Listening to community concerns in the COVID-19 infodemic: A WHO digital approach

Background

The Infodemic (too much information including false or misleading information in digital and physical environments) during the COVID-19 pandemic has led to confusion, risk-taking and behaviors that can amplify outbreaks, and reduce effectiveness of pandemic response efforts. To address this challenge, the WHO Information Network for Epidemics (EPI-WIN), in collaboration with research partners, developed a public health Infodemic intelligence analysis methodology for weekly analysis of digital media data to identify, categorize, and understand key concerns expressed in online conversations.

Methods

Thirty-five keyword-based searches (per language) using Meltwater Explore and Google Trends were created and grouped according to a set of pandemic public health taxonomy categories developed specifically for this analysis. The taxonomy has five thematic categories of conversation about COVID-19 and public health response: (1) the cause of the illness, (2) the illness, (3) the treatment, (4) the interventions and (5) Information.

Results

The two most recurring topics to attract increasing interest were Vaccines and Asymptomatic transmission followed by Immunity, Cause of the virus, Vulnerable communities and Reduction of movement, and Risk factors based on demographics and risk of misinformation.

Conclusions

The application of this taxonomy to online social listening week-on-week resulted in a better in-time understanding of the evolution and dynamics of high velocity conversations about COVID-19 globally during the pandemic and proposes a quantifiable approach to support planning of risk communication response.

Key messages

Describe widespread innovation in social listening methods for greater accountability to affected populations. Formulate insights into how digital media can be better utilized for more rapidly responding to the evolving needs of communities.

WHO public health research agenda for managing infodemics

Background

Following the World Health Organization's initial infodemic consultation in April 2020, a major infodemic conference was organised virtually in June-July 2020. Hundreds of experts participated to define science of infodemiology and build a public health research agenda that serves as a playbook for conducting relevant researches. Research Agenda provides guidance to invest in research and innovation so that we have better interventions and tools to understand, measure and respond to infodemics, and steer people towards timely, accessible, understandable information for good health choices.

Methods

The research agenda was developed during a virtual meeting, followed by research question prioritization exercise. It consisted of eight days spread out over four weeks. These were made up of: public preconference meeting; scientific conference, consisting of opening/closing plenary meetings either side of four separate “topic sprint” days; final public meeting to present the meeting outcomes.

After the meeting, a process took place to gather and rank research questions based on the research agenda created during the meeting.

Results

The following five streams and 65 research questions were developed. Measuring and monitoring the impact of infodemics during health emergencies Detecting and understanding the spread and impact of infodemics Responding and deploying interventions that protect against the infodemic and mitigate its harmful effects Evaluating infodemic interventions and strengthening resilience of individuals and communities to infodemics Promoting the development, adaptation and application of tools for managing infodemics.

Conclusions

Five streams with 65 research questions were developed and prioritized to structuralise infodemic management based on evidence. The conference yielded on the development of an infodemiology glossary, which can be used by the community of research.

Key messages

Discuss investments in research and innovation to enable a whole-of-society response to infodemics. Explain the practice of infodemic management as a discipline.

Effect of dietary sources of calcium and protein on hip fractures and falls in older adults in residential care: cluster randomised controlled trial

OBJECTIVE

To assess the antifracture efficacy and safety of a nutritional intervention in institutionalised older adults replete in vitamin D but with mean intakes of 600 mg/day calcium and <1 g/kg body weight protein/day.

DESIGN

Two year cluster randomised controlled trial.

SETTING

60 accredited residential aged care facilities in Australia housing predominantly ambulant residents.

PARTICIPANTS

7195 permanent residents (4920 (68%) female; mean age 86.0 (SD 8.2) years).

INTERVENTION

Facilities were stratified by location and organisation, with 30 facilities randomised to provide residents with additional milk, yoghurt, and cheese that contained 562 (166) mg/day calcium and 12 (6) g/day protein achieving a total intake of 1142 (353) mg calcium/day and 69 (15) g/day protein (1.1 g/kg body weight). The 30 control facilities maintained their usual menus, with residents consuming 700 (247) mg/day calcium and 58 (14) g/day protein (0.9 g/kg body weight).

MAIN OUTCOME MEASURES

Group differences in incidence of fractures, falls, and all cause mortality.

RESULTS

Data from 27 intervention facilities and 29 control facilities were analysed. A total of 324 fractures (135 hip fractures), 4302 falls, and 1974 deaths were observed. The intervention was associated with risk reductions of 33% for all fractures (121 v 203; hazard ratio 0.67, 95% confidence interval 0.48 to 0.93; P=0.02), 46% for hip fractures (42 v 93; 0.54, 0.35 to 0.83; P=0.005), and 11% for falls (1879 v 2423; 0.89, 0.78 to 0.98; P=0.04). The risk reduction for hip fractures and falls achieved significance at five months (P=0.02) and three months (P=0.004), respectively. Mortality was unchanged (900 v 1074; hazard ratio 1.01, 0.43 to 3.08).

CONCLUSIONS

Improving calcium and protein intakes by using dairy foods is a readily accessible intervention that reduces the risk of falls and fractures commonly occurring in aged care residents.

TRIAL REGISTRATION

Australian New Zealand Clinical Trials Registry ACTRN12613000228785.

Advances in real-time social listening for an adaptive public health response: WHO’s EARS platform

Background

COVID-19 pandemic was accompanied by an Infodemic (overabundance of information, including misinformation and disinformation, both online and offline); in response to this Infodemic, WHO launched the EARS platform (Early AI-assisted Response with Social Listening), showing real-time information about how people are talking about COVID-19 online. This information is intended to serve health information professionals to understand narratives and needs of the general public, in order to inform policy or communications decisions.

Methods

Data is collected daily from online conversations in publicly available sources, including Twitter, online forums, and blogs in English, French, Spanish and Portuguese, for 20 pilot countries. Once the data is collected, it is processed and classified into 39 categories, according to a set of pandemic public health taxonomy. The classification is made based on semi-supervised machine learning.

