Circulating tumour DNA in the evolving treatment landscape of locally advanced rectal cancer: where does it fit in?

Abstract

The management of locally advanced rectal cancer (LARC) requires multimodality treatment, typically with neoadjuvant chemoradiotherapy (CRT) followed by total mesorectal excision. However, the treatment landscape is rapidly evolving with total neoadjuvant therapy and non-operative management for selected patients emerging as other novel treatment approaches. With so many treatment options, there is a need for biomarkers to direct a more personalised treatment strategy for patients with LARC. In this review, we summarise the available data regarding the use of circulating tumour DNA (ctDNA) in patients with LARC, as both a marker of treatment response to neoadjuvant therapy and as a marker of minimal residual disease (MRD) after patients have completed definitive local treatment. To date, the ability of ctDNA status to predict for pathologic complete response at any timepoint during multimodality treatment has been variably reported. The most consistent finding across available studies is the ability of ctDNA to detect MRD after CRT and surgery, the presence of which confers a significantly poor prognosis, with increased risk of cancer recurrence and worse overall survival. It is yet to be determined if providing additional therapies to patients with MRD improves outcomes. The available studies assessing the potential utility of ctDNA in LARC are limited by significant heterogeneity in the choice of ctDNA assay, timepoint at which ctDNA was collected, treatment that patients received and length of follow-up, leading to uncertainties about how to implement it into daily clinical practice. As the treatment landscape evolves, larger randomised trials assessing the role of ctDNA in LARC are needed.

Introduction

Colorectal cancer represents a major global health burden, with approximately 30% of tumours arising from the rectum.¹ The standard treatment approach for patients with locally advanced rectal cancer (LARC), defined by clinical stage T3/T4 or N+, has evolved from surgery alone to a multimodality approach comprising neoadjuvant chemoradiotherapy (CRT) followed by total mesorectal excision (TME) surgery. CRT typically comprises 50.4 Gy of radiation, delivered over 5 weeks, administered with concurrent fluoropyrimidine based chemotherapy. In addition, most guidelines^2,3 support the use of adjuvant chemotherapy (AC) but a survival benefit from this approach has not been clearly demonstrated in the modern treatment era.^4,5

Approximately 15–27% of patients with LARC can achieve a pathologic complete response (pCR, ypT0N0) from CRT and may be spared potentially morbid surgery and a permanent stoma.^6,7 This prognostic marker is also associated with more favourable cancer-related outcomes compared to patients with residual disease at the time of surgery.^8,9

In addition, total neoadjuvant therapy (TNT) is emerging as a novel approach in the management of LARC, where both chemotherapy and either CRT or short course radiation (SRT, 25 Gy over 5 days) are provided prior to TME and this is now recognised as another standard treatment option.³ With an evolving list of treatment approaches, there is a need to identify biomarkers that can help guide a more individualised treatment approach for patients. Circulating tumour DNA (ctDNA) has been studied as one such potential biomarker.

ctDNA may be secreted into the bloodstream by living cancer cells or released at the time of cell death, and represents a very small fraction of all cell-free DNA. The ability to detect ctDNA, the so-called ‘liquid biopsy’, has garnered immense interest across multiple tumour types due to its potential abilities for molecular profiling of tumours, overcoming issues of tumour heterogeneity that may occur with a tissue biopsy, quick turnaround time and minimally invasive method. Its ability to identify the presence of minimal residual disease (MRD) has established its role as a tool to risk-stratify patients for recurrence, after apparently curative treatment in early-stage colon cancer.^10–12 ctDNA can be detected by either tumour-informed assays, where probes are designed according to mutations identified in a patient’s tumour tissue or by tumour-agnostic assays which use a fixed panel to assess for the presence of frequently altered genetic alterations within a particular cancer type.

Currently, ctDNA is not routinely used in the management of LARC. Here, we review the available evidence regarding the ability of ctDNA to predict treatment responses and long-term cancer-related outcomes and discuss its potential clinical applications in patients with LARC.

The use of ctDNA to predict response to neoadjuvant treatment

Patients with LARC who achieve a pCR following CRT have lower rates of local recurrence, distant metastasis and improved overall survival (OS) compared to non-pCR patients.^8,9 Importantly, such patients may not derive additional benefit from TME. pCR status can only be established after surgery, so the use of clinical complete response (cCR) following CRT, defined as undetectable tumour by magnetic resonance imaging (MRI) and endoscopy,¹³ has been used as a surrogate to guide selection of patients who may be appropriate for a ‘watch and wait’ that is, a non-operative approach. Several studies, with follow-up periods ranging from 43 to 60 months, have demonstrated that patients in cCR managed with a non-operative approach have either similar or better disease-free survival (DFS) and OS outcomes than those treated with surgery who achieved pCR.^13–16 In addition, the International Watch & Wait database¹⁷ has reported excellent survival outcomes in these patients, with 2-year local regrowth rates of 25.2% and 5-year disease-specific survival rates of 94%. Approximately 77% of patients with local regrowth underwent salvage surgery with curative intent.¹⁷

However, there are some reports of limited concordance between pCR and cCR status¹⁸ and therefore additional ways to predict treatment response prior to surgery are still needed to help direct treatment strategies. In addition, it may be important to identify patients responding poorly to CRT to allow for an escalation of the chosen treatment strategy. This may include offering TNT prior to TME, which has demonstrated higher rates of pCR (17–53%).^19–23 As such, ctDNA has been assessed as a potential biomarker to predict local treatment response in patients with LARC (Table 1; Figure 1).

