Paternal Exposure to Immunosuppressive and/or Biologic Agents and Birth Outcomes in Patients With Immune-Mediated Inflammatory Diseases.

BACKGROUND & AIMS: We conducted a retrospective cohort study to inform the safety of exposure to immunosuppressive and/or biologic agents around conception in expectant fathers with immune-mediated inflammatory diseases (IMIDs) on birth outcomes. METHODS: Using a deidentified administrative claims database (OptumLabs Data Warehouse), we identified 7453 expectant fathers with IMIDs (inflammatory bowel diseases, rheumatoid arthritis, psoriasis/psoriatic arthritis, and ankylosing spondylitis) linked to newborns with periconception medication exposure between 38 and 60 weeks before the newborn birth date (34-58 weeks prior for preterm newborns) and neonatal follow-up for 3 months after the birth date. Through logistic regression adjusting for paternal age and race (and, in a subset, for maternal age, race, presence of IMIDs, and nonsingleton births), we compared the risk of major congenital malformations (primary outcome) and preterm birth and low birth weight in fathers exposed to thiopurines (n = 461), methotrexate (n = 171), tumor necrosis factor (TNF) α antagonists (n = 1082), or non-TNF-targeting biologic agents (n = 132) vs fathers not exposed to any of these medications (n = 5607). RESULTS: As compared to unexposed fathers (3.4% prevalence of major congenital malformations), exposure to thiopurines (relative risk [RR], 1.12; 95% confidence interval [CI], 0.66-1.76), methotrexate (RR, 0.67; 95% CI, 0.21-1.55), TNF-α antagonists (RR, 1.14; 95% CI, 0.81-1.57), and non-TNF-targeting biologic agents (RR, 1.75; 95% CI, 0.80-3.24) was not associated with increased risk of major congenital malformations. No association was observed between paternal medication exposure and risk of preterm birth or low birth weight. Results were stable on subanalyses of linked father-mother-newborn triads. CONCLUSIONS: In a large cohort study of 7453 expectant fathers with IMIDs, exposure to immunosuppressive or biologic agents around conception was not associated with increased risk of adverse birth outcomes.

People with HIV have a higher risk of COVID-19 diagnosis but similar outcomes to the general population.

BACKGROUND: We investigated the effect of HIV on COVID-19 outcomes with attention to selection bias due to differential testing and comorbidity burden. METHODS: This was a retrospective cohort analysis using four hierarchical outcomes: positive SARS-CoV-2 test, COVID-19 hospitalization, intensive care unit (ICU) admission and hospital mortality. The effect of HIV status was assessed using traditional covariate-adjusted, inverse probability-weighted (IPW) analysis based on covariate distributions for testing bias (testing IPWs), HIV infection status (HIV-IPWs) and combined models. Among people living with HIV (PWH), we evaluated whether CD4 count and HIV plasma viral load (pVL) discriminated between those who did and those who did not develop study outcomes using receiver operating characteristic analysis. RESULTS: Between March and November 2020, 63 319 people were receiving primary care services at the University of California San Diego (UCSD), of whom 4017 were PWH. The PWH had 2.1 times the odds of a positive SARS-CoV-2 test compared with those without HIV after weighting for potential testing bias, comorbidity burden and HIV-IPW [95% confidence interval (CI): 1.6-2.8]. Relative to people without HIV, PWH did not have an increased rate of COVID-19 hospitalization after controlling for comorbidities and testing bias [adjusted incidence rate ratio (aIRR) = 0.5, 95% CI: 0.1-1.4]. PWH did not have a different rate of ICU admission (aIRR = 1.08, 95% CI: 0.31-3.80) or of in-hospital death (aIRR = 0.92, 95% CI: 0.08-10.94) in any examined model. Neither CD4 count nor pVL predicted any of the hierarchical outcomes among PWH. CONCLUSIONS: People living with HIV have a higher risk of COVID-19 diagnosis than those without HIV but the outcomes are similar in both groups.

No Benefit of Continuing 5-Aminosalicylates in Patients with Crohn's Disease Treated with Anti-metabolite Therapy.

