Fengqing Zhang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Fengqing Zhang has a total of 145 co-authors. Here are the top co-authors by number of shared publications:
- Jiangtao Gou
Department of Mathematics and Statistics, Villanova University, Villanova, PA 19085, United States of America
9 shared publications Xin Niu
Drexel Univ, Dept Psychol & Brain Sci, 3201 Chestnut St, Philadelphia, PA 19104 USA
9 shared publications- Evan M. Forman
Center for Weight, Eating, and Lifestyle Sciences (WELL Center), Drexel University
7 shared publications Hansoo Chang
Department of Psychological and Brain Sciences, Drexel University, Philadelphia, PA, USA
7 shared publications- Meghan L. Butryn
Department of Psychology, Drexel University, Philadelphia, Pennsylvania, USA
6 shared publications Alexei Taylor
Drexel University
5 shared publications
Latest Publications
- Open Access
- 17 September 2026
- Published by Elsevier in Contemporary Clinical Trials
- Vol. 170, 108473
Sample size estimation for a flexible phase II oncology trial design using two dependent endpoints with nested response metrics
- 17 September 2026
- Published by Elsevier in Contemporary Clinical Trials
- Vol. 170, 108473
Open AccessSample size estimation for a flexible phase II oncology trial design using two dependent endpoints with nested response metrics
- Open Access
- 5 July 2026
- Published by Elsevier in SSRN Electronic Journal
Sample size estimation for a flexible phase II oncology trial design using two dependent endpoints with nested response metrics
Phase II oncology trials are often designed using a single primary efficacy endpoint, even when multiple clinically relevant response metrics are available. When two endpoints are based on nested response metrics, such as objective response rate and disease control rate or progression-free survival at two ordered time points, success on the more stringent endpoint implies success on the broader endpoint. This structure induces a known dependence relationship that can be incorporated into sample size estimation and decision-rule construction. We develop a flexible phase II design framework for two primary endpoints with nested response metrics. The framework includes union-based designs for declaring activity on at least one endpoint, intersection-based designs for requiring activity on both endpoints, and multiplicity-adjusted designs for endpoint-specific inference with strong familywise error-rate control. For the multiplicity-adjusted setting, we propose Hochberg-type and Holm-type designs and provide sample size optimization algorithms based on the nested dependence structure. An application to a phase II trial of sintilimab in recurrent or progressive meningioma illustrates the approach using 12-month and 6-month progression-free survival. Compared with a Bonferroni design, the Holm-type and Hochberg-type designs can reduce the required sample size while maintaining prespecified error-rate and power requirements.
- 5 July 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessSample size estimation for a flexible phase II oncology trial design using two dependent endpoints with nested response metrics
Phase II oncology trials are often designed using a single primary efficacy endpoint, even when multiple clinically relevant response metrics are available. When two endpoints are based on nested response metrics, such as objective response rate and disease control rate or progression-free survival at two ordered time points, success on the more stringent endpoint implies success on the broader endpoint. This structure induces a known dependence relationship that can be incorporated into sample size estimation and decision-rule construction. We develop a flexible phase II design framework for two primary endpoints with nested response metrics. The framework includes union-based designs for declaring activity on at least one endpoint, intersection-based designs for requiring activity on both endpoints, and multiplicity-adjusted designs for endpoint-specific inference with strong familywise error-rate control. For the multiplicity-adjusted setting, we propose Hochberg-type and Holm-type designs and provide sample size optimization algorithms based on the nested dependence structure. An application to a phase II trial of sintilimab in recurrent or progressive meningioma illustrates the approach using 12-month and 6-month progression-free survival. Compared with a Bonferroni design, the Holm-type and Hochberg-type designs can reduce the required sample size while maintaining prespecified error-rate and power requirements.