Table 1.

Select published studies assessing the predictive and prognostic role of circulating tumour DNA in locally advanced rectal cancer.

Author, year, Ref	Study design and treatment	No of patients and characteristics	ctDNA measurement timepoints	ctDNA detection method	Key findings: treatment response	Key findings: recurrence or survival
Sclafani et al., 201824	Retrospective analysis of a phase II RCT NAC → CRT → TME → AC ± cetuximab	n = 97 61% male 64% stage III 36% stage II	Baseline	ddPCR assessing KRAS and BRAF mutations	No significant difference between baseline ctDNA-positive and -negative patients in terms of RECIST CR rate (15.4% versus 10%, OR: 1.63, 95% CI: 0.30–8.96, p = 0.57)	Median follow-up 37 m. No significant difference between baseline ctDNA-positive and -negative patients in terms of PFS (HR: 0.70, 95% CI: 0.29–1.70, p = 0.43) or OS (HR: 0.78, 95% CI: 0.31–1.96, p = 0.60)
Tie et al., 201925	Prospective, multicentre study CRT → TME → ± AC	n = 159 median age 62 years 67% male 78% stage III 22% stage II	Baseline Post-CRT Post-operative	NGS, tumour-informed assay	No significant association between ctDNA status at any timepoint or ctDNA clearance and pCR	Median follow-up 24 m. Baseline ctDNA status was not associated with RFS (p = 0.823). Post-CRT or post-operative ctDNA-positive patients had significantly worse 3 year-RFS than ctDNA-negative patients (post-CRT: 50% versus 85%, HR: 6.6, 95% CI: 2.6–17, p < 0.001; post-op: 33% versus 87%, HR: 13, 95% CI: 5.5–31, p < 0.001)
Appelt et al., 202026	Exploratory analysis of a phase III RCT CRT → TME ± AC	n = 146 Median age 64 years 64% male 88% stage III 12% stage II	Baseline	Tumour-agnostic assay assessing hypermethylation of NPY gene, measured by ddPCR	No correlation of baseline meth-ctDNA status with complete (p = 0.76) or major (p = 1.0) pathological TRG response	Detectable meth-ctDNA was associated with worse 5-year OS (47% versus 69%, p = 0.02) and 5-year freedom from distant metastasis (55% versus 72%, p = 0.01) after median follow-up of 10.6 years and 5.1 years, respectively
Khakoo et al., 202027	Prospective, single centre CRT → TME/organ preservation → ± AC	n = 47 Median age 59 years 62% male 89% stage III 11% stage I–II	Baseline Mid-CRT Post-CRT Post-operative or 3–6 monthly for organ preservation	ddPCR, tumour-informed assay	ctDNA status after CRT was associated with primary tumour response by MRI TRG. Poor responders (MRI TRG 3–5) were significantly more likely to have detectable ctDNA after CRT than good responders (MRI TRG 1–2), (33% versus 5%, p = 0.03). No significant difference in MRI TRG response between ctDNA positive versus ctDNA negative at any other timepoint.	Median follow-up 26.4 m MFS was significantly shorter in patients with: detectable ctDNA after completing CRT (HR: 7.1, 95% CI: 2.4–21.5, p < 0.001); persistently detectable ctDNA pre- and mid-CRT (HR: 3.8, 95% CI: 1.2–11.7, p = 0.02); persistently detectable ctDNA from pre- to post-CRT (HR: 11.5, 95% CI: 3.3–40.4, p < 0.001). DFS was shorter in those with detectable post-operative ctDNA (HR: 39.9, 95% CI: 4.0–399.5, p = 0.002).
Murahashi et al., 202028	Prospective, single centre ±NAC → SRT/CRT → TME/organ preservation	n = 85 Median age 60 years 76.5% male 68% stage III 32% stage II	Baseline Post-SRT/CRT Post-operative	Tumour agnostic, NGS	Neither baseline, pre-operative nor post-operative ctDNA status correlated with pathological response or complete clinical response. Change in ctDNA MAF from baseline to pre-operative was an independent predictor for pCR on multivariate analysis (OR 7.4, 95% CI: 1.2–144, p = 0.0276).	Median follow-up not reported. Post-operative ctDNA positivity was an independent prognostic marker for risk of recurrence after surgery (adjusted HR: 7.7, 95% CI: 1.6–42, p = 0.0127).
Pazdirek et al., 202029	Prospective, single centre. CRT → TME	n = 33 Average age 64 years 75% male 70% stage III 30% stage II	Baseline After 1 week of CRT	Denaturing capillary electrophoresis ± BEAMing (ddPCR)	No statistical correlation between baseline ctDNA status and ypTRG staging or ypTNM staging	Median follow-up not reported. Baseline positive ctDNA status was significantly associated with worse DFS (p = 0.015) and OS (p = 0.010). 3-year OS was 71.4% in baseline ctDNA-positive group compared to 91.2% in negative group.
