<div>Figure 1. Flowchart.</div><div>Procedures.</div><div>A comprehensive clinical assessment was systematically performed before
treatment start.</div><div>During the treatment period, clinical assessment of toxicities as well
as blood count, liver, renal and thyroid functions were assessed every 2
weeks for the first 3 months and then monthly. Patients were instructed
to monitor blood pressure at home (18,19). All adverse events were
prospectively graded according to the National Cancer Institute –
Common Toxicity Criteria for Adverse Events (NCI-CTCAE) version 5.0
(20). In case of grade 3 or 4 toxicity, pazopanib was suspended until
improvement to grade 1-2. CT scan and/or MRI tumour evaluations were
recommended every three months until progression.</div><div>The recommended starting dose of pazopanib was 800 mg/day. However, for
patients with a higher risk of toxicities (ECOG PS ≥ 2, age ≥ 75 years
old, albumin &lt; 30 g/L, cardiovascular background), a lower
starting dose (200 to 600 mg/day) could be prescribed at the discretion
of physician. Subsequently, doses could be adjusted in 200 mg increments
or decrements based on tolerance (21).</div><div>Pharmacokinetic assessments.</div><div>Blood samples were drawn at steady state every two weeks for the first
three months and then monthly. Blood was collected into 5 ml lithium
heparinized Vacutainer tubes at any time over the administration
interval. Samples were centrifuged at 3,000 rpm for five minutes at 4°C
and transferred to polypropylene tubes and kept at -20°C until assay.
Plasma pazopanib concentration was measured using high-performance
liquid chromatography coupled with UV detection. The intra- and
inter-assay coefficients of analytical variability were less than 8%
and 10%, respectively. The lower quantification limit was 1.2 mg/L. The
PAZ C<sub>min</sub> was estimated using a Bayesian method. The
population PK model of Yu et al. (11) was implemented in the Bayesian
estimator.</div><div>Study endpoints.</div><div>Regarding efficacy, the primary endpoint was PFS at 3 months in STS
patients. The main objective was to determine whether PAZ
C<sub>min</sub> ≥ 27 mg/L at day 15 (D15) was associated with a
longer PFS at 3 months. PFS was calculated as the time between the first
day of pazopanib to tumour progression or death. Tumour progression was
assessed using RECIST 1.1 criteria when measurable lesions were present
or was established by the referent oncologist based on clinical
findings, with retrospective confirmation by two oncologists (PBR and
JA). The cut-off of 3 months for the PFS was chosen for two reasons.
First, it corresponded to the first radiologic assessment, secondly it
was in accordance with the primary endpoint of PALETTE trial (6).</div><div>The secondary endpoints were overall survival (OS) in STS patients and
incidence of dose-limiting toxicity (DLT) during the first three months
in both STS and BS patients. OS was calculated as the time between the
first day of pazopanib to death (all causes included). DLT was defined
as any clinical or biological grade 3 or 4 toxicity leading to treatment
dose reduction, interruption (temporary stop) or permanent
discontinuation.</div><div>Statistical analysis.</div><div>Statistical analyses were performed with the software R (version 4.0.3).
Groups were compared with a Student’s t-test for quantitative variables,
and Chi2-test for qualitative variables. Based on the results of our
exploratory study and the PALETTE trial (6), we calculated that we would
need to enrol 82 patients to show a 35% difference of 3-month PFS (70%
vs 35% in patients with a C<sub>min</sub> ≥ 27 and &lt; 27
mg/L respectively) with a two-sided 5% significance level and an 80%
statistical power. Univariate and multivariable logistic regression
models were used to test the association of bio-clinical variables with
3-month PFS. Variables associated with significant p-value in univariate
analyses and potential confounders (initial daily dose, histological
subtype) were included into multivariable models, except tumour grade
(not applicable to some histological subtypes). Interaction tests
revealed no significant subgroup differences. Survival curves were
obtained with Kaplan-Meier estimates and compared with log-rank test.
Univariate and multivariable Cox proportional hazards regression was
used to identify variables associated with OS. The proportional hazards
assumption was checked for each model using graphical methods based on
Kaplan-Meier curves and the scaled Schoenfeld residuals. Univariate
logistic regression models were used to test the association of
bio-clinical variables with DLT.</div><div>All p-values were two-sided, and the level of significance was set at p
&lt; 0.05.</div><div>Our cohort study fulfils the STROBE criteria for the reporting of
observational studies in epidemiology (22).</div><div>RESULTS</div><div>Patients and treatment.</div><div>One hundred and eighteen patients, mostly with STS (81%), were eligible
for statistical analysis. Their baseline characteristics are outlined in
Table 1. The median treatment duration was 3.9 months (range 0.4-51.0).
At least one dose change was performed in 51 patients (43%), mainly
within the first 3 months (36%).</div>