When humans solve complex problems, they typically construct, reflect, and revise sequences of ideas, hypotheses, and beliefs until a final decision or conclusion is reached. Contrary to this, current machine learning models are mostly trained to map an input to one single and fixed output. In this paper, we investigate how we can equip models with the ability to represent, construct, and evaluate a second, third, and $k$-th thought within their prediction process. Drawing inspiration from Hegel’s dialectics, we propose and evaluate the thought flow concept which constructs a sequence of predictions. We present a self-correction mechanism which (a) is trained to estimate the model’s correctness and which (b) performs iterative prediction updates based on the gradient of the correctness prediction. We introduce our method focusing initially on question answering (QA) and carry out extensive experiments which demonstrate that (i) our method is able to correct its own predictions and that (ii) it can improve model performance by a large margin. In addition, we conduct a qualitative analysis of thought flow correction patterns and explore how thought flow predictions affect users’ human-AI collaboration in a crowdsourcing study. We find that (iii) thought flows improve user performance and are perceived as more natural, correct, and intelligent regarding single and/or top-3 predictions.

Background: Zilucoplan, a macrocyclic peptide complement component 5 inhibitor, sustained efficacy for up to 60 weeks of treatment, with a favourable safety profile in patients with acetylcholine receptor autoantibody-positive generalised myasthenia gravis in an interim analysis of RAISE-XT (NCT04225871). We evaluate the safety and efficacy of zilucoplan up to 96 weeks. Methods: RAISE-XT, a Phase 3, multicentre, open-label extension study, included patients who participated in the double-blind Phase 2 (NCT03315130) and Phase 3 (NCT04115293) zilucoplan studies. Patients self-administered daily subcutaneous zilucoplan 0.3mg/kg injections. Primary outcome was incidence of treatment-emergent adverse events (TEAEs). Secondary outcomes included change from baseline in Myasthenia Gravis Activities of Daily Living (MG-ADL) score. Results: At data cut-off (11 May 2023), median (range) exposure to zilucoplan was 1.8 (0.11–5.1) years (N=200). TEAEs occurred in 191 (95.5%) patients; the most common TEAE was COVID-19 (n=64; 32.0%). At Week 96, mean (standard error) change in MG-ADL score from double-blind study baseline was –6.33 (0.49) and –7.83 (0.60) for patients who received zilucoplan 0.3mg/kg and placebo in the double-blind studies, respectively. Conclusions: Zilucoplan demonstrated a favourable long-term safety profile. Efficacy was sustained for 96 weeks in patients who had previously received zilucoplan and who switched from placebo.

Background: Treatment of generalized myasthenia gravis (gMG) with reduced steroid dosages may minimize steroid-associated AEs. Corticosteroid dosage changes were not permitted during the 26-week, CHAMPION MG study of ravulizumab in adults with anti-acetylcholine receptor antibody-positive (AChRAb+) gMG. Participants who completed the study could receive ravulizumab in the open-label extension (OLE; NCT03920293); corticosteroid adjustments were permitted. Methods: Patients could receive intravenous ravulizumab (blind induction or bridging dose at Week 26 [OLE start] for those previously receiving placebo or ravulizumab, respectively, then 3000–3600 mg at Week 28 and every 8 weeks thereafter) for ≤4 years. Results: Among 161 patients (78 ravulizumab, 83 placebo) who entered the OLE and received ravulizumab for ≤164 weeks, 113 received oral or enteral corticosteroids during the OLE; the proportion treated with >10 mg/day corticosteroids decreased from 58% (n=66) at first OLE dose to 37% (n=42) (35 [31%] received ≤5 mg/day and 71 [63%] received ≤10 mg/day) at last reported dose. Fourteen patients (12%) discontinued corticosteroids. The mean (SD) corticosteroid dosage/patient decreased from 17.5 (11.9) mg/day at first OLE dose to 11.7 (10.9) mg/day at last assessment. Conclusions: Ravulizumab decreased corticosteroid use in patients with AChRAb+ gMG, suggesting a steroid-sparing role for ravulizumab.

In this paper, we provide a general framework for counting geometric structures in pseudo-random graphs. As applications, our theorems recover and improve several results on the finite field analog of questions originally raised in the continuous setting. The results present interactions between discrete geometry, geometric measure theory, and graph theory.

