This retrospective study included patients diagnosed with tissue characterized KRAS-mutant LA and with available plasma samples between November 2012 and February 2017 at our center. Staging of cases was done according to the American Joint Committee on Cancer Staging Manual, 8th edition.15 Plasma samples were collected before treatment and at different time points during the follow-up (in most cases, every 6 weeks during treatment and every 3 months during follow-up). Treatment response was evaluated with computed tomography (CT) according to RECISTv1.116 following local practice (every 8–9 weeks). The study was approved by the local ethics committee (CEIC 2012/4823/I). All participants gave written informed consent.

The primary objective of the study was the correlation between ctDNA- and tumor-based detection of mutations. Findings from plasma samples at the time of tumor biopsy were compared with those of the paired tumor biopsy. Agreement was considered positive when KRAS-mutant alleles were detected in plasma in at least 0.1% of frequency. Sensitivity, specificity, positive and negative predictive values were calculated considering tissue biopsy as a gold standard, and using as a negative control plasma samples from 11 cases characterized in tissue as KRAS wild type.

The association between the dynamics of the KRAS mutational load in plasma and radiological response and progression was analyzed for any treatment line received for each patient. The term “treatment line” refers to any administered treatment, including best supportive care. To analyze correlation with response and progression we used plasma samples obtained at the same time as a CT, whereas plasma samples obtained after the start of treatment and prior to the time of a CT were used for the analysis of prediction of response and progression. A positive correlation with/prediction of response was defined as the lack of detection or any level of decrease of KRAS mutational load in the plasma sample at the time (for correlation analysis) or prior to the time (for the prediction study) of the CT showed tumor response. A positive correlation with/prediction of progression was defined as the reappearance or any level of increase of KRAS mutational load in the plasma sample at the time (for correlation analysis) or prior to the time of (for the prediction study) the CT showed progression (Fig. 1).

Fig. 1.

Study design. The table shows the number of analyzed cases in each study time point and the reason of exclusion in not analyzed cases.

(0.26MB).

KRAS analysis in tumor tissue was performed in paraffin-embedded biopsies. DNA was extracted from macro-dissected sections of each tumor sample with the QIAamp DNA Mini kit (Qiagen). KRAS mutations (exon 2 and 3) were analyzed by Sanger sequencing using BigDyev3.1 (Applied Biosystems) on a 3500DX Genetic Analyzer (Applied Biosystems).

For ctDNA-based assessment of KRAS mutations, 10–15ml of blood were collected in K2EDTA tubes and processed within 4h of collection. The tubes were centrifuged twice at 1600×g for 15min; plasma was collected, aliquoted in 1ml fractions and frozen at −80°C. KRAS mutations were assessed by digital PCR (dPCR). DNA was extracted from 1ml of plasma using the MagMax (cfDNA) isolation kit (Applied Biosystems) with manual processing. Isolated DNA was quantified using the QubitdsDNA HS assay kit (Life Technologies). The KRAS-mutated allele frequency was quantified by dPCR with a QuantStudio3D dPCR system (Life Technologies) using wet-lab validated dPCR TaqMan Assays (Life Technologies). Used assays are listed in Supplementary Table 1. Mutation analysis was performed with the QuantStudio 3D AnalysisSuite™ Cloud Software (Life Technologies).

Circulating cell-free DNA (cfDNA) refers to the total circulating DNA and ctDNA only to the circulating tumoral DNA. KRAS mutational load was quantified with dPCR using the QuantStudio3D Analysis Suite™ software that expresses the allele frequency as the percentage of mutated copies referred to the total number of gene copies.

Concordance between tissue and plasma KRAS mutation status was calculated as the number of positive plasma samples out of the total number of tissue samples. Bivariate analyses were performed using the chi-square test. The Kaplan–Meier method was used to estimate progression free survival (PFS). All statistical analyses were performed with IBMR SPSSv.22. Significance was set at p≤0.05.

For this correlation study 51 paired tissue and plasma samples were available from 49 patients (Fig. 1). 49 patients were biopsied once and 2 patients twice (in these two cases, the second biopsy was performed at the time of progression). The median time from tumor biopsy to blood withdrawal was 19 days (range: −56 to 94) (Supplementary Table 2).

A positive correlation between plasma and tumor KRAS mutational status was observed in 42 cases (82.4%) (Fig. 1). The rate of detection of KRAS mutations in plasma was significantly higher in patients with more than one metastatic location (96.4% vs. 65.2%; p=0.004). The case with more than 1 metastatic location in which plasmatic KRAS was not detected, had only 2 metastatic locations (bone and liver). The median cfDNA obtained from the plasma samples was 3595.12 genomic equivalents (GE)/ml (range: 307.4–197,978.54). The rate of KRAS mutation detection was significantly lower in samples with a cfDNA burden below the median (76% vs. 100%; p=0.009). The presence of more than one metastatic location was significantly associated with a cfDNA burden above the median (p=0.028). The 6 cases with negative plasma-tissue correlation had the GE/ml below the median and only one metastatic location. Type of KRAS mutation (codon12 vs. codon13), presence of liver metastases, sex, smoking history, and time from tumor biopsy to blood withdrawal were not associated with the rate of detection of KRAS mutations in plasma (Table 2).

Circulating KRAS mutations were not detected in any of the 11 negative control cases. Sensitivity for KRAS detection in plasma was 0.82 and specificity 1. The positive and negative predictive values were 1 and 0.55 respectively.

