Hillary M. Heiling
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
Hillary M. Heiling has a total of 154 co-authors. Here are the top co-authors by number of shared publications:
- Allison Mary Deal
Lineberger Comprehensive Cancer Center, University of North Carolina at Chapel Hill, Chapel Hill, NC
5 shared publications Naim U. Rashid
Department of Biostatistics University of North Carolina at Chapel Hill Chapel Hill North Carolina USA
4 shared publicationsJen Jen Yeh
Lineberger Comprehensive Cancer Center University of North Carolina at Chapel Hill Chapel Hill North Carolina USA
3 shared publications- Joseph G. Ibrahim
University of North Carolina at Chapel Hill
3 shared publications - Quefeng Li
University of North Carolina at Chapel Hill
3 shared publications - Adam Brufsky
University of Pittsburgh, Pittsburgh, PA
2 shared publications
Latest Publications
- Open Access
- 20 July 2026
- Published by Elsevier in The Breast
- Vol. 89, 104881
Patient-reported outcomes from the TBCRC 022 study of neratinib and ado-trastuzumab emtansine for HER2-positive breast cancer brain metastases
- 20 July 2026
- Published by Elsevier in The Breast
- Vol. 89, 104881
Open AccessPatient-reported outcomes from the TBCRC 022 study of neratinib and ado-trastuzumab emtansine for HER2-positive breast cancer brain metastases
- 10 March 2025
- Published by Wiley in Statistics in Medicine
- Vol. 44 (6) , e10311
Efficient Computation of High‐Dimensional Penalized Piecewise Constant Hazard Random Effects Models
Identifying and characterizing relationships between treatments, exposures, or other covariates and time‐to‐event outcomes has great significance in a wide range of biomedical settings. In research areas such as multi‐center clinical trials, recurrent events, and genetic studies, proportional hazard mixed effects models (PHMMs) are used to account for correlations observed in clusters within the data. In high dimensions, proper specification of the fixed and random effects within PHMMs is difficult and computationally complex. In this paper, we approximate the proportional hazards mixed effects model with a piecewise constant hazard mixed effects survival model. We estimate the model parameters using a modified Monte Carlo expectation conditional minimization (MCECM) algorithm, allowing us to perform variable selection on both the fixed and random effects simultaneously. We also incorporate a factor model decomposition of the random effects in order to more easily scale the variable selection method to larger dimensions. We demonstrate the utility of our method using simulations, and we apply our method to a multi‐study pancreatic ductal adenocarcinoma gene expression dataset to select features important for survival.
- 10 March 2025
- Published by Wiley in Statistics in Medicine
- Vol. 44 (6) , e10311
Efficient Computation of High‐Dimensional Penalized Piecewise Constant Hazard Random Effects Models
Identifying and characterizing relationships between treatments, exposures, or other covariates and time‐to‐event outcomes has great significance in a wide range of biomedical settings. In research areas such as multi‐center clinical trials, recurrent events, and genetic studies, proportional hazard mixed effects models (PHMMs) are used to account for correlations observed in clusters within the data. In high dimensions, proper specification of the fixed and random effects within PHMMs is difficult and computationally complex. In this paper, we approximate the proportional hazards mixed effects model with a piecewise constant hazard mixed effects survival model. We estimate the model parameters using a modified Monte Carlo expectation conditional minimization (MCECM) algorithm, allowing us to perform variable selection on both the fixed and random effects simultaneously. We also incorporate a factor model decomposition of the random effects in order to more easily scale the variable selection method to larger dimensions. We demonstrate the utility of our method using simulations, and we apply our method to a multi‐study pancreatic ductal adenocarcinoma gene expression dataset to select features important for survival.
