Advancements in molecular-targeted therapies and biomarker detection have improved in recent years [1
]. Such technologies have been successfully applied to people living with metastatic colorectal cancer (mCRC). In particular, the treatment of mCRC has progressed significantly over the past 20 years, with the addition of new chemotherapy to fluoropyrimidine backbones along with the introduction of agents that target vascular endothelial growth factor (VEGF) and molecular aberrations of the cancer cells themselves. Such advancements in molecular-targeted therapies and biomarker detection have resulted in improvements in tumor response rates and overall survival [3
]. The presence of specific biomarkers in these tumors can predict the response to targeted therapies. For example, tumors with wild-type KRAS
(which account for 30%–50% of cases) [5
] are more likely to respond to EGFR-targeted therapies than tumors with mutations in either of these genes [13
]. Other examples of targeted therapies for mCRC include immune checkpoint inhibitors (ICIs) for microsatellite-instability-high (MSI-H) tumors [15
], tyrosine kinase inhibitors for BRAF
], and ERBB2-targeted therapies for ERBB2
]. Therefore, US treatment guidelines recommend testing for mutations in KRAS
, and BRAF
rearrangements; and MSI-high status in patients with advanced CRC [19
Resistance mutations may emerge after treatment with targeted therapies. For example, EGFR-targeted therapy may lead to alterations such as KRAS
, and EGFR
extracellular domain mutations or ERBB2
]. Some of these can be successfully treated with additional targeted therapies, such as trastuzumab-based therapies or trastuzumab deruxtecan (DS-8201) for ERBB2
]. However, for tumors lacking these mechanisms of resistance, repeated treatment with EGFR-targeted therapy is feasible and effective [20
Therefore, genomic assessment of mCRC after progression on initial treatment can guide therapy selection. While tissue biopsies have traditionally been used for such assessment, access to tumor tissue for biomarker testing may be challenging. Furthermore, archived tissue samples may not represent the current state of the tumor and it may be difficult to obtain a fresh tissue biopsy.
Recent advances in next-generation sequencing (NGS) of circulating tumor DNA (ctDNA) have made liquid biopsy a viable testing modality for sequencing mCRC after disease progression. Comprehensive genomic profiling of ctDNA can detect mutations, gene rearrangements, and amplification in genes that may be targets for subsequent therapy or provide information on resistance mechanisms to help clinicians avoid ineffective therapies [28
In this study, we sought to evaluate how clinicians apply ctDNA testing results when making treatment decisions for patients with mCRC that had progressed on prior therapy. Using a real-world database (GuardantINFORMTM, Guardant Health, Redwood City, CA, USA), we analyzed the treatment choices for patients with mCRC who underwent NGS-based ctDNA testing (Guardant360®, Guardant Health) after completing first- or second-line therapy. The goal was to evaluate the real-world application of ctDNA testing and its ability to identify clinically relevant genomic alterations and how such findings influence the subsequent treatment decisions and clinical outcomes for people living with mCRC.
The GuardantINFORM database is fully compliant and deidentified in accordance with the US patient confidentiality requirements set forth in Sections 164.514 (a)–(b)1ii of the Health Insurance Portability and Accountability Act (HIPAA) regarding the determination and documentation of statistically deidentified data. Institutional Review Board approval to conduct this study was not required because this study used deidentified patient records and did not involve the collection, use, or transmission of individually identifiable data.
2.2. Data Source and Extraction
We used secondary data from the GuardantINFORM database, which comprises anonymized genomic data for patients with advanced stage solid tumors in the US who underwent testing with a Guardant360 assay. This assay is certified according to Clinical Laboratory Improvement Amendments and accredited by the College of American Pathologists and has been approved by the US Food and Drug Administration (FDA) and New York State Department of Health for clinical testing of patients with advanced stage (stage III–IV) solid tumors. The assay utilizes hybrid capture technology and NGS to identify genomic alterations in 74 genes [30
]. The panel covers all coding exons in KRAS
, and BRAF
, and detects ERBB2
fusion, and MSI-high status.
In addition to the genetic data, the GuardantINFORM database includes structured commercial payer claims data collected from inpatient and outpatient facilities in academic and community settings. Claims data are provided through a commercial agreement with a data aggregator (the ‘parent dataset’), which collates high-quality industry-standard sources of anonymized patient level data and a further 150 fully privileged payer complete datasets that include a combination of open and closed claims. By percentage, the prescription claim coverage of the full parent dataset by payer type is: Commercial 78%, Medicare 0%, Medicare Advantage 16%, and Medicaid 6%. However, it does not capture clinical features that are not coded as claims, such as the results of tumor biomarker testing from other sources and the clinical response to cancer treatment. Deaths are sourced from third-party providers and aggregated with administrative claims data; for at least half of the CDC-reported deaths in the country, the parent dataset has an encounter within ≤1 month of the date of death. GuardantINFORM is refreshed every quarter. At the time of this study, GuardantINFORM contained records for more than 190,000 patients across 60 cancer types, representing approximately 7000 oncologists from US-based academic and community cancer practices.
