Urinary-based detection of MSL, HE4, and CA125 as an additional dimension for predictive and prognostic modelling in ovarian cancer.

e17567 Background: We have recently described a predictive/prognostic model for ovarian cancer, exploiting commonly available clinico-pathological parameters and the ovarian serum biomarkers mesothelin (MSL), human epididymis protein 4 (HE4) and cancer-antigen 125 (CA125). Considering urine as a prototype non-invasive sample, we investigated whether serum levels of these biomarkers are mirrored in urine and compared their clinical relevance in matched serum vs. urine samples. Methods: MSL, HE4 and CA125 were quantified by Lumipulse G chemiluminescent enzyme immunoassay (Fujirebio). In total, this study was based on 1080 biomarker detection runs from 188 serum and 172 urine samples from 192 ovarian cancer patients. Results: While absolute concentrations of MSL and CA125 were higher in serum than in urine, serum HE4 concentrations were considerably lower than urinary HE4 concentrations. Nonetheless, relative intra-patient levels of all three biomarkers strongly correlated between matched serum vs. urinary samples and were unrelated to BRCA1/2 mutational status. Consequently, prediction of surgical outcome or relapse/death by MSL, HE4 or CA125 was similarly efficient among urinary- vs. serum-based detection. HE4 provided the highest capacity in predicting surgical outcome and relapse/death among both body fluid types (urinary HE4: AUC=0.854; AUC=0.750, respectively). All clinically relevant findings of urinary MSL, HE4 and CA125 were reproducible among raw vs. creatinine-normalized datasets, suggesting that normalization may have subordinate priority for urine-based analysis. Conclusions: We report that the capacity of MSL, HE4 and CA125 for predicting surgical outcome and relapse/death is equivalent between serum vs. urine-based detection. Urinary biomarkers, in particular HE4, may provide an additional dimension for prognostic modeling in ovarian cancer.