Zhou et al., 202130	Prospective, multicentre CRT → NAC → TME → ± AC	n = 104 Median age 60 years 64% male 95% Clinical stage III 5% Clinical stage II	Baseline During CRT Pre-operative Post-operative	NGS, tumour-agnostic assay	The pre-operative ctDNA-positive rate was significantly lower in those with favourable pathological and radiological responses including pCR compared to non-pCR (0% versus 13.8%, p = 0.02), ypT0–2 compared to ypT3–4 (1.8% versus 21.2%, p = 0.002), ypTRG0–1 compared to ypTRG2–3 (0% versus 22.0%, p < 0.001), ymrEMVI negative versus ymrEMVI positive (3.3% versus 26.7%, p < 0.002). No significant association between ctDNA status at baseline or during CRT and parameters reflecting treatment response.	Median follow-up 18.8 m ctDNA positivity at all four timepoints was significantly associated with a shorter MFS: - Baseline. HR: undefined, p = 0.03 - During CRT. HR: 6.6 (95% CI: 1.2–35.5), p = 0.001 - Pre-operative. HR: 19.82 (95% CI: 2.03–193.7), p < 0.001 - Post-operative. HR: 25.3 (95% CI: 1.48–434.0), p < 0.001 Median VAF in ctDNA was also shown at all four timepoints to be significantly associated with shorter MFS.
Vidal et al., 202131	Biomarker sub-study of a phase II RCT NAC → CRT → TME	n = 72 Median age 60 years 64.5% male 72% cT3 69% cN2	Baseline Pre-operative	Tumour-agnostic assay, NGS.	No association demonstrated between baseline or pre-operative ctDNA status and pCR (p = 0.134 and p = 0.66, respectively).	Median follow-up 38 m. Detectable pre-operative ctDNA was significantly associated with disease recurrence, shorter DFS (HR: 4.029, 95% CI: 1.004–16.16, p = 0.033) and shorter OS (HR: 23, 95% CI: 2.4–212, p < 0.0001). Baseline ctDNA status was not significantly associated with DFS (p = 0.59) or OS (p = 0.38).
McDuff et al., 202132	Prospective, multicentre. CRT → ± NAC → TME → ± AC	n = 29 Median age 54 years 52% male 79% clinical stage III 21% stage II	Baseline Pre-operative Post-operative	Tumour-informed assay, ddPCR	Patients who had preoperative ctDNA-negative status compared to ctDNA-positive status had significantly higher rates of margin-negative, node-negative resection (88% versus 44%, p = 0.028) but no significant difference in rates of pCR (24% versus 11%, p = 0.63).	Median follow-up 20 m. Patients with detectable post-operative ctDNA had significantly poorer RFS (HR: 11.56, p = 0.007).
Wang et al., 202133	Prospective CRT → NAC → TME → AC	n = 119 Median age 57 years 71.4% male 66.4% stage IIIB 31.1% stage IIIC 2.5% other	Baseline During CRT Post-CRT Pre-operative Post-operative	Tumour-agnostic assay, NGS	Those who had cleared ctDNA from mid-CRT onwards had a lower probability of non-pCR than those without clearance (OR 0.11. 95% CI: 0.01–0.6, p = 0.04). ctDNA clearance rate during CRT was significantly lower (p = 0.0008) and rates of acquired mutations in ctDNA significantly higher (p = 0.02) in patients with higher pTRG scores.	Median follow-up 21.2 m. Detection of potential colorectal cancer driver genes in ctDNA after CRT was associated with significantly worse RFS (HR: 9.29, R 3.74–23.10, p < 0.001).
Liu et al., 202234	Associated translational study as part of an RCT CRT → TME → AC Or SRT → NAC → TME → AC	n = 60 Mean age 53 70% male 98% cT3–4 90% cN1–3	Baseline During NAT After NAT Pre-operatively	1. Tumour-informed assay using multiplex PCR 2. Tumour-agnostic assay 3. Low-depth sequencing for copy number alterations	ctDNA status measured using the tumour-informed assay neither during NAT nor after NAT were significantly associated with pCR. Amongst patients who were ctDNA positive at baseline, those who achieved clearance of ctDNA, as measured by the tumour-informed assay, following NAT were significantly less likely to have non-pCR/cCR status (56% versus 95%, p = 0.013).	Median follow-up 33.25 m Patients with detectable ctDNA after NAT using the tumour-informed assay had significantly increased risk of recurrence (HR: 27.38, 95% CI: 8.61–87.06, p < 0.0001) and worse OS (HR: 17.78, 95% CI: 1.94–162.6, p = 0.00054). ctDNA detected by the tumour-agnostic assay and by copy number alteration sequencing was also significantly associated with increased risk of recurrence (HR: 5.18, 95% CI: 1.76–15.2, p = 0.00086; HR: 9.24, 95% CI: 2.28–37.4, p = 0.00017, respectively).