- 30 June 2026
- Published by Drexel University Libraries
Spatial and temporal brain and biological aging trajectories and their impact on physical and mental health outcomes
Aging is a heterogeneous process that unfolds differently across brain regions and biological systems. Traditional biomarkers of aging, whether neuroimaging-derived brain age or biomarker-derived biological age, provide useful but limited perspectives because they are studied by calculating a single index to represent either brain or biological age. Importantly, while variability can be quantified at the population level, most existing approaches lack methods for quantifying uncertainty at the individual level, leaving researchers and clinicians uncertain about how confident to be in any single person's estimated brain or biological age. This dissertation develops a multidimensional aging framework that integrates brain and biological age estimates while utilizing CVAE-based spatiotemporal clustering approaches. Using UK Biobank as the primary dataset and MIDUS as an external validation cohort, we employed a deep neural network (Conditional Variational Autoencoders) to cluster brain ROI's and biological systems by both spatial and temporal patterns simultaneously to identify aging-relevant feature clusters across neuroimaging modalities and biological systems. For each identified aging biomarker cluster, stacked ensemble (SVR/GPR/DNN) and conformal inference methods will be used to quantify brain and biological age with individual-level prediction uncertainty. Results demonstrated that multidimensional clustering approaches identified distinct aging-related brain and biological domains with heterogeneous aging trajectories. Cluster-specific age estimates showed meaningful associations with physical, cognitive, and mental health outcomes, including cardiovascular disease, diabetes, dementia-related outcomes, depression, anxiety, and cognitive performance measures. In several analyses, cluster-specific aging measures provided more informative and localized associations than conventional single-index age-gap approaches. Conformal prediction analyses further demonstrated that uncertainty-aware aging measures could identify individuals with high-confidence accelerated aging profiles through positive out-of-bounds (OOB) classification. External validation analyses in MIDUS supported the generalizability of the developed framework across independent cohorts. By incorporating simultaneous spatial and temporal clustering, individual prediction intervals, and external replication, this work demonstrates a novel, rigorous, and interpretable framework for characterizing aging heterogeneity.
- 30 June 2026
- Published by Drexel University Libraries
Spatial and temporal brain and biological aging trajectories and their impact on physical and mental health outcomes
Aging is a heterogeneous process that unfolds differently across brain regions and biological systems. Traditional biomarkers of aging, whether neuroimaging-derived brain age or biomarker-derived biological age, provide useful but limited perspectives because they are studied by calculating a single index to represent either brain or biological age. Importantly, while variability can be quantified at the population level, most existing approaches lack methods for quantifying uncertainty at the individual level, leaving researchers and clinicians uncertain about how confident to be in any single person's estimated brain or biological age. This dissertation develops a multidimensional aging framework that integrates brain and biological age estimates while utilizing CVAE-based spatiotemporal clustering approaches. Using UK Biobank as the primary dataset and MIDUS as an external validation cohort, we employed a deep neural network (Conditional Variational Autoencoders) to cluster brain ROI's and biological systems by both spatial and temporal patterns simultaneously to identify aging-relevant feature clusters across neuroimaging modalities and biological systems. For each identified aging biomarker cluster, stacked ensemble (SVR/GPR/DNN) and conformal inference methods will be used to quantify brain and biological age with individual-level prediction uncertainty. Results demonstrated that multidimensional clustering approaches identified distinct aging-related brain and biological domains with heterogeneous aging trajectories. Cluster-specific age estimates showed meaningful associations with physical, cognitive, and mental health outcomes, including cardiovascular disease, diabetes, dementia-related outcomes, depression, anxiety, and cognitive performance measures. In several analyses, cluster-specific aging measures provided more informative and localized associations than conventional single-index age-gap approaches. Conformal prediction analyses further demonstrated that uncertainty-aware aging measures could identify individuals with high-confidence accelerated aging profiles through positive out-of-bounds (OOB) classification. External validation analyses in MIDUS supported the generalizability of the developed framework across independent cohorts. By incorporating simultaneous spatial and temporal clustering, individual prediction intervals, and external replication, this work demonstrates a novel, rigorous, and interpretable framework for characterizing aging heterogeneity.
- 26 June 2026
- Published by Drexel University Libraries
Predicting the risk of Alzheimer's disease and Parkinson's disease using neuroimaging biomarkers and clinical characteristics
Alzheimer's disease (AD) and Parkinson's disease (PD) are the two most prevalent neurodegenerative disorders. These conditions share overlapping features, which can complicate diagnosis. Additionally, many individuals with PD eventually develop Parkinson's disease dementia (PDD). Few studies have comprehensively integrated clinical and neuroimaging data using machine learning to improve the diagnosis of neurodegenerative diseases, and even fewer have employed prospective prediction approaches. This study addresses these gaps by applying supervised machine learning models that incorporate a broad set of clinical characteristics and neuroimaging biomarkers to prospectively predict AD and PD. It also aims to identify and compare the most predictive features for each disease and to evaluate how early these conditions can be distinguished. In parallel, this research uses clinical characteristics and neurocognitive measures to prospectively predict whether individuals with Parkinson's disease without dementia (PDND) will progress to PDD or maintain their current diagnosis. It further seeks to identify features that differentiate these disease trajectories and to assess how early these outcomes can be predicted. Data from the UK Biobank (UKB) were used to support these objectives. The models developed in this study demonstrated the ability to predict future AD, PD, and PDD above chance levels. The AD/PD prediction models showed greater variability and only performed at chance once when stratified into time windows. Feature importance analyses for the best-performing AD/HC, PD/HC, and PDND/PDD models indicated that comorbidities and neurocognitive test scores were key contributors, although their relative importance varied across tasks. This study did not find evidence that predictive performance for future disease risk varied over time for any classification task. Future research should continue to integrate diverse clinical features with multimodal neuroimaging biomarkers in predictive models. Such approaches may improve the early identification of neurodegenerative diseases and support more timely intervention and treatment.
