Correlation of Clinical Features to DNA Methylation-Based Prognostic Subtypes in Chordoma Patients

Background: Chordomas are rare aggressive primary bone cancers affecting the skull-base and spine. Despite standard of care treatment with radical surgical resection and radiotherapy, the overall survival at 10 years is only 40%, with most patients experiencing disease recurrence. Moreover, many experience devastating neurological morbidity that significantly impairs quality of life. Much of the research on chordomas points to a dichotomization in outcomes with some experiencing rapid recurrence and mortality despite maximal treatment, while others experience progression-free survival of up to 10 to 20 years. We have previously identified robust DNA methylation-based prognostic subtypes of chordomas, a poorer performing immune-infiltrated subtype and a better performing cellular subtype. Here, we have aimed to further characterize these prognostic subtypes with an extensively annotated clinical database to potentially identify factors that correlate with each subtype and, therefore, can be used as markers of chordoma subtype.

Methods: A total of 68 patients from a multi-institutional 20-year series were identified. These patients’ tumor samples had undergone whole genome DNA methylation profiling on the Illumina EPIC array. Extensive clinical data was extracted for all patients. Baseline clinical features, including clinical features (age, sex, pain, or neurological deficit at presentation, tumor size and diameter, tumor location), treatment details (extent of resection, complications, histological subtype, adjuvant radiotherapy), outcomes parameters (recurrence, metastasis, status of disease control) and imaging parameters (dural invasion, vascularity, soft-tissue extension, bony destruction) were analyzed using chi-squared or Kruskal–Wallis test.

Results: Of all the variables tested, age of onset (p-value = 0.012), location (skull base vs. spine vs. sacral: p-value = 0.0365) and histological subtype (classical vs. chondroid: p-value = 0.0132) were the only significant predictors of subgroup placement; older age, spinal location, and classical histological typing were predictors of immune infiltrated chordoma subtype, which has a poorer clinical performance. Sex, tumor size, degree of neurological deficit upon presentation, treatment experience such as extent of resection, complications or use of radiotherapy did not differ between the two groups. As expected, death from chordoma was significantly different between subtypes (p-value = 0.0056). Notably, imaging characteristics of the tumor did not correlate with subtype.

Conclusion: Overall, there are limited variables that correlate with methylation subtype after thorough assessment of clinical factors, meaning that the newly identified epigenetic chordoma subtypes cannot be reliably identified using clinical or imaging features of the patient and their tumor alone in the absence of molecular data. This is not unexpected as historically it has been very difficult to prognosticate chordomas using clinical factors and further underscores the value of molecular subgroups for this devastating disease. Currently, we are developing a model that utilizes both the DNA methylation subtype of chordoma along with clinical factors that were significant in this analysis to provide comprehensive composite 5-year and 10-year disease-specific survival risk values. We expect that chordoma prognostication using a multifactorial model considering DNA methylation subtype and clinical factors may allow us to most reliable prognosticate patients and thereby best tailor their treatment to their disease.

Publikationsverlauf

Artikel online veröffentlicht:
01. Februar 2023