Background: At the recommended dose of 60mg/day in mRCC, CABO is associated with frequent and/or severe toxicities in mRCC, which may be modified, along with clinical benefit, by interindividual variability in pharmacokinetics (PK). The objective of this multicentre ambispective study was to determine whether the PK of CABO could help to better predict efficacy and early dose-limiting toxicities (DLT) in mRCC. Methods: We collected plasma and determined the concentration of CABO in the first 3 months of initiation in all consecutive mRCC patients (pts) treated in 3 French oncology departments from Dec. 2016 to Mar. 2022. Correlation of plasma trough concentration (Cmin) with objective response rate (ORR), progression free survival (PFS), overall survival (OS), and DLT (any discontinuation or dose reduction in the first 3 months) was explored. An optimal therapeutic range was defined using ROC curves to predict efficacy and DLT. Cox proportional-hazards models were used to identify risk factors of DLT, progression and death. Results: 80 pts were included, 80% clear cell RCC, mean age 62.1 years (+/-13.3) and sex ratio (male/female) 3.2. CABO was administered as second-line, or third-or later-line in 36% and 55% of pts, respectively. Thirty-six (45 %) and 32 (40%) pts initiated CABO at 60mg and 40mg, respectively. After a median follow-up of 17.9 months (9.4-27.9), 24 (30%) were alive, 16 (20%) were still on CABO, and 46 (50%) had experienced DLT with a median to onset of 2 months (CI95%: 0.7-6.2). In the whole cohort, median Cmin was 614 ng/mL [IQR 439-827], with an interindividual variability in dose-normalized exposure of 47.5%. DLT was not associated with the starting dose (p=0.18) but was associated with higher baseline albumin (p=0.03) and higher Cmin (p=0.003). In multivariate analysis, only Cmin was associated with DLT (HR per 100-unit increase: 1.35 [1.03-1.77]; p=0.03). ORR was achieved in 25/79 (32%) evaluable pts, including 16/35 (46%) pts starting CABO at 60mg vs 9/44 (20%) in pts starting at 20/40mg (p=0.03). No difference was found in PFS and OS according to the starting dose (p=0.72 and p=0.49, respectively), nor with previously reported Cmin≥750ng/mL (vs <750, p=0.77 for PFS and p=0.59 for OS). According to the ROC curves, Cmin<350ng/mL tended to be associated with shorter PFS (p=0.0748) and OS (p=0.264), and that Cmin>800ng/mL was significantly associated with DLT (HR 1.56 [1.13-2.15]; p=0.009), although it was not associated with an increase in PFS (p=0.94) or OS (p=0.33). Conclusions: We show that Cmin of CABO is more robustly associated with DLT and PFS/OS than the starting dose, probably due to a high interindividual variability in exposure. Early plasma drug monitoring associated with a target Cmin of 350-800 ng/mL may be useful to optimize CABO in mRCC pts. Future studies should focus on validating this therapeutic range in external validation cohorts.