- 26 June 2026
- Published by Drexel University Libraries
Predicting the risk of Alzheimer's disease and Parkinson's disease using neuroimaging biomarkers and clinical characteristics
Alzheimer's disease (AD) and Parkinson's disease (PD) are the two most prevalent neurodegenerative disorders. These conditions share overlapping features, which can complicate diagnosis. Additionally, many individuals with PD eventually develop Parkinson's disease dementia (PDD). Few studies have comprehensively integrated clinical and neuroimaging data using machine learning to improve the diagnosis of neurodegenerative diseases, and even fewer have employed prospective prediction approaches. This study addresses these gaps by applying supervised machine learning models that incorporate a broad set of clinical characteristics and neuroimaging biomarkers to prospectively predict AD and PD. It also aims to identify and compare the most predictive features for each disease and to evaluate how early these conditions can be distinguished. In parallel, this research uses clinical characteristics and neurocognitive measures to prospectively predict whether individuals with Parkinson's disease without dementia (PDND) will progress to PDD or maintain their current diagnosis. It further seeks to identify features that differentiate these disease trajectories and to assess how early these outcomes can be predicted. Data from the UK Biobank (UKB) were used to support these objectives. The models developed in this study demonstrated the ability to predict future AD, PD, and PDD above chance levels. The AD/PD prediction models showed greater variability and only performed at chance once when stratified into time windows. Feature importance analyses for the best-performing AD/HC, PD/HC, and PDND/PDD models indicated that comorbidities and neurocognitive test scores were key contributors, although their relative importance varied across tasks. This study did not find evidence that predictive performance for future disease risk varied over time for any classification task. Future research should continue to integrate diverse clinical features with multimodal neuroimaging biomarkers in predictive models. Such approaches may improve the early identification of neurodegenerative diseases and support more timely intervention and treatment.
- Open Access
Prospective Association Between Weight Variability and Subsequent Long‐Term Weight Loss in the CALERIE Study
Objective: Higher short‐term weight variability (WV) predicts greater subsequent weight gain in weight‐stable adults and toddlers and lower long‐term weight loss among adults with overweight or obesity who previously lost weight. We extend prior work by examining the impact of naturally occurring WV on subsequent weight loss (weight at month 3 minus later weights) in individuals without obesity enrolled in the Comprehensive Assessment of the Long‐term Effects of Reducing Intake of Energy (CALERIE) study. Methods: Participants ( N = 143) followed a 25% caloric restriction diet over 2 years. WV was calculated as the root‐mean‐squared‐error of individuals' weights over the first 12 weeks. Multilevel models examined WV's prediction of subsequent weight loss at 6, 12, 18, and 24 months relative to month 3, adjusting for baseline BMI, 12‐week weight change, and demographic variables. Sensitivity analyses further included time‐varying adherence and eating behaviors (disinhibition, restraint, craving). Results: Higher WV was associated with less subsequent weight loss at 6 months ( B = −0.38, 95% CI [−0.63, −0.13], β = −0.16, p = 0.003). The interaction between time and WV was not significant; analyses suggested the association extends through year 1. Conclusions: Short‐term WV predicted lower subsequent weight loss in adults without obesity. Once its biobehavioral mechanisms are understood, WV might provide insight into humans' innate capacity for long‐term weight stability and inform future mechanistic research and clinical interventions. Trial Registration: ClinicalTrials.gov registration: NCT00427193
Open AccessProspective Association Between Weight Variability and Subsequent Long‐Term Weight Loss in the CALERIE Study
Objective: Higher short‐term weight variability (WV) predicts greater subsequent weight gain in weight‐stable adults and toddlers and lower long‐term weight loss among adults with overweight or obesity who previously lost weight. We extend prior work by examining the impact of naturally occurring WV on subsequent weight loss (weight at month 3 minus later weights) in individuals without obesity enrolled in the Comprehensive Assessment of the Long‐term Effects of Reducing Intake of Energy (CALERIE) study. Methods: Participants ( N = 143) followed a 25% caloric restriction diet over 2 years. WV was calculated as the root‐mean‐squared‐error of individuals' weights over the first 12 weeks. Multilevel models examined WV's prediction of subsequent weight loss at 6, 12, 18, and 24 months relative