**6.1 Advancing ctDNA detection accuracy**

*Advances in the Molecular Understanding of Colorectal Cancer*

wild type [66].

primary resistance [84].

mutant *KRAS* were associated with 0% disease control rate [85]. More recently, a large retrospective exploratory analysis used BEAMing technology to identify *KRAS*, *PIK3CA* and *BRAF* mutations in the plasma ctDNA of 503 patients who enrolled in the CORRECT trial of regorafenib, a multi-kinase inhibitor in refractory mCRC [66]. Tumour-associated *KRAS* mutations were readily detected with BEAMing of plasma DNA and were identified in 48% of patients who had previously received anti-EGFR therapy and whose archival tumour tissues were KRAS

Beyond *KRAS* mutations and amplifications, acquired genetic aberrations in other genes have been found to potentially lead to anti-EGFR therapy resistance, albeit in smaller subsets of patients. For example, emerging *EGFR* extracellular domain (ECD) mutations which lead to impaired antibody binding were found to be a resistance mechanism to anti-EGFR therapy in approximately 20% of patients. Interestingly, these mutations tend to arise later than *RAS* mutations during therapy, and patients with *EGFR* ECD mutations had greater and more durable response to anti-EGFR therapy than patients with *RAS* mutations [86]. Interestingly, a phase I trial of a third generation EGFR-targeting agent that binds multiple regions of the EGFR ECD demonstrated efficacy in patients with *EGFR* ECD mutations and acquired resistance to prior EGFR blockade [87]. Other genomic alterations linked to acquired resistance to EGFR blockade include *MET and ERBB2* amplifications [88, 89] and mutations in *NRAS*, *BRAF* and *PIK3CA* [6]. *ERBB2* amplification was found in the plasma in four out of eight *RAS* wild type patients who derived no clinical benefit from anti-EGFR treatment, suggesting that it may also be a source of

Another innovative study provided proof-of-principle that parallel analysis of patient-derived xenografts and ctDNA allowed the identification of resistance mechanisms to a pan-tropomyosin-related kinase (TRK) inhibitor in mCRC, with validation in preclinical models [90]. In interpreting these translational findings, it is important to note that typically, multiple complex molecular abnormalities emerge rather than a singular clone and an overlap exists between abnormalities

CtDNA genotyping has now paved the way for prospective clinical trials which aim to evaluate a range of targeted agents in mCRC and their resistance mechanisms. However, significant knowledge gaps exist in the field, including lack of standardisation of ctDNA techniques, clinical relevance of minority clones detected (for example, no threshold for *KRAS* allele frequency has been established to predict anti-EGFR therapy resistance) and it remains to be proven that changing treatment strategy according to ctDNA findings improves patient outcomes [6]. Challenges notwithstanding, it is foreseeable that in the near future, ctDNA genotyping may be used longitudinally to (i) identify *RAS* wild type patients with mCRC who may be suitable for anti-EGFR antibodies, (ii) dynamically assess treatment response, (iii) identify patients who are developing acquired resistance, (iv) delineate resistance mechanisms to therapy, and (v) discover new druggable

Despite growing enthusiasm, ctDNA in CRC remains largely unavailable for clinical application outside of the trial setting. Recently, there has been a surge of research to further investigate the utility of more sensitive and accurate technologies for ctDNA detection and analysis, and to further elucidate its clinical imple-

mentation and significance in the various settings of CRC management.

associated with primary and secondary resistance [6].

**64**

targets.

**6. Future directions**

Improved sensitivity techniques with the use of targeted-sequencing methods have been developed by several groups [9, 91]. For example, Lanman et al. validated the analytical and clinical use of a novel, ultra-high specific, digital sequencing technique (Guardant360) consisting of 54 clinically actionable cancer genes [91]. In 165 consecutively matched plasma and tumour tissue samples from patients with advanced cancer, this study demonstrated significantly improved sensitivity for Guardant360 in the plasma-derived cfDNA compared to that of tumour tissue. It also demonstrated the clinical success rate of the assay in 1000 consecutive plasma samples in the clinic (assay failure rate of 0.02%) due to its ability to eliminate false positives [91].

Other investigators have combined the use of DNA fragment sequencing by using molecular barcodes with relevant bioinformatics filtering steps to enhance sensitivity and specificity [30, 69, 92, 93]. In a study using cfDNA from mCRC patients, Mansukhani et al. showed that false positive mutation calls could be reduced by 98.6% when incorporating novel molecular barcodes for error correction and by applying custom solution hybrid capture enrichment [93].
