Multicentre validation of a machine learning model for predicting respiratory failure after noncardiac surgery.

BACKGROUND: Postoperative respiratory failure is a serious complication that could benefit from early accurate identification of high-risk patients. We developed and validated a machine learning model to predict postoperative respiratory failure, defined as prolonged (>48 h) mechanical ventilation or reintubation after surgery. METHODS: Easily extractable electronic health record (EHR) variables that do not require subjective assessment by clinicians were used. From EHR data of 307,333 noncardiac surgical cases, the model, trained with a gradient boosting algorithm, utilised a derivation cohort of 99,025 cases from Seoul National University Hospital (2013-9). External validation was performed using three separate cohorts A-C from different hospitals comprising 208,308 cases. Model performance was assessed by area under the receiver operating characteristic (AUROC) curve and area under the precision-recall curve (AUPRC), a measure of sensitivity and precision at different thresholds. RESULTS: The model included eight variables: serum albumin, age, duration of anaesthesia, serum glucose, prothrombin time, serum creatinine, white blood cell count, and body mass index. Internally, the model achieved an AUROC of 0.912 (95% confidence interval [CI], 0.908-0.915) and AUPRC of 0.113. In external validation cohorts A, B, and C, the model achieved AUROCs of 0.879 (95% CI, 0.876-0.882), 0.872 (95% CI, 0.870-0.874), and 0.931 (95% CI, 0.925-0.936), and AUPRCs of 0.029, 0.083, and 0.124, respectively. CONCLUSIONS: Utilising just eight easily extractable variables, this machine learning model demonstrated excellent discrimination in both internal and external validation for predicting postoperative respiratory failure. The model enables personalised risk stratification and facilitates data-driven clinical decision-making.

Mortality Risk within 14 Days after Coronavirus Disease 2019 Diagnosis in Dementia Patients: A Nationwide Analysis.

INTRODUCTION: The study evaluated the increased mortality risk within 14 days of coronavirus disease 2019 (COVID-19) diagnosis in dementia patients. METHODS: This retrospective study was conducted from February to April 2020 using the COVID-19 patients' database from the Korea Disease Control and Prevention Agency. The risk factors for early death within 14 days were determined using generalized logistic regression performed in a stepwise manner. Dementia patients diagnosed with COVID-19 were used for the study. The propensity score-matched cohort was included as controls. The differences in mortality within 14 days after COVID-19 diagnosis between the dementia patients and controls were evaluated. RESULTS: We enrolled 5,349 COVID-19 patients from the database; 224 had dementia as comorbidity. The mortality rate within 14 days after COVID-19 diagnosis in dementia patients and the controls was 23.7% versus 1.7%, respectively, before propensity score matching (PSM) (p < 0.001), and 23.7% versus 9.2% after PSM (p < 0.001). The hazard ratio (HR) for mortality within 14 days in COVID-19 patients with dementia was significant even after PSM (HR 5.104, 95% confidence interval 2.889-5.673, p < 0.001). The survival curve of dementia patients was steeply inclined within 14 days after COVID-19 diagnosis, resulting in 70.7% of all deaths in dementia patients. CONCLUSIONS: COVID-19 patients with dementia had a higher risk of early death within 14 days. Thus, prompt intervention is necessary for dementia patients after COVID-19 diagnosis.

Extracting Structured Genotype Information from Free-Text HLA Reports Using a Rule-Based Approach.

BACKGROUND: Human leukocyte antigen (HLA) typing is important for transplant patients to prevent a severe mismatch reaction, and the result can also support the diagnosis of various disease or prediction of drug side effects. However, such secondary applications of HLA typing results are limited because they are typically provided in free-text format or PDFs on electronic medical records. We here propose a method to convert HLA genotype information stored in an unstructured format into a reusable structured format by extracting serotype/allele information. METHODS: We queried HLA typing reports from the clinical data warehouse of Seoul National University Hospital (SUPPREME) from 2000 to 2018 as a rule-development data set (64,024 reports) and from the most recent year (6,181 reports) as a test set. We used a rule-based natural language approach using a Python regex function to extract the 1) number of patients in the report, 2) clinical characteristics such as indication of the HLA testing, and 3) precise HLA genotypes. The performance of the rules and codes was evaluated by comparison between the extracted results from the test set and a validation set generated by manual curation. RESULTS: Among 11,287 reports for development set and 1,107 for the test set describing HLA typing for a single patient, iterative rule generation developed 124 extracting rules and 8 cleaning rules for HLA genotypes. Application of these rules extracted HLA genotypes with 0.892-0.999 precision and 0.795-0.998 recall for the five HLA genes. The precision and recall of the extracting rules for the number of patients in a report were 0.997 and 0.994 and those for the clinical variable extraction were 0.997 and 0.992, respectively. All extracted HLA alleles and serotypes were transformed according to formal HLA nomenclature by the cleaning rules. CONCLUSION: The rule-based HLA genotype extraction method shows reliable accuracy. We believe that there are significant number of patients who takes profit when this under-used genetic information will be return to them.

