Identifying susceptibility of children and adolescents to the Omicron variant (B.1.1.529).

BACKGROUND: The Omicron variant (B.1.1.529) is estimated to be more transmissible than previous strains of SARS-CoV-2 especially among children, potentially resulting in croup which is a characteristic disease in children. Current coronavirus disease 2019 (COVID-19) cases among children might be higher because (i) school-aged children have higher contact rates and (ii) the COVID-19 vaccination strategy prioritizes the elderly in most countries. However, there have been no reports confirming the age-varying susceptibility to the Omicron variant to date. METHODS: We developed an age-structured compartmental model, combining age-specific contact matrix in South Korea and observed distribution of periods between each stage of infection in the national epidemiological investigation. A Bayesian inference method was used to estimate the age-specific force of infection and, accordingly, age-specific susceptibility, given epidemic data during the third (pre-Delta), fourth (Delta driven), and fifth (Omicron driven) waves in South Korea. As vaccine uptake increased, individuals who were vaccinated were excluded from the susceptible population in accordance with vaccine effectiveness against the Delta and Omicron variants, respectively. RESULTS: A significant difference between the age-specific susceptibility to the Omicron and that to the pre-Omicron variants was found in the younger age group. The rise in susceptibility to the Omicron/pre-Delta variant was highest in the 10-15 years age group (5.28 times [95% CI, 4.94-5.60]), and the rise in susceptibility to the Omicron/Delta variant was highest in the 15-19 years age group (3.21 times [95% CI, 3.12-3.31]), whereas in those aged 50 years or more, the susceptibility to the Omicron/pre-Omicron remained stable at approximately twofold. CONCLUSIONS: Even after adjusting for contact pattern, vaccination status, and waning of vaccine effectiveness, the Omicron variant of SARS-CoV-2 tends to propagate more easily among children than the pre-Omicron strains.

Epidemiology and Transmission Dynamics of Infectious Diseases and Control Measures.

The epidemiology and transmission dynamics of infectious diseases must be understood at the individual and community levels to improve public health decision-making for real-time and integrated community-based control strategies. Herein, we explore the epidemiological characteristics for assessing the impact of public health interventions in the community setting and their applications. Computational statistical methods could advance research on infectious disease epidemiology and accumulate scientific evidence of the potential impacts of pharmaceutical/nonpharmaceutical measures to mitigate or control infectious diseases in the community. Novel public health threats from emerging zoonotic infectious diseases are urgent issues. Given these direct and indirect mitigating impacts at various levels to different infectious diseases and their burdens, we must consider an integrated assessment approach, 'One Health', to understand the dynamics and control of infectious diseases.

Evaluation of Optimal Assessment Schedules for Surveillance After Definitive Locoregional Treatment of Locally Advanced Head and Neck Cancer: A Retrospective Cohort Study With Parametric Modeling of Event-Free Survival.