BACKGROUND AND AIMS: 5-aminosalicylates (5-ASA) are frequently used in the management of Crohn's disease (CD). We used a de-identified administrative claims database to compare patterns and outcomes of continuing versus stopping 5-ASA in patients with CD who escalated to anti-metabolite monotherapy. METHODS: Patients with CD on 5-ASA who were new users of anti-metabolite monotherapy and followed for at least 12 months from OptumLabs® Data Warehouse. Three patterns of 5-ASA use were identified: stopped 5-ASA, short-term 5-ASA (use for < 6 months after starting anti-metabolites), or persistent 5-ASA (use for > 6 months after starting anti-metabolites). Outcomes (need for corticosteroids, risk of CD-related hospitalization and/or surgery, treatment escalation to biologic therapy) were compared using Cox proportional hazard analysis adjusting for key covariates, with a 12-month immortal time period. RESULTS: Of 3036 patients with CD who were new-users of anti-metabolite monotherapy, 667 (21.9%), 626 (20.6%), and 1743 (57.4%) stopped 5-ASA, used 5-ASA transiently or persistently, respectively. Compared to patients who stopped 5-ASA after starting anti-metabolites, persistent 5-ASA use was associated with a higher risk of corticosteroid use (HR, 1.24 [1.08-1.42]), without an increase in risk of CD-related hospitalization (HR, 1.21 [0.98-1.49]), CD-related surgery (HR, 1.28 [0.90-1.80]) or treatment escalation (HR, 0.85 [0.62-1.20]). Sensitivity analyses using a 3-month window after initiation of anti-metabolites to classify patients as continuing vs. stopping 5-ASA showed similar results. Residual confounding by disease severity could not be excluded. CONCLUSION: 5-ASAs are frequently continued long-term even after escalation to anti-metabolite therapy in patients with CD but offer no clinical benefit over stopping 5-ASA.

Direct Oral Anticoagulants Versus Warfarin in the Treatment of Left Ventricular Thrombus.

BACKGROUND: Use of direct oral anticoagulants (DOACs) for the treatment of left ventricular (LV) thrombus has gained considerable interest. OBJECTIVE: We aimed to evaluate if DOACs are effective in the treatment of LV thrombus compared with warfarin. METHODS: We evaluated the medical records of patients diagnosed with a new LV thrombus at a tertiary medical center. The primary outcome was the composite of thrombus persistence, stroke, or systemic embolism. We adjusted for potential confounders using multiple logistic regression. The safety outcome was the composite of hemorrhagic stroke or bleeding requiring blood transfusion. RESULTS: A total of 129 patients were treated with warfarin and 22, with a DOAC. In unadjusted analysis, 54.3% of patients treated with warfarin met criteria for the efficacy outcome as compared with 40.9% of patients treated with a DOAC (P = 0.25). In adjusted analysis, no difference between groups was observed (odds ratio = 0.39; 95% CI = 0.14-1.06; P = 0.07 for DOAC vs warfarin). In all, 3.9% of patients treated with warfarin met safety criteria as compared with 4.5% of patients treated with a DOAC. A total of 8 patients in the warfarin group had a stroke or systemic embolism as compared with 0 patients in the DOAC group (P = 0.37). CONCLUSION AND RELEVANCE: Our data suggest that DOACs may be reasonable alternatives for treatment of LV thrombus. When added to the totality of available studies, this study demonstrates that the effectiveness of DOACs in LV thrombus remains uncertain. Randomized clinical trials are needed.

Epicardial FSTL1 reconstitution regenerates the adult mammalian heart.

The elucidation of factors that activate the regeneration of the adult mammalian heart is of major scientific and therapeutic importance. Here we found that epicardial cells contain a potent cardiogenic activity identified as follistatin-like 1 (Fstl1). Epicardial Fstl1 declines following myocardial infarction and is replaced by myocardial expression. Myocardial Fstl1 does not promote regeneration, either basally or upon transgenic overexpression. Application of the human Fstl1 protein (FSTL1) via an epicardial patch stimulates cell cycle entry and division of pre-existing cardiomyocytes, improving cardiac function and survival in mouse and swine models of myocardial infarction. The data suggest that the loss of epicardial FSTL1 is a maladaptive response to injury, and that its restoration would be an effective way to reverse myocardial death and remodelling following myocardial infarction in humans.

