Randomised Controlled Trial of a Customised Text Messaging and Activity Monitor Program for Lifestyle Improvement after Gestational Diabetes.

Gestational diabetes (GDM) is associated with a long-term risk of diabetes. We aimed to determine whether a text-messaging-based lifestyle support program would improve diabetes risk factors following GDM. Women with GDM were randomised following delivery to receive four text messages per week supporting a healthy lifestyle and parenting for 6 months, with feedback from an activity monitor (intervention), or to receive the activity monitor only (control). The primary outcome was a composite of weight, physical activity and dietary goals. There were 177 women randomised, with 88 intervention and 89 control participants. All the participants experienced COVID-19 lockdowns during the study. Six-month primary outcome data were obtained for 57 intervention participants and 56 controls. There were 7/57 (12%) intervention and 6/56 (11%) control participants who met the primary outcome (relative risk, 1.08; 95%CI, 0.63-1.85; p = 0.79). Two intervention participants met the dietary goals compared to none of the control participants (p = NS). The intervention participants were more likely to record >1000 steps/day (on 102 ± 59 vs. 81 ± 59 days, p = 0.03). When analysed monthly, this was not initially different but became significant 3-6 months post-partum. Interviews and surveys indicated that with the Intervention, healthier choices were made, but these were negatively impacted by COVID-19 restrictions. Participants found the messages motivational (74%) and the activity monitor useful (71%). In conclusion, no improvement in the diabetes risk factors occurred among the women receiving the text messaging intervention when affected by COVID-19 restrictions.

Hematopoietic stem cell quiescence and DNA replication dynamics maintained by the resilient ß-catenin/Hoxa9/Prmt1 axis.

ABSTRACT: Maintenance of quiescence and DNA replication dynamics are 2 paradoxical requirements for the distinct states of dormant and active hematopoietic stem cells (HSCs), which are required to preserve the stem cell reservoir and replenish the blood cell system in response to hematopoietic stress, respectively. Here, we show that key self-renewal factors, ß-catenin or Hoxa9, largely dispensable for HSC integrity, in fact, have dual functions in maintaining quiescence and enabling efficient DNA replication fork dynamics to preserve the functionality of hematopoietic stem and progenitor cells (HSPCs). Although ß-catenin or Hoxa9 single knockout (KO) exhibited mostly normal hematopoiesis, their coinactivation led to severe hematopoietic defects stemmed from aberrant cell cycle, DNA replication, and damage in HSPCs. Mechanistically, ß-catenin and Hoxa9 function in a compensatory manner to sustain key transcriptional programs that converge on the pivotal downstream target and epigenetic modifying enzyme, Prmt1, which protects the quiescent state and ensures an adequate supply of DNA replication and repair factors to maintain robust replication fork dynamics. Inactivation of Prmt1 phenocopied both cellular and molecular phenotypes of ß-catenin/Hoxa9 combined KO, which at the same time could also be partially rescued by Prmt1 expression. The discovery of the highly resilient ß-catenin/Hoxa9/Prmt1 axis in protecting both quiescence and DNA replication dynamics essential for HSCs at different key states provides not only novel mechanistic insights into their intricate regulation but also a potential tractable target for therapeutic intervention.

The relationship between body mass index and sleep in women with risk factors for gestational diabetes mellitus.

Background: Both obesity and sleep disorders are common among women during pregnancy. Although prior research has identified a relationship between obesity and sleep disorders, those findings are from women later in pregnancy. Objective: To explore the relationships between self-reported sleep duration, insufficient sleep and snoring with body mass index (BMI) among multiethnic women at risk of gestational diabetes mellitus (GDM)in early pregnancy. Methods: Cross-sectional study of baseline data from women at risk of GDM enrolled in the Treatment of BOoking Gestational diabetes Mellitus (TOBOGM) multicentre trial across 12 Australian/Austrian sites. Participants completed a questionnaire before 20 weeks' gestation to evaluate sleep. BMI <25 kg/m2 served as the reference group in multivariable logistic regression. Results: Among the 2865 women included, the prevalence of overweight and obesity classes I-III was 28%, 19%, 11% and 12%, respectively. There was no relationship between sleep duration and BMI. The risk of insufficient sleep >5 days/month was higher in class II and class III obesity (1.38 (1.03-1.85) and 1.34 (1.01-1.80), respectively), and the risk of snoring increased as BMI increased (1.59 (1.25-2.02), 2.68 (2.07-3.48), 4.35 (3.21-5.88) to 4.96 (3.65-6.74), respectively)). Conclusions: Obesity is associated with insufficient sleep among pregnant women at risk of GDM. Snoring is more prevalent with increasing BMI.

