Lennon, Ryan J.

ABSTRACT Ischemic events after contemporary percutaneous coronary intervention (PCI) are uncommon but carry high morbidity/mortality. Identifying patients at highest risk is critical to guide dual antiplatelet therapy (DAPT) intensity while minimizing bleeding. Current machine learning (ML) model‐developed risk scores do not incorporate pharmacogenetic data. To develop and externally validate ML models that integrate clinical, demographic, and CYP2C19 genetic information to predict 1‐year ischemic outcomes post‐PCI to refine clinical decision making. We analyzed 8317 patients from TAILOR‐PCI trial ( n = 4572) and the Precision PCI registry ( n = 3745). The outcome was a composite of cardiovascular death, myocardial infarction, stroke, and stent thrombosis at 1 year. Boruta feature selection identified 11 predictors. Multiple ML algorithms were trained in TAILOR‐PCI and validated in Precision PCI using cross‐validation and synthetic minority oversampling (SMOTE). Model performance was assessed by area under the receiver operating characteristic curve (AUC), sensitivity, and specificity. The best external performance was achieved with a support vector machine (SVM, polynomial kernel) (AUC 0.667; sensitivity 0.871; specificity 0.282), while XGBoost provided a more balanced profile (AUC 0.619; sensitivity 0.442; specificity 0.688). Variable importance for the SVM polynomial model demonstrated that all 11 included Boruta feature selected predictors had relatively high importance (> 75). ML models trained on large clinical trial and real‐world registry datasets can help identify the small subset of patients at high ischemic risk after PCI. This distinction is of relevance because ischemic events are rare, and most patients may safely de‐escalate DAPT to reduce bleeding risk while maintaining ischemic protection through a multimodal approach to risk stratification. Trial Registration: TAILOR‐PCI URL: https://clinicaltrials.gov/ct2/show/NCT01742117

Introduction: Identifying patients at risk for ischemic events after percutaneous coronary intervention (PCI) relies on traditional analysis of limited clinical and imaging variables. Machine learning (ML) has shown promise in effectively predicting cardiovascular risk in population studies. While existing ML models mainly predict mortality and incorporate clinical variables, there is a lack of tools that have utilized genetic data and that predict ischemic events. Aims: This study aims to develop and validate a ML model incorporating genotyping and clinical data to enhance prediction of ischemic outcomes for PCI patients utilizing large prospectively derived diverse datasets. Methods: Patients from the TAILOR-PCI trial (n=5302) were utilized for model development. 50% of the sample was utilized for Boruta feature selection and 50% for training and testing using cross validation. Features included demographics, medical history, medications, PCI characteristics, and genetic data (specifically, CYP2C19 *2, *3, *17 alleles). The primary endpoint was a composite of cardiovascular death, myocardial infarction, stroke, stent thrombosis, and severe recurrent ischemia at 12 months. Multiple ML classification algorithms, including Support Vector Machine (SVM) polynomial, Random Forest, Light gradient boost, XG Boost, among others, were benchmarked for their prediction performance on rare events. The top performing classifiers were externally validated on an independent dataset from the PRECISION PCI study (n=3,745). Results: Mean participant age of the training set was 64.2 ± 11.0 years, with 75.4% being male. During follow-up of 12 months, among 4,572 patients in the entire cohort 343 (7.5%) met the primary outcome. The SVM polynomial model demonstrated the highest area under the curve (AUC) of 0.67 for predicting the primary outcome with test dataset. The sensitivity, specificity, precision, and recall were 0.87, 0.28, 0.07, and 0.87 respectively (Figure). Peripheral arterial disease, body mass index, and age were among the top variables by feature importance. Conclusion: ML models incorporating both clinical and genetic data are feasible and highly promising in predicting major adverse cardiac events that may help guide use of anti-platelet drug therapy. The AUC values are reasonable given imbalances and misclassifications in datasets, and further model optimization with prospective utilization of the model will be paramount.