Results

Top 5 categories across regions are Covid-19 vaccine, Transmission settings, Personal measures, Testing and Industry (industry refers to the impact of the pandemic on the economy). We find that conversations around Covid-19 vaccines usually rank in the second or third position in all regions and represent 9%-12% of the conversation.

Conclusions

The configuration and application of the EARS platform has enabled progress towards more scalable and sustainable social listening to inform Infodemic management and response, compared to previous methods which were more manual, required data scientists in the team, or had fewer analytics capabilities. Future work will focus on gradually adding more data sources which can expand coverage and representativity.

Key messages

Discuss social listening methods for greater accountability to affected populations. Formulate insights into how digital media and information technology can be better utilized for more rapidly responding to the evolving needs of communities.

50 Global Actions to Manage the COVID-19 Infodemic: A WHO Framework

Issue

The World Health Organization describes an infodemic as an “overabundance of information - good or bad - that makes it difficult for people to make decisions for their health.”

Description of the problem

On April 7-8, 2020, the WHO Information Network for Epidemics (EPI-WIN) held a global online to crowdsource ideas from an interdisciplinary group of experts to form a novel COVID-19 infodemic response framework. The online consultation comprised of four plenary sessions and a brainstorming session conducted entirely online. Nearly 1500 individuals from over 100 countries and territories spanning social scientists, epidemiologists, staff from ministries of health and institutes of public health, registered for the consultation.

Results

A set of 50 proposed actions for a framework for managing infodemics in health emergencies was developed that will provide guidance for governments and public health institutions to take in five key areas of action that emerged from the consultation:

strengthening evidence and information

simplifying and explaining what is known fact-checking and addressing misinformation

amplifying messages and reaching the communities and individuals who need the information

quantifying and analysing the infodemic, including information flows, monitoring the acceptance of public health interventions, and assessing factors affecting behaviour at individual and population levels

strengthening systems for infodemic management in health emergencies

Lessons

Everyone has a role to play

Read the Call for Action

Sign the Call for Action

https://www.who.int/news/item/11-12-2020-call-for-action-managing-the-infode

Key messages

The confusion due to Infodemic can lead people to ignore public health measures and take risks that can cause serious harm.

Recognizing this WHO convened an interdisciplinary group of experts 7-8 April 2020 virtually to form a novel COVID-19 infodemic response framework.

Follow-up on new marker confirming the optimal status in the treatment of heart failure with reduced ejection fraction

Background

At the present time, there is no strong criterion to guarantee the optimal management for patients with heart failure and reduced ejection fraction (HFrEF). In the past, our group suggested the use of the size of the femoral vein (FV), measured by ultrasound as a marker of the fluid status in the venous compartment. Is this criterion still the best marker of optimal treatment of HFrEF at 3 years follow-up, especially for patients with significant co-morbidities (chronic obstructive pulmonary disease (COPD), end stage renal disease (ESRD) on hemodialysis (HD), cirrhosis of liver or sepsis?

Methods

Patients with HFrEF and co-morbidities as above were enrolled. All patients had echocardiography to confirm EF &lt;45% and underwent the ultrasound test to assess the size and expansibility of the femoral vein (SEFV). The SEFV is the ultrasound study of the FV examining its size and expansibility with cough. The location of the femoral artery (FA) and FV to be checked is the coronal plane immediately proximal to the bifurcation of the superficial and deep femoral artery. The normal size of FV is a little larger than of the FA. If the size of the FV is twice larger than the FA, the patient has fluid overload in the venous compartment (Figure 1). Then the patient was asked to cough in order to measure the size of the FV. If the FV did not increase its size with cough, the venous compartment was full. If the FV increased its size, the venous compartment was not full and could accommodate more fluid. In physical exam, the fluid overload is proved by the presence of extravascular fluid in the abdominal wall, ascites or leg edema or in the intravascular compartment by the presence of the jugular venous distention. During the 3 years of follow-up, the patients were seen in the office and had the SEFV at regular 6 months intervals. A small group of patients also underwent right heart catheterization to measure to the pulmonary capillary wedge pressure (PCWP).

Results

180 patients with HFrEF and significant comorbidities were enrolled. All patients were taught to follow a low Na diet, &lt;2000cc of fluid restriction and the guideline directed medical therapy. After about 3 years, 75% patients in the study group were asymptomatic, was not readmitted to the hospital for HF, and the size of the FV was within normal range. Their physical exam showed no fluid in the extravascular compartment. The PCWP became lower than 24mmHg in 18/20 who underwent the RHC. There was significant weight loss (15 lbs). In the control group, 60% of patients were asymptomatic and 50% were not readmitted for HF (p&lt;0.05).

Conclusions

With the SEFV test, the patients with HFrEF and significant comorbidities were accurately estimated for presence or absence of fluid overload. This SEFV test was especially sensitive to detect fluid overload in patients with multiple co-morbidities. Further randomized trials are needed to confirm the above preliminary results.

Funding Acknowledgement

Type of funding sources: None. Figure 1. Enlarged size of femoral vein

Slow flow or prolonged arterial phase in coronary arteries is the cause of ischemia or sudden death in patients with dilated cardiomyopathy

Background

In patients with dilated cardiomyopathy (DCM), the patients could be symptomatic or presented with recurrent fluid overload, syncope or sudden cardiac death (SCD) while the coronary arteries were patent. The aim of our study was to use the coronary flow abnormalities to stratify the high risk symptomatic versus the low risk asymptomatic patient with DCM.