Management of Poststroke Hyperglycemia: Results of the TEXAIS Randomized Clinical Trial.

BACKGROUND: Hyperglycemia in acute ischemic stroke reduces the efficacy of stroke thrombolysis and thrombectomy, with worse clinical outcomes. Insulin-based therapies are difficult to implement and may cause hypoglycemia. We investigated whether exenatide, a GLP-1 (glucagon-like peptide-1) receptor agonist, would improve stroke outcomes, and control poststroke hyperglycemia with minimal hypoglycemia. METHODS: The TEXAIS trial (Treatment With Exenatide in Acute Ischemic Stroke) was an international, multicenter, phase 2 prospective randomized clinical trial (PROBE [Prospective Randomized Open Blinded End-Point] design) enrolling adult patients with acute ischemic stroke ≤9 hours of stroke onset to receive exenatide (5 µg BID subcutaneous injection) or standard care for 5 days, or until hospital discharge (whichever sooner). The primary outcome (intention to treat) was the proportion of patients with ≥8-point improvement in National Institutes of Health Stroke Scale score (or National Institutes of Health Stroke Scale scores 0-1) at 7 days poststroke. Safety outcomes included death, episodes of hyperglycemia, hypoglycemia, and adverse event. RESULTS: From April 2016 to June 2021, 350 patients were randomized (exenatide, n=177, standard care, n=173). Median age, 71 years (interquartile range, 62-79), median National Institutes of Health Stroke Scale score, 4 (interquartile range, 2-8). Planned recruitment (n=528) was stopped early due to COVID-19 disruptions and funding constraints. The primary outcome was achieved in 97 of 171 (56.7%) in the standard care group versus 104 of 170 (61.2%) in the exenatide group (adjusted odds ratio, 1.22 [95% CI, 0.79-1.88]; P=0.38). No differences in secondary outcomes were observed. The per-patient mean daily frequency of hyperglycemia was significantly less in the exenatide group across all quartiles. No episodes of hypoglycemia were recorded over the treatment period. Adverse events of mild nausea and vomiting occurred in 6 (3.5%) exenatide patients versus 0 (0%) standard care with no withdrawal. CONCLUSIONS: Treatment with exenatide did not reduce neurological impairment at 7 days in patients with acute ischemic stroke. Exenatide did significantly reduce the frequency of hyperglycemic events, without hypoglycemia, and was safe to use. Larger acute stroke trials using GLP-1 agonists such as exenatide should be considered. REGISTRATION: URL: www.australianclinicaltrials.gov.au; Unique identifier: ACTRN12617000409370. URL: https://www.clinicaltrials.gov; Unique identifier: NCT03287076.

Diabetic Retinopathy Screening at the Point of Care (DR SPOC): detecting undiagnosed and vision-threatening retinopathy by integrating portable technologies within existing services.

INTRODUCTION: The aim of this study was to determine the prevalence of diabetic retinopathy (DR) in a low socioeconomic region of a high-income country, as well as determine the diagnostic utility of point-of-care screening for high-risk populations in tertiary care settings. RESEARCH DESIGN AND METHODS: This was a cross-sectional study of patients with diabetes attending foot ulcer or integrated care diabetes clinics at two Western Sydney hospitals (n=273). DR was assessed using portable, two-field, non-mydriatic fundus photography and combined electroretinogram/ pupillometry (ERG). With mydriatic photographs used as the reference standard, sensitivity and specificity of the devices were determined. Prevalence of DR and vision-threatening diabetic retinopathy (VTDR) were reported, with multivariate logistic regression used to identify predictors of DR. RESULTS: Among 273 patients, 39.6% had any DR, while 15.8% had VTDR, of whom 59.3% and 62.8% were previously undiagnosed, respectively. Non-mydriatic photography demonstrated 20.2% sensitivity and 99.5% specificity for any DR, with a 56.7% screening failure rate. Meanwhile, mydriatic photography produced high-quality images with a 7.6% failure rate. ERG demonstrated 72.5% sensitivity and 70.1% specificity, with a 15.0% failure rate. The RETeval ERG was noted to have an optimal DR cut-off score at 22. Multivariate logistic regression identified an eGFR of ≤29 mL/min/1.73 m2, HbA1c of ≥7.0%, pupil size of <4 mm diameter, diabetes duration of 5-24 years and RETeval score of ≥22 as strong predictors of DR. CONCLUSION: There is a high prevalence of vision-threatening and undiagnosed DR among patients attending high-risk tertiary clinics in Western Sydney. Point-of-care DR screening using portable, mydriatic photography demonstrates potential as a model of care which is easily accessible, targeted for high-risk populations and substantially enhances DR detection.

