of diverticulitis are lacking. polygenic risk scores have demonstrated promise in this area. In other diseases, the plasma proteome offers information that complements genetic risk profiles. It is unknown if a plasma proteomic signature exists for severe diverticulitis, and whether proteomics could improve models that account for genetic and lifestyle factors.
Methods: This UK Biobank study comprised two independent phases: a discovery phase to identify candidate proteins, and a validation phase to apply these proteins to risk prediction for severe diverticulitis (multiple inpatient admissions or operative management). In the discovery phase, linear and least absolute shrinkage and selection operator regression models were fit to identify proteins associated with a history of severe diverticulitis. A proteomic risk score was derived from these proteins, and associations with other conditions were explored using a phenome-wide association study (PheWAS). In the validation phase, a Cox proportional hazards model was fit with an outcome of incident severe diverticulitis in the follow-up period. Covariates were age, sex, body mass index, a healthy diet score, a polygenic risk score, and plasma proteome terms. Model performance was quantified through area under the receiver operating characteristics curve (AUC) and compared to a base model without genetic or proteomics terms through a likelihood ratio test. Associations were reported as hazard ratios with 95% confidence intervals.
Results: The discovery group included 651 cases of severe diverticulitis and 19,100 controls with mean [SD] age of 57 years [8] and 54% female. The validation group included 505 cases and 18,905 controls with mean [SD] age of 57 [8] and 55% female. There were 18 proteins associated with severe diverticulitis in the discovery phase. The full Cox model including the polygenic risk score and protein terms outperformed the base model (12-year AUC base model 0.67, full model 0.72, p < 0.01) (Figure 1A). The diverticulitis proteomic risk score was associated with renal and cardiometabolic conditions in the PheWAS (Figure 1B). Both the polygenic risk score (HR [95% CI] 1.35 [1.24 – 1.48]) and proteomic risk score (HR [95% CI] 1.32 [1.19 – 1.46]) remained associated with incident severe diverticulitis when controlling for demographics and dietary habits (Figure 1C).
Conclusion: Plasma proteomic profiles complemented a risk stratification approach based on genetic and lifestyle factors. Personalized multi-omic profiles may augment shared decision-making frameworks when considering operative options for diverticulitis.