Evaluation of Machine Learning Approaches for Predicting Warfarin Discharge Dose in Cardiac Surgery Patients: Retrospective Algorithm Development and Validation Study.

BACKGROUND: Warfarin dosing in cardiac surgery patients is complicated by a heightened sensitivity to the drug, predisposing patients to adverse events. Predictive algorithms are therefore needed to guide warfarin dosing in cardiac surgery patients. OBJECTIVE: This study aimed to develop and validate an algorithm for predicting the warfarin dose needed to attain a therapeutic international normalized ratio (INR) at the time of discharge in cardiac surgery patients. METHODS: We abstracted variables influencing warfarin dosage from the records of 1031 encounters initiating warfarin between April 1, 2011, and November 29, 2019, at St Michael's Hospital in Toronto, Ontario, Canada. We compared the performance of penalized linear regression, k-nearest neighbors, random forest regression, gradient boosting, multivariate adaptive regression splines, and an ensemble model combining the predictions of the 5 regression models. We developed and validated separate models for predicting the warfarin dose required for achieving a discharge INR of 2.0-3.0 in patients undergoing all forms of cardiac surgery except mechanical mitral valve replacement and a discharge INR of 2.5-3.5 in patients receiving a mechanical mitral valve replacement. For the former, we selected 80% of encounters (n=780) who had initiated warfarin during their hospital admission and had achieved a target INR of 2.0-3.0 at the time of discharge as the training cohort. Following 10-fold cross-validation, model accuracy was evaluated in a test cohort comprised solely of cardiac surgery patients. For patients requiring a target INR of 2.5-3.5 (n=165), we used leave-p-out cross-validation (p=3 observations) to estimate model performance. For each approach, we determined the mean absolute error (MAE) and the proportion of predictions within 20% of the true warfarin dose. We retrospectively evaluated the best-performing algorithm in clinical practice by comparing the proportion of cardiovascular surgery patients discharged with a therapeutic INR before (April 2011 and July 2019) and following (September 2021 and May 2, 2022) its implementation in routine care. RESULTS: Random forest regression was the best-performing model for patients with a target INR of 2.0-3.0, an MAE of 1.13 mg, and 39.5% of predictions of falling within 20% of the actual therapeutic discharge dose. For patients with a target INR of 2.5-3.5, the ensemble model performed best, with an MAE of 1.11 mg and 43.6% of predictions being within 20% of the actual therapeutic discharge dose. The proportion of cardiovascular surgery patients discharged with a therapeutic INR before and following implementation of these algorithms in clinical practice was 47.5% (305/641) and 61.1% (11/18), respectively. CONCLUSIONS: Machine learning algorithms based on routinely available clinical data can help guide initial warfarin dosing in cardiac surgery patients and optimize the postsurgical anticoagulation of these patients.

Evaluation of taste sensitivity in patients undergoing coronary artery bypass graft surgery.

Patients report changes in their perception of food tastes following cardiac surgery. This study was designed to explore changes in taste sensitivity following coronary artery bypass graft (CABG) surgery. Detection and recognition thresholds for sweet (sucrose), salty (sodium chloride), sour (citric acid), and bitter (quinine hydrochloride) were determined using the multiple forced-choice ascending concentration series method at baseline (presurgical), discharge, 5 weeks, and 16 weeks post-CABG. Demographic and gastrointestinal data were also obtained. Mixed-model analyses for repeated measures were performed using the baseline scores as reference. Thirty-three patients (mean age=61.8+/-8 years), consented to participate in the study between January 2003 and January 2006, with 13 completing all visits. Detection and recognition thresholds for sweet were significantly lower at discharge compared with baseline (1.7+/-1.2 vs 2.43+/-1.4 and 5.1+/-1.8 vs 5.5+/-1.3, respectively; P<0.05). This difference remained significant 4 months after surgery. Detection and recognition thresholds for salt also declined with time, with significant differences at 4 months post-surgery (2.3+/-2.0 vs 1.8+/-1.5; P<0.001 and 5.3+/-1.3 vs 4.2+/-2.2; P<0.05, respectively). The same trends were noted for the detection of sour and the recognition of bitter. Patients undergoing CABG demonstrated stable or improved taste sensitivity during the recovery period. Further studies aimed at clarifying the relationships between the biological state, taste sensitivity, reported taste changes, and food intake will help to clarify the clinical impact of taste changes and subsequently to guide clinical nutrition care.

Msh2 deficiency enhances somatic Apc and p53 mutations in Apc+/-Msh2-/- mice.

Inactivation of the adenomatous polyposis coli (Apc) gene by loss of the wild-type Apc allele (LOH) is a prerequisite for the development of intestinal adenomas in Msh2 proficient Min (Apc+/-Msh2+/+) mice. In contrast, adenomas from Msh2 deficient Min (Apc+/-Msh2-/-) mice are not usually associated with LOH. Given the role of Msh2 in post-replicative DNA repair, this study investigated whether Msh2 deficiency enhances somatic Apc and p53 mutations in Apc+/-Msh2-/- mice. Somatic Apc mutations (5/sample) were observed in the non-neoplastic intestinal mucosa from Apc+/-Msh2-/- mice but not from Min mice, suggesting that Msh2 deficiency is associated with a hypermutable state in the intestinal mucosa from Apc+/-Msh2-/- mice. Adenomas from Apc+/-Msh2-/- mice had a 2-fold higher rate of somatic Apc mutations (10/adenoma) than the non-neoplastic intestinal mucosa (5/sample), and did not demonstrate LOH. Truncating Apc mutations were observed in 82% of the adenomas from Apc+/-Msh2-/- mice and were not observed at all in the non-neoplastic intestinal mucosa. In contrast, in Min mice, all adenomas demonstrated LOH, had significantly less numbers of somatic Apc mutations (1.8 mutations/adenoma) compared with the adenomas from Apc+/-Msh2-/- mice, and harbored no truncating Apc mutations. These observations suggest that somatic Apc mutations, and not LOH, is a likely mechanism by which the Apc gene is inactivated in the development of adenomas in Apc+/-Msh2-/- mice in contrast to Min mice. Adenomas from Apc+/-Msh2-/- mice, but not from Min mice, also harbored somatic p53 mutations (mutation frequency of 45.5%), reflecting hypermutability associated with Msh2 deficiency. The nature and frequency of somatic Apc and p53 mutations in Apc+/-Msh2-/- mice suggest that many genomic sites, in addition to genes containing simple repeated sequences, are at risk of somatic mutations associated with Msh2 deficiency.