2.3. Study Population
For this study, we included all CRC patients who underwent testing between 1 June 2014 and 30 September 2021, with a Guardant360 assay performed within 90 days prior to initiating second or third-line systemic therapy, had a 6-month clean window of claims information (≥2 medical or pharmacy claims) without treatment prior to the first observed therapy date, and had claims data indicating the presence of metastatic disease. The clean window was implemented to ensure that the first observed systemic therapy date corresponds to first-line treatment. If patients underwent ctDNA testing more than once after initial treatment, data from the test conducted immediately after the earliest line of targeted therapy were analyzed; otherwise, the earlier test was used regardless of prior therapy.
2.4. Lines of Therapy
The patients’ treatment regimens were reconstructed over the study period based on the claims data. All systemic cancer therapies recommended for mCRC in the NCCN guidelines [19
] were included. The same regimens used within 90 days from the end of one to the start of the next administration date were combined and considered the same line of therapy. Any treatment modifications or additions to targeted therapies more than 60 days after treatment initiation were considered a new line of therapy, whereas changes to chemotherapies more than 15 days after treatment initiation were considered a new line of therapy. Changes in treatments involving fluorouracil or bevacizumab were not considered as new therapy. Discontinuing a medicine without adding a new medicine to a regimen was not considered a new line of therapy.
2.5. Genomic Profiling
Clinically informative genomic alterations were those classified as Therapeutic Levels 1, 2 or R1 by OncoKB [32
]; namely, KRAS
point mutations, BRAF
fusion, and MSI-high status. The Guardant360 assay used at the time of data collection did not include NTRK2
fusions, nor did it assess tumor mutation burden status.
For determination of actionable genomic profiles and molecularly informed therapy, only the results of Guardant360 tests were considered because the GuardantINFORM database did not include biomarker results for other testing modalities. Samples with detectable ctDNA but neither KRAS
mutations were classified as having an actionable molecular profile, and patients with such profiles were categorized as appropriate candidates for EGFR-targeted therapy except in the case of concurrent BRAF
V600E or primary tumor location in the right colon [33
]. Immune checkpoint inhibitors were considered molecularly informed only if MSI-high were detected. BRAF-directed therapies were considered informed for BRAF
V600E, ERBB2-targeted therapies were considered informed for ERBB2
amplification, and NTRK inhibitors were considered informed for NTRK1
fusions. Chemotherapy and/or VEGF-targeted therapy were classified as molecularly informed in the absence of ERBB2
fusion, and MSI-high or the presence of a mutation in KRAS
without other actionable biomarkers. Non-targeted therapy was also considered informed for patients with right-sided tumors and ctDNA profiles lacking RAS
mutations or other actionable alterations. Throughout this manuscript, “targeted therapy” refers to systemic anti-cancer agents that are targeted primarily against aberrant proteins expressed by cancer cells or to immune checkpoint inhibitors; VEGF-targeted therapy is not included in this definition.
2.6. Patient Outcomes
As a surrogate for progression-free survival, we evaluated the time to treatment discontinuation (TTD) for each line of therapy, which was defined as the time from the first day therapy to the estimated last day of therapy, the last claim activity date, or death, whichever occurred first. Patients with a last claim activity date prior to the end of therapy or <90 days after the end of therapy were censored due to lack of follow-up information. We also evaluated overall survival, which was defined as the time from the ctDNA report date prior to the start of second- or third-line therapy (for second- or third-line therapy patients, respectively) until death. Patients were censored at last claim activity date.
2.7. Statistical Analysis
All patients in the study population were included in the baseline descriptive statistics. Means, medians, and standard deviations were calculated for continuous variables whereas counts and proportions were calculated for categorical variables. Proportions were compared using the two proportion z-test or Fisher’s Exact test for small samples.
The TTD for anti-cancer therapies administered after ctDNA testing was evaluated using the Kaplan–Meier method for the following four subgroups of patients: (1) patients with a clinically actionable genomic profile who received the matched therapy, (2) patients with a clinically actionable genomic profile who did not receive a matched therapy, (3) patients without a clinically actionable genomic profile, and (4) patients in whom ctDNA was not detected. Log-rank tests were used to compare TTD across these subgroups. In addition, a Cox proportional hazards model was performed with these four subgroups to generate pairwise hazard ratios (HRs) with 95% confidence intervals. A p-value of <0.05 was considered statistically significant.
