Patients diagnosed with mCRC between April 2014 and November 2016 were enrolled in the current study. All patients were treated at Saint Margit Hospital (Budapest, Hungary) or at the National Institute of Oncology (Budapest, Hungary).

Inclusion criteria were the following: (1) patients with mCRC scheduled for first-line chemotherapy combined with monoclonal antibody therapy (bevacizumab or cetuximab) based on multidisciplinary board decision; (2) every patient was required to have at least one metabolically measurable metastatic lesion in the liver (> = 2 cm); (3) patients’ performance status should be less than ECOG two; (4) life expectancy of >8 weeks; and (5) signed informed consent.

Patients with (1) a history of allergic reactions to intravenous iodinated contrast agents, or (2) suffering from claustrophobia or (3) uncontrolled diabetes were excluded from our study. Those patients were also excluded who previously received chemo- or targeted therapy for their metastatic disease.

The study was approved by National Institute of Pharmacy and Nutrition (OGYÉI) and the Ethics Committee of the Medical Research Council (ETT-TUKEB) and complied with the Helsinki Declaration.

Chemotherapy plus monoclonal antibody was administered according to current Hungarian guidelines.

Genetic testing of patients for somatic mutation in KRAS and NRAS is routinely applied in our patients [10]. Tumor tissue samples were investigated in some cases from primary site (if the metastases appeared at the same time with primary tumor) or from metastatic sites (if these metastases were metachron).

Methods used for testing the KRAS / NRAS mutations Genomic DNA from formalin-fixed, paraffin-embedded tissue (FFPET) – after deparaffinization – was extracted with the cobas® DNA Sample Preparation Kit, ROCHE.

Kras exon 2 (codon 12 and 13) and exon 3 (codon 61) mutation analysis was performed usingcobas® KRAS Mutation Test (Roche), on the cobas z 480 analyzer (Roche). Sensitivity of the method was 5%, specificity: 99%.

For screening the Nras exon 2 (codon 12 and 13) and exon 3 (codon 59 and 61) mutations, we used an in-house assay based on melting curve analysis on the LightCycler 2.0 instrument (Roche). Primers and FRET probes were purchased from IDT. Analytical sensitivity was 10%, specificity: 100%.

For screening the Kras exon 4 (codon 117 and 146) and NRAS exon 4 (codon 117 and 146) mutations, we used a HRM detection-based in-house assay, where primers and probes were from IDT, LC green was from BioFire Defense LTD, and the assay was carried out on the cobas z 480 analyzer (Roche). Analytical sensitivity was 10%, specificity: 100%.

Patients were categorized according to KRAS or NRAS mutation status into 2 groups: mutant RAS and wild-type RAS. Monoclonal antibody therapy was selected accordingly.

Monoclonal antibodies are dosed as follows: bevacizumab 5 mg/kg intravenously .or cetuximab 500 mg/m2, intravenously. Bevacizumab and cetuximab was applied in 35 patients and 18 patients, respectively. In addition to the monoclonal antibody therapy, patients were treated with chemotherapy: either FOLFOX4 (oxaliplatin 100 mg/m2/2h on day1, leucovorin 200 mg/m2/2h on day 1–2, 400 mg/m2/10min 5-FU on day1–2 and 5-FU continuous infusion 1200 mg/m2/46h) or FOLFIRI (irinotecan 180 mg/m2/90min on day1, leucovorin 200 mg/m2/2h on day 1–2, 400 mg/m2/10min 5-FU on day 1–2. and 5-FU continuous infusion 1200 mg/m2/46h) regimens were administered.

Treatment cycles were repeated every 14 days. All patients were treated with the same regimen as the first applied, until disease progression or if excessive toxicity was noted. After disease progression different second and third line treatment regimens were applied, according to the physician’s choice.

