Defense Date


Document Type


Degree Name

Doctor of Philosophy



First Advisor

Roy T. Sabo, Ph.D.

Second Advisor

Robert Perera, Ph.D.

Third Advisor

Nitai Mukhopadhyay, Ph.D.

Fourth Advisor

Amir Toor, M.D.

Fifth Advisor

Robert Kirkpatrick, Ph.D.


Predicting the trajectory of lymphoid recovery following myeloablative hematopoietic stem cell transplantation (SCT) can help guide subsequent therapeutic decisions, since poor recovery has been associated with graft-versus-host disease (GVHD), relapse and mortality. Previous attempts at classifying patients depended on absolute criteria being set prior to modeling absolute lymphocyte counts (ALCs) over time. Having an empirical clinical decision support tool for objectively determining the trajectory an individual might take during their recovery would be advantageous. We propose using growth-based trajectory modeling (GBTM) and growth mixture modeling (GMM), which utilize machine learning algorithms to empirically identify latent groupings of data. Due to lymphocyte reconstitution having complicated and decidedly non-linear longitudinal trends, cubic B-splines were used in place of the polynomials traditionally used in the GBTM and GMM approaches due to their flexibility and ability to fit complex trends. Our methods rely on objective measurements to prospectively predict the course an individual might take following stem cell transplantation. The adapted GBTM and GMM methods classified existing patients into groups with varying rates and magnitudes of lymphoid recovery into distinct groups of trajectories and we demonstrated how this method could be used in practice to prospectively identify lymphoid recovery of a new patient. Our method has the potential to aid clinicians in identifying patients who may be predisposed to varying clinical outcomes based on immune reconstitution, so the immunosuppression intensity may be appropriately adjusted.


© Brielle A. Forsthoffer

Is Part Of

VCU University Archives

Is Part Of

VCU Theses and Dissertations

Date of Submission


Available for download on Wednesday, July 29, 2026

Included in

Biostatistics Commons