AC, adjuvant chemotherapy; CRT, chemoradotherapy; ctDNA, circulating tumour DNA; DFS, disease-free survival; MAF, mutant allele fraction; MFS, metastasis-free survival; MRI, magnetic resonance imaging; NAC, neoadjuvant chemotherapy; NGS, next-generation sequencing; OS, overall survival; pCR, pathological complete response; PFS, progression-free survival; RCT, randomised controlled trial; RFS, reoccurrence-free survival; SRT, short course radiation; TME, total mesorectal excision; TRG, tumour regression grade; VAF, variant allele frequency.

Figure 1.

Prognostic and predictive roles of ctDNA in locally advanced rectal cancer. Source: Image created with BioRender.com.

CRT, chemoradiotherapy; ctDNA, circulating tumour DNA; pCR, pathologic complete response; LARC, locally advanced rectal cancer; MRD, minimal residual disease.

CtDNA status during and post-neoadjuvant CRT

Available studies have reported conflicting data regarding the ability of ctDNA status during or following CRT to predict pCR. Understanding the potential role of ctDNA in this setting is complicated by the heterogeneity in study design, including whether or not patients received neoadjuvant chemotherapy (NAC), the ctDNA assay used and the timepoints at which ctDNA was assessed. Overall, most studies have not demonstrated a significant association between ctDNA status and pCR.^25,27,31,32

In the largest study to date, Tie et al.²⁵ prospectively analysed multiple plasma samples for ctDNA in 159 patients with LARC undergoing neoadjuvant CRT followed by TME. Using the tumour-informed SafeSeqS next-generation sequencing (NGS) assay to identify tumour-specific mutations, ctDNA was detectable in 77% of samples at baseline, 8.3% following CRT and in 12% post-operatively. Although patients who had undetectable ctDNA after CRT had higher rates of pCR than ctDNA-positive patients, this was not statistically significant (21% versus 9%, p = 0.46). The conversion of ctDNA status from positive at baseline to negative after CRT was also not associated with pCR (pCR versus non-pCR, 95% versus 88%, p = 0.46).

Khakoo et al.²⁷ used a tumour-informed ddPCR assay in 47 patients with LARC undergoing neoadjuvant CRT followed by either TME or organ preservation and detected ctDNA in 74% of patients at baseline, 21% mid-CRT, 21% post-CRT and in 13% post-operatively. Post-CRT ctDNA status was not associated with either pT or pN stage (p = 0.25, p = 0.21, respectively) but in all three patients who developed pCR, ctDNA was detectable at baseline and became undetectable from mid-CRT onwards. They were the first group to report an association between MRI response to CRT and ctDNA status, identifying that poor responders [defined as MRI tumour regression grade (TRG) of 3–5] were significantly more likely to have detectable ctDNA after CRT than good responders (33% versus 5%, p = 0.03).

As part of a biomarker sub-study of a phase II randomised controlled trial (RCT), Vidal et al.³¹ analysed ctDNA in 72 patients with LARC undergoing a TNT approach. All patients received 3 months of neoadjuvant fluoropyrimidine-based doublet chemotherapy ± aflibercept, followed by neoadjuvant CRT and TME. ctDNA was assessed using a tumour-agnostic NGS assay at baseline and pre-operatively. A significant relationship between pre-operative ctDNA status and pCR, ypT or ypN status could not be demonstrated (p = 0.134, p = 0.8969, p = 0.586, respectively).

A small study by McDuff et al.,³² evaluated ctDNA status using a tumour-informed ddPCR assay in 29 patients with LARC undergoing CRT and surgery. In addition, 45% of patients received NAC prior to CRT. ctDNA status was assessed at baseline, pre-operatively and post-operatively. Patients with undetectable ctDNA prior to surgery had higher rates of pCR compared to ctDNA-positive patients but this was not statistically significant (24% versus 11%, p = 0.63). However, they did achieve higher rates of margin-negative, node-negative surgical resection (88% versus 44%, p = 0.028) than ctDNA-positive patients.

Conversely, Zhou et al.³⁰ were the first group to report a significant association between pre-operative ctDNA status and pCR. In their study of 104 patients with LARC, patients were randomised 1:1 to receive three cycles of capecitabine either alone or in combination with oxaliplatin, prior to surgery, with the first two of those cycles delivered concurrently with 5 weeks of radiotherapy (RT). ctDNA was assessed using NGS and found to be detectable in 75% at baseline, 15.6% during CRT, 10.5% pre-operatively and 6.7% post-operatively. They identified that pre-operative (post-CRT) ctDNA positivity was significantly lower in those who achieved pCR versus non-pCR (0% versus 13.8%, p = 0.02), ypT0-2 versus ypT3-4 status (1.8% versus 21.2%, p = 0.0002), favourable versus unfavourable pTRG scores (0% versus 22.0%, p < 0.001) and in those with favourable versus unfavourable radiologic findings as measured by extramural vascular invasion negativity on MRI (3.3% versus 26.7%, p < 0.002). These results highlight that ctDNA negativity may not be an adequate marker to select patients for a non-operative approach given that 86.2% of non-pCR patients did not have detectable post-CRT ctDNA.