The literature on green tea consumption and glucose metabolism has reported conflicting findings. This cross-sectional study examined the association of green tea consumption with abnormal glucose metabolism among 3000 rural residents aged 40–60 years in Khánh Hòa province in Vietnam. Multinomial logistic regression analysis was conducted to examine the association of green tea consumption (0, < 200, 200–< 400, 400–< 600 or ≥ 600 ml/d) with prediabetes and diabetes (based on the American Diabetes Association criteria). Linear regression analysis was performed to examine the association between green tea consumption and the log-transformed homeostatic model assessment of insulin resistance (HOMA-IR) (a marker of insulin resistance) and the log-transformed homeostatic model assessment of β-cell function (HOMA-β) (a marker of insulin secretion). The OR for prediabetes and diabetes among participants who consumed ≥ 600 ml/d v. those who did not consume green tea were 1·61 (95 % CI = 1·07, 2·42) and 2·04 (95 % CI = 1·07, 3·89), respectively. Higher green tea consumption was associated with a higher level of log-transformed HOMA-IR (Pfor trend = 0·04) but not with a lower level of log-transformed HOMA-β (Pfor trend = 0·75). Higher green tea consumption was positively associated with the prevalence of prediabetes, diabetes and insulin resistance in rural Vietnam. The findings of this study indicated prompting the need for further research considering context in understanding the link between green tea consumption and glucose metabolism, especially in rural settings in low- and middle-income countries.

Background: Efgartigimod is a human IgG1 antibody Fc-fragment that reduces IgG levels through FcRn blockade. A key efficacy indicator in the treatment of IgG autoantibody-mediated generalized myasthenia gravis (gMG) is improvement in MG-ADL score. Methods: The ADAPT phase 3 trial evaluated safety and efficacy of efgartigimod in patients with gMG, including reaching and maintaining of minimal symptom expression (MSE; defined as an MG-ADL total score of 0 or 1). Results: 167 patients (AChR-Ab+, n=129; AChR-Ab-, n=38) were randomized to receive treatment cycles of 4 weekly infusions of efgartigimod or placebo. Significantly more AChR-Ab+ efgartigimod-treated patients achieved MSE during cycle 1 compared to placebo-treated patients (40.0% [n=26/65] vs 11.1% [n=7/63; P<0.0001]). In cycle 2, 31.4% (n=16/51) of AChR-Ab+ patients in the efgartigimod cohort achieved MSE compared to none in the placebo cohort. MG-ADL score improved by ≥6 points in 56.9% of AChR-Ab+ efgartigimod-treated patients compared to 20.6% of placebo-treated patients in cycle 1. Most patients achieved MSE by week 4 of a cycle, paralleling early reduction in IgG levels, and MSE duration ranged from 1 to ≥10 weeks. Adverse events were predominantly mild to moderate. Conclusions: Efgartigimod treatment resulted in more patients with AChR-Ab+ gMG achieving both MSE and clinically meaningful MG-ADL improvements.

Background: Efgartigimod, a human IgG1 antibody Fc-fragment, reduces IgG levels through neonatal Fc receptor blockade. Patients with anti-acetylcholine receptor antibody–negative (AChR-Ab–) generalized myasthenia gravis (gMG) comprise 15%-20% of the gMG population and have limited approved treatment options. We evaluated long-term safety and efficacy of efgartigimod in AChR-Ab– patients from ADAPT/ADAPT+ (open-label extension). Methods: ADAPT evaluated safety and efficacy of efgartigimod versus placebo in AChR-Ab+ (n=129) and Ab– (n=38) patients with gMG. This integrated analysis includes 37 AChR-Ab– patients who received ≥1 dose of efgartigimod in ADAPT/ADAPT+ through October 2020 (median[range] follow-up: 453[85-721] days). Responder status was defined as ≥2-point (MG-ADL) and ≥3-point (QMG) improvement for ≥4 consecutive weeks (with first improvement 1 week after last infusion). Results: Among AChR-Ab– patients in ADAPT (cycle 1), 68.4% (13/19) efgartigimod-treated were MG-ADL responders (placebo, 63.2% [12/19]), and 52.6% (10/19) were QMG responders (placebo, 36.8% [7/19]). In the integrated ADAPT/ADAPT+ analysis (cycle 1), AChR-Ab– patients improved from baseline in MG-ADL/QMG scores, with consistent improvements across multiple subsequent cycles. No clinically meaningful differences in safety or efficacy outcomes between AChR-Ab+ and Ab– patients occurred. Conclusions: Long-term treatment (median >1 year) with efgartigimod was well tolerated and associated with clinically meaningful improvements in MG-ADL/QMG scores in AChR-Ab– patients.