Background: Remote patient monitoring (RPM) is an emerging focus in health care, and specialized programs may reduce medical costs, supplement in-office visits, and improve patient satisfaction. In this study, we describe the development, feasibility, and early outcomes of an RPM program for patients with decompensated cirrhosis. Methods: Forty-six patients were offered enrollment at the time of hospital discharge in the cirrhosis RPM program (CiRPM), of which 41 completed at least 30 days of monitoring. Participants were mailed remote monitoring equipment and a tablet to be used for patient-reported outcomes. Alerts were continuously monitored by virtual nursing staff who could perform targeted interventions. A cohort of historical controls (n = 74) was created for comparison using inverse probability of treatment weighting. Results: Patients were enrolled in the program for a mean of 83.9 days, with 28 (68%) completing the full 90-day program. Participants uploaded vital signs and responded to symptom-based questionnaires on 93% of the monitored days. On end-of-program surveys, over 75% of patients expressed satisfaction with the program. Gender, age, and MELD-Na were similar between CiRPM and weighted control groups. The 90-day readmission rate was 34% in CiRPM and 47% in weighted controls. In the CiRPM group, 12% of subjects had 2 or more admissions, compared to 37% in the weighted control group. Conclusion: This study demonstrates the feasibility of a cirrhosis-specific RPM program. Overall, patient satisfaction and utilization of the CiRPM was high. Future studies are needed to confirm the impact of RPM on the reduction of hospital readmissions in decompensated cirrhosis.

Introduction Hospitalizations, proton-pump inhibitors (PPI), and systemic antibiotics increase the risk of Clostridioides difficile infection (CDI) in cirrhosis. We compare the risk of CDI with hepatic encephalopathy (HE) medications, hypothesizing that rifaximin may decrease CDI risk. Materials and Methods A retrospective study of hospitalized HE patients treated with lactulose and/or rifaximin at Mayo Clinic Minnesota, Florida, and Arizona from 2008 to 2013 was conducted. Data on demographics, hospitalizations, antibiotics, and PPI use and CDI were gathered. Univariate and multivariable cox models were constructed. Results We found 1,112 hospitalizations in 1,055 unique patients (55 had 1 subsequent readmission, 1 patient had 2); 428 (40.6%) patients were women (median age: 58 years [interquartile range: 52–65]). CDI developed after 66/1,112 (5.9%) hospitalizations within 12 months post-discharge. Lactulose was administered in 21 (31.8%), rifaximin in 5 (7.6%), both in 40 (60.6%) hospitalizations. Systemic antibiotics were used in 28 (42.4%) patients and PPIs in 60 (90.9%) patients.Univariate analysis using medication (with lactulose alone as the reference group) showed rifaximin was not significantly associated with CDI compared with lactulose (hazard ratio [HR]: 1.57, 95% confidence interval [CI]: 0.57–4.33, p = 0.39). Use of both medications was not significant compared with lactulose (HR: 1.41, 95% CI: 0.84–2.38, p = 0.19). Results were similar after controlling for confounders. Multivariable analysis based on length of stay, age, and gender showed no differences between rifaximin versus lactulose and both versus lactulose. Conclusion There is no significant difference between lactulose and rifaximin on CDI development in HE patients. However, CDI should still be considered when managing diarrhea in HE patients.

Objectives Chronic pancreatitis (CP) is a chronic fibroinflammatory condition of the pancreas difficult to diagnose in early stages. Novel biomarkers useful to facilitate early diagnosis or treatment responses may be found in biofluids. Although saliva can be easily and noninvasively collected from patients, useful salivary biomarkers from CP patients have not yet been identified. Methods Here, we analyzed the proteome by quantitative proteomics, cytokine/chemokine levels by Luminex analysis, prostaglandin E2 (PGE2) levels by a mass spectrometry-based assay, and bacterial species diversity by 16S ribosomal ribonucleic acid sequencing in saliva samples from confirmed CP patients and healthy controls. Results Our results indicate the presence of various differentially expressed proteins, cytokines/chemokines, and a loss of oral bacterial diversity in the saliva of CP patients. The PGE2 levels trend toward elevation in CP patients. Area under the receiver operating characteristic curve models for proteomic, cytokine, and PGE2 assays ranged from 0.59 to 0.90. Conclusions Collectively, our studies identify a range of putative CP biomarkers and alterations in human saliva requiring further validation. The biomarker discovery approaches we used might lead to identification of biomarkers useful for CP diagnosis and monitoring.