PET/CT scans were carried out at baseline (scan-1) and on day 21 (scan-2), after two cycles of combined chemotherapy. Patients were examined with PET/CT (Siemens Biograph TruePoint 6 HD, Siemens, Knoxville, US), according to routine oncological protocols. Patients fasted for at least 6 h (except diabetic patients, who fasted for 4 h) before examination. Uptake time was 60 ± 5 min in case of both scan-1 and scan-2. Low-dose, whole body CT scan (120 keV, 60 mA) preceded the PET imaging, which was started 7–10 min after the intravenous administration of 3.7 MBq FDG per kilogram body weight. PET raw data were iteratively reconstructed with proper correction for decay, dead-time, scatter, randoms and tissue attenuation with the help of the CT to display standardized (to body weight and injected activity) uptake values (SUV). PET/CT data were analyzed by two independent nuclear medicine specialists.

PET/CT parameters which were measured in case of the metastatic liver lesions included maximum standardized uptake values (SUVmax), total lesion glycolysis (TLG), standardized added metabolic activity (SAM), and normalized standardized added metabolic activity (NSAM). The SUVmax normalized to body weight was measured by the PMOD software (v3.310, Zürich, Switzerland). SAM seeks to determine the total metabolic activity above background due to tumor uptake while avoiding partial volume effect. It was calculated by the same formula as used by Mertens et al. [11]. Briefly, a first volume of interest (VOI, ie.VOI1) was drawn around the metastatic lesion in the liver. A second VOI (VOI2) was delineated around VOI1, directed to a small zone of homogeneous background. SAM was calculated as follows:

$$ \mathrm=\mathrm\ \mathrm\ \mathrm1-\left(\mathrm\ \mathrm\times \mathrm\ \mathrm1\right), $$

where mean BG represents mean background activity, which was derived using the following formula:

$$ \mathrm\ \mathrm=\frac\ \mathrm\ \mathrm2-\mathrm\ \mathrm\ \mathrm1\ }\ \mathrm2-\mathrm\ \mathrm1} $$

in which total SUV is the product of the mean SUV and the respective volume. In patients with multiple liver metastases, SAM was calculated as the sum of the individual SAMs of the lesions.

Table 1 shows all metabolic parameters and their calculation methods for which the relation with OS and PFS was investigated.

Metabolic response was categorized according to the adapted EORTC (European Organization for Research and Treatment of Cancer) PET criteria [12]. The highest pre- and posttreatment SUVmax, the percentage change of SUVmax (Table 1), percentage change of SAM, NSAM and TLG were also calculated. Patients with a reduction in SUVmax more than 25% were classified as responders, meanwhile, when an increase above 25% was found, patients were categorized as non-responders. Different thresholds were applied (30%, respectively) for SAM and TLG to classify the response rates, according to Mertens et al. [11]. As can be seen in Table 2, stable metabolic disease was also defined.

The analysis of the study data followed the principle of intention-to-treat. All applied statistical tests were two-sided and p-values<0.05 were considered significant.

OS and PFS was calculated from the first therapeutic cycle until date of death and CT confirmed progression, respectively. Right censoring was applied as per the last date of follow-up for the patients alive or for patients who did not show progression at the last follow up or at the end of the study. Log-rank analyses were used to assess the relationship between the clinical characteristics and PFS and OS. Cox regression was carried out with stepwise selection of variables. Variables with a p value smaller than <0.25 were selected in the model. 95% Wald and Likelihood confidence intervals were calculated. ZPH test was used together with a time varying coefficient plots [fitted penalized B-spline curve with 95% CI)] to check for non-proportional hazards. In case of non-proportionality, the variables were taken up in the model as time-varying variable. Separate Kaplan-Meier for assessing OS with a two-sided log rank test were calculated for selected variables in the Cox model.

The analyses to assess predicting factors for OS and PFS were conducted in four steps, because one model with too much variables may not be able to reveal all interesting variables. Hence, four separate, per domain, Cox regression models were fitted: first, a model with all demographic variables, then a second model with disease specific background factors, and a third where the metabolic factors were investigated. Finally, all the remaining best predicting variables were put together in one model (supplementary material). All analyses were conducted with SAS version 9.4 (SAS Institute Ltd., Budapest, Hungary) and Statistica v13.2 (StatSoft Inc., Budapest, Hungary).