Patient experiences of continuous glucose monitoring and sensor-augmented insulin pump therapy for diabetes: A systematic review of qualitative studies.

AIMS: Blood glucose control is central to the management of diabetes, and continuous glucose monitoring (CGM) improves glycemic control. We aimed to describe the perspectives of people with diabetes using CGM. MATERIALS AND METHODS: We performed a systematic review of qualitative studies. RESULTS: Fifty-four studies involving 1845 participants were included. Six themes were identified: gaining control and convenience (reducing pain and time, safeguarding against complications, achieving stricter glucose levels, and sharing responsibility with family); motivating self-management (fostering ownership, and increasing awareness of glycemic control); providing reassurance and freedom (attaining peace of mind, and restoring social participation); developing confidence (encouraged by the endorsement of others, gaining operational skills, customizing settings for ease of use, and trust in the device); burdened with device complexities (bewildered by unfamiliar technology, reluctant to rely on algorithms, overwhelmed by data, frustrated with malfunctioning and inaccuracy, distressed by alerts, and bulkiness of machines interfering with lifestyle); and excluded by barriers to access (constrained by cost, lack of suppliers). CONCLUSIONS: CGM can improve self-management and confidence in patients managing diabetes. However, the technical issues, uncertainty in readings, and cost may limit the uptake. Education and training from the health professionals may help to reduce the practical and psychological burden for better patient outcomes.

Usage of electronic health record (eHR) viewer among healthcare professionals (HCProfs): A territory-wide study of 3972 participants.

BACKGROUND: The Electronic Health Record Sharing System (eHRSS) is an electronic platform for two-way communication between the public and private sectors in Hong Kong. The authorised healthcare professionals (HCProfs) could access and upload patients' health records on the eHR Viewer in the eHRSS. This study aims to evaluate the usage of eHR viewer among the HCProfs from the private sector by 1) examining the correlation of various factors and the data access of eHR viewer; 2) investigating the trend on data access and upload to eHR viewer by time period and domain. METHODS: A total of 3972 HCProfs from private hospitals, group practice, and solo practice were included in the study. Regression analysis was performed to identify the correlation between various factors and the data access to eHR viewer. Trends on accessing and data uploading to eHR viewer by time period and domains were evaluated. Trends on data uploading to eHR viewer by time period and domains were presented in the line chart as well. FINDINGS: All types of HCProfs had a higher likelihood of accessing the eHR viewer as compared to those from private hospitals. HCProfs with specialities (apart from anaesthesia) had a higher possibility of accessing the eHR viewer than general practitioners without specialities. HCProfs participating in the Public-Private Partnership (PPP) Programme and the eHealth System (Subsidies) (eHS(S)) were more likely to access the eHR viewer. The overall trend of accessing eHR viewer was rising notably from 2016 to 2022, every domain showed rising trends, especially the laboratory domain with a 5-times increase between 2016 and 2022. CONCLUSION: HCProfs with speciality were more likely to access the eHR viewer (except anaesthesiology), compared with the general practitioners. Participation in the PPP programmes and eHS(S) also increased the access rate of the eHR viewer. Besides, the use of eHR viewer (accessing and data uploading) would be influenced by social policy and the epidemic. Future research should focus on the impact of government programmes on eHRSS adoption.

Effect of Mobile Phone Text Messaging Self-Management Support for Patients With Diabetes or Coronary Heart Disease in a Chronic Disease Management Program (SupportMe) on Blood Pressure: Pragmatic Randomized Controlled Trial.