Importance: In clinical practice, assessment schedules are often arbitrarily determined after definitive treatment of head and neck cancer (HNC), producing heterogeneous and inconsistent surveillance plans. Objective: To establish an optimal assessment schedule for patients with definitively treated locally advanced HNC, stratified by the primary subsite and HPV status, using a parametric model of standardized event-free survival curves. Design, Setting, and Participants: This was a retrospective study including 2 tertiary referral hospitals and a total of 673 patients with definitive locoregional treatment of locally advanced HNC (227 patients with nasopharyngeal cancer [NPC]; 237 patients with human papillomavirus-positive oropharyngeal cancer [HPV+ OPC]; 47 patients with HPV-negative [HPV-] OPC; 65 patients with hypopharyngeal cancer [HPC]; and 97 patients with laryngeal cancer [LC]). Patients had received primary treatment in 2008 through 2019. The median (range) follow-up duration was 57.8 (6.4-158.1) months. Data analyses were performed from April to October 2021. Main Outcomes and Measures: Tumor recurrence and secondary malignant neoplasms. Event-free survival was defined as the period from the end of treatment to occurrence of any event. Event-free survival curves were estimated using a piecewise exponential model and divided into 3 phases of regular follow-up. A 5% event rate criterion determined optimal follow-up time point and interval. Results: The median (range) age of the 673 patients at HNC diagnosis was 58 (15-83) years; 555 (82.5%) were men; race and ethnicity were not considered. The event rates of NPC, HPV+ OPC, HPV- OPC, HPC, and LC were 18.9% (43 of 227), 14.8% (35 of 237), 36.2% (17 of 47), 44.6% (29 of 65), and 30.9% (30 of 97), respectively. Parametric modeling demonstrated optimal follow-up intervals for HPC, LC, and NPC, respectively, every 2.1, 3.2, and 6.1 months; 3.7, 5.6, and 10.8 months; and 9.1, 13.8, and 26.5 months until 16.5, 16.5 to 25.0, and 25.0 to 99.0 months posttreatment (open follow-up thereafter). For HPV- OPC, assessment was recommended every 2.7, 4.8, and 11.8 months until 16.5, 16.5 to 25.0, and 25 to 99 months posttreatment, respectively. In contrast, HPV+ OPC optimal intervals were every 7.7, 13.7, and 33.7 months until 16.5, 16.5 to 25.0, and 25 to 99 months posttreatment, respectively. Five, 4, 12, 15, and 10 follow-up visits were recommended for NPC, HPV+ OPC, HPV- OPC, HPC, and LC, respectively. Conclusions and Relevance: This retrospective cohort study using parametric modeling suggests that the HNC assessment schedules should be patient tailored and evidence based to consider primary subsites and HPV status. Given limited health care resources and rising detection rates and costs of HNC, the guidelines offered by these findings could benefit patients and health systems and aid in developing future consensus guidelines.

SLIDE: A surrogate fairness constraint to ensure fairness consistency.

As they take a crucial role in social decision makings, AI algorithms based on ML models should be not only accurate but also fair. Among many algorithms for fair AI, learning a prediction ML model by minimizing the empirical risk (e.g., cross-entropy) subject to a given fairness constraint has received much attention. To avoid computational difficulty, however, a given fairness constraint is replaced by a surrogate fairness constraint as the 0-1 loss is replaced by a convex surrogate loss for classification problems. In this paper, we investigate the validity of existing surrogate fairness constraints and propose a new surrogate fairness constraint called SLIDE, which is computationally feasible and asymptotically valid in the sense that the learned model satisfies the fairness constraint asymptotically and achieves a fast convergence rate. Numerical experiments confirm that the SLIDE works well for various benchmark datasets.

Age-Varying Susceptibility to the Delta Variant (B.1.617.2) of SARS-CoV-2.

Importance: The Delta variant (B.1.617.2) is estimated to be more transmissible than previous strains of SARS-CoV-2, especially among children and adolescents. However, to our knowledge, there are no reports confirming this to date. Objective: To gain a better understanding of the association of age with susceptibility to the Delta variant of SARS-CoV-2. Design, Setting, and Participants: This decision analytic model used an age-structured compartmental model using the terms symptom onset (S), exposure (E), infectious (I), and quarantine (Q) (SEIQ) to estimate the age-specific force of infection, combining age-specific contact matrices and observed distribution of periods between each stage of infection (E to I [ie, latent period], I given S, and S to Q [ie, diagnostic delay]) developed in a previous contact tracing study. A bayesian inference method was used to estimate the age-specific force of infection (S to E) and, accordingly, age-specific susceptibility. The age-specific susceptibility during the third wave (ie, before Delta, from October 15 to December 22, 2020, when the COVID-19 vaccination campaign was not yet launched) and the fourth wave (ie, the Delta-driven wave, from June 27 to August 21, 2021) in Korea were compared. As vaccine uptake increased, individuals who were vaccinated were excluded from the susceptible population in accordance with vaccine effectiveness against the Delta variant. This nationwide epidemiologic study included individuals who were diagnosed with COVID-19 during the study period in Korea. Data were analyzed from September to November 2021. Exposures: Age group during the third wave (ie, before Delta) and fourth wave (ie, Delta-driven) of the COVID-19 pandemic in South Korea. Main Outcomes and Measures: Age-specific susceptibility during the third and fourth waves was estimated. Results: Among 106 866 confirmed COVID-19 infections (including 26 597 infections and 80 269 infections during the third and fourth waves of COVID-19 in Korea, respectively), a significant difference in age-specific susceptibility to the Delta vs pre-Delta variant was found in the younger age group. After adjustment for contact pattern and vaccination status, the increase in susceptibility to the Delta vs pre-Delta variant was estimated to be highest in the group aged 10 to 15 years, approximately doubling (1.92-fold increase [95% CI, 1.86-fold to 1.98-fold]), whereas in the group aged 50 years or more, susceptibility to the Delta vs pre-Delta variant remained stable at an approximately 1-fold change (eg, among individuals aged 50-55 years: 0.997-fold [95% CI, 0.989-fold to 1.001-fold). Conclusions and Relevance: In this study, the Delta variant of SARS-CoV-2 was estimated to propagate more easily among children and adolescents than pre-Delta strains, even after adjusting for contact pattern and vaccination status.