Background: The 26-week double-blind, randomized, placebo-controlled period (RCP) of the CHAMPION MG study (NCT03920293) demonstrated ravulizumab’s efficacy and tolerability in anti-acetylcholine receptor antibody-positive (AChR Ab+) generalized myasthenia gravis (gMG). Methods: In the ongoing open-label extension (OLE), patients receive intravenous ravulizumab (blind loading dose in placebo-treated patients or bridging dose in ravulizumab-treated patients, then 3000–3600 mg according to body weight every 8 weeks) for ≤4 years. Data from RCP baseline up to Week 60 were analyzed. Results: Ravulizumab’s long-term efficacy (n=161) and safety (n=169) were assessed. Patients who switched from placebo in the RCP to ravulizumab in the OLE (n=83) showed rapid improvement (least squares mean, 95%CI) in Myasthenia Gravis-Activity of Daily Living (MG-ADL) and Quantitative Myasthenia Gravis (QMG) total scores, which were maintained through 34 weeks (MG-ADL: −1.7, −2.7 to −0.8; QMG: −3.1, −4.2 to −1.9). Improvements achieved by ravulizumab-treated patients (n=78) in the RCP were sustained through 60 weeks (MG-ADL: −4.0, −4.8 to −3.1; QMG: −4.1, −5.4 to −2.9). Ravulizumab was well tolerated; no meningococcal infections were reported. Four deaths unrelated to study treatment occurred. Conclusions: Ravulizumab demonstrated sustained improvements in MG symptoms and was well tolerated for up to 60 weeks in adults with AChR Ab+ gMG.

Background: Efgartigimod is a human IgG1 antibody Fc-fragment that reduces total and pathogenic IgG autoantibody levels through FcRn blockade. ADAPT was a phase 3 trial evaluating efgartigimod in patients with generalized myasthenia gravis (gMG). Patients who completed ADAPT could enroll in ADAPT+ (open-label extension). Methods: Efgartigimod (10 mg/kg intravenous) was administered in cycles of 4 weekly infusions, with subsequent cycles initiated based on clinical evaluation. ADAPT+ evaluated long-term safety and tolerability of efgartigimod in patients with gMG. Efficacy was assessed utilizing MG-ADL and QMG scores. Results: Of 167 patients from ADAPT, 151 (90%) entered ADAPT+, and 145 received ≥1 cycle as of January 2022. Over 217.55 patient-years of follow-up (mean duration per patient, 548 days), incidence of adverse events did not increase with subsequent cycles. AChR-Ab+ patients with ≥1 year of follow-up across ADAPT/ADAPT+ (n=95) received a median (range) 5.0 (0.4–7.6) cycles per year. All AChR-Ab+ patients (n=111) demonstrated consistent improvements (mean change [SE], week 3 of cycle 1) in MG-ADL (-5.0 [0.33]; up to 14 cycles) and QMG (-4.7 [0.41]; up to 7 cycles) scores during each cycle. Conclusions: These ADAPT+ analyses suggest long-term efgartigimod treatment is well tolerated and efficacious. Additional final data cut analyses will be presented at CNSF 2023.

The paper is concerned with common shock models of claim triangles. These are usually constructed as linear combinations of shock components and idiosyncratic components. Previous literature has discussed the unbalanced property of such models, whereby the shocks may over- or under-contribute to some observations. The literature has also introduced corrections for this. The present paper discusses “auto-balanced” models, in which all shock and idiosyncratic components contribute to observations such that their proportionate contributions are constant from one observation to another. The conditions for auto-balance are found to be simple and applicable to a wide range of model structures. Numerical illustrations are given.

Objectives: Healthcare-associated infections (HAIs) are one of the greatest challenges and concerns in Vietnam and around the world. Many studies have shown that HAIs may result in an increase in hospital length of stay, antibiotic use, multidrug-resistant organism (MDROs) infections, treatment costs, and mortality. Therefore, in the past 5 years, the Department of Infection Control of Cho Ray Hospital has carried out many infection and prevention control (IPC) activities to reduce the rates of HAI and MDRO infection. We evaluated IPC activities and results achieved in these efforts at Cho Ray Hospital during 2017–2021. Methods: We described the implemented IPC activities and retrospectively collected data from HAIs surveillance reports during 2017–2021 for 3 intensive care units (ICUs): ICU-B, ICU-D, and the NICU. Results: In the past 5 years, we implemented synchronous IPC activities, including promoting hand hygiene training and surveillance, environmental cleaning surveillance, carrying out improvement projects such as a ventilator-associated pneumonia (VAP) prevention bundle, an MDRO prevention bundle, and an environmental cleaning quality improvement project. Many positive results were achieved, although a slight increase in the HAI incidence occurred in 2021 due to the COVID-19 pandemic. Overall, the hand hygiene compliance rate increased from 49.7% to 83.8%. The rate of HAIs per 1,000 patient days decreased steadily from 5.4 to 2.4. The VAP rate fell from 30.5 to 17.2 per 1,000 patient days, and the central-line–associated bloodstream infection (CLABSI) rate decreased gradually from 5.4 to 2.4 per 1,000 patient days. The catheter-associated urinary tract infection (CAUTI) rate decreased from 2.9 to 0.9 per 1,000 patient days, and the MDRO infection rate decreased significantly from 32.7 to 11.3 per 1,000 patient days. Conclusions: The synchronous implementation of HAI prevention bundles promoting hand hygiene and environmental cleaning achieved significant effects in the efforts to decrease HAIs and MDROs in the ICUs of Cho Ray Hospital.