Plasma and Urine Free Glycosaminoglycans as Monitoring and Predictive Biomarkers in Metastatic Renal Cell Carcinoma: A Prospective Cohort Study

Abstract

Purpose: No liquid biomarkers are approved in metastatic renal cell carcinoma (mRCC) despite the need to predict and monitor response noninvasively to tailor treatment choices. Urine and plasma free glycosaminoglycan profiles (GAGomes) are promising metabolic biomarkers in mRCC. The objective of this study was to explore if GAGomes could predict and monitor response in mRCC.

Patients and methods: We enrolled a single-center prospective cohort of patients with mRCC elected for first-line therapy (ClinicalTrials.gov identifier: NCT02732665) plus three retrospective cohorts (ClinicalTrials.gov identifiers: NCT00715442 and NCT00126594) for external validation. Response was dichotomized as progressive disease (PD) versus non-PD every 8-12 weeks. GAGomes were measured at treatment start, after 6-8 weeks, and every third month in a blinded laboratory. We correlated GAGomes with response and developed scores to classify PD versus non-PD, which were used to predict response at treatment start or after 6-8 weeks.

Results: Fifty patients with mRCC were prospectively included, and all received tyrosine kinase inhibitors (TKIs). PD correlated with alterations in 40% of GAGome features. We developed plasma, urine, and combined glycosaminoglycan progression scores that monitored PD at each response evaluation visit with the area under the receiving operating characteristic curve (AUC) of 0.93, 0.97, and 0.98, respectively. For internal validation, the scores predicted PD at treatment start with the AUC of 0.66, 0.68, and 0.74 and after 6-8 weeks with the AUC of 0.76, 0.66, and 0.75. For external validation, 70 patients with mRCC were retrospectively included and all received TKI-containing regimens. The plasma score predicted PD at treatment start with the AUC of 0.90 and at 6-8 weeks with the AUC of 0.89. The pooled sensitivity and specificity were 58% and 79% at treatment start. Limitations include the exploratory study design.

Conclusion: GAGomes changed in association with mRCC response to TKIs and may provide biologic insights into mRCC mechanisms of response.

FIG 1.

Study design. (A) Patient flow and samples collected at each visit. (B) Workflow used to select free GAGome features correlated with response, to develop GAG progression scores, and to internally and externally validate their use to predict response at two different time points, treatment start and after 6-8 weeks. GAG, glycosaminoglycan; GAGome, glycosaminoglycan profile; non-PD, nonprogressive disease; PD, progressive disease.

FIG 2.

Correlation of detectable plasma and urine free GAGome features (as static measurements or as change from treatment start) in 39 mRCC patients with PD (n = 17) versus non-PD (n = 37) during response evaluation follow-up visits. The posterior probability density of the log OR for PD per unit change of the free GAGome feature (in standard deviation from the mean value) is plotted together with the mean log OR and the 95% credible interval (thick black line). The ROPE is marked by the two vertical dashed lines. A free GAGome feature was deemed compatible with PD or non-PD if its 95% credible interval did not fall inside the ROPE by > 5%. CS, chondroitin sulfate; GAGome, glycosaminoglycan profile; HA, hyaluronic acid; HS, heparan sulfate; mRCC, metastatic renal cell carcinoma; non-PD, nonprogressive disease; OR, odds ratio; PD, progressive disease; ROPE, region of practical equivalence.

FIG 3.

Development of plasma, urine, and combined GAG progression score in mRCC at the response evaluation follow-up visits and their corresponding receiver operating characteristic curves (n = 39 patients, 17 PD v 37 non-PD for plasma; 35 patients, 13 PD v 33 non-PD for urine and combined). GAG, glycosaminoglycan; mRCC, metastatic renal cell carcinoma; non-PD, nonprogressive disease; PD, progressive disease.

FIG 4.

(Top) Waterfall plot for the percentage change from the nadir in the sum of diameters of target lesions in the first 3-month response evaluation color-coded by plasma, urine, and combined GAG progression score (n = 22 patients for plasma and 20 for urine and combined). Unequivocal new lesions are marked with a star. (Bottom) Plasma, urine, and combined GAG progression scores by radiologic response according to RECIST 1.1 at the first 3-month response evaluation visit (n_PD = 6, n_SD = 12, and n_PR = 4). GAG, glycosaminoglycan; PD, progressive disease; PR, partial response; SD, stable disease.

FIG 5.

Internal validation of plasma, urine, and combined GAG progression scores to predict PD versus non-PD at treatment start (top, n = 50, 17 PD v 33 non-PD, n_plasma = 49, n_urine = 43, n_combined = 42) and at 6 weeks (bottom, n = 47, 15 PD v 32 non-PD, n_plasma = 46, n_urine = 41, n_combined = 40) and their corresponding receiver operating characteristic curves. GAG, glycosaminoglycan; non-PD, nonprogressive disease; PD, progressive disease.

FIG 6.

External validation of the plasma GAG progression score to predict PD versus non-PD at treatment start (n = 70, 13 PD v 57 non-PD) and at 8 weeks (n = 69, 12 PD v 57 non-PD) across three cohorts, and their corresponding receiver operating characteristic curves. GAG, glycosaminoglycan; non-PD, nonprogressive disease; PD, progressive disease.