Three studies^28,33,34 have identified significant relationships between ctDNA clearance during neoadjuvant treatment and pCR. All three utilised NAC and were studied in Asian populations. Murahashi et al.²⁸ were the first group to report such a relationship. In their study of 85 patients with LARC, patients were treated with either SRT or CRT and 45.9% also received NAC. This was followed by either surgery or organ preservation. ‘Responders’ (24.7% of patients) were defined as those who achieved either a pCR or were in cCR and successfully managed for at least 12 months with a non-operative approach. Using a tumour-agnostic NGS assay, ctDNA was detected in 57.6% of patients at baseline and in 22.3% pre-operatively. Patients with undetectable pre-operative ctDNA were more likely to be ‘responders’ than ctDNA-positive patients but this was not statistically significant (24.2% versus 10.5%, p = 0.33). However, for those with detectable ctDNA at baseline and/or pre-operatively, the change in ctDNA mean allele frequency of ⩾80% versus <80% was found to be significantly associated with ‘responders’ and remained an independent predictor on multivariate analysis (MVA) (adjusted OR: 7.4, 95% CI: 1.2–144, p = 0.0276). The interpretation of the results from this study is complicated by the lack of standardisation of pre-operative therapy and the low rate of baseline ctDNA detected.

Liu et al.³⁴ evaluated ctDNA status in 60 patients with LARC. As part of the STELLAR trial,²² patients were randomised to receive either neoadjuvant CRT followed by TME and six cycles of AC (capecitabine and oxaliplatin) or SRT followed by four cycles of NAC, TME and two cycles of AC. They assessed for the presence of ctDNA at baseline, during neoadjuvant therapy, after neoadjuvant therapy and pre-operatively using three approaches: a tumour-informed assay, a tumour-agnostic assay and low-depth sequencing for copy number alterations. Although ctDNA status using the tumour-informed assay was significantly associated with ypN0 status during neoadjuvant therapy (p = 0.007) and with ypT0-2 status after neoadjuvant therapy (p = 0.04), at neither timepoint was it associated with pCR (both p > 0.99). However, patients who achieved clearance of ctDNA (defined as a ratio of post-neoadjuvant therapy ctDNA fraction to baseline ctDNA fraction of <2%) were significantly more likely to achieve either pCR or cCR status (44% versus 5%, p = 0.013).

Wang et al.³³ explored the potential of using ctDNA in combination with MRI to predict pCR in 119 patients with LARC, who received CRT, followed by one cycle of NAC (capecitabine and irinotecan), TME then a further five cycles of AC. Using a tumour-agnostic NGS assay, they detected ctDNA in 84% of patients at baseline and identified that those who achieved ctDNA clearance (defined as being unable to detect the mutation with the highest variant allele frequency) from mid-CRT onwards had a significantly lower probability of non-pCR than those without clearance (OR: 0.11. 95% CI: 0.01–0.6, p = 0.04). Using a risk score predictive model, they identified that using both ctDNA (features of baseline ctDNA, ctDNA clearance and acquired mutation status) and mrTRG had increased performance in predicting pCR compared with ctDNA alone or mrTRG alone.

Baseline ctDNA status and tumour response

Some studies have assessed the role of ctDNA status at other timepoints during treatment to predict pCR status. Those that have assessed the role of ctDNA status at baseline have largely been negative, including those by Tie, Vidal, Murahashi, Zhou, Wang and Liu.^{25,28,30,31,33,34}

In addition, Appelt et al.²⁶ performed an exploratory analysis of baseline methylated ctDNA (meth-ctDNA) in 146 patients receiving CRT and surgery as part of an RCT³⁵ assessing two different doses of RT. They assessed for hypermethylation of the NPY gene by ddPCR and found no correlation between baseline meth-ctDNA status and complete (p = 0.76) or major (p = 1.0) pathological TRG response. Furthermore, Pazdirek et al.²⁹ analysed ctDNA by denaturing capillary electrophoresis and ddPCR in 33 patients with LARC and could not identify any statistical correlation between baseline ctDNA status and ypTRG response.

In terms of radiological response, Khakoo et al.²⁷ did not find an association between ctDNA status at baseline and primary tumour response assessed by mrTRG (p = 0.74). In addition, Sclafani et al.²⁴ performed a retrospective analysis of patients with LARC treated as part of the randomised phase II EXPERT-C trial.³⁶ They assessed ctDNA at baseline using ddPCR to detect KRAS and BRAF mutations in 97 patients. There was no significant difference in RECIST CR rate between baseline ctDNA-positive versus -negative patients (15.4% versus 10%, OR: 1.63, 95% CI: 0.30–8.96, p = 0.57).

Post-operative ctDNA status and tumour response

Studies that have assessed for an association between post-operative ctDNA status and pCR have not identified a significant relationship.^25,27,28,30 Tie et al.²⁵ reported a significant association between post-operative ctDNA status and both ypT3-4 status (p = 0.01) and pN+ status (p = 0.05) but not pCR status (p = 0.37). Khakoo et al.²⁷ and Zhou et al.³⁰ reported significant associations between post-operative ctDNA status and ypN+ status (p = 0.02) and ypT3-4 status (p = 0.03) respectively but not pCR status.