Optimal extent of resection for glioblastoma according to site, extension, and size: a population-based study in the temozolomide era.

The effect of the extent of resection (EOR) on prognosis in glioblastoma may differ depending on various conditions. We evaluated the prognostic impact of the EOR for glioblastoma according to the tumor site, extension, and size. Data from glioblastoma patients who underwent gross total resection (GTR), subtotal resection (STR), or open biopsy between 2005 and 2014 were retrieved from the Surveillance, Epidemiology, and End Results database. Univariate and multivariate analyses for overall survival (OS) were performed. Between 2005-2009 and 2010-2014, the proportion of GTR and STR performed increased from 41.4 to 42.3% and 33.0 to 37.1%, respectively. EOR only affected OS in the 3 years after diagnosis. Median survival in the GTR (n = 4155), STR (n = 3498), and open biopsy (n = 2258) groups was 17, 13, and 13 months, respectively (p < .001). STR showed no significant difference in OS from open biopsy (p = .33). GTR increased OS for midline-crossing tumors. Although STR was more frequently performed than GTR for tumors ≥ 6 cm in size, GTR significantly increased the OS rate relative to STR for tumors 6-8 cm in size (p = .001). For tumors ≥ 8 cm, STR was comparable to GTR (p = .61) and superior to open biopsy (p = .05). GTR needs to be performed more frequently for glioblastoma measuring ≥ 6 cm or that have crossed the midline to increase OS. STR was marginally superior to open biopsy when the tumor was ≥ 8 cm.

Recurrence outcomes after omission of postoperative radiotherapy following breast-conserving surgery for ductal carcinoma in situ of the breast: a multicenter, retrospective study in Korea (KROG 16-02).

PURPOSE: To evaluate the loco-regional recurrence (LRR) rate after breast-conserving surgery without postoperative radiotherapy (RT) for ductal carcinoma in situ (DCIS) of the breast. METHODS: Between 2000 and 2010, 311 DCIS patients from 9 institutions were analyzed retrospectively. The median age was 47 (range, 20-82). The median tumor size was 7 mm (range, 0.01-76). Margin width was <1 cm in 85 patients (27.3%), and nuclear grade was high in 37 patients (11.9%). Two hundred and three patients (65.3%) received tamoxifen. RESULTS: With a median follow-up of 74 months (range, 5-189), there were 11 local recurrences (invasive carcinoma in 6 and DCIS in 5) and 1 regional recurrence. The 7-year LRR rate was 3.8%. On univariate analysis, age and margin width were significant risk factors influencing LRR (p = 0.017 and 0.014, respectively). When age and margin width were combined among 211 patients whose margin width were available, the 7-year LRR rates were as follows (p < 0.001): (1) 0% in patients with age >50 years and any margin width status (n = 64), (2) 1.2% in age ≤50 years and margin width ≥1 cm (n = 93), (3) 13.1% in age ≤50 years and margin width <1 cm (n = 54). CONCLUSIONS: The LRR rate was very low in selected DCIS patients treated with breast-conserving surgery without postoperative RT. However, adjuvant RT should be considered for those with age ≤50 years and margin width <1 cm.

Change in carbohydrate antigen 19-9 level as a prognostic marker of overall survival in locally advanced pancreatic cancer treated with concurrent chemoradiotherapy.