Nonconvex Sparse Regularization for Deep Neural Networks and Its Optimality.

Recent theoretical studies proved that deep neural network (DNN) estimators obtained by minimizing empirical risk with a certain sparsity constraint can attain optimal convergence rates for regression and classification problems. However, the sparsity constraint requires knowing certain properties of the true model, which are not available in practice. Moreover, computation is difficult due to the discrete nature of the sparsity constraint. In this letter, we propose a novel penalized estimation method for sparse DNNs that resolves the problems existing in the sparsity constraint. We establish an oracle inequality for the excess risk of the proposed sparse-penalized DNN estimator and derive convergence rates for several learning tasks. In particular, we prove that the sparse-penalized estimator can adaptively attain minimax convergence rates for various nonparametric regression problems. For computation, we develop an efficient gradient-based optimization algorithm that guarantees the monotonic reduction of the objective function.

Contact-adjusted Immunity Levels against SARS-CoV-2 in Korea and Prospects for Achieving Herd Immunity.

The proportion of population vaccinated cannot be directly translated into the herd immunity. We have to account for the age-stratified contact patterns to calculate the population immunity level, since not every individual gathers evenly. Here, we calculated the contact-adjusted population immunity against severe acute respiratory syndrome coronavirus 2 in South Korea using age-specific incidence and vaccine uptake rate. We further explored options to achieve the theoretical herd immunity with age-varying immunity scenarios. As of June 21, 2021, when a quarter of the population received at least one dose of a coronavirus disease 2019 (COVID-19) vaccine, the contact-adjusted immunity level was 12.5% under the social distancing level 1. When 80% of individuals aged 10 years and over gained immunity, we could achieve a 58.2% contact-adjusted immunity level. The pros and cons of vaccinating children should be weighed since the risks of COVID-19 for the young are less than the elderly, and the long-term safety of vaccines is still obscure.

Radiological assessment schedule for 1p/19q-codeleted gliomas during the surveillance period using parametric modeling.

BACKGROUND: There have been no evidence-based guidelines on the optimal schedule for the radiological assessment of 1p/19q-codeleted glioma. This study aimed to recommend an appropriate radiological evaluation schedule for 1p/19q-codeleted glioma during the surveillance period through parametric modeling of the progression-free survival (PFS) curve. METHODS: A total of 234 patients with 1p/19q-codeleted glioma (137 grade II and 97 grade III) who completed regular treatment were retrospectively reviewed. The patients were stratified into each layered progression risk group by recursive partitioning analysis. A piecewise exponential model was used to standardize the PFS curves. The cutoff value of the progression rate among the remaining progression-free patients was set to 10% at each scan. RESULTS: Progression risk stratification resulted in 3 groups. The optimal magnetic resonance imaging (MRI) interval for patients without a residual tumor was every 91.2 weeks until 720 weeks after the end of regular treatment following the latent period for 15 weeks. For patients with a residual tumor after the completion of adjuvant radiotherapy followed by chemotherapy, the optimal MRI interval was every 37.5 weeks until week 90 and every 132.8 weeks until week 361, while it was every 33.6 weeks until week 210 and every 14.4 weeks until week 495 for patients with a residual tumor after surgery only or surgery followed by radiotherapy only. CONCLUSIONS: The optimal radiological follow-up schedule for each progression risk stratification of 1p/19q-codeleted glioma can be established from the parametric modeling of PFS.