to month 3, adjusting for baseline BMI, 12‐week weight change, and demographic variables. Sensitivity analyses further included time‐varying adherence and eating behaviors (disinhibition, restraint, craving). Results: Higher WV was associated with less subsequent weight loss at 6 months ( B = −0.38, 95% CI [−0.63, −0.13], β = −0.16, p = 0.003). The interaction between time and WV was not significant; analyses suggested the association extends through year 1. Conclusions: Short‐term WV predicted lower subsequent weight loss in adults without obesity. Once its biobehavioral mechanisms are understood, WV might provide insight into humans' innate capacity for long‐term weight stability and inform future mechanistic research and clinical interventions. Trial Registration: ClinicalTrials.gov registration: NCT00427193
- 29 April 2026
- Published by Wiley in International Journal of Eating Disorders
- Vol. 59 (8) , 1800-1813
Detecting Binge Eating Risk With Naturalistic Data and Machine Learning: A Comparative Observational Study
Objective: Negative affect is a primary antecedent for binge eating (BE). Just‐in‐time adaptive interventions (JITAIs) prompt the use of therapy skills when at risk for maladaptive behaviors. JITAIs show promise for improving emotion regulation skill use and decreasing BE. Identifying momentary negative affect has relied on ecological momentary assessment (EMA). EMA surveys are typically delivered 4–6 times/day via smartphone and are self‐report; accordingly, EMA is limited by temporal granularity, participant insight, and adherence. Wearable sensors which continuously, passively measure physiological correlates of affect may be able to detect risk for negative affect‐related BE while overcoming limitations of EMA. This study compared machine learning models using EMA‐ and sensor‐measured negative affect to predict BE. Method: Thirty adults with recurrent BE wore smartwatches to measure heart rate and electrodermal activity and reported affect and BE on EMA for 4 weeks (preregistration: https://www.researchprotocols.org/2023/1/e47098/ ). Support vector machines, random forest, and neural network models were trained using the first 3 weeks and evaluated using the last week of EMA data, sensor data, and combined EMA and sensor data. Results: The best‐performing EMA‐only model for predicting BE had macro‐averaged accuracy of 0.64 (AUROC = 0.69, sensitivity = 0.47, specificity = 0.81), the sensor‐only model had macro‐averaged accuracy of 0.64 (AUROC = 0.78, sensitivity = 0.92, specificity = 0.35), and the combined model had macro‐averaged accuracy of 0.61 (AUROC = 0.68, sensitivity = 0.90, specificity = 0.33); model macro‐averaged accuracies did not significantly differ. Discussion: Psychophysiological sensor data demonstrate comparable accuracy to EMA for predicting BE, setting the stage for low‐burden JITAIs targeting negative affect‐related BE.
- 29 April 2026
- Published by Wiley in International Journal of Eating Disorders
- Vol. 59 (8) , 1800-1813
Detecting Binge Eating Risk With Naturalistic Data and Machine Learning: A Comparative Observational Study
Objective: Negative affect is a primary antecedent for binge eating (BE). Just‐in‐time adaptive interventions (JITAIs) prompt the use of therapy skills when at risk for maladaptive behaviors. JITAIs show promise for improving emotion regulation skill use and decreasing BE. Identifying momentary negative affect has relied on ecological momentary assessment (EMA). EMA surveys are typically delivered 4–6 times/day via smartphone and are self‐report; accordingly, EMA is limited by temporal granularity, participant insight, and adherence. Wearable sensors which continuously, passively measure physiological correlates of affect may be able to detect risk for negative affect‐related BE while overcoming limitations of EMA. This study compared machine learning models using EMA‐ and sensor‐measured negative affect to predict BE. Method: Thirty adults with recurrent BE wore smartwatches to measure heart rate and electrodermal activity and reported affect and BE on EMA for 4 weeks (preregistration: https://www.researchprotocols.org/2023/1/e47098/ ). Support vector machines, random forest, and neural network models were trained using the first 3 weeks and evaluated using the last week of EMA data, sensor data, and combined EMA and sensor data. Results: The best‐performing EMA‐only model for predicting BE had macro‐averaged accuracy of 0.64 (AUROC = 0.69, sensitivity = 0.47, specificity = 0.81), the sensor‐only model had macro‐averaged accuracy of 0.64 (AUROC = 0.78, sensitivity = 0.92, specificity = 0.35), and the combined model had macro‐averaged accuracy of 0.61 (AUROC = 0.68, sensitivity = 0.90, specificity = 0.33); model macro‐averaged accuracies did not significantly differ. Discussion: Psychophysiological sensor data demonstrate comparable accuracy to EMA for predicting BE, setting the stage for low‐burden JITAIs targeting negative affect‐related BE.