Methods

Consecutive patients with DCM were enrolled. Twenty patients with normal ejection fraction (EF) without coronary artery disease served as control. The study patients were checked for symptoms (fluid overload, syncope, SCD) and re-admission. All patients underwent a new coronary angiographic technique with injection of contrast until all the coronary arteries were completely filled. As the injection of contrast stopped, the blood in white color moved in and the blood movement could be clearly observed. The angiogram was recorded from the entry of blood flow until all the contrast was cleared. During the review, the investigators downloaded, selected the angiogram from the electronic medical record, tapped on the Key Image and used the Up and Down arrow to move the images, frame-by-frame. Each frame represented a 0.06-second recording. The duration of the arterial phase was calculated starting the time when the blood entered the ostium of the index artery until all the contrast disappeared from the distal vasculature. At the same time, an AI program was trained to measure the length of the arterial phase by Machine learning, supervised and unsupervised Deep Learning and Convoluted Neural Networks. (Figure 1) The AI programs compared the time when the arteries were full with contrast until there was no contrast left in the distal vasculature. (Figure 2)

Results

One hundred patients with DCM were consecutively enrolled. Twenty patients served as control. In the control group with normal flow and EF, the duration of the arterial phase was 24–30 frames (1.44 to 2 seconds). In the study group, seventy patients had extremely prolonged arterial phase (average of 120 frames or &gt;8 seconds (p&lt;0.05). These patients were very symptomatic and had recurrent hospitalizations. Thirty patients had normal arterial phase of &lt;2 seconds. These patients had shorter length of stay (&lt;3 days), became asymptomatic after only 2 days of treatment and had rare readmission. (p&lt;0.05) The AI programs confirmed the results of the arterial phases calculated manually by junior investigators.

Conclusions

In patients with DCM, the extreme prolonged arterial phase caused ischemia in the myocardium even there was no coronary artery disease. This ischemic burden triggered recurrent ventricular dysfunction, arrhythmia, syncope and SCD. The patients with normal arterial phase became asymptomatic after optimal medical treatment. With these results, more effective prevention and management could be achieved for high risk symptomatic patients with high mortality and readmissions.

Funding Acknowledgement

Type of funding sources: None. U Net architectureArterial Phase

Coronary artery cavitation as a trigger for atherosclerotic plaque progression: a numerical and computational fluid dynamic demonstration

Background

Coronary cavitation is supposed to be generated by both concentric and eccentric coronary artery stenosis creating microbubbles which exploded when the fluid pressure was lower than the vapor pressure at a local thermodynamic state.

Aims

To assess, using computational fluid dynamic analysis (CFD), the potential role of cavitation in inducing coronary artery endothelial damage and promote atherosclerotic plaque progression.

Methods

We retrospectively reviewed the procedural records of consecutive patients evaluated between 1st January 2013 and 1st January 2014 with an isolated hemodynamically significant Left Main (LM) disease. Each bifurcation was reconstructed on the patient-specific geometries derived from the CCTA applying patient-specific hemodynamic features. Vapour has been modelled as discrete vapour bubbles and its trajectory determined using a Lagrangian frame of reference. Cavitation started with micro-cavitation nuclei which subsequently grow into bubbles undergoing different physical processes determined in a stochastic Monte-Carlo approximation.

Results

Among the 12 patients analysed [8 males, mean age 68.2±12.8 years old], the mean LM stenosis was 72.3±3.6%. In all subjects, LM stenoses induced cavitation which propagates downstream the vessel. The higher concentration of vapour region was detected before the carina (within 0.8 to 1.3 cm from the stenosis). The mean bubbles radius observed before the carina was 4.2±1.4 μm; their impact with the endothelial surface generated a mean peak pressure of 3.9±0.5 MPa determining a local shockwave (Figure 1).

Conclusion

The collapse of micro-bubbles alongside the endothelium generated micro-shockwaves determining repeated dynamic load measurable as an instantaneous pressure-peaks able to induce endothelial injury or dysfunction.

Funding Acknowledgement

Type of funding sources: None. Figure 1. (A) The simulation illustrates the vapour fraction iso-surfaces and the scattered bubble plots as predicted by the Langrangian model. Notably, most cavitation bubbles that form at the inlet of left main bifurcation do not collapse immediately but they are transported towards the vessel determining several interactions with the endothelium. (B) Graphical representation of the bubble radius modification and related pressure transmitted to the endothelium if the bubble collapse happened near to this last one. The magnification in boxes (B1-B4) evidences the dynamic modification of bubbles. The re-entry jet causing the collapse is evidenced with a red arrow.

Cavitation on top of collision breaks the cover of coronary plaques and triggers acute coronary syndrome

Background

Coronary injuries are hypothesized to be caused by the cavitation phenomenon (explosion of air bubbles) which is seen frequently in industrial pipes. Based on hydraulics principles applied to the coronary circulation. during distal negative suctioning in diastole, if the coronary static pressure decreases below the vapor pressure (VP), bubbles will form. They explode when the coronary static pressure recovers &gt; the VP during systole. These explosions create jet waves weakening and rupturing the cover of the coronary plaques, triggering acute coronary syndrome (ACS). How could these events be observed, recorded and compared?

Methods

Coronary angiograms of patients with ACS and stable coronary artery disease (CAD) (control) were selected. The arteries were recorded at 15 frames per second and saved in the electronic health records and reviewed image by image. After the index artery was completely filled with contrast, the following images showed the blood in white moving in on a background of black contrast. The flow could be laminar, turbulent (mixing of blood in white and contrast in black), antegrade or RETROGRADE (black column traveling backward). At the same time, an artificial intelligence (AI) program was used to detect and identify the flow.