Fast convergence rates of deep neural networks for classification.

We derive the fast convergence rates of a deep neural network (DNN) classifier with the rectified linear unit (ReLU) activation function learned using the hinge loss. We consider three cases for a true model: (1) a smooth decision boundary, (2) smooth conditional class probability, and (3) the margin condition (i.e., the probability of inputs near the decision boundary is small). We show that the DNN classifier learned using the hinge loss achieves fast rate convergences for all three cases provided that the architecture (i.e., the number of layers, number of nodes and sparsity) is carefully selected. An important implication is that DNN architectures are very flexible for use in various cases without much modification. In addition, we consider a DNN classifier learned by minimizing the cross-entropy, and show that the DNN classifier achieves a fast convergence rate under the conditions that the noise exponent and margin exponent are large. Even though they are strong, we explain that these two conditions are not too absurd for image classification problems. To confirm our theoretical explanation, we present the results of a small numerical study conducted to compare the hinge loss and cross-entropy.

Left-censored dementia incidences in estimating cohort effects.

We estimate the dementia incidence hazard in Germany for the birth cohorts 1900 until 1954 from a simple sample of Germany's largest health insurance company. Followed from 2004 to 2012, 36,000 uncensored dementia incidences are observed and further 200,000 right-censored insurants included. From a multiplicative hazard model we find a positive and linear trend in the dementia hazard over the cohorts. The main focus of the study is on 11,000 left-censored persons who have already suffered from the disease in 2004. After including the left-censored observations, the slope of the trend declines markedly due to Simpson's paradox, left-censored persons are imbalanced between the cohorts. When including left-censoring, the dementia hazard increases differently for different ages, we consider omitted covariates to be the reason. For the standard errors from large sample theory, left-censoring requires an adjustment to the conditional information matrix equality.

Radiological assessment schedule for high-grade glioma patients during the surveillance period using parametric modeling.

BACKGROUND: An optimal radiological surveillance plan is crucial for high-grade glioma (HGG) patients, which is determined arbitrarily in daily clinical practice. We propose the radiological assessment schedule using a parametric model of standardized progression-free survival (PFS) curves. METHODS: A total of 277 HGG patients (178 glioblastoma [GBM] and 99 anaplastic astrocytoma [AA]) from a single institute who completed the standard treatment protocol were enrolled in this cohort study and retrospectively analyzed. The patients were stratified into each layered risk group by genetic signatures and residual mass or through recursive partitioning analysis. PFS curves were estimated using the piecewise exponential survival model. The criterion of a 10% progression rate among the remaining patients at each observation period was used to determine the optimal radiological assessment time point. RESULTS: The optimal follow-up intervals for MRI evaluations of isocitrate dehydrogenase (IDH) wild-type GBM was every 7.4 weeks until 120 weeks after the end of standard treatment, followed by a 22-week inflection period and every 27.6 weeks thereafter. For the IDH mutated GBM, scans every 13.2 weeks until 151 weeks are recommended. The optimal follow-up intervals were every 22.8 weeks for IDH wild-type AA, and 41.2 weeks for IDH mutated AA until 241 weeks. Tailored radiological assessment schedules were suggested for each layered risk group of the GBM and the AA patients. CONCLUSIONS: The optimal schedule of radiological assessments for each layered risk group of patients with HGG could be determined from the parametric model of PFS.

Transmission onset distribution of COVID-19.