In order to avoid the heterogeneity, data were analyzed excluding on one hand the two biopsies performed at the time of progression, and on the other hand cases with stage III disease. In the 49 cases biopsied at the initial diagnosis, a positive correlation between plasma and tumor KRAS mutational status was observed in 40 (81.6%). As shown in the overall population, the rate of detection of KRAS mutations in plasma was significantly higher in cases with more than one metastatic location (96.2% vs. 55.6%; p=0.002). No significant differences in the detection rate were found when cases were analyzed according to the different clinical characteristics. When the 6 cases with stage III excluded from the analysis, the rate of KRAS detection in plasma in cases with a stage IV disease stage was 80.4%. This rate was higher in cases with more than one metastatic location (96.4% vs. 55.6%; p=0.001).

In the analysis of correlation and prediction of response and progression to treatment, plasma samples from 32 patients, who received a total of 52 treatment lines, were included. The median number of treatment lines analyzed per patient was 1 (range: 0–5). Most cases were treated with chemotherapy (72%). In the five patients who did not receive any treatment, plasma samples were collected both at diagnosis and disease progression. The radiological evaluation of 50% of the treatment lines was classified as progressive disease. Clinical characteristics of cases are summarized in Supplementary Table 3, and details on analyzed treatment lines of each case in Supplementary Table 4.

Overall, the mean percentage of circulating KRAS mutant alleles was lower at the time of radiological response (0.4%; standard deviation (SD): 0.87) than at pre-treatment (3.32%; SD: 5.46) and higher at radiological progression (4.63%; SD: 7.69) (Supplementary Figure 1).

17 (33%) of the 52 treatment lines were included in the analysis of correlation with radiological response (Fig. 1). In 15 (88%) of the 17 cases, there was a positive correlation between the dynamics of the KRAS mutational load and radiological response (Fig. 1). The KRAS mutational load decreased in 6 cases and was not detectable in 9 at response evaluation. One of the 2 cases without a positive correlation had stage III disease and the other one had bone as the sole metastatic location.

42 (80%) of the 52 treatment lines were included in the analysis of correlation with radiological progression (Fig. 1). In 30 (71%) of cases, there was a positive correlation between the dynamics of KRAS mutational load and radiological progression. The KRAS mutational load increased in 21 cases and reappeared in 9. No significant differences in the agreement rate with response or progression were observed when cases were analyzed according to the different clinical characteristics.

The mean reduction of ctDNA at the time of radiological response compared to initial levels in the 16 cases with radiological response and available plasma was 64.7% (range: −100 to 140), whereas the mean increase of ctDNA at the time of radiological progression compared to ctDNA levels at the moment of radiological response (or before treatment initiation in case of non-responders) in the 42 progressing cases with available plasma was 3104% (range: −100 to 44,059) (Fig. 2).

Fig. 2.

Waterfall plot showing disease burden and ctDNA dynamics in the 44 treatment lines with available plasma and measurable disease. Cases with progression disease driven by appearance of new lesions are shown as 100% increase in disease burden.

(0.09MB).

In seven cases within this cohort, no plasma sample was available at the time of tumor biopsy and were thus not evaluable for the ctDNA- and tumor-based correlation study, and only second or subsequent treatment lines could be evaluated. When these 7 cases were excluded from the analysis, the rate of concordance between the dynamics of plasma and radiological response and progression were 93.3% and 68.6% respectively.

15 (29%) of the 52 treatment lines from 32 patients were included in the analysis of prediction of radiological response (Fig. 1). In 14 (93%) of the 15 cases, the dynamics of the KRAS mutational load were able to predict radiological response (Fig. 1). The KRAS mutational load decreased in 8 cases and was undetectable in 6 at response. The case where radiological response was not predicted had bone as the sole metastatic location. In cases where radiological response was predicted, the median time elapsed between obtaining the plasma for the analysis of prediction of response and the CT that stated the radiological response was 33 days (range: 14–67).

19 (36%) of the 52 treatment lines were included in the analysis of prediction of radiological progression (Fig. 1). In 12 (63%) of the 19 cases, the dynamics of KRAS mutational load were able to predict radiological progression (Fig. 1). The KRAS mutational load increased in 8 cases and reappeared in 4. In cases where radiological progression was predicted, time elapsed between obtaining the plasma for the analysis of prediction of progression and the CT that stated the disease progression was 50 days (range: 13–120). No significant differences in the rate of prediction of radiological response or progression were shown when cases were analyzed according to the different clinical characteristics (Supplementary Table 5).

The ratio of prediction of radiological response and progression when the 7 cases without available plasma sample at the time of tumor diagnostic was 92.3% and 61.1% respectively.

PFS was significantly longer for those patients with a ctDNA clearance after treatment initiation, compared to those with a ctDNA persistence during treatment (7.7 versus 3.2 months; HR: 0.44, p=0.004) (Fig. 3). No differences in PFS were shown depending on pre-treatment ctDNA levels. Graphic representation of cases in which plasmatic KRAS kinetics were predictive and not predictive of clinical outcome is shown in Fig. 4 and Supplementary Figure 2, respectively.

Fig. 3.

Kaplan–Meier plot showing progression-free survival according to the ctDNA clearance in plasma during treatment. ctDNA, circulant tumor DNA; mPFS, median progression free survival; HR, hazard ratio.

(0.07MB).

Fig. 4.

Dynamics of KRAS mutational load in three patients with an initial radiological response followed by disease progression (A–C) and one non-responder patient (D).

(0.36MB).