medwireNews: Methylation levels of cell-free DNA in pretreatment plasma samples from people with muscle-invasive bladder cancer may predict pathologic response to neoadjuvant chemotherapy, suggests an analysis of prospective SWOG S1314 data.
Thomas Flaig, from the University of Colorado Cancer Center in Aurora, USA, presented the findings at the 2022 ASCO Annual Meeting in Chicago, Illinois, USA.
For the study, his team used the Infinium MethylationEPIC BeadChip array to interrogate over 850,000 methylation sites across the genome in plasma samples taken from 72 patients with urothelial carcinoma (T2 to T4a, N0, M0) on day 1 of cycle 1 of neoadjuvant chemotherapy with either gemcitabine plus cisplatin or dose-dense methotrexate, vinblastine, doxorubicin, and cisplatin.
This showed that no single methylation locus correlated with response at radical cystectomy, but the 37 responders clustered separately to the 35 nonresponders when the 500 most differentially methylated DNA loci were considered.
Using machine learning, the researchers created a methylation-based neoadjuvant chemotherapy response score (mR-score) which significantly correlated with response to both regimens. Overall, the score predicted response with 63.6% accuracy.
And this increased to 72.0% when the score was applied to 57 on-treatment samples taken before the start of cycle 2.
Flaig and colleagues then combined the pretreatment and on-treatment mR-scores with circulating bladder DNA levels, which they found also correlated with the depth of response and were independent of mR-score, to create a risk stratification model.
Together, this risk stratification model correctly predicted neoadjuvant chemotherapy response for 79% of patients.
Flaig therefore concluded: “Machine learning can be applied to the cell-free DNA methylomes to generate a biomarker signature, which can be predictive of a response to therapy.”
He added that the risk stratification “approach may aid in the treatment selection if further validated in this setting.”
medwireNews is an independent medical news service provided by Springer Healthcare Ltd. © 2022 Springer Healthcare Ltd, part of the Springer Nature Group