Results

There were 104 patients with ACS enrolled and 20 patients with stable CAD as control. First, in the ACS group, 84 lesions (80%) were in the end of the proximal segment of the left anterior descending artery (LAD) and mid-segment of the right coronary artery (RCA). 20 lesions (19%) were at the distal RCA. Second, during diastole, 95% of the flow were laminar. The flow became turbulent at the beginning of systole. The turbulence was caused by the COLLISION of the antegrade flow (end of diastole) and the retrograde flow (at the beginning of systole). These collisions were seen in 95% at the location of vulnerable plaques of patients with ACS. In the control patients, there were only 2 cases (10%) with collision. Third, in the 20 patients with lesions at the distal RCA, the lesions were seen to be located at the areas of recirculating flow, at the ostium of the posterior descending artery (PDA) or proximal to the origin of the PDA. The cause of turbulence was most likely due to cavitation on top of collision. The cavitation happened because of continuous steady forward flow (of the PDA) in the myocardium during systole, while at the proximal RCA the blood flew forward more slowly. (Fig.1) The DSICREPANCY of velocities at the proximal and distal RCA allowed the formation of an empty gap (bubble of air). When the flow reversed during systole, this retrograde flow slammed on the bubble which collapsed violently, injured, ruptured the cover of the plaque and started ACS.

Conclusions

Rupture of bubbles (cavitation) on top of collision was most likely the cause of injury to the cover of vulnerable plaques, triggering ACS. Understanding the mechanism will help to better manage ACS.

Funding Acknowledgement

Type of funding sources: None. Cavity formation and collisionFormation of cavitation at the PDA

Normal ranges of global left ventricular myocardial work indices in adults: a meta-analysis

Background

Non-invasive global myocardial work recently emerged as new parameter to characterize left ventricle function with potential advantages over both ejection fraction and global longitudinal strain.

Purpose

We aimed to perform a meta-analysis of normal ranges of non-invasive left ventricular myocardial work (MW) indices including global constructive work (GCW), global work index (GWI), global wasted work (GWW), and global work efficiency (GWE) and to identify confounding factors that may contribute to variance in reported measures.

Methods

The authors searched four databases, Pubmed, Scopus, Embase, and Cochrane Library through January 2021 using the key terms “myocardial work”,“global constructive work”, “global wasted work”, “global work index”, “global work efficiency”. Studies were included if the articles reported LV myocardial work using 2D transthoracic echocardiography in healthy normal subjects, either in the control group or comprising the entire study cohort. The weighted mean was estimated by using the random effect model with a 95% confidence interval. Heterogeneity across studies was assessed using the I2 test. Publication bias was examined by funnel plot and Egger's regression test.

Results

The search yielded 476 articles. After abstract and full text screening we included 13 datasets with 1665 patients for meta-analysis. The reported normal mean values of GCW and GWI among the studies were 2278 (95% CI, 2167 to 23878; I2=95%), and 2.010 (95% CI, 1922 to 2098, I2=97%), respectively. The mean GWE was 96.0 (95% CI, 95.6% to 96.5; I2=92%), and the mean GWW was 79.7% (95% CI, 68.8% to 90.7%; I2=90%) (Figure). Furthermore, age and gender did not significantly contribute to variations in normal values. No evidence of significant publication bias was observed in the funnel plots and the Egger test.

Conclusion

In this meta-analysis, we provide echocardiographic reference ranges for non-invasive indices of MW. These normal values should serve as a template for clinical and research use for this promising technology.

Funding Acknowledgement

Type of funding sources: None.

Effectiveness of a multicomponent strategy for implementing guidelines for treating tobacco use in Vietnam Commune Health Centers

INTRODUCTION

Strategies are needed to increase implementation of evidence-based tobacco dependence treatment (TDT) in health care systems in low-and middle-income countries (LMICs).

METHODS

We conducted a two-arm cluster randomized controlled trial to compare the effectiveness of two strategies for implementing TDT guidelines in community health centers (n=26) in Vietnam. Arm 1 included training and a tool kit (e.g., reminder system) to promote and support delivery of the 4As (Ask about tobacco use, Advise to quit, Assess readiness, Assist with brief counseling) (Arm 1). Arm 2 included Arm 1 components plus a system to refer smokers to a community health worker (CHW) for more intensive counseling (4As+R). Provider surveys were conducted at baseline, six- and 12-months to assess the hypothesized effect of the strategies on provider and organizational-level factors. The primary outcome was provider adoption of the 4As.

RESULTS

Adoption of the 4As increased significantly across both study arms (all p<.001). Perceived organizational priority for TDT, compatibility with current workflow, and provider attitudes, norms and self-efficacy related to TDT also improved significantly across both arms. In Arm 2 sites, 41% of smokers were referred to a CHW for additional counseling.

CONCLUSION

The study demonstrated the effectiveness of a multicomponent and multilevel strategy (i.e., provider and system) for implementing evidence-based TDT in the Vietnam public health system. Combining provider-delivered brief counseling with opportunities for more in-depth counseling offered by a trained CHW may optimize outcomes and offers a potentially scalable model for increasing access to TDT in health care systems like Vietnam.

IMPLICATIONS

Improving implementation of evidence-based tobacco dependence treatment (TDT) guidelines is a necessary step towards reducing the growing burden of non-communicable disease (NCDs) and premature death in LMICs. The findings provide new evidence on the effectiveness of multilevel strategies for adapting and implementing TDT into routine care in Vietnam, and offers a potentially scalable model for meeting FCTC Article 14 goals in other LMICs with comparable public health systems. The study also demonstrates that combining provider-delivered brief counseling with referral to a community health worker for more in-depth counseling and support can optimize access to evidence-based treatment for tobacco use.

The structural basis of Salmonella A2B5 toxin neutralization by antibodies targeting the glycan-receptor binding subunits

Many bacterial pathogens secrete A₍₂₎B₅ toxins comprising two functionally distinct yet complementary “A” and “B” subunits to benefit the pathogens during infection. The lectin-like pentameric B subunits recognize specific sets of host glycans to deliver the toxin into target host cells. Here, we offer the molecular mechanism by which neutralizing antibodies, which have the potential to bind to all glycan-receptor binding sites and thus completely inhibit toxin binding to host cells, are inhibited from exerting this action. Cryogenic electron microscopy (cryo-EM)-based analyses indicate that the skewed positioning of the toxin A subunit(s) toward one side of the toxin B pentamer inhibited neutralizing antibody binding to the laterally located epitopes, rendering some glycan-receptor binding sites that remained available for the toxin binding and endocytosis process, which is strikingly different from the counterpart antibodies recognizing the far side-located epitopes. These results highlight additional features of the toxin-antibody interactions and offer important insights into anti-toxin strategies.