PURPOSE: To investigate the significance of carbohydrate antigen 19-9 (CA19-9) levels for survival in locally advanced pancreatic cancer (LAPC) treated with concurrent chemoradiotherapy (CCRT). METHODS/PATIENTS: We retrospectively reviewed data from 97 LAPC patients treated with CCRT between 2000 and 2013. CA19-9 levels (initial and post-CCRT) and their changes [{(post-CCRT CA19-9 level - initial CA19-9 level)/(initial CA19-9 level)} × 100] were analyzed for overall survival. A cut-off point of 37 U/mL was used to analyze initial and post-CCRT CA19-9 levels. In order to define an optimal cut-off point for change in CA19-9 level, the maxstat package of R was applied. RESULTS: Median overall survival was 14.7 months (95% CI 13.4-16.0), and the 2-year survival rate was 16.5%. The estimated optimal cut-off point of CA19-9 level change was 94.4%. On univariate analyses, CA19-9 level change between initial and post-CCRT was significantly correlated with overall survival (median survival time 9.7 vs 16.3 months, p < 0.001). Multivariate analyses confirmed that CA19-9 level change from initial to post-CCRT was the only prognostic factor (p < 0.001). CONCLUSIONS: Change in CA19-9 level between initial and post-CCRT was a significant prognostic marker for overall survival in LAPC treated with CCRT. A CA19-9 level increase >94.4% might serve as a surrogate marker for poor survival in patients with LAPC undergoing CCRT, and the prognostic power surpassed other CA19-9 variables including initial and post-CCRT values.

Future prospects of health management systems using cellular phones.

BACKGROUND: Cellular phones enable communication between healthcare providers and patients for prevention, diagnosis, and treatment of diseases. However, few studies have examined the user-friendliness or effectiveness of cellular phone-based medical informatics (CPBMI) for healthcare. MATERIALS AND METHODS: This study investigated the use of CPBMI to identify its current status within the medical field, advantages and disadvantages, practicability, clinical effectiveness, costs, and cost-saving potential. RESULTS: CPBMI was validated in terms of practicality and provision of medical benefits. It is critical to use CPBMI in accordance with the different features of each disease and condition. Use of CPBMI is expected to be especially useful for patients with chronic disease. CONCLUSIONS: We discussed the current status of the clinical use, benefits, and risks of CPBMI. CPBMI and information technology-based health management tools are anticipated to become useful and effective components of healthcare management in the future.

Hunting for the vulnerability in chondrosarcoma by tracing metabolic and genetic links.

In this Backstory, we narrate our journey in establishing a multidisciplinary team for sarcoma research and uncovering vulnerabilities in chondrosarcoma cells associated with their NAD+ dependencies for survival.1 Our findings hold promise for exploitation, yielding a synergistic cytotoxic effect when combined with systemic therapy.

Privacy-Preserving Prediction of Postoperative Mortality in Multi-Institutional Data: Development and Usability Study.

BACKGROUND: To circumvent regulatory barriers that limit medical data exchange due to personal information security concerns, we use homomorphic encryption (HE) technology, enabling computation on encrypted data and enhancing privacy. OBJECTIVE: This study explores whether using HE to integrate encrypted multi-institutional data enhances predictive power in research, focusing on the integration feasibility across institutions and determining the optimal size of hospital data sets for improved prediction models. METHODS: We used data from 341,007 individuals aged 18 years and older who underwent noncardiac surgeries across 3 medical institutions. The study focused on predicting in-hospital mortality within 30 days postoperatively, using secure logistic regression based on HE as the prediction model. We compared the predictive performance of this model using plaintext data from a single institution against a model using encrypted data from multiple institutions. RESULTS: The predictive model using encrypted data from all 3 institutions exhibited the best performance based on area under the receiver operating characteristic curve (0.941); the model combining Asan Medical Center (AMC) and Seoul National University Hospital (SNUH) data exhibited the best predictive performance based on area under the precision-recall curve (0.132). Both Ewha Womans University Medical Center and SNUH demonstrated improvement in predictive power for their own institutions upon their respective data's addition to the AMC data. CONCLUSIONS: Prediction models using multi-institutional data sets processed with HE outperformed those using single-institution data sets, especially when our model adaptation approach was applied, which was further validated on a smaller host hospital with a limited data set.

Decoupling NAD+ metabolic dependency in chondrosarcoma by targeting the SIRT1-HIF-2α axis.