- Open Access
- 6 July 2024
- Published by Elsevier in Annals of Oncology
- Vol. 35 (11) , 993-1002
Neratinib and ado-trastuzumab emtansine for pretreated and untreated human epidermal growth factor receptor 2 (HER2)-positive breast cancer brain metastases: Translational Breast Cancer Research Consortium trial 022
- 6 July 2024
- Published by Elsevier in Annals of Oncology
- Vol. 35 (11) , 993-1002
Open AccessNeratinib and ado-trastuzumab emtansine for pretreated and untreated human epidermal growth factor receptor 2 (HER2)-positive breast cancer brain metastases: Translational Breast Cancer Research Consortium trial 022
- Open Access
glmmPen: High Dimensional Penalized Generalized Linear Mixed Models
Generalized linear mixed models (GLMMs) are widely used in research for their ability to model correlated outcomes with non-Gaussian conditional distributions. The proper selection of fixed and random effects is a critical part of the modeling process since model misspecification may lead to significant bias. However, the joint selection of fixed and random effects has historically been limited to lower-dimensional GLMMs, largely due to the use of criterion-based model selection strategies. Here we present the R package glmmPen, one of the first to select fixed and random effects in higher dimension using a penalized GLMM modeling framework. Model parameters are estimated using a Monte Carlo Expectation Conditional Minimization (MCECM) algorithm, which leverages Stan and RcppArmadillo for increased computational efficiency. Our package supports the Binomial, Gaussian, and Poisson families and multiple penalty functions. In this manuscript we discuss the modeling procedure, estimation scheme, and software implementation through application to a pancreatic cancer subtyping study. Simulation results show our method has good performance in selecting both the fixed and random effects in high dimensional GLMMs.
Open AccessglmmPen: High Dimensional Penalized Generalized Linear Mixed Models
Generalized linear mixed models (GLMMs) are widely used in research for their ability to model correlated outcomes with non-Gaussian conditional distributions. The proper selection of fixed and random effects is a critical part of the modeling process since model misspecification may lead to significant bias. However, the joint selection of fixed and random effects has historically been limited to lower-dimensional GLMMs, largely due to the use of criterion-based model selection strategies. Here we present the R package glmmPen, one of the first to select fixed and random effects in higher dimension using a penalized GLMM modeling framework. Model parameters are estimated using a Monte Carlo Expectation Conditional Minimization (MCECM) algorithm, which leverages Stan and RcppArmadillo for increased computational efficiency. Our package supports the Binomial, Gaussian, and Poisson families and multiple penalty functions. In this manuscript we discuss the modeling procedure, estimation scheme, and software implementation through application to a pancreatic cancer subtyping study. Simulation results show our method has good performance in selecting both the fixed and random effects in high dimensional GLMMs.
- Open Access
Efficient Computation of High-Dimensional Penalized Generalized Linear Mixed Models by Latent Factor Modeling of the Random Effects
Modern biomedical datasets are increasingly high dimensional and exhibit complex correlation structures. Generalized Linear Mixed Models (GLMMs) have long been employed to account for such dependencies. However, proper specification of the fixed and random effects in GLMMs is increasingly difficult in high dimensions, and computational complexity grows with increasing dimension of the random effects. We present a novel reformulation of the GLMM using a factor model decomposition of the random effects, enabling scalable computation of GLMMs in high dimensions by reducing the latent space from a large number of random effects to a smaller set of latent factors. We also extend our prior work to estimate model parameters using a modified Monte Carlo Expectation Conditional Minimization algorithm, allowing us to perform variable selection on both the fixed and random effects simultaneously. We show through simulation that through this factor model decomposition, our method can fit high dimensional penalized GLMMs faster than comparable methods and more easily scale to larger dimensions not previously seen in existing approaches.
Open AccessEfficient Computation of High-Dimensional Penalized Generalized Linear Mixed Models by Latent Factor Modeling of the Random Effects
Modern biomedical datasets are increasingly high dimensional and exhibit complex correlation structures. Generalized Linear Mixed Models (GLMMs) have long been employed to account for such dependencies. However, proper specification of the fixed and random effects in GLMMs is increasingly difficult in high dimensions, and computational complexity grows with increasing dimension of the random effects. We present a novel reformulation of the GLMM using a factor model decomposition of the random effects, enabling scalable computation of GLMMs in high dimensions by reducing the latent space from a large number of random effects to a smaller set of latent factors. We also extend our prior work to estimate model parameters using a modified Monte Carlo Expectation Conditional Minimization algorithm, allowing us to perform variable selection on both the fixed and random effects simultaneously. We show through simulation that through this factor model decomposition, our method can fit high dimensional penalized GLMMs faster than comparable methods and more easily scale to larger dimensions not previously seen in existing approaches.