Introduction: Peripheral arterial disease (PAD) is a strong predictor of systemic atherosclerosis with substantial morbidity and mortality. Medical management in PAD and coronary artery disease (CAD) has changed in recent eras with the use of high intensity statin and dual antiplatelet therapies. Hypothesis: Test whether mortality and major adverse cardiovascular events (MACE) (death and MI) has decreased over eras among PAD patients undergoning revascularization with percutaneous coronary intervention (PCI). Methods: Longitudinal cohort study of 549 patients who underwent PCI and non-invasive lower arterial study between 1997 and 2008 as identified in the Mayo Clinic PCI registry were stratified into 2 time eras: 1997-2003 (era1), and 2004-2008 (era2). PAD was assessed using the highest resting ABI and stratified according to established categories; < 0.90 (definite PAD); 1.00-1.39 (normal); and ≥ 1.40 defined a non-compressible artery. Kaplan-Meier methods were used to analyze follow-up survival. Results: Stratified by ABI over both time eras, 225 patients had PAD, 94 were non-compressible, and 142 were normal. Overall mortality rates were higher in era2 non-compressible and PAD patients than in non-PAD patients when compared to era1 (58%, 33%, 26%, vs 49% 32%, 19%, p<0.001 for all within-era comparisons) (Figure 1). MACE-free survival were higher in era2 non-compressible and PAD patients than in non-PAD patients when compared to the era1 (64%, 41%, 29% vs 59% 39%, 25%, p≤0.005 for all within-era comparisons). There were significant differences in medical therapy between era2 vs. era1 with higher discharge use of aspirin (98%; 94%), beta-blockers (88%; 66%), any lipid lowering drugs (86%; 76%), ACE inhibitors (67%; 56%), and diuretics (52%; 40%). Conclusions: Over the past two eras, the overall mortality has increased among patients with PAD and non-compressible vessels in era2 vs. era1 despite higher utilization of statin and antiplatelet therapies.

Introduction: Relationship of QOL and frailty are poorly defined. Hypothesis: To study the prevalence and prognosis of poor QOL in frail patients. Methods: Predismissal standardized tests for frailty and QOL in 2 cohorts were administered. In cohort 1, 629 patients ≥ 65 years underwent percutaneous coronary intervention (PCI) from 2005-2008, frailty (Fried) and QOL [SF-36 and SAQ] were ascertained. Cohort 2 [921 patients, 535 PCI, ≥55 years] underwent cardiac catheterization from 2014-18, frailty (Rockwood) and QOL (Likert) were determined. Results: In cohort 1, 19%; and 20% in cohort 2 were frail. Lower median SAQ (58.9 vs. 82.2); physical (29.5 vs. 43.9) and mental (49.2 vs. 57.4) scores of SF-36 in cohort 1 higher rates of fair/poor health (56% vs 18%) in cohort 2 (all comparisons, p<0.001) was seen in frail patients. As compared to without, frail patients were 5 times more likely (59% vs. 11%, p<0.001) in cohort 1 and 7 times more likely (56% vs. 8%) in cohort 2 to have poor QOL. Adjusted three year all-cause death and death/myocardial infarction (MI) was significantly higher for patients undergoing PCI with frailty; [HR (95% CI) death, 4.20 (2.63, 6.68) and death or MI HR 2.72 (1.91, 3.87), p<0.001) and with poor QOL [HR death 2.47 (1.59, 3.84) and death or MI 1.61 (1.16, 2.24) p<0.001). There was no significant interaction between frailty and QOL (p=0.64) indicating their independent prognostic influence. Conclusions: Poor QOL is commonly seen with frailty. Both frailty and poor QOL had significant and independent association with long-term survival.

AbstractBackgroundCurrent approaches in tracking Clostridioides difficile infection (CDI) and individualizing patient management are incompletely defined.MethodsWe recruited 468 subjects with CDI at Mayo Clinic Rochester between May and December 2016 and performed whole-genome sequencing (WGS) on C. difficile isolates from 397. WGS was also performed on isolates from a subset of the subjects at the time of a recurrence of infection. The sequence data were analyzed by determining core genome multilocus sequence type (cgMLST), with isolates grouped by allelic differences and the predicted ribotype.ResultsThere were no correlations between C. difficile isolates based either on cgMLST or ribotype groupings and CDI outcome. An epidemiologic assessment of hospitalized subjects harboring C. difficile isolates with ≤2 allelic differences, based on standard infection prevention and control assessment, revealed no evidence of person-to-person transmission. Interestingly, community-acquired CDI subjects in 40% of groups with ≤2 allelic differences resided within the same zip code. Among 18 subjects clinically classified as having recurrent CDI, WGS revealed 14 with initial and subsequent isolates differing by ≤2 allelic differences, suggesting a relapse of infection with the same initial strain, and 4 with isolates differing by &gt;50 allelic differences, suggesting reinfection. Among the 5 subjects classified as having a reinfection based on the timing of recurrence, 3 had isolates with ≤2 allelic differences between them, suggesting a relapse, and 2 had isolates differing by &gt;50 allelic differences, suggesting reinfection.ConclusionsOur findings point to potential transmission of C. difficile in the community. WGS better differentiates relapse from reinfection than do definitions based on the timing of recurrence.