BACKGROUND: Maintaining engagement and support for patients with chronic diseases is challenging. SMS text messaging programs have complemented patient care in a variety of situations. However, such programs have not been widely translated into routine care. OBJECTIVE: We aimed to examine the implementation and utility of a customized SMS text message-based support program for patients with type 2 diabetes (T2D), coronary heart disease, or both within a chronic disease integrated care program. METHODS: We conducted a 6-month pragmatic parallel-group, single-blind randomized controlled trial that recruited people with T2D or coronary heart disease. Intervention participants received 4 semipersonalized SMS text messages per week providing self-management support to supplement standard care. Preprogrammed algorithms customized content based on participant characteristics, and the messages were sent at random times of the day and in random order by a fully automated SMS text messaging engine. Control participants received standard care and only administrative SMS text messages. The primary outcome was systolic blood pressure. Evaluations were conducted face to face whenever possible by researchers blinded to randomization. Participants with T2D were evaluated for glycated hemoglobin level. Participant-reported experience measures were evaluated using questionnaires and focus groups and summarized using proportions and thematic analysis. RESULTS: A total of 902 participants were randomized (n=448, 49.7% to the intervention group and n=454, 50.3% to the control group). Primary outcome data were available for 89.5% (807/902) of the participants. At 6 months, there was no difference in systolic blood pressure between the intervention and control arms (adjusted mean difference=0.9 mm Hg, 95% CI -1.1 to 2.1; P=.38). Of 642 participants with T2D, there was no difference in glycated hemoglobin (adjusted mean difference=0.1%, 95% CI -0.1% to 0.3%; P=.35). Self-reported medication adherence was better in the intervention group (relative risk=0.82, 95% CI 0.68-1.00; P=.045). Participants reported that the SMS text messages were useful (298/344, 86.6%) and easily understood (336/344, 97.7%) and motivated change (217/344, 63.1%). The lack of bidirectional messaging was identified as a barrier. CONCLUSIONS: The intervention did not improve blood pressure in this cohort, possibly because of high clinician commitment to improved routine patient care as part of the chronic disease management program as well as favorable baseline metrics. There was high program engagement, acceptability, and perceived value. Feasibility as part of an integrated care program was demonstrated. SMS text messaging programs may supplement chronic disease management and support self-care. TRIAL REGISTRATION: Australian New Zealand Clinical Trials Registry ACTRN12616001689460; https://anzctr.org.au/Trial/Registration/TrialReview.aspx?id=371769&isReview=true. INTERNATIONAL REGISTERED REPORT IDENTIFIER (IRRID): RR2-10.1136/bmjopen-2018-025923.

FS-BAN: Born-Again Networks for Domain Generalization Few-Shot Classification.

Conventional Few-shot classification (FSC) aims to recognize samples from novel classes given limited labeled data. Recently, domain generalization FSC (DG-FSC) has been proposed with the goal to recognize novel class samples from unseen domains. DG-FSC poses considerable challenges to many models due to the domain shift between base classes (used in training) and novel classes (encountered in evaluation). In this work, we make two novel contributions to tackle DG-FSC. Our first contribution is to propose Born-Again Network (BAN) episodic training and comprehensively investigate its effectiveness for DG-FSC. As a specific form of knowledge distillation, BAN has been shown to achieve improved generalization in conventional supervised classification with a closed-set setup. This improved generalization motivates us to study BAN for DG-FSC, and we show that BAN is promising to address the domain shift encountered in DG-FSC. Building on the encouraging findings, our second (major) contribution is to propose Few-Shot BAN (FS-BAN), a novel BAN approach for DG-FSC. Our proposed FS-BAN includes novel multi-task learning objectives: Mutual Regularization, Mismatched Teacher, and Meta-Control Temperature, each of these is specifically designed to overcome central and unique challenges in DG-FSC, namely overfitting and domain discrepancy. We analyze different design choices of these techniques. We conduct comprehensive quantitative and qualitative analysis and evaluation over six datasets and three baseline models. The results suggest that our proposed FS-BAN consistently improves the generalization performance of baseline models and achieves state-of-the-art accuracy for DG-FSC. Project Page: yunqing-me.github.io/Born-Again-FS/.

Associations between adoption of eHealth management module and optimal control of HbA1c in diabetes patients.