Chondrosarcomas represent the second most common primary bone malignancy. Despite the vulnerability of chondrosarcoma cells to nicotinamide adenine dinucleotide (NAD+) depletion, targeting the NAD+ synthesis pathway remains challenging due to broad implications in biological processes. Here, we establish SIRT1 as a central mediator reinforcing the dependency of chondrosarcoma cells on NAD+ metabolism via HIF-2α-mediated transcriptional reprogramming. SIRT1 knockdown abolishes aggressive phenotypes of chondrosarcomas in orthotopically transplanted tumors in mice. Chondrosarcoma cells thrive under glucose starvation by accumulating NAD+ and subsequently activating the SIRT1-HIF-2α axis. Decoupling this link via SIRT1 inhibition unleashes apoptosis and suppresses tumor progression in conjunction with chemotherapy. Unsupervised clustering analysis identifies a high-risk chondrosarcoma patient subgroup characterized by the upregulation of NAD+ biosynthesis genes. Finally, SIRT1 inhibition abolishes HIF-2α transcriptional activity and sensitizes chondrosarcoma cells to doxorubicin-induced cytotoxicity, irrespective of underlying pathways to accumulate intracellular NAD+. We provide system-level guidelines to develop therapeutic strategies for chondrosarcomas.

Predisposal of Interferon Regulatory Factor 1 Deficiency to Accumulate DNA Damage and Promote Osteoarthritis Development in Cartilage.

OBJECTIVE: Interferon regulatory factor 1 (IRF1) is a transcriptional regulator conventionally associated with immunomodulation. Recent molecular analyses mapping DNA binding sites of IRF1 have suggested its potential function in DNA repair. However, the physiologic significance of this noncanonical function remains unexplored. Here, we investigated the role of IRF1 in osteoarthritis (OA), a condition marked by senescence and chronic joint inflammation. METHODS: OA progression was examined in wild-type and Irf1-/- mice using histologic assessments and microcomputed tomography analysis of whole-joint OA manifestations and behavioral assessments of joint pain. An integrated analysis of assay for transposase-accessible chromatin with sequencing and whole transcriptome data was conducted for the functional assessment of IRF1 in chondrocytes. The role of IRF1 in DNA repair and senescence was investigated by assaying γ-H2AX foci and senescence-associated beta-galactosidase activity. RESULTS: Our genome-wide investigation of IRF1 footprinting in chondrocytes revealed its primary occupancies in the promoters of DNA repair genes without noticeable footprint patterns in those of interferon-responsive genes. Chondrocytes lacking IRF1 accumulated irreversible DNA damage under oxidative stress, facilitating their entry into cellular senescence. IRF1 was down-regulated in the cartilage of human and mouse OA. Although IRF1 overexpression did not elicit an inflammatory response in joints or affect OA development, genetic deletion of Irf1 caused enhanced chondrocyte senescence and exacerbated post-traumatic OA in mice. CONCLUSION: IRF1 offers DNA damage surveillance in chondrocytes, protecting them from oxidative stress associated with OA risk factors. Our study provides a crucial and cautionary perspective that compromising IRF1 activity renders chondrocytes vulnerable to cellular senescence and promotes OA development.

Causal relationship between particulate matter 2.5 and diabetes: two sample Mendelian randomization.

Backgrounds: Many studies have shown particulate matter has emerged as one of the major environmental risk factors for diabetes; however, studies on the causal relationship between particulate matter 2.5 (PM2.5) and diabetes based on genetic approaches are scarce. The study estimated the causal relationship between diabetes and PM2.5 using two sample mendelian randomization (TSMR). Methods: We collected genetic data from European ancestry publicly available genome wide association studies (GWAS) summary data through the MR-BASE repository. The IEU GWAS information output PM2.5 from the Single nucleotide polymorphisms (SNPs) GWAS pipeline using pheasant-derived variables (Consortium = MRC-IEU, sample size: 423,796). The annual relationship of PM2.5 (2010) were modeled for each address using a Land Use Regression model developed as part of the European Study of Cohorts for Air Pollution Effects. Diabetes GWAS information (Consortium = MRC-IEU, sample size: 461,578) were used, and the genetic variants were used as the instrumental variables (IVs). We performed three representative Mendelian Randomization (MR) methods: Inverse Variance Weighted regression (IVW), Egger, and weighted median for causal relationship using genetic variants. Furthermore, we used a novel method called MR Mixture to identify outlier SNPs. Results: From the IVW method, we revealed the causal relationship between PM2.5 and diabetes (Odds ratio [OR]: 1.041, 95% CI: 1.008-1.076, P = 0.016), and the finding was substantiated by the absence of any directional horizontal pleiotropy through MR-Egger regression (ß = 0.016, P = 0.687). From the IVW fixed-effect method (i.e., one of the MR machine learning mixture methods), we excluded outlier SNP (rs1537371) and showed the best predictive model (AUC = 0.72) with a causal relationship between PM2.5 and diabetes (OR: 1.028, 95% CI: 1.006-1.049, P = 0.012). Conclusion: We identified the hypothesis that there is a causal relationship between PM2.5 and diabetes in the European population, using MR methods.