- 1 March 2024
- Published by Oxford University Press (OUP) in European Journal of Cardio-Thoracic Surgery
- Vol. 65 (3)
Complete anatomic segmentectomy shows improved oncologic outcomes compared to incomplete anatomic segmentectomy
To compare oncologic outcomes after segmentectomy with division of segmental bronchus, artery, and vein (complete anatomic segmentectomy) versus segmentectomy with division of < 3 segmental structures (incomplete anatomic segmentectomy). We conducted a single-center, retrospective analysis of patients undergoing segmentectomy from 03/2005-05/2020. Operative reports were audited to classify procedures as complete or incomplete anatomic segmentectomy. Patients who underwent neoadjuvant therapy or pulmonary resection beyond indicated segments were excluded. Survival was estimated with Kaplan-Meier models and compared using log-rank tests. Cox proportional hazards models were used to estimate hazard ratios for death. Cumulative incidence functions for locoregional recurrence were compared with Gray’s test, with death considered a competing event. Cox and Fine-Gray models were used to estimate cause-specific and subdistribution hazard ratios, respectively, for locoregional recurrence. Of 390 cases, 266 (68.2%) were complete and 124 were incomplete anatomic segmentectomy. Demographics, pulmonary function, tumor size, stage, and perioperative outcomes did not significantly differ between groups. Surgical margins were negative in all but one case. Complete anatomic segmentectomy was associated with improved lymph node dissection (5 vs. 2 median nodes sampled; p<0.001). Multivariable analysis revealed reduced incidence of locoregional recurrence (CSHR=0.42; 95% CI 0.22-0.80; SDHR=0.43; 95% CI 0.23-0.81), and non-significant improvement in overall survival (HR = 0.66; 95% CI: 0.43-1.00) after complete vs. incomplete anatomic segmentectomy. This single-center experience suggests complete anatomic segmentectomy provides superior locoregional control and may improve survival relative to incomplete anatomic segmentectomy. We recommend surgeons perform complete anatomic segmentectomy and lymph node dissection whenever possible.
- 1 March 2024
- Published by Oxford University Press (OUP) in European Journal of Cardio-Thoracic Surgery
- Vol. 65 (3)
Complete anatomic segmentectomy shows improved oncologic outcomes compared to incomplete anatomic segmentectomy
To compare oncologic outcomes after segmentectomy with division of segmental bronchus, artery, and vein (complete anatomic segmentectomy) versus segmentectomy with division of < 3 segmental structures (incomplete anatomic segmentectomy). We conducted a single-center, retrospective analysis of patients undergoing segmentectomy from 03/2005-05/2020. Operative reports were audited to classify procedures as complete or incomplete anatomic segmentectomy. Patients who underwent neoadjuvant therapy or pulmonary resection beyond indicated segments were excluded. Survival was estimated with Kaplan-Meier models and compared using log-rank tests. Cox proportional hazards models were used to estimate hazard ratios for death. Cumulative incidence functions for locoregional recurrence were compared with Gray’s test, with death considered a competing event. Cox and Fine-Gray models were used to estimate cause-specific and subdistribution hazard ratios, respectively, for locoregional recurrence. Of 390 cases, 266 (68.2%) were complete and 124 were incomplete anatomic segmentectomy. Demographics, pulmonary function, tumor size, stage, and perioperative outcomes did not significantly differ between groups. Surgical margins were negative in all but one case. Complete anatomic segmentectomy was associated with improved lymph node dissection (5 vs. 2 median nodes sampled; p<0.001). Multivariable analysis revealed reduced incidence of locoregional recurrence (CSHR=0.42; 95% CI 0.22-0.80; SDHR=0.43; 95% CI 0.23-0.81), and non-significant improvement in overall survival (HR = 0.66; 95% CI: 0.43-1.00) after complete vs. incomplete anatomic segmentectomy. This single-center experience suggests complete anatomic segmentectomy provides superior locoregional control and may improve survival relative to incomplete anatomic segmentectomy. We recommend surgeons perform complete anatomic segmentectomy and lymph node dissection whenever possible.
