Discussion
To the best of our knowledge, there have been only 3 studies to demonstrate that rises and declines in plasma ctDNA levels predicted radiographic tumor progression and response, respectively, in MSS or MSI-H mCRC treated with immunotherapy (9-11). In one study, changes in plasma ctDNA levels 4 weeks from initiation of immunotherapy was predictive of radiographic response, while undetectable ctDNA at week 8 from immunotherapy was associated with prolonged progression-free survival and overall survival in another study (9,11).
Our case series supports the use of plasma ctDNA as a dynamic marker of response to immunotherapy in mCRC. In all cases, a decrease or clearance of ctDNA corroborated radiographic tumor response, while rises in plasma ctDNA levels corroborated radiographic tumor progression to immunotherapy. In cases where colorectal tumors produced CEA, ctDNA levels aligned with changes in CEA, and furthermore, both were in concordance with radiographic response assessments. Our findings are consistent with the literature showing that plasma ctDNA correlates with conventional measures of tumor response to systemic therapy (by CEA and imaging) in mCRC (12). However, unlike CEA, wherein measurable levels can be found in normal individuals but a proportion of colorectal tumors can be non-secretors of CEA, ctDNA as a measure of minimal residual disease (MRD) is a binary metric whose presence indicates that a large burden of residual metastatic cancer cells remain in one's body (13). It is therefore not surprising that there is growing evidence to suggest that ctDNA demonstrates superior sensitivity to CEA to detect recurrence in non-metastatic CRC and radiographic progression events in mCRC (13-16). In mCRC, changes in ctDNA levels from pretreatment to cycle 2 was a better predictor of radiologic response to chemotherapy than CEA at this early time point (17). In resected, stage I-III CRC, ctDNA detection offers median lead times of up to 11 months for radiographic detection of recurrence (18). A recent large observational cohort demonstrated that in those with postoperative ctDNA positivity following curative-intent surgery for stage I-IV CRC, clearance of MRD with adjuvant chemotherapy was also a positive prognostic indicator (19).
Importantly, we report a novel finding that plasma ctDNA levels predicted radiographic response (or lack of) to immunotherapy in CEA non-producers. CEA represents the only conventional and recognized blood-based test for monitoring response to systemic therapy in mCRC. Notably, up to 34% of patients with CRC are CEA non-producers, which represents a clinically relevant proportion of patients without a marker to monitor systemic treatment response short of interval surveillance imaging (6). In Case 3, plasma mSEPT9 ctDNA levels were detectable at baseline but subsequently cleared over the course of treatment with pembrolizumab and corresponded to a complete radiographic response by cycle 4 of immunotherapy (Fig. 3). CEA was uninformative in this case as the patient had low values of CEA throughout the course of immunotherapy (non-producer). Conversely, in Case 4, CEA levels were fairly low and would not have indicated the degree of rapid disease progression that was detected by interval imaging and a substantial rise in plasma ctDNA levels from baseline to cycle 4 of immunotherapy. This patient died from progressive disease shortly thereafter.
When purposed for detection of tumor mutations through targeted next-generation sequencing (NGS), ctDNA can be used to track dynamic changes in mutations in response to systemic therapy to detect presence of resistance mutations (12). However, in our series we have used a tumor-agnostic, methylated ctDNA marker (mSEPT9) and therefore its assessment of systemic tumor burden is not influenced by tumor mutational profile or prior therapies. When purposed for MRD detection, the timing of therapy to the collection of blood for ctDNA measurement can influence ctDNA levels. For example, surgery can lead to elevation in ctDNA levels lasting as long as 4 weeks from surgery, which is why many groups have recommended postoperative measures of ctDNA to not occur until 4 weeks after CRC surgery (12,20). Chemotherapy can also cause changes in ctDNA as early as 48 h of chemotherapy infusion, although in another series most changes in ctDNA occurred by cycle 2 of chemotherapy (17,21). Early spikes in ctDNA levels within the first few days of chemotherapy infusion may represent a rapid release of ctDNA from lysing tumors (17). The collection of blood samples in our 4 cases for ctDNA assessments occurred prior to the infusion of immunotherapy and before the first dose of oral chemotherapy on each day 1 of the respective cycle of therapy.
Although promising and durable responses have been observed with immune checkpoint blockade in MSI-H mCRC, it should be noted that the median time to onset of response is 2.2 months (range 1.8 to 18.8 months) (1). Therefore, during this initial critical period up until the first interval assessment scan, having a convenient and minimally invasive biomarker that can reliably monitor tumor response to immunotherapy can be of immense benefit to the physician while easing patient anxiety, particularly in the absence of high CEA levels to begin with. This holds especially true in non-metastatic settings where neoadjuvant immunotherapy has been used in MSI-H locally advanced rectal cancer given increasing evidence that these tumors are fairly resistant to standard chemotherapy and chemoradiation (22). Neoadjuvant immunotherapy has been explored in early-stage colon cancers as well (23). Here, serial plasma ctDNA assessments might provide an early indicator of benefit to neoadjuvant immunotherapy and potentially allow a timely transition to salvage chemoradiation or surgery, while preserving a curative intent treatment pathway in the event of lack of tumor response to immunotherapy.
Our group of mCRC cases demonstrates that longitudinal plasma ctDNA assessments can predict true progressors to immunotherapy, even in the context of a CEA non-producing tumor. This is noteworthy, given that there are increasing efforts to consistently detect the phenomenon of pseudoprogression to immunotherapy (24). As evidence accumulates in support of ctDNA being a more sensitive measure of changes in tumor burden than conventional approaches in CRC, it would be prudent to further investigate its role in the detection of pseudoprogression and true progression in larger, prospective cohorts of mCRC patients treated with immunotherapy (7). There are ongoing prospective studies seeking to evaluate the impact of tumor response assessments using plasma ctDNA assays in patients with mCRC and other advanced solid tumors treated with immunotherapy that will hopefully provide more insight in this context (NCT04761783).
Plasma ctDNA represents a minimally invasive tool with promising potential as a measure of tumor burden and tumor response to systemic therapies in CRC. In this case series, we demonstrate that blood-based assessments of mSEPT9 ctDNA predicted radiographic response to immunotherapy in patients with mCRC that were MSS or MSI-H. Interestingly, plasma ctDNA was a predictor of response to immunotherapy even in colorectal tumors that were CEA non-producers. Our findings add to ongoing efforts to establish the role of plasma ctDNA in monitoring response to immunotherapy in CRC. Future studies are warranted to investigate the potential of plasma ctDNA to detect hyperprogressors or differentiate true progression from pseudoprogression in the context of immunotherapy-based treatments in CRC.