Nguyen et al. find that toxin-neutralizing antibodies targeting glycan-receptor binding B subunits can be split into two classes based on their epitope locations. They describe how these two classes exhibit significantly different neutralizing efficacies, a feature that appears to be shared among A₍₂₎B₅ toxins, and thus they provide insights into anti-toxin strategies.

Genetic surveillance in the Greater Mekong subregion and South Asia to support malaria control and elimination

Background

National Malaria Control Programmes (NMCPs) currently make limited use of parasite genetic data. We have developed GenRe-Mekong, a platform for genetic surveillance of malaria in the Greater Mekong Subregion (GMS) that enables NMCPs to implement large-scale surveillance projects by integrating simple sample collection procedures in routine public health procedures.

Methods

Samples from symptomatic patients are processed by SpotMalaria, a high-throughput system that produces a comprehensive set of genotypes comprising several drug resistance markers, species markers and a genomic barcode. GenRe-Mekong delivers Genetic Report Cards, a compendium of genotypes and phenotype predictions used to map prevalence of resistance to multiple drugs.

Results

GenRe-Mekong has worked with NMCPs and research projects in eight countries, processing 9623 samples from clinical cases. Monitoring resistance markers has been valuable for tracking the rapid spread of parasites resistant to the dihydroartemisinin-piperaquine combination therapy. In Vietnam and Laos, GenRe-Mekong data have provided novel knowledge about the spread of these resistant strains into previously unaffected provinces, informing decision-making by NMCPs.

Conclusions

GenRe-Mekong provides detailed knowledge about drug resistance at a local level, and facilitates data sharing at a regional level, enabling cross-border resistance monitoring and providing the public health community with valuable insights. The project provides a rich open data resource to benefit the entire malaria community.

Funding

The GenRe-Mekong project is funded by the Bill and Melinda Gates Foundation (OPP11188166, OPP1204268). Genotyping and sequencing were funded by the Wellcome Trust (098051, 206194, 203141, 090770, 204911, 106698/B/14/Z) and Medical Research Council (G0600718). A proportion of samples were collected with the support of the UK Department for International Development (201900, M006212), and Intramural Research Program of the National Institute of Allergy and Infectious Diseases.

O-222 An artificial intelligence model that was trained on pregnancy outcomes for embryo viability assessment is highly correlated with Gardner Score

Study question

Do artificial intelligence (AI) models used to assess embryo viability (based on pregnancy outcomes) also correlate with known embryo quality measures such as Gardner score?

Summary answer

An AI for embryo viability assessment also correlated with Gardner score, further substantiating the use of AI for assessment and selection of good quality embryos.

What is known already

The Gardner score consists of three separate components of embryo morphology that are graded individually, then combined to give a final score describing Day 5 embryo (blastocyst) quality. Evidence suggests the Gardner score has some correlation with clinical pregnancy. We hypothesized that an AI model trained to evaluate likelihood of clinical pregnancy based on fetal heartbeat (in clinical use globally) would also correlate with components of the Gardner score itself. We also compared the ability of the AI and Gardner score to predict pregnancy outcomes.

Study design, size, duration

This study involved analysis of a prospectively collected dataset of single static Day 5 embryo images with associated Gardner scores and AI viability scores. The dataset comprised time-lapse images of 1,485 embryos (EmbryoScope) from 638 patients treated at a single in vitro fertilization (IVF) clinic between November 2019 and December 2020. The AI model was not trained on data from this clinic.

Participants/materials, setting, methods

Average patient age was 35.4 years. Embryologists manually graded each embryo using the Gardner method, then subsequently used the AI to obtain a score between 0 (predicted non-viable, unlikely to lead to a pregnancy) and 10 (predicted viable, likely to lead to a pregnancy). Correlation between the AI viability score and Gardner score was then assessed.

Main results and the role of chance

The average AI score was significantly correlated with the three components of the Gardner score: expansion grade, inner cell mass (ICM) grade, and trophectoderm grade. Average AI score generally increased with advancing blastocyst developmental stage.

Blastocysts with expansion grades of ≥ 3 are generally considered suitable for transfer. This study showed that embryos with expansion grade 3 had lower AI scores than those with grades 4-6, consistent with a reduced pregnancy rate. AI correlation with trophectoderm grade was more significant than with ICM grade, consistent with studies demonstrating that trophectoderm grade is more important than ICM in determining clinical pregnancy likelihood.

The AI predicted Gardner scores of ≥ 2BB with an accuracy of 71.7% (sensitivity 75.1%, specificity 45.9%), and an AUC of 0.68. However, when used to predict pregnancy outcome, the AI performed 27.9% better than the Gardner score (accuracies of 49.8% and 39.0% respectively).

Even though the AI was highly correlated with the Gardner score, the improved efficacy for predicting pregnancy suggests that a) the AI provides an advantage in standardization of scoring over the manual and subjective Gardner method, and b) the AI is likely identifying and evaluating morphological features of embryo quality that are not captured by the Gardner method.

Limitations, reasons for caution

The Gardner score is not a linear score, creating challenges with setting a suitable threshold relating to the prediction of pregnancy. The 2BB treshold was chosen based on literature (Munné et al 2019) and verified by experienced embryologists. This correlative study may also require additional confirmatory studies on independent datasets.

Wider implications of the findings

The correlation between AI scores and known features of embryo quality (Gardner score) substantiates the use of the AI for embryo assessment. The AI score provides further insight into components of the Gardner score, and may detect morphological features related to clinical pregnancy beyond those evaluated by the Gardner method.

