Causal Inference of CNS-regulated Hormones in COVID-19: A Bidirectional Two-sample Mendelian Randomization Study (preprint)

Objectives: We assessed the causal association of three COVID-19 phenotypes with insulin-like growth factor 1 (IGF-1), estrogen, testosterone, dehydroepiandrosterone (DHEA), thyroid-stimulating hormone (TSH), thyrotropin-releasing hormone (TRH), luteinizing hormone (LH), and follicle-stimulating hormone (FSH). Methods: We used a bidirectional two-sample univariate and multivariable Mendelian randomization (MR) analysis to evaluate the direction, specificity, and causality of the association between CNS-regulated hormones and COVID-19 phenotypes. Genetic instruments for CNS-regulated hormones were selected from the largest publicly available genome-wide association studies in the European population. Summary-level data on COVID-19 severity, hospitalization, and susceptibility were obtained from the COVID-19 host genetic initiative. Results: DHEA was associated with increased risks of very severe respiratory syndrome (OR=4.21, 95% CI: 1.41-12.59), consistent with the results in multivariate MR (OR=3.72, 95% CI: 1.20-11.51), and hospitalization (OR = 2.31, 95% CI: 1.13-4.72) in univariate MR. LH was associated with very severe respiratory syndrome (OR=0.83; 95% CI: 0.71-0.96) in univariate MR. Estrogen was negatively associated with very severe respiratory syndrome (OR=0.09, 95% CI: 0.02-0.51), hospitalization (OR=0.25, 95% CI: 0.08-0.78), and susceptibility (OR=0.50, 95% CI: 0.28-0.89) in multivariate MR. Conclusions: We found strong evidence for the causal relationship of DHEA, LH, and estrogen with COVID-19 phenotypes.

A Generalized Multinomial Probabilistic Model for SARS-CoV-2 Infection Prediction and Public Health Intervention Assessment in an Indoor Environment (preprint)

SARS-CoV-2 Omicron has become the predominant variant globally. Current infection models are limited by the need for large datasets or calibration to specific contexts, making them difficult to cater for different settings. To ensure public health decision-makers can easily consider different public health interventions (PHIs) over a wide range of scenarios, we propose a generalized multinomial probabilistic model of airborne infection to systematically capture group characteristics, epidemiology, viral loads, social activities, environmental conditions, and PHIs, with assumptions made on social distancing and contact duration, and estimate infectivity over short time-span group gatherings. This study is related to our 2021 work published in Nature Scientific Reports that modelled airborne SARS-CoV-2 infection (Han, Lam, Li, et al., 2021). It is differentiated from former works on probabilistic infection modelling in terms of the following: (1) predicting new cases arising from more than one infectious in a gathering, (2) incorporating additional key infection factors, and (3) evaluating the effectiveness of multiple PHIs on SARS-CoV-2 infection simultaneously. Although our results reveal that limiting group size has an impact on infection, improving ventilation has a much greater positive health impact. Our model is versatile and can flexibly accommodate other scenarios by allowing new factors to be added, to support public health decision-making.