The effect of residential greenness during pregnancy on infant neurodevelopment using propensity score weighting: A prospective mother-infant paired cohort study.

BACKGROUND: While prior studies have suggested an association between green spaces and infant neurodevelopment, the causal effect of green space exposure during pregnancy has not been fully investigated. This study aimed to identify with causal inference the effect of exposure to residential greenness during pregnancy on infants' mental-psychomotor development and the role of maternal education in modifying this association. METHODS: We prospectively collected data of pregnant women and their infants from Mothers and Children Environmental Health cohort study. Based on residential addresses, we compiled information on the percent of green space using different buffer distances (100 m, 300 m, and 500 m) and air pollution (PM2.5). Infant neurodevelopment was measured at 6 months of age using the Korean Bayley Scales of Infant Development II Mental Developmental Index (MDI) and Psychomotor Developmental Index (PDI). Generalized propensity scores (GPSs) were estimated from machine-learning (ML) algorithms. We deduced causal inference through GPS adjustment and weighting approaches. Further analyses confirmed whether the association was altered by maternal academic background. RESULTS: A total of 845 mother-infant pairs from the cohort study were included. We found that exposure to green spaces was robustly associated with infants' mental development. For example, an increase in % green space within 300 m increased the MDI by 14.32 (95 % confidence interval [CI], 3.44-25.2) in the weighting approach. Additionally, the association was even more noticeable for mothers with college degrees or above: an increase in % green space within 300 m increased the MDI by 23.69 (95 % CI, 8.53-38.85) and the PDI by 22.45 (95 % CI, 2.58-42.33) in the weighting approach. This association did not appear in mothers without college degrees. CONCLUSION: Exposure to green spaces during pregnancy showed a beneficial relationship with infant mental development. Maternal academic background could modify the impact of green space exposure on infant neurodevelopment.

COVID-19, maternal, and neonatal outcomes: National Mother-Child Cohort (NMCC) of K-COV-N cohort in South Korea.

Throughout the COVID-19 pandemic, pregnant women have been classified as a vulnerable population. However, the evidence on the effect of infection during pregnancy on maternal and neonatal outcomes is still uncertain, and related research comprising a large population of pregnant women in Asian countries is limited. We constructed a national cohort including mothers and children (369,887 pairs) registered in the Prevention Agency-COVID-19-National Health Insurance Service (COV-N), from January 1, 2020 to March 31, 2022. We performed propensity score matchings and generalized estimation equation models to estimate the effect of COVID-19 on maternal and neonatal outcomes. In summary, we found little evidence of the effect of COVID-19 infection during pregnancy on maternal and neonatal outcomes; however, a relationship between COVID-19 infection in the second trimester and postpartum hemorrhages was discovered (Odds ratio (OR) of Delta period: 2.26, 95% Confidence intervals (CI): 1.26, 4.05). In addition, neonatal intensive care unit (NICU) admissions increased due to COVID-19 infection (pre-Delta period: 2.31, 95% CI: 1.31, 4.10; Delta period: 1.99, 95% CI: 1.47, 2.69; Omicron period: 2.36, 95% CI: 1.75, 3.18). Based on the national retrospective cohort study data, this study investigated the effects of COVID-19 infection on maternal and neonatal outcomes in Korea from the pre-Delta to the initial Omicron epidemic periods. Our evidence suggests that the timely and successful policies of the government and academia in response to COVID-19 infections in newborns in Korea may cause an increase in NICU admissions, but nonetheless, they prevent adverse maternal and neonatal outcomes simultaneously.

Sarcoma Immunotherapy: Confronting Present Hurdles and Unveiling Upcoming Opportunities.