Trial registration number

Not applicable

P–202 Past embryo viability is not always a good predictor of future pregnancy: dynamic viability suggests video has limited benefit over static images for AI assessment

Study question

Does embryo quality/viability change over time, suggesting the use of video for AI-based embryo quality assessment has limited benefit over single point-in-time images?

Summary answer

AI assessment of single static embryo images at multiple time-points indicates embryo viability is dynamic, and past viability is a limited predictor of future pregnancy.

What is known already

Artificial Intelligence (AI) has been applied to the problem of embryo quality (viability) assessment using either video or single static images. However, whether historical data within video provide an additional advantage over single static images of embryos (at the time of transfer) for assessing embryo viability is not known. This applies to both manual and AI-based embryo assessment. If embryo viability changes over time prior to transfer, then the implication is that the assessment of future pregnancy using historical embryo data from videos would provide limited additional value over single static images taken immediately prior to transfer.

Study design, size, duration

Retrospective dataset of single embryo images taken at up-to three time-points prior to transfer: Early Day 5, Late Day 5 (8 hours later), and Early Day 6 (16 hours later), with corresponding fetal heartbeat (pregnancy) outcomes. The AI assessed the viability of each embryo at its available timepoints. Viability prediction was compared with pregnancy outcome to assess viability predictiveness at each timepoint prior to transfer, and assess the variability of viability over time.

Participants/materials, setting, methods

Single static images of 173 embryos were taken using time-lapse incubators from a single IVF clinic. 116 embryos were viable (led to a pregnancy) and 57 were non-viable (did not lead to a pregnancy). The AI was trained on thousands of Day 5 static embryo images taken from multiple IVF laboratories and countries, but was not trained on data from this clinic.

Main results and the role of chance

When embryos were assessed as viable by the AI immediately prior to transfer (no delay), the AI accuracy (sensitivity) in predicting pregnancy was 88.1% (59/67) for Early Day 5, 84.8% (28/33) for Late Day 5 and 87.5% (14/16) for Early Day 6. When the delay between AI assessment and transfer is 8 hours, 16 hours and 24 hours, the the accuracy drops to 66.7% (22/33), 31.3% (5/16) and 12.5% (2/16), respectively.

These results indicate that the viability of the embryo is dynamic, and therefore time series analysis, i.e. using video, may not be well suited for embryo viability assessment because past viability is not necessarily a good predictor of future viability or pregnancy outcome. The viability of the embryo immediately prior to transfer, from a single static image, is a reliable predictor of viability. This is consistent with the current clinical practice of using Gardner score end-point assessment for embryo quality.

Results also suggest significant benefits from using time-lapse with AI, where AI continually assesses embryo viability over time using static images. The time point at which the embryo should be transferred to maximize pregnancy outcome is when the embryo has the greatest AI viability score.

Limitations, reasons for caution

Although evidence suggests past embryo viability is a limited predictor of future pregnancy, a side-by-side comparison of video versus single static image AI assessment would further verify that the historical or change in embryo development or viability has minimal impact on embryo viability assessment at the time prior to transfer.

Wider implications of the findings: Time-lapse and AI can beneficially change the way embryos are assessed. Continual AI monitoring of embryos enables optimization of which embryo to transfer and when, to ultimately improve pregnancy outcomes for patients. The findings also suggest that static end-point AI assessment is sufficient for predicting embryo implantation potential.

Trial registration number

Not applicable

P–782 A natural language processing approach of global survey results on what the embryologist thinks and faces

Study question

What are the major problems faced by embryologists at 1) Clinic level, 2) Professional level, 3) Personal level, and 4) What are their career goals?

Summary answer

Embryologists, essential professionals of Fertility Centres, are less satisfied in many quantifiable aspects, but they love their profession and have many aspirational goals.

What is known already

IVF success depends in part on embryologists’ skills. The need to recognize clinical embryology as a specialty and clinical embryologists’ educational level, responsibilities, and workload have been addressed by a few national societies. However, data are lacking from the embryologists’ viewpoint at a global level about their profession. Qualitative data-analysis methods provide thick, rich descriptions of subjects’ thoughts, feelings, and lived experiences but can be time-consuming, labor-intensive, and prone to bias.

Study design, size, duration

A questionnaire was prepared using SurveyMonkey online software (SurveyMonkey, Inc., USA) and distributed to IVF lab professionals through embryology societies, online social media, and email databases. The questionnaire consisted of open-ended questions focused on identifying problems faced by embryologists at the clinic, in the profession, and in a personal level, as well as questions about their career outlook. The survey was active from May 2016 until February 2017. From 73 countries, 720 responses were obtained.

Participants/materials, setting, methods

Using natural language processing (NLP), the top 15 most frequently used keywords were identified and correlated with each other. Stronger correlation (≥0.5) between semantically similar words expressing a strong signal from each answer, and their usage was further analyzed for positive versus negative sentiment. By normalizing the frequency of positive/negative samples for each keyword as a percentage, “sentiment wheels” were produced, identifying the key concepts that respondents answered and quantifying how they felt about them.

Main results and the role of chance

The responses received were from 80% private, 17% public and 3% other ART settings distributed all over the world. From the embryologists’ viewpoints reported and after the NLP processing it was shown that the common topics related to strong negative sentiments were: embryologists’ remuneration (0.6) at the Clinic level; certification (0.7), recognition (0.5), respect (0.5), learn (0.5) and experience (0.5) at the Professional level; and remuneration (0.7), emotional (0.5) dealing (0.5) at the Personal level. Renumeration was reported and strongly related to embryologists’ viewpoint at both the clinic and personal level in combination with the need for certification, recognition and ongoing development at the Professional level. Moreover, the NLP processing demonstrated that the common topics on career goal analysis related to strong positive sentiments were: teaching (0.7), education (0.7), and continuation (0.5) all three topics are compatible with a professional orientation open to ongoing development and practice advancement. The NLP and the manual data analysis project an image of the typical embryologist as a knowledge seeking professional who is deeply dedicated to the job but feels the need for professional development and suffers some lack of recognition and feels in some cases not fairly treated as an employee.

