Survival and Treatment Response in CTCL Can be Predicted Using Volumetric Lymph Node Size and Dynamic Changes in Lymph Node Size and Skin Score

Background

Lymph node (LN) involvement is a well-established prognostic indicator in cutaneous T-cell lymphoma (CTCL); however, current radiologic criteria, which use a cutoff of 1.5 cm for the longest LN diameter (LDi), are based on data from nodal lymphomas and have not been validated in CTCL. Recently, volumetric tumor measurement has emerged as a novel predictor of survival and treatment response outcomes in a variety of cancer subtypes. Additionally, tumor kinetic modeling, which simultaneously describes tumor growth (g) and regression (d), has emerged as a promising computational method for predicting outcomes in cancer patients.

The Aims of this study were:

1. To create a new, validated lymph node assessment method for patients with CTCL using novel LN volumetric measurements correlated with survival outcomes
2. To apply kinetic modeling to volumetric LN data and clinically-observed skin involvement in patients with CTCL, to determine patients’ prognoses and predict response to therapy.

Methods

We modeled LN assessment from an open-label randomized controlled phase 3 trial testing mogamulizumab versus vorinostat in previously treated CTCL (MAVORIC trial), in which patient data was collected prospectively using serially-collected mSWAT (modified Severity-Weighted Assessment Tool) data and CT scans. A total of 370 eligible patients were randomly assigned to receive mogamulizumab (n=184) or vorinostat (n=186), comprising the intention-to-treat population.

For each patient included in the study, baseline and follow-up CT scans were analyzed using the Weasis imaging platform, with the LNs deemed to be the most enlarged or morphologically abnormal undergoing a semi-automated segmentation process to determine volumes: quality assurance studies were performed for each target node, followed by calculation of uni/bidimensional and volumetric measurements; an algorithm then matched LNs across scans at various time points, with the resulting data saved to a central database. LN volume was correlated survival data using the Kaplan-Meier method. To determine the volume cutoff that best stratified patients by OS, receiver operating characteristic (ROS) analysis was applied using Harrell’s concordance statistic.

Kinetic modeling was applied to each patient, thereby estimating g and d for LN volume and mSWAT at each time point. Patients were divided into two groups (for LN volume) and four groups (for mSWAT scores) based on g and correlation with overall survival (OS), time to treatment failure (TTF), and progression-free survival (PFS) were evaluated using Kaplan-Meier curves. Tumor doubling times were estimated using the equation dt = 0.693/g.

Results

We demonstrated that the current standard-of-care cutoff of LDi >1.5 cm does not correlate with OS when using standard assessment criteria. Volumetric LN measurements, however, were found to predict OS for a cohort of baseline CT scans from 119 patients, with the most appropriate volume cutoff being 3900 mm3. This cutoff showed similar predictive value when patient data was adjusted for disease stage. Finally, the cutoff was validated with a cohort of 161 patients, comprising 246 additional scans; the previous analysis was repeated and the cutoff of 3900 mm3 was found to correlate with OS.

When analyzing the same cohort of patient data using kinetic modeling, it was found that volumetric LN g is significantly associated with OS and TTF. Additionally, g for skin involvement (mSWAT) was determined to be significantly associated with PFS and TTF. Overall, these results show that static volumetric LN measurements, dynamic changes in LN size, and dynamic changes in mSWAT score can all be used as novel prognostic tools for patients with CTCL.