Sarcomas are rare and heterogeneous mesenchymal neoplasms originating from the bone or soft tissues, which pose significant treatment challenges. The current standard treatment for sarcomas consists of surgical resection, often combined with chemo- and radiotherapy; however, local recurrence and metastasis remain significant concerns. Although immunotherapy has demonstrated promise in improving long-term survival rates for certain cancers, sarcomas are generally considered to be relatively less immunogenic than other tumors, presenting substantial challenges for effective immunotherapy. In this review, we examine the possible opportunities for sarcoma immunotherapy, noting cancer testis antigens expressed in sarcomas. We then cover the current status of immunotherapies in sarcomas, including progress in cancer vaccines, immune checkpoint inhibitors, and adoptive cellular therapy and their potential in combating these tumors. Furthermore, we discuss the limitations of immunotherapies in sarcomas, including a low tumor mutation burden and immunosuppressive tumor microenvironment, and explore potential strategies to tackle the immunosuppressive barriers in therapeutic interventions, shedding light on the development of effective and personalized treatments for sarcomas. Overall, this review provides a comprehensive overview of the current status and potential of immunotherapies in sarcoma treatment, highlighting the challenges and opportunities for developing effective therapies to improve the outcomes of patients with these rare malignancies.

A COVID-19 mortality prediction model for Korean patients using nationwide Korean disease control and prevention agency database.

The experience of the early nationwide COVID-19 pandemic in South Korea led to an early shortage of medical resources. For efficient resource allocation, accurate prediction of the prognosis or mortality of confirmed patients is essential. Therefore, the aim of this study was to develop an accurate model for predicting COVID-19 mortality using epidemiolocal and clinical variables and for identifying a high-risk group of confirmed patients. Clinical and epidemiolocal variables of 4049 patients with confirmed COVID-19 between January 20, 2020 and April 30, 2020 collected by the Korean Disease Control and Prevention Agency were used. Among the 4049 total confirmed patients, 223 patients died, while 3826 patients were released from isolation. Patients who had the following risk factors showed significantly higher risk scores: age over 60 years, male sex, difficulty breathing, diabetes, cancer, dementia, change of consciousness, and hospitalization in the intensive care unit. High accuracy was shown for both the development set (n = 2467) and the validation set (n = 1582), with AUCs of 0.96 and 0.97, respectively. The prediction model developed in this study based on clinical features and epidemiological factors could be used for screening high-risk groups of patients and for evidence-based allocation of medical resources.

ANNO: A General Annotation Tool for Bilingual Clinical Note Information Extraction.

OBJECTIVE: This study was conducted to develop a generalizable annotation tool for bilingual complex clinical text annotation, which led to the design and development of a clinical text annotation tool, ANNO. METHODS: We designed ANNO to enable human annotators to support the annotation of information in clinical documents efficiently and accurately. First, annotations for different classes (word or phrase types) can be tagged according to the type of word using the dictionary function. In addition, it is possible to evaluate and reconcile differences by comparing annotation results between human annotators. Moreover, if the regular expression set for each class is updated during annotation, it is automatically reflected in the new document. The regular expression set created by human annotators is designed such that a word tagged once is automatically labeled in new documents. RESULTS: Because ANNO is a Docker-based web application, users can use it freely without being subjected to dependency issues. Human annotators can share their annotation markups as regular expression sets with a dictionary structure, and they can cross-check their annotated corpora with each other. The dictionary-based regular expression sharing function, cross-check function for each annotator, and standardized input (Microsoft Excel) and output (extensible markup language [XML]) formats are the main features of ANNO. CONCLUSIONS: With the growing need for massively annotated clinical data to support the development of machine learning models, we expect ANNO to be helpful to many researchers.

Reducing severe cutaneous adverse and type B adverse drug reactions using pre-stored human leukocyte antigen genotypes.