Latent Tuberculosis Screening Using Electronic Health Record Data.

Screening for latent tuberculosis infection is recommended for foreign-born persons in the United States. We used proxy data from electronic health records to determine that 17.5% of foreign-born outpatients attending the UC San Diego Health clinic (San Diego, CA, USA) underwent screening. Ending the global tuberculosis epidemic requires improved screening.

Prognostic impact of total and tyrosine phosphorylated GIV/Girdin in breast cancers.

Gα-interacting vesicle-associated protein (GIV, aka Girdin) is a guanine exchange factor (GEF) for the trimeric G protein Gαi and a bona fide metastasis-related gene that serves as a platform for amplification of tyrosine-based signals via G-protein intermediates. Here we present the first exploratory biomarker study conducted on a cohort of 187 patients with breast cancer to evaluate the prognostic role of total GIV (tGIV) and tyrosine phosphorylated GIV (pYGIV) across the various molecular subtypes. A Kaplan-Meier analysis of recurrence-free survival showed that the presence of tGIV, either cytoplasmic or nuclear, carried poor prognosis, but that nuclear tGIV had a greater prognostic impact (P = 0.007 in early and P = 0.0048 in late clinical stages). Activated pYGIV in the cytoplasm had the greatest prognostic impact in late clinical stages (P = 0.006). Furthermore, we found that the prognostic impacts of cytoplasmic pYGIV and nuclear tGIV were additive (hazard ratio 19.0548; P = 0.0002). Surprisingly, this additive effect of nuclear tGIV/cytoplasmic pYGIV was observed in human epidermal growth factor receptor 2-positive tumors (hazard ratio 16.918; P = 0.0005) but not in triple-negative breast cancers. In triple-negative breast cancers, tGIV and cytoplasmic pYGIV had no prognostic impact; however, membrane-association of pYGIV carried a poor prognosis (P = 0.026). Both tGIV and pYGIV showed no correlation with clinical stage, tumor size, pathologic type, lymph node involvement, and BRCA1/2 status. We conclude that immunocytochemical detection of pYGIV and tGIV can serve as an effective prognosticator. On the basis of the differential prognostic impact of tGIV/pYGIV within each molecular subtype, we propose a diagnostic algorithm. Further studies on larger cohorts are essential to rigorously assess the effectiveness and robustness of this algorithm in prognosticating outcome among patients with breast cancer.-Dunkel, Y., Diao, K., Aznar, N., Swanson, L., Liu, L., Zhu, W., Mi, X.-Y., Ghosh, P. Prognostic impact of total and tyrosine phosphorylated GIV/Girdin in breast cancers.

Probing nuclear pore complex architecture with proximity-dependent biotinylation.

Proximity-dependent biotin identification (BioID) is a method for identifying protein associations that occur in vivo. By fusing a promiscuous biotin ligase to a protein of interest expressed in living cells, BioID permits the labeling of proximate proteins during a defined labeling period. In this study we used BioID to study the human nuclear pore complex (NPC), one of the largest macromolecular assemblies in eukaryotes. Anchored within the nuclear envelope, NPCs mediate the nucleocytoplasmic trafficking of numerous cellular components. We applied BioID to constituents of the Nup107-160 complex and the Nup93 complex, two conserved NPC subcomplexes. A strikingly different set of NPC constituents was detected depending on the position of these BioID-fusion proteins within the NPC. By applying BioID to several constituents located throughout the extremely stable Nup107-160 subcomplex, we refined our understanding of this highly conserved subcomplex, in part by demonstrating a direct interaction of Nup43 with Nup85. Furthermore, by using the extremely stable Nup107-160 structure as a molecular ruler, we defined the practical labeling radius of BioID. These studies further our understanding of human NPC organization and demonstrate that BioID is a valuable tool for exploring the constituency and organization of large protein assemblies in living cells.