- Open Access
- 29 January 2024
- Published by Oxford University Press (OUP) in Biometrics
- Vol. 80 (1)
Efficient computation of high-dimensional penalized generalized linear mixed models by latent factor modeling of the random effects
Modern biomedical datasets are increasingly high-dimensional and exhibit complex correlation structures. Generalized linear mixed models (GLMMs) have long been employed to account for such dependencies. However, proper specification of the fixed and random effects in GLMMs is increasingly difficult in high dimensions, and computational complexity grows with increasing dimension of the random effects. We present a novel reformulation of the GLMM using a factor model decomposition of the random effects, enabling scalable computation of GLMMs in high dimensions by reducing the latent space from a large number of random effects to a smaller set of latent factors. We also extend our prior work to estimate model parameters using a modified Monte Carlo Expectation Conditional Minimization algorithm, allowing us to perform variable selection on both the fixed and random effects simultaneously. We show through simulation that through this factor model decomposition, our method can fit high-dimensional penalized GLMMs faster than comparable methods and more easily scale to larger dimensions not previously seen in existing approaches.
- 29 January 2024
- Published by Oxford University Press (OUP) in Biometrics
- Vol. 80 (1)
Open AccessEfficient computation of high-dimensional penalized generalized linear mixed models by latent factor modeling of the random effects
Modern biomedical datasets are increasingly high-dimensional and exhibit complex correlation structures. Generalized linear mixed models (GLMMs) have long been employed to account for such dependencies. However, proper specification of the fixed and random effects in GLMMs is increasingly difficult in high dimensions, and computational complexity grows with increasing dimension of the random effects. We present a novel reformulation of the GLMM using a factor model decomposition of the random effects, enabling scalable computation of GLMMs in high dimensions by reducing the latent space from a large number of random effects to a smaller set of latent factors. We also extend our prior work to estimate model parameters using a modified Monte Carlo Expectation Conditional Minimization algorithm, allowing us to perform variable selection on both the fixed and random effects simultaneously. We show through simulation that through this factor model decomposition, our method can fit high-dimensional penalized GLMMs faster than comparable methods and more easily scale to larger dimensions not previously seen in existing approaches.
- Open Access
- 1 December 2023
- Published by The R Foundation in The R Journal
- Vol. 15 (4) , 106-128
glmmPen: High Dimensional Penalized Generalized Linear Mixed Models
Generalized linear mixed models (GLMMs) are widely used in research for their ability to model correlated outcomes with non-Gaussian conditional distributions. The proper selection of fixed and random effects is a critical part of the modeling process, where model misspecification may lead to significant bias. However, the joint selection of fixed and random effects has historically been limited to lower dimensional GLMMs, largely due to the use of criterion-based model selection strategies. Here we present the R package glmmPen, one of the first to select fixed and random effects in higher dimension using a penalized GLMM modeling framework. Model parameters are estimated using a Monte Carlo expectation conditional minimization (MCECM) algorithm, which leverages Stan and RcppArmadillo for increased computational efficiency. Our package supports the Binomial, Gaussian, and Poisson families and multiple penalty functions. In this manuscript we discuss the modeling procedure, estimation scheme, and software implementation through application to a pancreatic cancer subtyping study. Simulation results show our method has good performance in selecting both the fixed and random effects in high dimensional GLMMs.
- 1 December 2023
- Published by The R Foundation in The R Journal
- Vol. 15 (4) , 106-128
Open AccessglmmPen: High Dimensional Penalized Generalized Linear Mixed Models
Generalized linear mixed models (GLMMs) are widely used in research for their ability to model correlated outcomes with non-Gaussian conditional distributions. The proper selection of fixed and random effects is a critical part of the modeling process, where model misspecification may lead to significant bias. However, the joint selection of fixed and random effects has historically been limited to lower dimensional GLMMs, largely due to the use of criterion-based model selection strategies. Here we present the R package glmmPen, one of the first to select fixed and random effects in higher dimension using a penalized GLMM modeling framework. Model parameters are estimated using a Monte Carlo expectation conditional minimization (MCECM) algorithm, which leverages Stan and RcppArmadillo for increased computational efficiency. Our package supports the Binomial, Gaussian, and Poisson families and multiple penalty functions. In this manuscript we discuss the modeling procedure, estimation scheme, and software implementation through application to a pancreatic cancer subtyping study. Simulation results show our method has good performance in selecting both the fixed and random effects in high dimensional GLMMs.