In January 2021, the eHealth App was launched in Hong Kong by the Hong Kong government to support the Electronic Health Record Sharing System (eHRSS). A Health Management Module in the eHealth App introduced new functions to record blood pressure, blood sugar, and heart rate, and downloading and sharing records. This study aims to compare the level of glycaemic control between users of the eHealth App and non-users. Type 2 diabetes patients who have joined the eHRSS with existing haemoglobin A1c (HbA1c) level records are recruited. Correlations between predictors and optimal HbA1c control (<7%) are examined using logistic regression analyses. A total of 109,823 participants are included, with 76,356 non-users of eHealth App, 31,723 users of eHealth App, and 1744 users of the eHealth Management Module together with the App. We collect HbA1c values from Jan 2021 to May 2022, and they are 6 months after the use of the App on average. Users of the eHealth Management Module are found to have more optimal HbA1c levels across all subgroups, with the strongest effect observed in younger females (aOR = 1.66, 95% CI = 1.27-2.17). eHealth App usage is also positively associated with optimal HbA1c levels, particularly amongst younger females (aOR = 1.17, 95% CI = 1.08-1.26). Overall, users of eHealth App and eHealth Management Module demonstrate more optimal HbA1c levels when compared with non-users, particularly among younger adults and females. These findings support its potential adoption in diabetes patients. Future studies should examine the impact of eHealth interventions on other clinical targets and diabetes complications.

Translation of nurse-initiated protocols to manage fever, hyperglycaemia and swallowing following stroke across Europe (QASC Europe): A pre-test/post-test implementation study.

Introduction: Poor adoption of stroke guidelines is a problem internationally. The Quality in Acute Stroke Care (QASC) trial demonstrated significant reduction in death and disability with facilitated implementation of nurse-initiated. Methods: This was a multi-country, multi-centre, pre-test/post-test study (2017-2021) comparing post implementation data with historically collected pre-implementation data. Hospital clinical champions, supported by the Angels Initiative conducted multidisciplinary workshops discussing pre-implementation medical record audit results, barriers and facilitators to FeSS Protocol implementation, developed action plans and provided education, with ongoing support co-ordinated remotely from Australia. Prospective audits were conducted 3-month after FeSS Protocol introduction. Pre-to-post analysis and country income classification comparisons were adjusted for clustering by hospital and country controlling for age/sex/stroke severity. Results: Data from 64 hospitals in 17 countries (3464 patients pre-implementation and 3257 patients post-implementation) showed improvement pre-to-post implementation in measurement recording of all three FeSS components, all p < 0.0001: fever elements (pre: 17%, post: 51%; absolute difference 33%, 95% CI 30%, 37%); hyperglycaemia elements (pre: 18%, post: 52%; absolute difference 34%; 95% CI 31%, 36%); swallowing elements (pre: 39%, post: 67%; absolute difference 29%, 95% CI 26%, 31%) and thus in overall FeSS Protocol adherence (pre: 3.4%, post: 35%; absolute difference 33%, 95% CI 24%, 42%). In exploratory analysis of FeSS adherence by countries' economic status, high-income versus middle-income countries improved to a comparable extent. Discussion and conclusion: Our collaboration resulted in successful rapid implementation and scale-up of FeSS Protocols into countries with vastly different healthcare systems.

Multimodal Mutual Information Maximization: A Novel Approach for Unsupervised Deep Cross-Modal Hashing.

In this article, we adopt the maximizing mutual information (MI) approach to tackle the problem of unsupervised learning of binary hash codes for efficient cross-modal retrieval. We proposed a novel method, dubbed cross-modal info-max hashing (CMIMH). First, to learn informative representations that can preserve both intramodal and intermodal similarities, we leverage the recent advances in estimating variational lower bound of MI to maximizing the MI between the binary representations and input features and between binary representations of different modalities. By jointly maximizing these MIs under the assumption that the binary representations are modeled by multivariate Bernoulli distributions, we can learn binary representations, which can preserve both intramodal and intermodal similarities, effectively in a mini-batch manner with gradient descent. Furthermore, we find out that trying to minimize the modality gap by learning similar binary representations for the same instance from different modalities could result in less informative representations. Hence, balancing between reducing the modality gap and losing modality-private information is important for the cross-modal retrieval tasks. Quantitative evaluations on standard benchmark datasets demonstrate that the proposed method consistently outperforms other state-of-the-art cross-modal retrieval methods.

Factors Associated With the Acceptance of an eHealth App for Electronic Health Record Sharing System: Population-Based Study.