Subgroup analyses were performed for patients who received EGFR-targeted therapy or immune checkpoint inhibitors, and the TTD was compared between patients whose ctDNA profiles supported the choice of the targeted treatment (matched) and those whose profiles did not support such treatment (unmatched).
All analyses were conducted using SAS software package 9.4 (SAS Institute, Cary, NC, USA).
For patients with mCRC and disease progression, we analyzed a real-world medical claims database combined with ctDNA test results to assess the types of therapeutic decisions made based on ctDNA profiling and the clinical outcomes associated with those decisions.
The assay used in this analysis detected ctDNA in 93.7% of the samples, a similar result to that previously reported for plasma samples from a large cohort of mCRC patients tested in routine clinical practice [34
]. Clinically actionable ctDNA profiles were detected in approximately half (507/1064, 47.7%) of all patients tested, not only among those who had received targeted therapy immediately prior to testing (97/138, 70.3%) but also in those with disease progression after non-targeted therapy (410/926, 44.3%).
The frequencies of informative genomic alterations detected in this study were generally consistent with those described in previous studies [35
mutations were the most common (424 of 997 samples with ctDNA detected, 42.5%), followed by BRAF
V600E (7.3%), NRAS
(5.2%), and ERBB2
amplification (2.8%). MSI-high assessment was added to the ctDNA assay in 2018, but the database includes patients tested from 2014; therefore, the observed proportion of samples with MSI-high status (2.0%) most likely underestimates the true prevalence, which is expected to be up to 5% [40
]. Among patients with detectable ctDNA, the proportion of patients with KRAS
mutations was lower in those who had previously received targeted therapy (21.5%) than in those who had not (45.8%), presumably due to prior selection for RAS
wild-type profiles in consideration of EGFR-targeted regimens, which were the most commonly used targeted therapies prior to ctDNA testing. However, there was a trend for more NRAS
mutations among patients exposed to prior targeted treatment (8.9%) vs. patients not exposed to prior targeted therapy (4.6%) (p
= 0.063). As expected, prior targeted therapy, which was overwhelmingly EGFR-targeted, increased the proportion of cancers with BRAF
V600E (12.6% vs. 6.5%, p
= 0.019) and ERBB2
amplification (5.9% vs. 2.3%, p
Targeted therapies were more likely to be administered to patients with actionable ctDNA profiles (238/507, 46.9%) than to those without actionable profiles (100/557, 18.0%), but only slightly more than one-third of patients with actionable profiles received appropriately matched therapy (38.1%, 193/507). Excluding patients in whom ctDNA could not be detected, molecularly informed therapy, targeted or not, was administered to 60.5% of patients (603/997). However, because the claims database used in this study does not include results of other biomarker tests, we cannot exclude the possibility that some treatment decisions were molecularly informed by other assays.
In one recent study of 78 patients with mCRC, alterations that could be targeted by available therapies were detected using ctDNA in 59 (75.6%), and 45 satisfied OncoKB criteria in a manner similar to that employed in our study [36
]. In another study of 94 patients who underwent ctDNA testing, non-synonymous mutations were detected in 74 patients (79%), of whom 69 (73%) had potentially actionable alterations [37
]. In a study that used cell-free DNA (cfDNA) samples from 128 patients with mCRC, somatic alterations were detected in 100 (78%), of which 50% had potentially actionable alterations [41
]. However, even under controlled study circumstances, targeted therapy may not be administered to all potentially eligible patients. For example, in the MOSCATO 01 trial in which NGS was used to guide treatment decisions in a cohort of 948 patients with advanced cancers who underwent tumor biopsy, a molecular portrait was developed for 843. An actionable alteration was found in 411 patients, but only 199 (48.4%) received a targeted therapy matched to the alteration [42
Our findings support the clinical utility of matched therapy for mCRC patients with actionable molecular profiles detected by ctDNA. Median TTD, a surrogate for progression-free survival, was longest for patients with actionable profiles who received matched targeted treatment (5.3 months) than in those with actionable profiles who received unmatched treatment (3.9 months). Patients without actionable ctDNA profiles or those in whom ctDNA was not detected had the shortest median TTD (3.5 months each). These results are consistent with those of other studies demonstrating the importance of matching treatment to tissue-based molecular profiles [36
]. The survival benefit of molecularly matched therapy based on tissue-based testing was previously reported [43
]. However, the utility of ctDNA genotyping for molecularly matched therapy in real-world clinical practice, especially for mCRC patients following prior targeted therapies, has not been fully addressed. The clinical outcomes of 35 mCRC patients from a single institution suggested that matched therapy based on ctDNA results achieved a higher tumor response rate than unmatched therapy, with a trend for longer progression-free survival [37
]. Our study, which included patients from academic and community-based settings, confirmed the usefulness of ctDNA genotyping for identification of patients who would benefit from targeted therapies.