Integrating clinical research into electronic health record workflows to support a learning health system.

Objective: Integrating clinical research into routine clinical care workflows within electronic health record systems (EHRs) can be challenging, expensive, and labor-intensive. This case study presents a large-scale clinical research project conducted entirely within a commercial EHR during the COVID-19 pandemic. Case Report: The UCSD and UCSDH COVID-19 NeutraliZing Antibody Project (ZAP) aimed to evaluate antibody levels to SARS-CoV-2 virus in a large population at an academic medical center and examine the association between antibody levels and subsequent infection diagnosis. Results: The project rapidly and successfully enrolled and consented over 2000 participants, integrating the research trial with standing COVID-19 testing operations, staff, lab, and mobile applications. EHR-integration increased enrollment, ease of scheduling, survey distribution, and return of research results at a low cost by utilizing existing resources. Conclusion: The case study highlights the potential benefits of EHR-integrated clinical research, expanding their reach across multiple health systems and facilitating rapid learning during a global health crisis.

Identification of intercellular cell adhesion molecule 1 (ICAM-1) as a hypoglycosylation marker in congenital disorders of glycosylation cells.

Many human inherited disorders cause protein N-glycosylation defects, but there are few cellular markers to test gene complementation for such defects. Plasma membrane glycoproteins are potential biomarkers because they may be reduced or even absent in plasma membranes of glycosylation-deficient cells. We combined stable isotope labeling by amino acids in cell culture (SILAC) with linear ion trap mass spectrometry (LTQ Orbitrap(TM)) to identify and quantify membrane proteins from wild-type CHO and glycosylation-deficient CHO (Lec9) cells. We identified 165 underrepresented proteins from 1447 unique quantified proteins, including 18 N-glycosylated plasma membrane proteins. Using various methods, we found that intercellular cell adhesion molecule 1 (ICAM-1) was reduced in Lec9 cells and in fibroblasts from 31 congenital disorder of glycosylation (CDG) patients compared with normal controls. Mannose supplementation of phosphomannose isomerase-deficient CDG-Ib (MPI-CDG) cells and complementation with PMM2 in PMM2-deficient CDG-Ia (PMM2-CDG) cells partially corrected hypoglycosylation based on increased ICAM-1 presence on the plasma membrane. These data indicate that ICAM-1 could be a useful hypoglycosylation biomarker to assess gene complementation of CDG-I patient cells and to monitor improved glycosylation in response to therapeutic drugs.

Lipoprotein(a) Testing Trends in a Large Academic Health System in the United States.

Background Despite its high prevalence and clinical significance, clinical measurement of lipoprotein(a) is rare but has not been systematically quantified. We assessed the prevalence of lipoprotein(a) testing overall, in those with various cardiovascular disease (CVD) conditions and in those undergoing cardiac testing across 6 academic medical centers associated with the University of California, in total and by year from 2012 to 2021. Methods and Results In this observational study, data from the University of California Health Data Warehouse on the number of individuals with unique lipoprotein(a) testing, unique CVD diagnoses (using International Classification of Diseases, Tenth Revision [ICD-10], codes), and other unique cardiac testing were collected. The proportion of total individuals, the proportion of individuals with a given CVD diagnosis, and the proportion of individuals with a given cardiac test and lipoprotein(a) testing any time during the study period were calculated. From 2012 to 2021, there were 5 553 654 unique adults evaluated in the University of California health system, of whom 18 972 (0.3%) had lipoprotein(a) testing. In general, those with lipoprotein(a) testing were more likely to be older, men, and White race, with a greater burden of CVD. Lipoprotein(a) testing was performed in 6469 individuals with ischemic heart disease (2.9%), 836 with aortic stenosis (3.1%), 4623 with family history of CVD (3.3%), 1202 with stroke (1.7%), and 612 with coronary artery calcification (6.1%). For most conditions, the prevalence of testing in the same year as the diagnosis of CVD was relatively stable, with a small upward trend over time. Lipoprotein(a) testing was performed in 10 753 individuals (1.8%) who had lipid panels, with higher rates with more specialized testing, including coronary computed tomography angiography (6.8%) and apolipoprotein B (63.0%). Conclusions Lipoprotein(a) testing persists at low rates, even among those with diagnosed CVD, and remained relatively stable over the study period.