BACKGROUND: In the second stage of the Electronic Health Record Sharing System (eHRSS) development, a mobile app (eHealth app) was launched to further enhance collaborative care among the public sector, the private sector, the community, and the caregivers. OBJECTIVE: This study aims to investigate the factors associated with the downloading and utilization of the app, as well as the awareness, perception, and future improvement of the app. METHODS: We collected 2110 surveys; respondents were stratified into 3 groups according to their status of enrollment in the eHRSS. The primary outcome measure was the downloading and acceptance of the eHealth app. We collected the data on social economics factors, variables of the Technology Acceptance Model and Theory of Planned Behavior. Any factors identified as significant in the univariate analysis (P<.20) will be included in a subsequent multivariable regression analysis model. All P values ≤.05 will be considered statistically significant in multiple logistic regression analysis. The structural equation modeling was performed to identify interactions among the variables. RESULTS: The respondents had an overall high satisfaction rate and a positive attitude toward continuing to adopt and recommend the app. However, the satisfaction rate among respondents who have downloaded but not adopted the app was relatively lower, and few of them perceived that the downloading and acceptance processes are difficult. A high proportion of current users expressed a positive attitude about continuing to adopt and recommend the app to friends, colleagues, and family members. The behavioral intention strongly predicted the acceptance of the eHealth app (ß=.89; P<.001). Attitude (ß=.30; P<.001) and perceived norm; ß=.37; P<.001) played important roles in determining behavioral intention, which could predict the downloading and acceptance of the eHealth app (ß=.14; P<.001). CONCLUSIONS: Despite the high satisfaction rate among the respondents, privacy concerns and perceived difficulties in adopting the app were the major challenges of promoting eHealth. Further promotion could be made through doctors and publicity. For future improvement, comprehensive health records and tailored health information should be included.

The prevalence of metabolic dysfunction-associated fatty liver disease and its association on adverse pregnancy outcomes in women with gestational diabetes mellitus.

AIMS: To determine the prevalence of metabolic dysfunction-associated fatty liver disease (MAFLD) and its association on adverse pregnancy outcomes in women with gestational diabetes mellitus (GDM). METHODS: 380 women with GDM had a FibroScan® performed between 24 and 32 weeks. A Controlled Attenuation Parameter (CAP) ≥ 233.5 dB/m signified MAFLD. Gestational hypertension, pre-eclampsia and eclampsia contributed to a composite of adverse maternal outcomes. A composite of adverse neonatal outcomes included pre-term birth, hypoglycaemia, small/large-for-gestational age and admission to neonatal high dependency. Multiple logistic regression was used to determine independent associations with MAFLD. RESULTS: 147 (38.7 %) women had MAFLD. Higher pre-pregnancy BMI (median 28.4, IQR 24.7-33.0 kg/m2 vs median 24.0, IQR 21.7-26.5 kg/m2, p < 0.01) and rates of insulin therapy (62.2 % vs 40.7 %, p < 0.01) occurred in MAFLD over non-MAFLD women. MAFLD was independently associated with obesity (aOR 4.73 [95 % CI 2.53-8.86]) and insulin therapy (aOR 1.79 [95 % CI 1.12-2.85]). MAFLD and non-MAFLD women had comparable rates of adverse maternal (7.7 % vs 9.5 %, p = 0.56) and neonatal outcomes (75.4 % vs 68.8 %, p = 0.18). CONCLUSIONS: Within this GDM cohort, almost 40% had FibroScan®-detected MAFLD in mid-to-late gestation. MAFLD was not associated with worse pregnancy outcomes.

Antenatal FibroScan® assessment for metabolic-associated fatty liver in pregnant women at risk of gestational diabetes from a multiethnic population: a pilot study.

