This was a prospective, single-centre, biomarker-driven study recruiting patients that were treated at the Christie NHS Foundation Trust for colorectal cancer with liver metastases. Ethical approval was obtained from the local ethics committee (see supplementary information). All patients gave written informed consent to participate in the study.
Study Participants
Eligible participants included those with histologically-proven colorectal cancer; liver metastases measuring at least 30mm in the longest axis; 18 years of age or older; a World Health Organization (WHO) performance status of 0 to 2; were planned to commence primary therapy with oxaliplatin plus 5-fluorouracil (5-FU) or capecitabine; white cell blood count ≥4×109/l; platelet count ≥100×109/l; serum total bilirubin concentration ≤1.5×upper limit of normal (ULN); serum alkaline phosphatase concentration ≤5×ULN and; a calculated glomerular filtration rate ≥50ml per minute.
Patients were excluded if MRI was contra-indicated due to standard criteria relating to metal implants or allergy to MRI contrast; use of adjuvant chemotherapy within 12 months prior to study enrolment; a personal medical history including any non-colorectal malignancy within 5 years of study enrolment; concurrent use of other investigational medicinal product or; pregnant or breast-feeding women.
Study Drugs
Patients were treated with either oxaliplatin plus 5-FU (oxaliplatin 85mg/m2 of body surface area [BSA] plus folinic acid 350mg and 5-FU 400mg/m2 on day 1 followed by 5-FU 2,400mg/m2 intravenous infusion [46 hours] every two-week cycle) or oxaliplatin plus capecitabine (oxaliplatin 130mg/m2 on day 1 and capecitabine 1,000mg/m2 on day 1 to 14 every 3-week cycle) for a maximum of 6 cycles.
Clinical Endpoints
Clinical endpoints included progression-free survival (PFS) and overall survival (OS). Progressive disease was defined as the time interval from the date of study registration to the date of either clinical or radiological progression or death. On imaging, progressive disease was measured using the response evaluation criteria in solid tumours (RECIST) version 1.1 (26). OS was defined as the time interval from the date of study registration to the date of death. All patients were followed up until they reached the PFS efficacy endpoint; no censoring was present in the dataset.
Computed tomography (CT) was performed every 8 weeks as part of standard tumour assessment. As part of standard treatment, plasma carcinoembryonic antigen (CEA) and lactate dehydrogenase (LDH) concentrations were measured at the start of each cycle of chemotherapy. Both can be used to predict prognosis and response to treatment in metastatic colorectal cancer (27,28).
Biomarker Schedule
A detailed description of the methodology used to for the imaging and circulating biomarkers is provided in the supplementary information.
Study time points for dynamic contrast-enhanced MRI (DCE-MRI) and diffusion weighted MRI (DW-MRI) included pre-treatment, cycle 1 day 2, cycle 1 day 8, cycle 2 day 2 of chemotherapy and following 12 weeks of chemotherapy. At pre-treatment, MRI scans were carried out twice, at least 24 hours apart, to determine the repeatability of the imaging biomarkers. Regions of interest (ROIs) within the liver were defined manually by a trained operator, in order to determine whole tumour volume (WTV) from T1- and T2-weighted images as well as the DCE-MRI images. Parameters derived from DCE-MRI included the transfer coefficient (Ktrans), volume of extravascular extracellular space (νe) and vascular plasma volume (νp). For DWI-MRI, the apparent diffusion coefficient (ADC) was derived.
Blood samples for circulating tumour cells (CTCs) and a panel of plasma-derived circulating protein biomarkers were collected at the same time points as MRI including Ang2, VEGF-A, VEGF-C, VEGF-D, VEGFR1, VEGFR2, IL6, IL8, Tie2, KGF, PlGF, FGFb, HGF, PDGFbb, SDF1b, E-selectin, M65 and VCAM-1.
Statistical Analysis
The target recruitment for the study was 20 patients. All biomarkers were assessed for normality and transformed when necessary. To identify whether biomarker concentrations changed significantly from pre-treatment to cycle 2 day 2, paired Student’s t-tests were performed. Cycle 2 day 2 of chemotherapy was selected for significance testing in order to determine the early effects of cytotoxic chemotherapy. A correlation network analysis was performed to examine the relationship between multiple biomarkers without the requirement to conduct multiple sequential analyses (23,25). This was done based on Pearson correlations and build from the qgraph package in R.
Cox proportional hazard regression was used for survival analysis, respecting the proportionality and linearity assumptions. Kaplan Meier curves were constructed using dichotomized data (longitudinal increase versus decrease in biomarker concentration), and the median PFS and OS intervals in each group were calculated. Statistical significance was determined using p-values, with a cut off of 0.025 being considered statistically significant in order to reduce the impact of multiple testing. More stringent adjustment for multiple comparisons was not considered due to the limited sample size. Analysis was carried out using R 3.5.0.