Proteins and an inflammatory network expressed in colon tumors.

The adenomatous polyposis coli (APC) protein is crucial to homeostasis of normal intestinal epithelia because it suppresses the ß-catenin/TCF pathway. Consequently, loss or mutation of the APC gene causes colorectal tumors in humans and mice. Here, we describe our use of multidimensional protein identification technology (MudPIT) to compare protein expression in colon tumors to that of adjacent healthy colon tissue from Apc(Min/+) mice. Twenty-seven proteins were found to be up-regulated in colon tumors and 25 were down-regulated. As an extension of the proteomic analysis, the differentially expressed proteins were used as "seeds" to search for coexpressed genes. This approach revealed a coexpression network of 45 genes that is up-regulated in colon tumors. Members of the network include the antibacterial peptide cathelicidin (CAMP), Toll-like receptors (TLRs), IL-8, and triggering receptor expressed on myeloid cells 1 (TREM1). The coexpression network is associated with innate immunity and inflammation, and there is significant concordance between its connectivity in humans versus mice (Friedman: p value = 0.0056). This study provides new insights into the proteins and networks that are likely to drive the onset and progression of colon cancer.

The Wnt/planar cell polarity protein-tyrosine kinase-7 (PTK7) is a highly efficient proteolytic target of membrane type-1 matrix metalloproteinase: implications in cancer and embryogenesis.

PTK7 is an essential component of the Wnt/planar cell polarity (PCP) pathway. We provide evidence that the Wnt/PCP pathway converges with pericellular proteolysis in both normal development and cancer. Here, we demonstrate that membrane type-1 matrix metalloproteinase (MT1-MMP), a key proinvasive proteinase, functions as a principal sheddase of PTK7. MT1-MMP directly cleaves the exposed PKP(621)↓LI sequence of the seventh Ig-like domain of the full-length membrane PTK7 and generates, as a result, an N-terminal, soluble PTK7 fragment (sPTK7). The enforced expression of membrane PTK7 in cancer cells leads to the actin cytoskeleton reorganization and the inhibition of cell invasion. MT1-MMP silencing and the analysis of the uncleavable L622D PTK7 mutant confirm the significance of MT1-MMP proteolysis of PTK7 in cell functions. Our data also demonstrate that a fine balance between the metalloproteinase activity and PTK7 levels is required for normal development of zebrafish (Danio rerio). Aberration of this balance by the proteinase inhibition or PTK7 silencing results in the PCP-dependent convergent extension defects in the zebrafish. Overall, our data suggest that the MT1-MMP-PTK7 axis plays an important role in both cancer cell invasion and normal embryogenesis in vertebrates. Further insight into these novel mechanisms may promote understanding of directional cell motility and lead to the identification of therapeutics to treat PCP-related developmental disorders and malignancy.

Internal cleavages of the autoinhibitory prodomain are required for membrane type 1 matrix metalloproteinase activation, although furin cleavage alone generates inactive proteinase.

The functional activity of invasion-promoting membrane type 1 matrix metalloproteinase (MT1-MMP) is elevated in cancer. This elevated activity promotes cancer cell migration, invasion, and metastasis. MT1-MMP is synthesized as a zymogen, the latency of which is maintained by its prodomain. Excision by furin was considered sufficient for the prodomain release and MT1-MMP activation. We determined, however, that the full-length intact prodomain released by furin alone is a potent autoinhibitor of MT1-MMP. Additional MMP cleavages within the prodomain sequence are required to release the MT1-MMP enzyme activity. Using mutagenesis of the prodomain sequence and mass spectrometry analysis of the prodomain fragments, we demonstrated that the intradomain cleavage of the PGD/L(50) site initiates the MT1-MMP activation, whereas the (108)RRKR(111)/Y(112) cleavage by furin completes the removal and the degradation of the autoinhibitory prodomain and the liberation of the functional activity of the emerging enzyme of MT1-MMP.