- Open Access
- 12 October 2025
- Published by Association for Computing Machinery (ACM)
- p. 1
Investigating Pace of Biological Aging as a Mental and Physical Health Biomarker in Youth with Perinatally-Acquired HIV
- 12 October 2025
- Published by Association for Computing Machinery (ACM)
- p. 1
Open AccessInvestigating Pace of Biological Aging as a Mental and Physical Health Biomarker in Youth with Perinatally-Acquired HIV
- Open Access
- 8 August 2025
- Published by Elsevier in Contemporary Clinical Trials
- Vol. 156, 108012
Using multiple biomarkers for patient enrichment in two-stage clinical designs
- 8 August 2025
- Published by Elsevier in Contemporary Clinical Trials
- Vol. 156, 108012
Open AccessUsing multiple biomarkers for patient enrichment in two-stage clinical designs
- Open Access
- 16 July 2025
- Published by Elsevier in Developmental Cognitive Neuroscience
- Vol. 75, 101597
Integrating multilevel, multidomain and multimodal neuroimaging factors to predict early alcohol exposure trajectories using explainable AI
- 16 July 2025
- Published by Elsevier in Developmental Cognitive Neuroscience
- Vol. 75, 101597
Open AccessIntegrating multilevel, multidomain and multimodal neuroimaging factors to predict early alcohol exposure trajectories using explainable AI
- Open Access
Integrating Multilevel, Multidomain and Multimodal Neuroimaging Factors to Predict Early Alcohol Exposure Trajectories Using Explainable AI
Alcohol consumption tends to increase from childhood to adolescence, and risk factors at the individual, family, and environmental level (multilevel, multidomain factors), as well as changes in brain structure and function, have been associated with the likelihood of developing alcohol use disorder (AUD) or binge drinking later in life. Most studies have focused on limited subsets of multilevel or neuroimaging factors, typically emphasizing the risk of alcohol initiation, binge drinking, or AUD in cross-sectional designs rather than exploring longitudinal alcohol consumption trajectories. Our study addresses these gaps by examining a comprehensive set of multilevel, multidomain factors and multimodal brain imaging features to prospectively predict early alcohol sipping trajectories over time with large data from the Adolescent Brain Cognitive Development Study. We applied machine learning methods to baseline individual, family, and environmental factors as well as structural and functional brain connectivity features, analyzing these predictors separately and in combination. Key findings reveal that functional connectivity and multilevel factors distinguish youth with an increasing alcohol sipping trajectory from those who initially experimented with alcohol but reduced their consumption over time. Moreover, we found important structural and functional features that predicted those who increasingly sipped over time versus the ones who did not engage in alcohol experimentation. Stable interactions between age, socioeconomical status and positive attitudes towards drinking could predict a pattern of increasing alcohol sipping over time. These trends could inform how individual, family, and environmental factors together with brain imaging features impact the development of different alcohol sipping trajectories over time.
Open AccessIntegrating Multilevel, Multidomain and Multimodal Neuroimaging Factors to Predict Early Alcohol Exposure Trajectories Using Explainable AI
Alcohol consumption tends to increase from childhood to adolescence, and risk factors at the individual, family, and environmental level (multilevel, multidomain factors), as well as changes in brain structure and function, have been associated with the likelihood of developing alcohol use disorder (AUD) or binge drinking later in life. Most studies have focused on limited subsets of multilevel or neuroimaging factors, typically emphasizing the risk of alcohol initiation, binge drinking, or AUD in cross-sectional designs rather than exploring longitudinal alcohol consumption trajectories. Our study addresses these gaps by examining a comprehensive set of multilevel, multidomain factors and multimodal brain imaging features to prospectively predict early alcohol sipping trajectories over time with large data from the Adolescent Brain Cognitive Development Study. We applied machine learning methods to baseline individual, family, and environmental factors as well as structural and functional brain connectivity features, analyzing these predictors separately and in combination. Key findings reveal that functional connectivity and multilevel factors distinguish youth with an increasing alcohol sipping trajectory from those who initially experimented with alcohol but reduced their consumption over time. Moreover, we found important structural and functional features that predicted those who increasingly sipped over time versus the ones who did not engage in alcohol experimentation. Stable interactions between age, socioeconomical status and positive attitudes towards drinking could predict a pattern of increasing alcohol sipping over time. These trends could inform how individual, family, and environmental factors together with brain imaging features impact the development of different alcohol sipping trajectories over time.