Predicting Survival Outcomes With PET Radiomics in Diffuse Large B-Cell Lymphoma

By: Julia Fiederlein
Posted: Tuesday, May 31, 2022

Luca Ceriani, PhD, of Ente Ospedaliero Cantonale, Bellinzona, Switzerland, and colleagues investigated the prognostic value of PET radiomics in patients with diffuse large B-cell lymphoma who were treated with rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisolone (R-CHOP). The post hoc analysis of the SAKK 38/07 trial, which was published in Hematological Oncology, provided proof of principle for the application of radiomic models in this clinical setting.

“The radiomics score allowed [for the] risk classification of patients with significantly different progression-free, cause-specific, and overall survival in both cohorts [and showed] better predictive accuracy [with] respect to clinical international indices,” the investigators remarked. “PET-derived radiomics may improve the prediction of outcomes in patients with diffuse large B-cell lymphoma.”

A total of 107 radiomics features were extracted from the baseline PET scans of 133 patients who were treated with R-CHOP every 14 days in the SAKK 38/07 trial. A least absolute shrinkage and selection operator Cox regression analysis was performed to assess the impact of the international prognostic indices; main clinical parameters; standard PET metrics; and 52 radiomics uncorrelated features on progression-free, cause-specific, and overall survival outcomes. A linear combination of the resulting parameters generated the prognostic radiomics score. The efficacy of the radiomics score was validated in an independent cohort of 107 patients who received R-CHOP every 21 days.

The investigators identified four radiomics features that seemed to predict progression-free survival outcomes in the SAKK 38/07 trial. In both the testing (P < .001) and validation (P < .001) sets, the derived radiomics score demonstrated a significant capability to predict progression-free survival outcomes. The radiomics score also seemed to be significantly associated with cause-specific and overall survival in the testing (P < .001 and P < .001, respectively) and validation (P < .0001 and P = .004, respectively) sets.

Disclosure: The study authors reported no conflicts of interest.

Hematological Oncology

---