Colorectal Cancer Recurrence Prediction Using a Tissue-Free Epigenomic Minimal Residual Disease Assay

Abstract

Purpose: Posttreatment detection of ctDNA is strongly predictive of recurrence. Most minimal/molecular residual disease assays require prior tissue testing to guide ctDNA analysis, resulting in lengthy time to initial results and unevaluable patients.

Experimental design: We assessed a tissue-free assay (Guardant Reveal) that bioinformatically evaluates >20,000 epigenomic regions for ctDNA detection in 1,977 longitudinally collected postoperative plasma samples from 342 patients with resected colorectal cancer.

Results: We observed sensitive and specific detection of minimal/molecular residual disease associated with clinically meaningful differences in recurrence-free intervals at each time point evaluated with a median lead time of 5.3 months. The longitudinal sensitivity in stage II or higher colon cancer was 81%. Sensitivity increased with serial measurement and varied by recurrence site: higher for liver (100%) versus lung (53%) and peritoneal (40%). Sensitivity among patients with rectal cancer was 60% owing to a high proportion of lung metastases. Specificity was 98.2% among 1,461 posttreatment samples (99.1% among those with follow-up longer than the upper IQR of the lead time observed in this study).

Conclusions: Our data demonstrate the potential clinical utility of ctDNA as a tool to improve the management of stage II and higher colorectal cancer with a methodology that is noninvasive, accessible, and allows for rapid evaluation to inform clinical decisions.

Figure 1.

Longitudinal detection of ctDNA is associated with a high risk of recurrence and lead time before standard radiographic detection and is more common in patients with high-risk clinicopathologic features. A, Kaplan–Meier estimates of RFI based on whether ctDNA was ever or never detected during longitudinal surveillance after the completion of all treatments (samples collected before adjuvant chemotherapy completion excluded). B, The median lead time among 23 samples with ctDNA detected before clinical relapse was 5.3 months (IQR, 3.0–16.4 months) but ranged up to 28.7 months. C, Postoperative ctDNA was detected at any time point in 15.3% of patients and strongly associated with T and N stage and trended toward significance among patients with lymphovascular or perineural invasion. ACT, adjuvant chemotherapy; LVI, lymphovascular invasion; MSI, microsatellite instability; MSS, microsatellite stability; PNI, perineural invasion.

Figure 2.

Heatmaps illustrating DMR. More than 20,000 unique DMR and control regions are included in the Reveal assay. A representation of ∼1,700 common DMR in colorectal cancer is illustrated (A) for patients without recurrence and (B) for patients with recurrence. Each row represents a DMR, and each column represents a plasma sample. For this figure, methylation data from the most recent posttreatment sample were used for illustrative purposes. The blue shading denotes the relative number of methylated molecules at each region normalized by universally methylated control regions (darker colors indicate a higher concentration of methylated molecules). A bioinformatic algorithm classifies each sample as ctDNA detected (1) or not detected (0) based on a predefined statistical likelihood threshold that the patterns of methylation are tumor derived (i.e., positive samples have higher concentrations of DMR). The median number of DMR per positive sample in this study was 994 (IQR, 792–1,203).

Figure 3.

Sensitivity to detect recurrence increases with serial sampling and varies by site of recurrence. A, Kaplan–Meier estimate of cumulative sensitivity to detect ctDNA in 41 patients with stage II or higher disease who experienced recurrence during surveillance follow-up analyzed based on whether the patient had colon or rectal cancer. Events were recorded at the first postoperative detection of ctDNA. Patients without ctDNA detected were censored at the time of their last sample collection before or at the time of recurrence if a recurrence sample was available. B, Sensitivity to detect recurrence varied by site of recurrence and increased with longitudinal sampling following the day-28 postsurgery time point (comparison by the two-tailed Fisher exact test). In many cases, ctDNA was detectable the clinical detection of relapse. Patients with multiple sites of metastasis included lung + renal (ctDNA⁺), liver + lung (ctDNA⁺), peritoneum + lymph node (ctDNA⁺), lung + locoregional (ctDNA⁻), and pleura + lymph node (ctDNA⁻). C, The sensitivity by site of metastasis in this study is similar to what has been reported in three published studies of tissue-informed ctDNA assays in resected colorectal cancer that have data on ctDNA detection by site of metastasis (–27). In Reinert 2019, “carcinosis” is categorized under “multiple,” and one patient with lung metastasis with ctDNA detected 5 months after recurrence after systemic treatment had started is not included as ctDNA detected in this figure. D, Recurrences in rectal cancer were more likely to involve the lung only compared with recurrence from colon cancer (55% vs. 19%; P = 0.02; comparison by the two-tailed Fisher exact test), but sensitivity by the site of metastasis was similar for colon and rectal cancers (E). F, Time to recurrence from surgery was longer for metastases to the lung vs. liver, which might contribute to differences in the proportion of patients with recurrent disease who had detectable ctDNA at the day-28 time point (comparison by the Mann–Whitney test). CNS, central nervous system.

Figure 4.

Risk of recurrence based on ctDNA detection 28 days postsurgery for patients with stage II or higher colorectal cancer. A, Forest plot of the Cox proportional-hazards analysis of clinical factors and their association with RFI. *, P < 0.05; **, P < 0.0001. B–D, Kaplan–Meier estimates of RFI for (B) colorectal cancer combined, (C) colon cancer only, and (D) rectal cancer only. Forty-eight percent of patients with colon cancer and 65% of patients with rectal cancer received subsequent adjuvant chemotherapy. ACT, adjuvant chemotherapy; LVI, lymphovascular invasion; Neg, negative; Pos, positive.

Figure 5.

Risk of recurrence based on ctDNA detection following adjuvant chemotherapy in patients with stage II or higher colorectal cancer. Kaplan–Meier estimates of RFI for (A) the first available sample postadjuvant chemotherapy completion (median, 53 days) and (B) based on paired ctDNA status before and after chemotherapy. Patients who consistently tested negative for ctDNA had significantly improved RFI compared with those who cleared ctDNA in pairwise analysis (HR, 0.22; 95% CI, 0.03–1.52; P = 0.0041). Forty-two percent of patients had rectal cancer, and 58% had colon cancer; 25% stage II, 68% stage III, and 7% clinical stage III/pathologic stage IV.