Development of a COVID-19 Application Ontology for the ACT Network.

Clinical data networks that leverage large volumes of data in electronic health records (EHRs) are significant resources for research on coronavirus disease 2019 (COVID-19). Data harmonization is a key challenge in seamless use of multisite EHRs for COVID-19 research. We developed a COVID-19 application ontology in the national Accrual to Clinical Trials (ACT) network that enables harmonization of data elements that that are critical to COVID-19 research. The ontology contains over 50,000 concepts in the domains of diagnosis, procedures, medications, and laboratory tests. In particular, it has computational phenotypes to characterize the course of illness and outcomes, derived terms, and harmonized value sets for SARS-CoV-2 laboratory tests. The ontology was deployed and validated on the ACT COVID-19 network that consists of nine academic health centers with data on 14.5M patients. This ontology, which is freely available to the entire research community on GitHub at https://github.com/shyamvis/ACT-COVID-Ontology, will be useful for harmonizing EHRs for COVID-19 research beyond the ACT network.

Development of a Coronavirus Disease 2019 (COVID-19) Application Ontology for the Accrual to Clinical Trials (ACT) network.

Clinical data networks that leverage large volumes of data in electronic health records (EHRs) are significant resources for research on coronavirus disease 2019 (COVID-19). Data harmonization is a key challenge in seamless use of multisite EHRs for COVID-19 research. We developed a COVID-19 application ontology in the national Accrual to Clinical Trials (ACT) network that enables harmonization of data elements that are critical to COVID-19 research. The ontology contains over 50 000 concepts in the domains of diagnosis, procedures, medications, and laboratory tests. In particular, it has computational phenotypes to characterize the course of illness and outcomes, derived terms, and harmonized value sets for severe acute respiratory syndrome coronavirus 2 laboratory tests. The ontology was deployed and validated on the ACT COVID-19 network that consists of 9 academic health centers with data on 14.5M patients. This ontology, which is freely available to the entire research community on GitHub at https://github.com/shyamvis/ACT-COVID-Ontology, will be useful for harmonizing EHRs for COVID-19 research beyond the ACT network.

Mass spectrometry-based label-free quantitative proteomics.

In order to study the differential protein expression in complex biological samples, strategies for rapid, highly reproducible and accurate quantification are necessary. Isotope labeling and fluorescent labeling techniques have been widely used in quantitative proteomics research. However, researchers are increasingly turning to label-free shotgun proteomics techniques for faster, cleaner, and simpler results. Mass spectrometry-based label-free quantitative proteomics falls into two general categories. In the first are the measurements of changes in chromatographic ion intensity such as peptide peak areas or peak heights. The second is based on the spectral counting of identified proteins. In this paper, we will discuss the technologies of these label-free quantitative methods, statistics, available computational software, and their applications in complex proteomics studies.

Biochemical evidence of the interactions of membrane type-1 matrix metalloproteinase (MT1-MMP) with adenine nucleotide translocator (ANT): potential implications linking proteolysis with energy metabolism in cancer cells.

Invasion-promoting MT1-MMP (membrane type-1 matrix metalloproteinase) is a key element in cell migration processes. To identify the proteins that interact and therefore co-precipitate with this proteinase from cancer cells, we used the proteolytically active WT (wild-type), the catalytically inert E240A and the C-end truncated (tailless; DeltaCT) MT1-MMP-FLAG constructs as baits. The identity of the pulled-down proteins was determined by LC-MS/MS (liquid chromatography tandem MS) and then confirmed by Western blotting using specific antibodies. We determined that, in breast carcinoma MCF cells (MCF-7 cells), ANT (adenine nucleotide translocator) efficiently interacted with the WT, E240A and DeltaCT constructs. The WT and E240A constructs also interacted with alpha-tubulin, an essential component of clathrin-mediated endocytosis. In turn, tubulin did not co-precipitate with the DeltaCT construct because of the inefficient endocytosis of the latter, thus suggesting a high level of selectivity of our test system. To corroborate these results, we then successfully used the ANT2-FLAG construct as a bait to pull-down MT1-MMP, which was naturally produced by fibrosarcoma HT1080 cells. We determined that the presence of the functionally inert catalytic domain alone was sufficient to cause the proteinase to interact with ANT2, thus indicating that there is a non-proteolytic mode of these interactions. Overall, it is tempting to hypothesize that by interacting with pro-invasive MT1-MMP, ANT plays a yet to be identified role in a coupling mechanism between energy metabolism and pericellular proteolysis in migrating cancer cells.