- Open Access
- 14 June 2023
- Published by Wiley in Cancer Medicine
- Vol. 12 (15) , 16502-16516
Physical activity maintenance among young adult cancer survivors in an mHealth intervention: Twelve‐month outcomes from the IMPACT randomized controlled trial
Background: Most physical activity (PA) interventions in young adult cancer survivors (YACS) have focused on short‐term outcomes without evaluating longer‐term outcomes and PA maintenance. This study examined the effects of an mHealth PA intervention at 12 months, after 6 months of tapered contacts, relative to a self‐help group among 280 YACS.Methods: YACS participated in a 12‐month randomized trial that compared self‐help and intervention groups. All participants received an activity tracker, smart scale, individual videochat session, and access to a condition‐specific Facebook group. Intervention participants also received lessons, tailored feedback, adaptive goal setting, text messages, and Facebook prompts for 6 months, followed by tapered contacts. Accelerometer‐measured and self‐reported PA (total [primary outcome], moderate‐to‐vigorous [MVPA], light, steps, sedentary behaviors) were collected at baseline, 6, and 12 months. Generalized estimating equation analyses evaluated group effects on outcomes from baseline to 12 months.Results: From baseline to 12 months, there were no between‐ or within‐group differences in accelerometer‐measured total PA min/week, while increases in self‐reported total PA were greater in the intervention versus self‐help group (mean difference = +55.8 min/week [95% CI, 6.0–105.6], p = 0.028). Over 12 months, both groups increased accelerometer‐measured MVPA (intervention: +22.5 min/week [95% CI, 8.8–36.2] vs. self‐help: +13.9 min/week [95% CI, 3.0–24.9]; p = 0.34), with no between‐group differences. Both groups maintained accelerometer‐measured and self‐reported PA (total, MVPA) from 6 to 12 months. At 12 months, more intervention participants reported meeting national PA guidelines than self‐help participants (47.9% vs. 33.1%, RR = 1.45, p = 0.02).Conclusion: The intervention was not more effective than the self‐help group at increasing accelerometer‐measured total PA over 12 months. Both groups maintained PA from 6 to 12 months. Digital approaches have potential for promoting sustained PA participation in YACS, but additional research is needed to identify what strategies work for whom, and under what conditions.
- 14 June 2023
- Published by Wiley in Cancer Medicine
- Vol. 12 (15) , 16502-16516
Open AccessPhysical activity maintenance among young adult cancer survivors in an mHealth intervention: Twelve‐month outcomes from the IMPACT randomized controlled trial
Background: Most physical activity (PA) interventions in young adult cancer survivors (YACS) have focused on short‐term outcomes without evaluating longer‐term outcomes and PA maintenance. This study examined the effects of an mHealth PA intervention at 12 months, after 6 months of tapered contacts, relative to a self‐help group among 280 YACS.Methods: YACS participated in a 12‐month randomized trial that compared self‐help and intervention groups. All participants received an activity tracker, smart scale, individual videochat session, and access to a condition‐specific Facebook group. Intervention participants also received lessons, tailored feedback, adaptive goal setting, text messages, and Facebook prompts for 6 months, followed by tapered contacts. Accelerometer‐measured and self‐reported PA (total [primary outcome], moderate‐to‐vigorous [MVPA], light, steps, sedentary behaviors) were collected at baseline, 6, and 12 months. Generalized estimating equation analyses evaluated group effects on outcomes from baseline to 12 months.Results: From baseline to 12 months, there were no between‐ or within‐group differences in accelerometer‐measured total PA min/week, while increases in self‐reported total PA were greater in the intervention versus self‐help group (mean difference = +55.8 min/week [95% CI, 6.0–105.6], p = 0.028). Over 12 months, both groups increased accelerometer‐measured MVPA (intervention: +22.5 min/week [95% CI, 8.8–36.2] vs. self‐help: +13.9 min/week [95% CI, 3.0–24.9]; p = 0.34), with no between‐group differences. Both groups maintained accelerometer‐measured and self‐reported PA (total, MVPA) from 6 to 12 months. At 12 months, more intervention participants reported meeting national PA guidelines than self‐help participants (47.9% vs. 33.1%, RR = 1.45, p = 0.02).Conclusion: The intervention was not more effective than the self‐help group at increasing accelerometer‐measured total PA over 12 months. Both groups maintained PA from 6 to 12 months. Digital approaches have potential for promoting sustained PA participation in YACS, but additional research is needed to identify what strategies work for whom, and under what conditions.