OBJECTIVES: The distribution of the transmission onset of COVID-19 relative to the symptom onset is a key parameter for infection control. It is often not easy to study the transmission onset time, as it is difficult to know who infected whom exactly when. METHODS: We inferred transmission onset time from 72 infector-infectee pairs in South Korea, either with known or inferred contact dates, utilizing the incubation period. Combining this data with known information of the infector's symptom onset, we could generate the transmission onset distribution of COVID-19, using Bayesian methods. Serial interval distribution could be automatically estimated from our data. RESULTS: We estimated the median transmission onset to be 1.31 days (standard deviation, 2.64 days) after symptom onset with a peak at 0.72 days before symptom onset. The pre-symptomatic transmission proportion was 37% (95% credible interval [CI], 16-52%). The median incubation period was estimated to be 2.87 days (95% CI, 2.33-3.50 days), and the median serial interval to be 3.56 days (95% CI, 2.72-4.44 days). CONCLUSIONS: Considering that the transmission onset distribution peaked with the symptom onset and the pre-symptomatic transmission proportion is substantial, the usual preventive measures might be too late to prevent SARS-CoV-2 transmission.

Smooth Function Approximation by Deep Neural Networks with General Activation Functions.

There has been a growing interest in expressivity of deep neural networks. However, most of the existing work about this topic focuses only on the specific activation function such as ReLU or sigmoid. In this paper, we investigate the approximation ability of deep neural networks with a broad class of activation functions. This class of activation functions includes most of frequently used activation functions. We derive the required depth, width and sparsity of a deep neural network to approximate any Hölder smooth function upto a given approximation error for the large class of activation functions. Based on our approximation error analysis, we derive the minimax optimality of the deep neural network estimators with the general activation functions in both regression and classification problems.

Logistic LASSO regression for the diagnosis of breast cancer using clinical demographic data and the BI-RADS lexicon for ultrasonography.

PURPOSE: The aim of this study was to compare the performance of image analysis for predicting breast cancer using two distinct regression models and to evaluate the usefulness of incorporating clinical and demographic data (CDD) into the image analysis in order to improve the diagnosis of breast cancer. METHODS: This study included 139 solid masses from 139 patients who underwent a ultrasonography-guided core biopsy and had available CDD between June 2009 and April 2010. Three breast radiologists retrospectively reviewed 139 breast masses and described each lesion using the Breast Imaging Reporting and Data System (BI-RADS) lexicon. We applied and compared two regression methods-stepwise logistic (SL) regression and logistic least absolute shrinkage and selection operator (LASSO) regression-in which the BI-RADS descriptors and CDD were used as covariates. We investigated the performances of these regression methods and the agreement of radiologists in terms of test misclassification error and the area under the curve (AUC) of the tests. RESULTS: Logistic LASSO regression was superior (P<0.05) to SL regression, regardless of whether CDD was included in the covariates, in terms of test misclassification errors (0.234 vs. 0.253, without CDD; 0.196 vs. 0.258, with CDD) and AUC (0.785 vs. 0.759, without CDD; 0.873 vs. 0.735, with CDD). However, it was inferior (P<0.05) to the agreement of three radiologists in terms of test misclassification errors (0.234 vs. 0.168, without CDD; 0.196 vs. 0.088, with CDD) and the AUC without CDD (0.785 vs. 0.844, P<0.001), but was comparable to the AUC with CDD (0.873 vs. 0.880, P=0.141). CONCLUSION: Logistic LASSO regression based on BI-RADS descriptors and CDD showed better performance than SL in predicting the presence of breast cancer. The use of CDD as a supplement to the BI-RADS descriptors significantly improved the prediction of breast cancer using logistic LASSO regression.

Detection of gene-environment interactions in a family-based population using SCAD.

Gene-environment interaction (GxE) is emphasized as one potential source of missing genetic variation on disease traits, and the ultimate goal of GxE research is prediction of individual risk and prevention of complex diseases. However, there are various challenges in statistical analysis of GxE. In this paper, we focus on the three methodological challenges: (i) the high dimensions of genes; (ii) the hierarchical structure between interaction effects and their corresponding main effects; and (iii) the correlation among subjects from family-based population studies. In this paper, we propose an algorithm that approaches all three challenges simultaneously. This is the first penalized method focusing on an interaction search based on a linear mixed effect model. For verification, we compare the empirical performance of our new method with other existing methods in simulation study. The results demonstrate the superiority of our method under overall simulation setup. In particular, the outperformance obviously becomes greater as the correlation among subjects increases. In addition, the new method provides a robust estimate for the correlation among subjects. We also apply the new method on Genetics of Lipid Lowering Drugs and Diet Network study data. Copyright © 2017 John Wiley & Sons, Ltd.

