KEY POINTS
- This prospective pilot enrolled 10 patients with prostate cancer receiving MRI-guided SBRT to 36.25 Gy in five every-other-day fractions with daily online adaptive planning and a 3-mm PTV margin.
- Urine was collected at simulation, before the fifth fraction after a cumulative 29 Gy, and again 2–4 months after treatment. MicroRNA sequencing was performed first, followed by qPCR confirmation of selected candidate signals.
- Four of 10 patients developed CTCAE grade 2 urinary frequency or urgency, all requiring tamsulosin. Twenty-seven urine samples were suitable for sequencing, and 111 microRNAs entered differential-expression analysis.
- Differences between patients with and without urinary toxicity were greatest during treatment: 11 microRNAs were differentially expressed at the mid-treatment time point, compared with three at baseline and seven during early follow-up.
- miR-95-3p showed one of the most reproducible temporal signals, decreasing during MR-SBRT by both sequencing and qPCR and subsequently returning toward baseline. qPCR showed a p<0.001 decrease during treatment.
- miR-26a-5p showed a significant interaction between mid-treatment expression and toxicity after multiple-testing correction (q=0.042). miR-204-3p also showed different trajectories in toxicity versus toxicity-free patients, although this association did not remain significant after false-discovery correction.
- miR-375-3p was consistently lower in patients with higher-risk prostate cancer and was reproduced by both sequencing and qPCR. Pathway enrichment among toxicity-associated miRNAs implicated MAPK, NOD-like receptor, and IL-1 signaling, but interpretation is limited by the very small cohort, multiple comparisons, early-toxicity-only assessment, and complete confounding between urinary toxicity and tamsulosin use.
CLINICAL TAKEAWAY
Urine microRNA profiling is technically feasible during prostate MR-SBRT and may eventually provide a non-invasive way to detect biological susceptibility to urinary toxicity. At present this is proof-of-concept only: with 10 patients, medication confounding, and multiple biomarker comparisons, none of these signatures is ready for clinical risk stratification.