BACKGROUND: Metabolic-associated fatty liver disease (MAFLD) is a leading cause of chronic liver disease with increasingly recognised associations with gestational diabetes (GDM), including within the antenatal period. AIMS: To assess the relationship between MAFLD in pregnancy and development of GDM. METHODS: Fifty pregnant women were enrolled before 24 weeks gestation from a multiethnic obstetrics service in Sydney, Australia. Two FibroScan® assessments were performed, one prior to 24 weeks and one after 30 weeks gestation, to assess hepatic steatosis and stiffness. A control attenuated parameter (CAP) score ≥ 233.5 dB/m signified MAFLD. GDM was determined by an antenatal 75-g oral glucose tolerance testing. RESULTS: Six (12%) women had evidence of FibroScan®-detected MAFLD in early pregnancy, while none had abnormal hepatic stiffness. Sixteen (32%) women developed GDM. No significant difference was observed in GDM rates (50% vs 29.5%; P = 0.37) between those with MAFLD in early pregnancy and those without. At the second scan (completed by 34 women), those who developed GDM had a lower observed mean increase in CAP scores (11.1 ± 23.3 dB/m vs -14.9 ± 26.0 dB/m; P = 0.004) and lower maternal weight gain (0.6 ± 0.2 kg/week vs 0.4 ± 0.2 kg/week; P = 0.04). CONCLUSIONS: There was no statistically significant association between FibroScan®-detected MAFLD in early pregnancy and subsequent development of GDM in this pilot study. Maternal weight gain may be associated with changes in the CAP scores, which reflect steatosis, during pregnancy.

Shell Theory: A Statistical Model of Reality.

The foundational assumption of machine learning is that the data under consideration is separable into classes; while intuitively reasonable, separability constraints have proven remarkably difficult to formulate mathematically. We believe this problem is rooted in the mismatch between existing statistical techniques and commonly encountered data; object representations are typically high dimensional but statistical techniques tend to treat high dimensions a degenerate case. To address this problem, we develop a dedicated statistical framework for machine learning in high dimensions. The framework derives from the observation that object relations form a natural hierarchy; this leads us to model objects as instances of a high dimensional, hierarchal generative processes. Using a distance based statistical technique, also developed in this paper, we show that in such generative processes, instances of each process in the hierarchy, are almost-always encapsulated by a distinctive-shell that excludes almost-all other instances. The result is shell theory, a statistical machine learning framework in which separability constraints (distinctive-shells) are formally derived from the assumed generative process.

Toward Scalable and Unified Example-Based Explanation and Outlier Detection.

When neural networks are employed for high-stakes decision-making, it is desirable that they provide explanations for their prediction in order for us to understand the features that have contributed to the decision. At the same time, it is important to flag potential outliers for in-depth verification by domain experts. In this work we propose to unify two differing aspects of explainability with outlier detection. We argue for a broader adoption of prototype-based student networks capable of providing an example-based explanation for their prediction and at the same time identify regions of similarity between the predicted sample and the examples. The examples are real prototypical cases sampled from the training set via a novel iterative prototype replacement algorithm. Furthermore, we propose to use the prototype similarity scores for identifying outliers. We compare performance in terms of the classification, explanation quality and outlier detection of our proposed network with baselines. We show that our prototype-based networks extending beyond similarity kernels deliver meaningful explanations and promising outlier detection results without compromising classification accuracy.

Aspirin for the prevention of pre-eclampsia in women with pre-existing diabetes: Systematic review.

BACKGROUND: There is a lack of evidence for pre-eclampsia prophylaxis with aspirin in women with pre-existing diabetes mellitus (DM). AIMS: To examine the evidence for aspirin in pre-eclampsia prophylaxis in women with pre-existing DM. MATERIAL AND METHODS: An electronic search using Ovid MEDLINE, Embase, CinicalTrials.gov and the Cochrane CENTRAL register of controlled trials through to February 2021 was performed. Reference lists of identified studies, previous review articles, clinical practice guidelines and government reports were manually searched. Randomised controlled trials (RCTs) of aspirin vs placebo for pre-eclampsia prophylaxis were included. Articles were manually reviewed to determine if cohorts included women with DM. The systematic review was performed following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses statement. Data from included trials were extracted independently by two authors who also independently assessed risk of bias as per the Cochrane Handbook criteria version 5.1.0. Data were analysed using Rev-Man 5.4. RESULTS: Forty RCTs were identified, of which 11 included a confirmed subset of women with DM; however, data were insufficient for meta-analysis. Meta-analysis of 930 women with DM, from individual patient data included in a systematic review and unpublished data from one of the 11 RCTs, showed a non-significant difference in the outcome of pre-eclampsia in participants treated with aspirin compared to placebo (odds ratio 0.58; 95% CI 0.20-1.71; P = 0.33). CONCLUSIONS: Pre-eclampsia risk reduction with aspirin prophylaxis in women with pre-existing DM may be similar to women without pre-existing DM. However, randomised data within this meta-analysis were insufficient, warranting the need for further studies within this high-risk group of women.