Limitations, reasons for caution

The data obtained is limited. Only one natural language processing model was used to analyze the results. Different analysts using other methods may have different results. For these reasons, the results should be interpreted with caution.

Wider implications of the findings: It is important to focus on the lab as an organization and not just a service for the patients in treatment at the moment. The NLP results ultimately obtained may help streamline professional satisfaction efforts, and guide future quality management strategies

Trial registration number

Not applicable

P–228 AI-based assessment of embryo viability correlates with features of embryo ploidy

Study question

Do AI models used to assess embryo viability (based on pregnancy outcome) also correlate with known embryo quality measures such as ploidy status?

Summary answer

An AI for embryo viability assessment correlated with ploidy status, and with karyotypic features of aneuploidy, supporting its use for embryo selection.

What is known already

One factor that can influence pregnancy success is the genetic status of the embryo. PGT-A is commonly used to test for embryo ploidy, with the aim of identifying karyotypically normal embryos (euploid embryos), for preferential transfer. There is evidence suggesting that transfer of euploid embryos produces favorable clinical outcomes over aneuploid embryos.

Given the AI model was trained to evaluate clinical pregnancy, it was hypothesized that the score might also correlate with ploidy status, and with different types of aneuploidies. Little is known about morphological correlations with embryo ploidy status, so we also sought to explore this relationship.

Study design, size, duration

This study involved analysis of a retrospective dataset of single static Day 5 embryo (blastocyst) images with associated PGT-A results and AI viability scores. The dataset comprised images of 5,469 embryos from 2,615 consecutive patients treated at five US IVF clinics between February 2015 and April 2020. The AI was trained on thousands of Day 5 embryo images from multiple IVF laboratories in multiple countries, but was not trained on data used in this study.

Participants/materials, setting, methods

Average patient age was 36.2 years, and average embryo cohort size was 2.1/patient. PGT-A analysis was performed on embryos at time of evaluation. The dataset comprised 3,251 (59.4%) euploid embryos, 1,815 (33.2%) aneuploid embryos, and 403 (7.4%) mosaic embryos. The AI was retrospectively used to provide a score between 0 (predicted non-viable) and 10 (predicted viable) for each image. Correlation between the AI viability score and euploid, mosaic and aneuploid embryos was then assessed.

Main results and the role of chance

Results showed a statistically significant correlation between AI viability score and PGT-A outcome, consistent with a relationship between pregnancy outcome and ploidy status. The average score for euploid embryos was 8.20, which was significantly higher than the average score for aneuploid embryos of 7.80 (p &lt; 0.0001).

There was a significant linear increase in confidence score from full aneuploid embryos, through mosaic embryos (average score 7.97), to full euploid embryos (mosaic threshold of 20–80%). High mosaic embryos tended to have a lower average score (7.60) than low mosaic embryos (7.96), consistent with correlation of viability (pregnancy outcome) with the degree of mosaicism. AI viability score also correlated with ploidy features believed to affect pregnancy outcomes. Trisomic changes had higher average scores than monosomic changes. Segmental changes had higher average scores than full gain or loss. The AI score differentiated euploid from aneuploid status more efficiently in embryos with poorer morphology than those with good morphology.

Whilst there was an evident correlation between pregnancy outcome and ploidy status, the AI was only weakly predictive of euploidy, with an accuracy of 57.3% using an AI viability score threshold of 7.5/10.This suggests pregnancy-related morphological features are somewhat correlated with embryo ploidy, but not completely.

Limitations, reasons for caution

The PGT-A technique is held to have some limitations for evaluating ploidy status, therefore it would be of benefit to perform additional confirmatory studies on independent datasets. It would be of interest to conduct prospective studies evaluating correlations between the AI’s evaluation of morphology and pregnancy outcome with ploidy status.

Wider implications of the findings: The AI score correlated with genetic features of embryos that are known to correlate with pregnancy, which further supports the efficacy and use of AI for embryo viability assessment. The AI identified morphological features that are somewhat predictive of ploidy status, with potential application to embryos of poorer Gardner score.

Trial registration number

none

Constraints on Cosmic Strings Using Data from the Third Advanced LIGO-Virgo Observing Run

We search for gravitational-wave signals produced by cosmic strings in the Advanced LIGO and Virgo full O3 dataset. Search results are presented for gravitational waves produced by cosmic string loop features such as cusps, kinks, and, for the first time, kink-kink collisions. A template-based search for short-duration transient signals does not yield a detection. We also use the stochastic gravitational-wave background energy density upper limits derived from the O3 data to constrain the cosmic string tension Gμ as a function of the number of kinks, or the number of cusps, for two cosmic string loop distribution models. Additionally, we develop and test a third model that interpolates between these two models. Our results improve upon the previous LIGO-Virgo constraints on Gμ by 1 to 2 orders of magnitude depending on the model that is tested. In particular, for the one-loop distribution model, we set the most competitive constraints to date: Gμ≲4×10^{-15}. In the case of cosmic strings formed at the end of inflation in the context of grand unified theories, these results challenge simple inflationary models.

Constraints on Cosmic Strings Using Data from the Third Advanced LIGO-Virgo Observing Run

We search for gravitational-wave signals produced by cosmic strings in the Advanced LIGO and Virgo full O3 dataset. Search results are presented for gravitational waves produced by cosmic string loop features such as cusps, kinks, and, for the first time, kink-kink collisions. A template-based search for short-duration transient signals does not yield a detection. We also use the stochastic gravitational-wave background energy density upper limits derived from the O3 data to constrain the cosmic string tension Gμ as a function of the number of kinks, or the number of cusps, for two cosmic string loop distribution models. Additionally, we develop and test a third model that interpolates between these two models. Our results improve upon the previous LIGO-Virgo constraints on Gμ by 1 to 2 orders of magnitude depending on the model that is tested. In particular, for the one-loop distribution model, we set the most competitive constraints to date: Gμ≲4×10^{-15}. In the case of cosmic strings formed at the end of inflation in the context of grand unified theories, these results challenge simple inflationary models.