Diagnostic Value of Combining Tumor and Inflammatory Markers in Lung Cancer.

BACKGROUND: Despite major advances in lung cancer treatment, early detection remains the most promising way of improving outcomes. To detect lung cancer in earlier stages, many serum biomarkers have been tested. Unfortunately, no single biomarker can reliably detect lung cancer. We combined a set of 2 tumor markers and 4 inflammatory or metabolic markers and tried to validate the diagnostic performance in lung cancer. METHODS: We collected serum samples from 355 lung cancer patients and 590 control subjects and divided them into training and validation datasets. After measuring serum levels of 6 biomarkers (human epididymis secretory protein 4 [HE4], carcinoembryonic antigen [CEA], regulated on activation, normal T cell expressed and secreted [RANTES], apolipoprotein A2 [ApoA2], transthyretin [TTR], and secretory vascular cell adhesion molecule-1 [sVCAM-1]), we tested various sets of biomarkers for their diagnostic performance in lung cancer. RESULTS: In a training dataset, the area under the curve (AUC) values were 0.821 for HE4, 0.753 for CEA, 0.858 for RANTES, 0.867 for ApoA2, 0.830 for TTR, and 0.552 for sVCAM-1. A model using all 6 biomarkers and age yielded an AUC value of 0.986 and sensitivity of 93.2% (cutoff at specificity 94%). Applying this model to the validation dataset showed similar results. The AUC value of the model was 0.988, with sensitivity of 93.33% and specificity of 92.00% at the same cutoff point used in the validation dataset. Analyses by stages and histologic subtypes all yielded similar results. CONCLUSIONS: Combining multiple tumor and systemic inflammatory markers proved to be a valid strategy in the diagnosis of lung cancer.

Meta-markers for the differential diagnosis of lung cancer and lung disease.

UNLABELLED: Misdiagnosis of lung cancer remains a serious problem due to the difficulty of distinguishing lung cancer from other respiratory lung diseases. As a result, the development of serum-based differential diagnostic biomarkers is in high demand. In this study, 198 clinical serum samples from non-cancer lung disease and lung cancer patients were analyzed using nLC-MRM-MS for the levels of seven lung cancer biomarker candidates. When the candidates were assessed individually, only SERPINEA4 showed statistically significant changes in the serum levels. The MRM results and clinical information were analyzed using a logistic regression analysis to select model for the best 'meta-marker', or combination of biomarkers for differential diagnosis. Also, under consideration of statistical interaction, variables having low significance as a single factor but statistically influencing on meta-marker model were selected. Using this probabilistic classification, the best meta-marker was determined to be made up of two proteins SERPINA4 and PON1 with age factor. This meta-marker showed an enhanced differential diagnostic capability (AUC=0.915) for distinguishing the two patient groups. Our results suggest that a statistical model can determine optimal meta-markers, which may have better specificity and sensitivity than a single biomarker and thus improve the differential diagnosis of lung cancer and lung disease patients. BIOLOGICAL SIGNIFICANCE: Diagnosing lung cancer commonly involves the use of radiographic methods. However, an imaging-based diagnosis may fail to differentiate lung cancer from non-cancerous lung disease. In this study, we examined several serum proteins in the sera of 198 lung cancer and non-cancerous lung disease patients by multiple-reaction monitoring. We then used a combination of variables to generate a meta-marker model that is useful as a differential diagnostic biomarker.

Erratum: Diagnostic Value of Combining Tumor and Inflammatory Markers in Lung Cancer.

[This corrects the article on p. 187 in vol. 21, PMID: 27722145.].