Summary

Based on the currently available data, there is insufficient evidence to support the role of using the presence or absence of detectable ctDNA at any timepoint as a predictor of pCR. Although the clearance of ctDNA has been shown to significantly increase the chance of pCR in a small number of studies, this chance may still be lower than 50% and thus should not be used in isolation to select patients for organ preservation. Patients who fail to clear their ctDNA have a very low chance of pCR and this finding may help to identify those who should not be recommended for organ preservation.

The use of ctDNA in LARC after CRT and after surgery as a prognostic biomarker

Regardless of which combination of treatment strategies is chosen, once treatment for LARC has been completed, patients undergo close surveillance to detect local or distant disease recurrence. Most guidelines recommend regular clinical, endoscopic, imaging and carcinoembryonic antigen surveillance for 5 years.^2,3

The presence of ctDNA after CRT and after surgery and its relationship with disease recurrence and survival have been assessed in several studies (Table 1, Figure 1). The results are largely consistent, with many studies identifying a significantly negative prognostic value of detectable ctDNA at these timepoints.^{25,27,28,30–34}

In the first and largest study to date,²⁵ with a median follow-up of 24 months, detectability of ctDNA at both post-CRT and post-operative timepoints were significantly associated with worse 3-year recurrence-free survival (RFS) compared to ctDNA-negative patients (post-CRT 50% versus 85%, HR: 6.6, 95% CI: 2.6–17, p < 0.001; post-operative 33% versus 87%, HR: 13, 95% CI: 5.5–31, p < 0.001). The use of AC in the study was at the discretion of treating clinicians, who were blinded to the ctDNA results. After correcting for the use of AC and relevant clinicopathological variables, positive post-operative ctDNA status was the strongest predictor of poor RFS (HR: 6.0, p < 0.001).

After a median follow-up of 26.4 months, Khakoo et al.²⁷ demonstrated that compared to ctDNA-negative patients, patients with detectable ctDNA after CRT had significantly shorter metastasis-free survival (MFS) (HR: 7.1, 95% CI: 2.4–21.5, p < 0.001). In addition, post-operative ctDNA positivity was significantly associated with shorter DFS (HR: 39.9, 95% CI: 4.0–399.5, p = 0.002). In the small number of patients (N = 15) who were managed with an organ preservation approach, there was no significant difference in local RFS by ctDNA status measured at the end of CRT (HR: 5.8, 95% CI: 0.9–35.3, p = 0.06).

Murahashi et al.²⁸ detected ctDNA in 21/59 patients (36%) after surgery. Using a cut-off for mutant allele fraction of 0.5%, patients with positive post-operative ctDNA status had significantly worse RFS compared to ctDNA negative patients (HR: 20, 95% CI: 1.4–163). This remained significant on MVA (HR: 7.7, 95% CI: 5.6–72, p = 0.01).

Similarly, with a median follow-up of 18.8 months, Zhou et al.³⁰ demonstrated that detectable ctDNA during CRT, post-CRT and post-operatively were all significantly associated with shorter MFS [HRs: 6.6 (95% CI: 1.2–35.5, p = 0.001), 19.8 (95% CI: 2.03–193.7, p < 0.001) and 25.3 (95% CI: 1.48–434.0, p < 0.001), respectively] compared to ctDNA-negative patients. McDuff et al. also demonstrated that post-operative ctDNA positivity was significantly associated with poor RFS (HR: 11.6, p = 0.007). In Wang et al.’s study,³³ patients with detectable ctDNA after CRT had significantly worse RFS (HR: 9.3, 95% CI: 3.7–23.1, p < 0.001).

In the first study to demonstrate an association between ctDNA status and OS in LARC, Vidal et al.³¹ detected ctDNA in 7/45 (15.6%) patients after NAC and CRT. After a median follow-up of 38 months, detectable pre-operative ctDNA was significantly associated with shorter DFS (HR: 4.0, 95% CI: 1.0–16.2, p = 0.03) and OS (HR: 23, 95% CI: 2.4–212, p < 0.0001). Kaplan–Meier estimates for 3-year and 4-year OS were 67% versus 97% and 33% versus 97%, respectively. This study did not examine the prognostic role of post-operative ctDNA.

In Liu et al.’s study,³⁴ 14/60 (23.3%) of patients had detectable ctDNA using their tumour-informed assay after completing either NAC and SRT or CRT. With a median follow-up of 33.3 months, these patients had significantly increased risk of recurrence (HR: 27.4, 95% CI: 8.6–87.1, p < 0.0001), worse local RFS (HR: 20.6, 95% CI: 2.3–187.4, p = 0.0002), distant MFS (HR: 17.4, 95% CI: 4.6–65.3, p < 0.0001) and OS (HR: 17.8, 95% CI: 1.9–162.6, p = 0.0005). After MVA, ctDNA positivity remained an independent predictor of worse RFS (HR: 18.0, 95% CI: 5.3–60.9, p < 0.001). In addition, ctDNA detection assessed by other methods including a tumour-agnostic assay and low-depth sequencing for copy number alterations were also both significantly associated with worse RFS.