One important role of ctDNA testing prior to later-line therapies is identification of resistance mutations. For example, testing identified patients who are less likely to respond to EGFR-targeted therapy. We observed that TTD following EGFR-targeted therapy was significantly prolonged when treatment was matched to the ctDNA profile. Furthermore, even though the absence of detectable ctDNA in plasma samples of mCRC patients is associated with a better prognosis [47
], the TTD was longer in patients with a clinically actionable genomic profile who received matched therapy than in ctDNA-negative patients. This underscores the efficacy of matched targeted therapy based on ctDNA genotyping.
Although ctDNA testing identified patients who are less likely to benefit from EGFR-targeted treatments, this class of therapy appeared to be under-utilized in this cohort. A small proportion of patients received EGFR-targeted therapies immediately prior to ctDNA testing (125/1064, 11.7%), although it is expected that about 40% should be eligible [9
]. After ctDNA testing, 248 patients with molecular and clinical profiles supporting EGFR-targeted therapies received other types of therapies instead. For most of these, there was no prior history of EGFR-targeted treatment. Although chemotherapy with or without anti-VEGF therapy can be effective in this setting, clinicians and patients should be aware of alternative treatment options. It is also important to consider that ctDNA can identify patients who should not receive EGFR-targeted therapy. The median TTD was shorter for patients treated with EGFR-targeted agents with unfavorable ctDNA results (77/233 treated; 3.2 months) compared with patients whose treatment was matched to the ctDNA profile (5.5 months).
Most patients treated with immune checkpoint inhibitors did not have MSI-high detected by ctDNA (62/78, 79.5%). There was a trend toward longer median TTD among patients treated with immunotherapy and MSI-high samples (6 months) compared to those without MSI-high (4.2 months). However, this was not statistically significant due to the small number of patients with MSI-high and treated with immune checkpoint inhibitors. The ctDNA assay used in this study introduced MSI-high assessment several years into the observation period and did not include all potential genomic information that could inform the suitability of immunotherapy. For example, POLD1
are not in the panel. Mutations in these genes, which occur in <1% of cases, are associated with response to immune checkpoint inhibitors [40
]. Furthermore, the assay used in this study did not assess for the presence of high tumor mutation burden, which can be detected in approximately 3% of patients with mCRC without MSI-high [40
]. The US FDA approved pembrolizumab for the treatment of solid tumors with a high tumor mutation burden (defined as ≥10 mutations/megabase), as determined by an FDA-approved test, in patients with disease progression after prior treatment and who have no satisfactory alternative treatment options [49
]. However, standard treatment was not exhausted by over one-quarter of patients without MSI-high (18/62) who received an immune checkpoint inhibitor. Furthermore, even among mCRC patients with a high tumor mutation burden in the absence of MSI-high, the effectiveness of immunotherapy remains controversial [39
There are some limitations of this study. In particular, the database lacked complete clinical information, such as performance status of the patients, physician-assessed tumor response, and results of concurrent biomarker testing that may have informed clinical decisions. Furthermore, because all treatments were identified using healthcare claims data, the database did not evaluate participation in clinical trials. Despite these factors, the genomic landscape observed in this study is generally consistent with that reported in other studies [51
This study was designed to determine the proportion of patients for whom the potential use of targeted therapy would be appropriate following ctDNA testing. Therefore, our definition of molecularly informed therapy was intentionally narrow. Many of the treatment decisions made in this real-world setting may be valid, considering additional clinical information that was not available when these patients were treated. Tumor molecular information is necessary to make informed treatment decisions for patients with mCRC, but it should not be the only consideration.
In summary, this analysis of a real-world database demonstrates the clinical value of plasma-based comprehensive genomic profiling in mCRC patients with disease progression after one or two lines of prior therapy. Analysis of ctDNA identifies patients who are candidates for new or further targeted therapy, regardless of prior treatment history. Most importantly, in standard clinical practice, patients who received targeted treatments that matched their ctDNA test results had more favorable treatment outcomes than patients who received unmatched therapies. The modest improvement in TTD with the use of matched therapy for actionable alterations underscores the need for continued improvement in the identification of predictive biomarkers and the development of therapeutic agents to target such biomarkers. Future analyses of real-world clinical practice databases will help to further characterize the evolving role of ctDNA testing and its potential to influence the clinical outcomes of people living with advanced CRC.