BACKGROUND: Several type B adverse drug reactions (ADRs), especially severe cutaneous adverse reactions (SCARs), are associated with particular human leukocyte antigen (HLA) genotypes. However, pre-stored HLA information obtained from other clinical workups has not been used to prevent ADRs. We aimed to simulate the preemptive use of pre-stored HLA information in electronic medical records to evaluate whether this information can prevent ADRs. METHODS: We analyzed the incidence and the risk of ADRs for selected HLA alleles (HLA-B*57:01, HLA-B*58:01, HLA-A*31:01, HLA-B*15:02, HLA-B*15:11, HLA-B*13:01, HLA-B*59:01, and HLA-A*32:01) and seven drugs (abacavir, allopurinol, carbamazepine, oxcarbazepine, dapsone, methazolamide, and vancomycin) using pre-stored HLA information of transplant patients based on the Pharmacogenomics Knowledge Base guidelines and experts' consensus. RESULTS: Among 11,988 HLA-tested transplant patients, 4092 (34.1%) had high-risk HLA alleles, 4583 (38.2%) were prescribed risk drugs, and 580 (4.8%) experienced type B ADRs. Patients with HLA-B*58:01 had a significantly higher incidence of type B ADR and SCARs associated with allopurinol use than that of patients without HLA-B*58:01 (17.2% vs. 11.9%, odds ratio [OR] 1.53 [95% confidence interval {CI} 1.09-2.13], p = 0.001, 2.3% versus 0.3%, OR 7.13 [95% CI 2.19-22.69], p < 0.001). Higher risks of type B ADR and SCARs were observed in patients taking carbamazepine or oxcarbazepine if they had one of HLA-A*31:01, HLA-B*15:02, or HLA-B*15:11 alleles. Vancomycin and dapsone use in HLA-A*32:01 and HLA-B*13:01 carriers, respectively, showed trends toward increased risk of type B ADRs. CONCLUSION: Utilization of pre-stored HLA data can prevent type B ADRs including SCARs by screening high-risk patients.

Multi-center validation of machine learning model for preoperative prediction of postoperative mortality.

Accurate prediction of postoperative mortality is important for not only successful postoperative patient care but also for information-based shared decision-making with patients and efficient allocation of medical resources. This study aimed to create a machine-learning prediction model for 30-day mortality after a non-cardiac surgery that adapts to the manageable amount of clinical information as input features and is validated against multi-centered rather than single-centered data. Data were collected from 454,404 patients over 18 years of age who underwent non-cardiac surgeries from four independent institutions. We performed a retrospective analysis of the retrieved data. Only 12-18 clinical variables were used for model training. Logistic regression, random forest classifier, extreme gradient boosting (XGBoost), and deep neural network methods were applied to compare the prediction performances. To reduce overfitting and create a robust model, bootstrapping and grid search with tenfold cross-validation were performed. The XGBoost method in Seoul National University Hospital (SNUH) data delivers the best performance in terms of the area under receiver operating characteristic curve (AUROC) (0.9376) and the area under the precision-recall curve (0.1593). The predictive performance was the best when the SNUH model was validated with Ewha Womans University Medical Center data (AUROC, 0.941). Preoperative albumin, prothrombin time, and age were the most important features in the model for each hospital. It is possible to create a robust artificial intelligence prediction model applicable to multiple institutions through a light predictive model using only minimal preoperative information that can be automatically extracted from each hospital.

Comprehensive DNA repair gene expression analysis and its prognostic significance in acute myeloid leukemia.

BACKGROUND: Deficiency in DNA damage response (DDR) pathway and accumulation of DNA damage increases mutation rates resulting in genomic instability and eventually increases the risk of cancer. The aim of our study was to investigate expressions of DNA repair genes as new prognostic biomarkers in acute myeloid leukemia (AML). METHODS: We utilized The Cancer Genome Atlas AML project (TCGA-LAML cohort, 15 acute promyelocytic leukemia (APL) and 155 non-APL AML) for the expression data of DNA repair genes. For validation, clinical samples (Ewha study group, 9 APL and 72 non-APL AML patients) were analyzed for the expression of 22 DNA repair genes using a custom RT2 Profiler PCR Array. RESULTS: APL patients presented significantly lower expression of DNA repair genes than non-APL AML patients in both study groups. Among non-APL AML patients, high expression levels of PARP1, XRCC1, and RAD51 were associated with poor overall survival (OS) probability in both study groups. Furthermore, Cox regression analysis showed that increased expression levels of PARP1, XRCC1, RAD51, BRCA1 and MRE11A could be independent risk factors for OS in the Ewha study group. Among non-APL patients of the Ewha study group, the OS probability of DDR-overexpressed group with at least one gene or more showing Z score greater than 1.5 was poorer than that of DDR non-overexpressed group. CONCLUSION: In the current study, the DNA repair gene expression profile of APL patients was different from that of non-APL AML patients. Overexpression of DNA repair genes could be a poor prognostic biomarker in non-APL AML.