GraphDTA: predicting drug-target binding affinity with graph neural networks

SUMMARY

The development of new drugs is costly, time consuming and often accompanied with safety issues. Drug repurposing can avoid the expensive and lengthy process of drug development by finding new uses for already approved drugs. In order to repurpose drugs effectively, it is useful to know which proteins are targeted by which drugs. Computational models that estimate the interaction strength of new drug-target pairs have the potential to expedite drug repurposing. Several models have been proposed for this task. However, these models represent the drugs as strings, which is not a natural way to represent molecules. We propose a new model called GraphDTA that represents drugs as graphs and uses graph neural networks to predict drug-target affinity. We show that graph neural networks not only predict drug-target affinity better than non-deep learning models, but also outperform competing deep learning methods. Our results confirm that deep learning models are appropriate for drug-target binding affinity prediction, and that representing drugs as graphs can lead to further improvements.

AVAILABILITY OF IMPLEMENTATION

The proposed models are implemented in Python. Related data, pre-trained models and source code are publicly available at https://github.com/thinng/GraphDTA. All scripts and data needed to reproduce the post hoc statistical analysis are available from https://doi.org/10.5281/zenodo.3603523.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

MO622IMPAIRED PROTEIN-BOUND UREMIC TOXIN EXCRETION SUGGESTS TUBULAR DYSFUNCTION IN DIABETIC NEPHROPATHY

Background and Aims

Kidney tubular damage is an important prognostic determinant in diabetic nephropathy (DN) (1). Proximal tubular secretion is a vital homeostatic function that is responsible for excretion of waste, such as protein bound uremic toxins (PBUTs) that are minimally eliminated via glomerular filtration. PBUTs are potentially harmful waste products of endogenous metabolism that are efficiently excreted by tubular secretion via organic anion transporters (OATs). Currently available diagnostic methods cannot accurately detect tubular dysfunction. We hypothesize that renal PBUT clearance may be a sensitive tubular function marker. Here, we measured PBUTs in long-term streptozotocin (STZ)-induced murine diabetic nephropathy (DN), which was characterized by severe tubular atrophy and interstitial fibrosis (2).

Method

Diabetes mellitus was induced in C57Bl/6 mice by a single intraperitoneal injection of 200 mg/kg STZ which resulted in substantial kidney damage after 6 months evaluated by histopathology and conventional markers (2). Indoxyl sulfate (IS), hippuric acid (HA), kynurenic acid (KA), kynurenine (Kyn), p-cresol glucuronide (pCG), p-cresol sulfate (pCS) and indole acetic acid (IAA) were measured in plasma and urine after 6 and 8 months by LC-MS/MS.

Results

Among the PBUTs with the highest OAT affinity, viz. IS, HA and KA, plasma concentrations were 2.2-, 2.3- and 1.5-fold higher (p=0.005, 0.0006, 0.03; Fig 1A-C) after 6 months and 1.9-, 2.1- and 2-fold higher (p=0.008, 0.0005, 0.001; Fig 2A-C) after 8 in DN, and urinary excretions (normalized for plasma concentrations) were 3.3-, 2.3- and 3.0-fold lower (p=0.012, 0.16, 0.03; Fig 1G-I) after 6 months and 2.5-, 1.6-, 2.3-fold lower (p=0.028, 0.046, 0.0005; Fig 2G-I) after 8 months in DN. Other PBUTS, viz. IAA, Kyn, pCS and pCG, were not significantly affected.

Conclusion

Our findings suggest that OAT function is compromised in murine long-term DN. Renal clearance of IS, HA and KA may be a marker of tubular function in DN. Future studies should focus on correlations with histology and validation in other species.

Connective Tissue Growth Factor Is Overexpressed in Explant Lung Tissue and Broncho-Alveolar Lavage in Transplant-Related Pulmonary Fibrosis

Background

Connective tissue growth factor (CTGF) is an important mediator in several fibrotic diseases, including lung fibrosis. We investigated CTGF-expression in chronic lung allograft dysfunction (CLAD) and pulmonary graft-versus-host disease (GVHD).

Materials and Methods

CTGF expression was assessed by quantitative real-time polymerase chain reaction (qPCR) and immunohistochemistry in end-stage CLAD explant lung tissue (bronchiolitis obliterans syndrome (BOS), n=20; restrictive allograft syndrome (RAS), n=20), pulmonary GHVD (n=9). Unused donor lungs served as control group (n=20). Next, 60 matched lung transplant recipients (BOS, n=20; RAS, n=20; stable lung transplant recipients, n=20) were included for analysis of CTGF protein levels in plasma and broncho-alveolar lavage (BAL) fluid at 3 months post-transplant, 1 year post-transplant, at CLAD diagnosis or 2 years post-transplant in stable patients.

Results

qPCR revealed an overall significant difference in the relative content of CTGF mRNA in BOS, RAS and pulmonary GVHD vs. controls (p=0.014). Immunohistochemistry showed a significant higher percentage and intensity of CTGF-positive respiratory epithelial cells in BOS, RAS and pulmonary GVHD patients vs. controls (p<0.0001). BAL CTGF protein levels were significantly higher at 3 months post-transplant in future RAS vs. stable or BOS (p=0.028). At CLAD diagnosis, BAL protein content was significantly increased in RAS patients vs. stable (p=0.0007) and BOS patients (p=0.042). CTGF plasma values were similar in BOS, RAS, and stable patients (p=0.74).

Conclusions

Lung CTGF-expression is increased in end-stage CLAD and pulmonary GVHD; and higher CTGF-levels are present in BAL of RAS patients at CLAD diagnosis. Our results suggest a potential role for CTGF in CLAD, especially RAS, and pulmonary GVHD.