MRI Reliably Captures Bone Marrow Metrics in Myelofibrosis

Purpose

Biopsy remains the clinical standard for evaluating bone marrow of patients with myelofibrosis (MF). However, randomly sampling a small volume of bone marrow, typically from the ilium, fails to capture heterogeneity of disease across anatomic sites. The invasive, painful nature of the procedure limits patient acceptance of repeated biopsies to monitor disease status and response to therapy. Here, we present a non-invasive, MRI technique for evaluating bone marrow in MF.

Participants and Methods

We analyzed bone marrow of 66 study participants (45 with MF; 15 with essential thrombocytosis or polycythemia vera; and 6 healthy controls). We assessed participant bone marrow across three anatomic sites (vertebral bodies, ilium, and femoral heads), using three quantitative MRI sequences: proton density fat fraction (PDFF) for fat content; apparent diffusion coefficient (ADC) for cellular effect on water mobility; and magnetization transfer (MT) for macromolecular structure and fiber deposition. We correlated these MRI metrics with patient prognosis (dynamic international prognostic scoring system; DIPSS) and iliac biopsy data for cellularity and fibrosis (MF grade).

Results

Participants with MF had elevated MT ratio (MTR) and ADC and lower PDFF compared to healthy participants. Non-MF MPN participants demonstrated the same trend, though to a lesser extent. Individual MRI metrics correlated strongly across anatomic sites (r_pearson from 0.57 to 0.89, p-value < 0.01), with notable heterogeneity both within and among different regions. DIPSS risk groups had limited correlation with MRI metrics. To understand to what extent quantitative MRI biomarkers identify fibrotic marrow, we developed a multivariate logistic regression model to stratify the 45 participants with MF by early fibrosis (MF grades 0-1) and overt fibrosis (MF grades 2-3). Of the seven total anatomy-MRI metric combinations, three had the greatest contributions toward grading marrow fibrosis: PDFF in the vertebral bodies and ilium along with ADC in the ilium. We developed the logistic regression model to predict early and overt fibrosis with these three parameters using a 70%/30% train/test split with 5-fold cross validation. Based on this model, quantitative MRI metrics are associated with the severity of fibrosis in the bone marrow (test set performance: accuracy = 84.6%, sensitivity to predict overt fibrosis = 88.9%, specificity = 75%, area under the receiver operator curve = 0.94).

Conclusion

Quantitative bone marrow MRI reliably captures relevant metrics of MF bone marrow, including cellularity and fat replacement by fibrosis, and may be useful in clinical drug trials and patient management for MF.