No benefit of continuing vs stopping 5-aminosalicylates in patients with ulcerative colitis escalated to anti-metabolite therapy.

BACKGROUND: Whilst continuation of 5-aminosalicylates (5-ASA) after escalation to biologic therapy is considered ineffective in patients with ulcerative colitis (UC), their role in patients escalated to anti-metabolites is unclear. AIM: To compared patterns and outcomes of continuing vs stopping 5-ASA in patients with UC who escalated to anti-metabolite monotherapy, using a de-identified administrative claims database. METHODS: We identified patients with UC who were new users of anti-metabolite monotherapy who were receiving 5-ASA, and were followed for at least 12 months after starting anti-metabolite therapy. We evaluated patterns of 5-ASA use (stopped 5-ASA, short-term 5-ASA use for <6 months after starting anti-metabolites, persistent 5-ASA use for >6 months after starting anti-metabolites). We compared outcomes (risk of UC-related hospitalisation and/or surgery, need for corticosteroids, treatment escalation to biologic therapy) by pattern of 5-ASA use, using Cox proportional hazard analysis adjusting for age, sex, race, comorbidity burden, and hospitalisation or emergency department visit, abdominal surgery and corticosteroid use in the previous 12 months (as measures of disease severity), with a 12-month immortal time period. RESULTS: Of 4068 patients with UC who were new-users of anti-metabolite monotherapy, 578 (14.2%), 782 (19.2%) and 2708 (66.6%) stopped 5-ASA, used 5-ASA transiently or persistently, respectively. Compared to patients who stopped 5-ASA after starting anti-metabolites, persistent 5-ASA use was associated with a higher risk of UC-related hospitalisation (HR, 1.40 [1.07-1.83]) and corticosteroid use (HR, 1.48 [1.28-1.70]), without an increase in risk of UC-related surgery (HR, 1.32 [0.86-2.00]) or treatment escalation (HR, 0.80 [0.53-1.20]). Sensitivity analyses using a 3 months window after initiation of anti-metabolites to classify patients as continuing vs stopping 5-ASA showed similar results. Residual confounding by disease severity could not be excluded. CONCLUSION: 5-ASA are usually continued long-term even after escalating to anti-metabolite therapy in patients with UC without clinical benefit.

Improving risk prediction in heart failure using machine learning.

BACKGROUND: Predicting mortality is important in patients with heart failure (HF). However, current strategies for predicting risk are only modestly successful, likely because they are derived from statistical analysis methods that fail to capture prognostic information in large data sets containing multi-dimensional interactions. METHODS AND RESULTS: We used a machine learning algorithm to capture correlations between patient characteristics and mortality. A model was built by training a boosted decision tree algorithm to relate a subset of the patient data with a very high or very low mortality risk in a cohort of 5822 hospitalized and ambulatory patients with HF. From this model we derived a risk score that accurately discriminated between low and high-risk of death by identifying eight variables (diastolic blood pressure, creatinine, blood urea nitrogen, haemoglobin, white blood cell count, platelets, albumin, and red blood cell distribution width). This risk score had an area under the curve (AUC) of 0.88 and was predictive across the full spectrum of risk. External validation in two separate HF populations gave AUCs of 0.84 and 0.81, which were superior to those obtained with two available risk scores in these same populations. CONCLUSIONS: Using machine learning and readily available variables, we generated and validated a mortality risk score in patients with HF that was more accurate than other risk scores to which it was compared. These results support the use of this machine learning approach for the evaluation of patients with HF and in other settings where predicting risk has been challenging.