High-Performance Cellular Radiosensitivity Assay Combined to a Polygenic Score to Predict Late Post Radiotherapy Overreaction in Cancer Prostate Patients: A Proof of Concept

Purpose/Objective(s)

Radioinduced toxicity is a multifactorial problem, related not only to the delivered dose, but also to an intrinsic process within tissues responding to cellular injury. Recently, a novel polygenic risk score (PRS) that for the first time incorporates SNP-SNP interactions (PRSi) was shown to be able to predict late toxicities in the REQUITE prostate cancer cohort (Franco et al, RO 2021). On the other way, recent studies have stressed the role of the ATM protein (pATM) in the response to radiation on lymphocytes from blood and skin fibroblasts (Pereira et al. IJROBP 2018, Deneuve et al. Cancers 2021) and have confirmed that pATM is a relevant biomarker. The main objective of this analysis was to evaluate and validate the prediction assay RadioDTect® (RDT, based on pATM concentration, Deneuve et al, 2021) combined with Single Nucleotide Polymorphisms (SNPs) towards gastrointestinal (GI) and genitourinary (GU) toxicities after radiotherapy (RT) for localized prostate cancer.

Materials/Methods

We considered 48 prostate cancer patients treated with exclusive RT (VMAT) at a single institution at 74-80Gy with conventional fractionation. For all patients, toxicity was prospectively scored before RT and every 6 months till 3 years after the RT end combining CTCAE scale and patient-reported questionnaires. The endpoint of this analysis was any (GI OR GU) grade≥2 toxicity at 3 years after RT. Blood samples were prospectively collected at least 6 months after RT and used to have the RDT and extract germline DNA. We tested a pool of 15 SNPs associated with late RT toxicity from the literature with uni- and multi-variate logistic analysis. We used the multivariable logistic regression predicted probabilities to compute a PRS (using z-standardization of the predicted probability). We investigated the value of a possible RDT+PRS combined biomarker and compared it with the results from dosimetry-based toxicity prediction (including the maximum dose to the rectum and the bladder).

Results

18/48 patients scored with any grade≥2 late toxicity. Patients with grade≥2 late toxicity were classified using RDT with a previously determined cutoff of 57.8 ng/mL with AUC=0.63 (p=0. 04). 3/15 SNPs from the literature confirmed their association with late toxicity (rs11122573 from Kerns et al. J Natl Cancer Inst 2020, rs2293054 and rs845552 from De Langhe et al. RO 2014). The PRS from these 3 SNPs predicted late toxicity with AUC=0.71 (p=0.01). Combining PRS and RDT led to an improvement of AUC of 7% (p=0.04, AUC=0.78). Adding dosimetric information improved AUC of 2% (p=0.5, AUC=0.8).

Conclusion

The results of this proof of concept allowed us to confirm that biological assays tested in a population of prostate cancer patients treated with VMAT overperformed existing models for toxicity prediction.