Prognostic value of ctDNA at baseline

Although the negative prognostic role of detectable ctDNA after CRT and surgery is largely consistent across studies of patients with LARC, there are conflicting reports about its significance in treatment naïve patients. Most studies have not identified a significant association between baseline ctDNA status and RFS or MFS^{24,25,27,31,34} or OS.^24,31,34

Conversely, Zhou et al.³⁰ identified a significant association between baseline ctDNA positivity and shorter MFS (p < 0.05), as did Pazdirek et al.²⁹ for DFS (p = 0.015) and OS (p = 0.010).

Appelt et al.²⁶ reported that patients with detectable meth-ctDNA at baseline had significantly worse distant MFS and OS at 5 years (55% versus 72%, p = 0.01; 47% versus 69%, p = 0.02, respectively). This remained significant after correcting for other prognostic factors upon MVA (distant MFS HR: 2.2, 95% CI: 1.2–4.1, p = 0.01; OS HR: 2.1, 95% CI: 1.2–1.5, p = 0.007).

Wang et al.³³ did not identify an overall association between ctDNA status at baseline and RFS but did find that the detection of TP53 or KRAS mutations using ctDNA at baseline were associated with a high recurrence risk (p < 0.001, p = 0.02, respectively).

Summary

The prognostic importance of detectable ctDNA after CRT and after surgery has now been demonstrated consistently in several studies of patients with LARC. Compared to the lack of evidence supporting its role as a marker of pathologic response, ctDNA may therefore represent a better marker of systemic MRD, rather than of localised disease. However, one must consider that most of the studies to date have been with small numbers of patients with relatively short periods of follow-up. In addition, how the presence of MRD should alter subsequent management of patients with LARC is not yet clear.

Potential future applications of ctDNA within the evolving treatment landscape for LARC

For patients undergoing neoadjuvant CRT followed by surgery, could (de-)escalating the adjuvant treatment strategy based on MRD status improve patient outcomes?

Selecting the patients with LARC who will derive benefit from the use of AC, in addition to the choice of which chemotherapy agent(s) to use, remain unclear. Prior to the routine use of neoadjuvant CRT before surgery, earlier studies had demonstrated a survival benefit from the use of single-agent fluoropyrimidine AC.³⁷ However, more modern studies using neoadjuvant CRT prior to surgery have not identified a clear survival benefit from this adjuvant approach.⁵ While the addition of oxaliplatin has been shown to improve DFS,⁴ an OS benefit is yet to be proven. However, the use of AC is still recommended by most guidelines.^2,3

Assessing MRD status in patients with LARC may allow clinicians to risk-stratify patients and identify those most likely to benefit from AC while also identifying those who may be less likely to benefit from unnecessary treatment (Figure 2). Such an approach has recently been studied in patients with stage II colon cancer,¹² where a ctDNA-guided management approach resulted in significantly lower rates of AC administration while maintaining non-inferiority in terms of RFS outcomes, compared to standard of care management.

Figure 2.

Potential ways that the role of ctDNA could be studied in patients being treated for LARC. For patients who have received induction chemotherapy and then planned to receive neoadjuvant radiotherapy, the remaining neoadjuvant treatment could be (de)-intensified according to postinduction chemotherapy ctDNA status. Similarly, for patients whose first modality of treatment is radiation, consolidation chemotherapy could be (de)-intensified according to post-radiation ctDNA status. Patients with detectable ctDNA following TNT and surgery are at high risk of recurrence and the potential role of additional novel therapies in such patients has not yet been studied. The potential role of ctDNA in the surveillance of patients who achieve a complete clinical response and do not proceed to surgery has also not yet been well studied. For patients receiving neoadjuvant CRT and surgery, a risk-adapted approach to adjuvant therapy according to post-operative ctDNA status is being studied in clinical trials but results are yet to be published.

Source: Image created with BioRender.com.

CRT, chemoradiotherapy; ctDNA, circulating tumour DNA; LARC, locally advanced rectal cancer; SRT, short course radiation; TNT, total neoadjuvant therapy.

DYNAMIC-RECTAL (ACTRN12617001560381) is an Australian multicentre prospective randomised controlled study of patients with LARC who have completed CRT and surgery. It is comparing a ctDNA-guided approach, where patients receive AC if they have evidence of either MRD as assessed by ctDNA or high pathologic risk, to a standard of care approach where patients receive AC depending on the standard pathologic risk alone. This study has completed recruitment and results are eagerly awaited.

Can ctDNA be used to assess response and adjust treatment in patients receiving TNT?

As the treatment landscape for patients with LARC evolves, many questions regarding the potential role of ctDNA arise. Utilising chemotherapy prior to surgery (TNT) has recently been demonstrated in randomised trials to significantly improve rates of pCR and DFS,^19,22,23 with one study²² also demonstrating an OS benefit. As such, TNT is now recognised as a standard treatment approach according to NCCN guidelines.³ In addition, the sequence of CRT and NAC may be important, with two studies demonstrating significantly higher rates of pCR when NAC is used after (consolidation), rather than prior to (induction) CRT.^20,21

There are several ways in which the utility of ctDNA could be assessed in future TNT trials (Figure 2). First, in patients receiving induction chemotherapy, could CRT be intensified if ctDNA is not cleared? Similarly, if patients have rapid ctDNA clearance, could they be spared from CRT? Omitting RT from multimodality treatment in patients with LARC is currently not-standard. However, given the risk of late RT-induced toxicities including bowel dysfunction, faecal incontinence, genitourinary dysfunction and pelvic fractures,³⁸ there is an interest in designing randomised trials that include arms where RT is not provided.^39,40 One randomised study³⁹ of 495 patients with LARC compared 4–6 cycles of NAC (mFOLFOX6) followed by surgery and 6–8 cycles of AC (mFOLFOX6) to five cycles of NAC (infusional 5FU), three of which were given concurrently with 46.0–50.4 Gy of RT, followed by surgery and a further seven cycles of AC (infusional 5FU). Long-term outcomes did not differ significantly between the two groups, including DFS (HR: 0.94; 95% CI: 0.63–1.42, p = 0.77), rates of locoregional recurrence (HR: 0.95; 95% CI: 0.43–2.12, p = 0.91) and OS (HR: 1.06, 95% CI: 0.57–1.96, p = 0.86).

In patients receiving CRT first, could the subsequent management plan be adjusted according to whether or not ctDNA is cleared (Figure 2)? This may be particularly relevant in patients with rising ctDNA levels which could reflect distant micro-metastatic disease and therefore may warrant intensification of systemic therapy and reconsideration for whether continuing local therapy (RT, surgery) is still appropriate.

Could ctDNA be utilised to assess for local recurrence in patients receiving a non-operative/‘Watch and Wait’ approach?

Patients with LARC who do not proceed to surgery due to achieving cCR from CRT or otherwise due to concerns about operative risk or patient preference, typically undergo close surveillance to detect early recurrence that could still managed with a curative approach. Whether or not ctDNA could be a useful adjunct in the surveillance of these patients has not yet been well studied. Results from observational studies by Khakoo et al.²⁷ (median follow-up 26.4 m) and Murahashi et al.²⁸ (median follow-up not reported) suggest that patients in cCR could undergo non-operative management. Given that the time to local regrowth is typically within 24 months,¹⁷ and the lower sensitivity of ctDNA in detection local compared to distant relapse, future trials with sufficient periods of follow-up may help to better assess the role of ctDNA in surveillance for local recurrence in patients not undergoing surgery.

A recent single institution phase II study⁴¹ has investigated the role of the immune checkpoint inhibitor, dostarlimab as neoadjuvant therapy for patients with mismatch repair deficient (dMMR) LARC. Of the 12 patients who completed 6 months of therapy, after a median follow-up of 12 months, they observed a remarkable cCR rate of 100% and no patients required subsequent CRT or surgery. If this impressive result is validated in an independent larger patient cohort with longer follow-up, this organ-sparing treatment approach will be paradigm changing for this unique molecular subgroup of LARC. Another recent study⁴² examined the role of neoadjuvant pembrolizumab in 35 localised dMMR solid tumours, including 27 patients with colorectal cancer, and observed a pCR rate of 65% among 17 patients who underwent surgery. This study however did observe six progression events, four of which were colorectal patients. The ctDNA correlative analysis suggested that early decrease in ctDNA during pembrolizumab was a predictor of lack of progression. We keenly await further correlative ctDNA results from ongoing or future neoadjuvant immune checkpoint inhibitor studies to define the role of ctDNA for dMMR LARC, such as a response monitoring tool to help select patients for the organ preservation approach and guide treatment duration, or a surveillance strategy for those who have achieved cCR after immune checkpoint inhibition.

Issues with implementing ctDNA in routine clinical practice

Although ctDNA is being increasingly used in clinical studies, it is not yet part of routine clinical care for patients with LARC. While its predictive and prognostic value continues to be elucidated, the heterogeneity between studies in terms of choice of ctDNA assay, timepoint at which ctDNA was collected, treatment received and length of follow-up create uncertainty about how to implement its use in routine patient care. This highlights the importance of incorporating ctDNA analysis in ongoing or future trials in LARC to further understand the ctDNA dynamics during the various phases of neoadjuvant treatment and during surveillance to further inform its potential clinical utility.

Conclusions

The available data are conflicting regarding the ability of ctDNA to predict pCR from CRT and therefore to identify patients who would be most suitable for non-operative management. The most consistent evidence regarding its use relates to its ability to detect MRD in patients who have completed CRT and surgery and predict for poor cancer-related outcomes. However, how to best manage patients with evidence of MRD is unclear until results from randomised ctDNA-guided management trials are available. To gain further understanding of how to use ctDNA in the management of patients with LARC, especially in the era of TNT and organ preservation approaches, dedicated, large, randomised trials assessing its utility are still needed and these will need to adapt to the constantly evolving LARC treatment landscape.