Molecular residual disease and efficacy of adjuvant chemotherapy in patients with colorectal cancer

Abstract

Despite standard-of-care treatment, more than 30% of patients with resectable colorectal cancer (CRC) relapse. Circulating tumor DNA (ctDNA) analysis may enable postsurgical risk stratification and adjuvant chemotherapy (ACT) treatment decision-making. We report results from GALAXY, which is an observational arm of the ongoing CIRCULATE-Japan study (UMIN000039205) that analyzed presurgical and postsurgical ctDNA in patients with stage II-IV resectable CRC (n = 1,039). In this cohort, with a median follow-up of 16.74 months (range 0.49-24.83 months), postsurgical ctDNA positivity (at 4 weeks after surgery) was associated with higher recurrence risk (hazard ratio (HR) 10.0, P < 0.0001) and was the most significant prognostic factor associated with recurrence risk in patients with stage II or III CRC (HR 10.82, P < 0.001). Furthermore, postsurgical ctDNA positivity identified patients with stage II or III CRC who derived benefit from ACT (HR 6.59, P < 0.0001). The results of our study, a large and comprehensive prospective analysis of ctDNA in resectable CRC, support the use of ctDNA testing to identify patients who are at increased risk of recurrence and are likely to benefit from ACT.

Fig. 1. Study design and population.

a, Overview of CIRCULATE-Japan study, illustrating the observational GALAXY protocol with sample collection schema. b, CONSORT (Consolidated Standards of Reporting Trials) diagram illustrating patient inclusion and exclusion criteria for sub-analyses.

Fig. 2. ctDNA-based MRD testing is predictive of survival outcomes in postsurgical patients with CRC.

a, Kaplan–Meier estimates for DFS stratified by ctDNA-negative and ctDNA-positive status 4 weeks after surgery (pathological stages I–IV). HRs and 95% CIs were calculated using the Cox proportional hazard model. P values were calculated using the two-sided log-rank test. 6M-DFS, 6-month DFS; 12M-DFS, 12-month DFS; 18M-DFS, 18-month DFS. b, Forest plot depicting the multivariate analysis for recurrence in patients with pathological stage II–III CRC. Various prognostic factors and their association with DFS, as indicated by HR, were analyzed across the cohort using the two-sided Wald chi-squared test. The unadjusted HRs (squares) and 95% CIs (horizontal lines) are shown for each prognostic factor. Vertical dotted line, the null hypothesis. Number of events = 102; global P value (log-rank) = 2.4188 × 10⁻²⁷; Akaike information criterion = 1,183.72; concordance index = 0.81. c, Correlation between ctDNA status and CEA status 12 weeks after surgery.

Fig. 3. ctDNA-based MRD testing is predictive of response to ACT in postsurgical patients with CRC.

a,b, Kaplan–Meier estimates for DFS stratified by observation and ACT in patients with high-risk pathological stage II or stage III disease and ctDNA positivity 4 weeks after surgery (a) (HR was adjusted by sex and performance status) and ctDNA negativity 4 weeks after surgery (b) (HR was adjusted by age, pathological stage, MSI and performance status). HRs and 95% CIs were calculated using the Cox proportional hazard model. P values were calculated using the two-sided log-rank test.

Fig. 4. ctDNA-based treatment response monitoring is possible in postsurgical patients with CRC.

Kaplan–Meier estimates for DFS according to ctDNA dynamics from 4 weeks to 12 weeks after surgery. Landmark analysis was performed 12 weeks after surgery. HRs and 95% CIs were calculated using the Cox proportional hazard model. P values were calculated using the two-sided log-rank test.

Fig. 5. Cumulative incidence of ctDNA clearance in patients with pathological stage I–IV CRC.

a, Kaplan–Meier estimates indicating cumulative clearance in patients stratified by ACT and observation. Differences in cumulative incidence were investigated using two-sided Gray’s test. b, Kaplan–Meier estimates for DFS for ctDNA-positive patients 4 weeks after surgery according to their ctDNA clearance status on ACT. HRs and 95% CIs were calculated using the Cox proportional hazard model. P values were calculated using the two-sided log-rank test. HR was adjusted by sex, performance status and pathological stage.

Extended Data Fig. 1. Genes Selected for ctDNA assay.

A total of 8,374 genes were selected for 1,039 patients.

Extended Data Fig. 2. ctDNA is Elevated in MRD-Positive Patients.

Average ctDNA concentration, in mean tumor molecules/mL of plasma (MTM/mL), across stages (stage I (red, stage II (blue), stage III (green), and stage IV (grey) at the following time points was assessed at: A. Prior to surgery, B. 4 weeks post-surgery (MRD timepoint), C. Patients who cleared their ctDNA at the MRD time point were observed to have lower presurgical ctDNA levels. D. Patients were further stratified by ACT induced ctDNA clearance beyond the 4-weeks post-surgical MRD time point across stages. Patients who cleared ctDNA during ACT had lower ctDNA at the MRD time point. Box plots were generated using ggplot2 package v3.3.6 in R v4.2.1. Center line, median MTM/ml; bottom of box, 25% quantile; top of box, 75% quantile; data points represent outliers. P-values were calculated using the two-sided t-test.

Extended Data Fig. 3. ctDNA-based MRD testing is Prognostic of Survival Outcomes in Post-Surgical CRC Patients.

Kaplan-Meier estimates for disease-free survival stratified by ctDNA-negative (blue) and ctDNA-positive (red) status at 4-weeks post-surgery in patients by pathological stage: A. Stage I, B. Stage II, C. Stage III, and D. Stage IV CRC. Hazard ratio (HR) and 95% confidence interval (CI) were calculated by the Cox proportional hazard model. P-value was calculated using the two-sided log-rank test.

Extended Data Fig. 4. Presurgical ctDNA-based MRD testing is not Prognostic of Survival Outcomes in CRC Patients.

Kaplan-Meier estimates for disease-free survival stratified by ctDNA-negative (blue) and ctDNA-positive (red) status before surgery. Hazard Ratio (HR) and 95% confidence interval (CI) were calculated by the Cox proportional hazard model. P-value was calculated using the two-sided log-rank test.

Extended Data Fig. 5. ctDNA-based MRD testing is Predictive of Response to Adjuvant Chemotherapy in Post-Surgical Patients with Colorectal Cancer.

Kaplan-Meier estimates for disease-free survival stratified by observation (red) and adjuvant chemotherapy (blue) in patients with ctDNA positivity at 4-week after surgery, by stage: A. High-risk stage II patients; B. Stage III patients C. Stage IV patients. Hazard ratio (HR) was adjusted by sex, and performance status. Hazard ratio and 95% confidence interval (CI) were calculated by the Cox proportional hazard model. P-value was calculated using the two-sided log-rank test D. Forest plot depicting the multivariate analysis for recurrence in pathological-stage II-IV ctDNA-positive patients. Various prognostic factors and their association with disease-free survival, as indicated by hazard ratio were analyzed across the cohort using two-sided Wald chi-square test. The unadjusted hazard ratio (squares) and 95% CIs (horizontal lines) are shown for each prognostic factor. Vertical dotted line indicates the null hypothesis.

Extended Data Fig. 6. Landmark analysis at 8 weeks representing ctDNA-based MRD testing to be Predictive of Response to ACT in Post-Surgical CRC Patients.

Kaplan-Meier estimates for disease-free survival stratified by observation (red) and adjuvant chemotherapy (blue) in pathological high-risk stage II/Stage III CRC patients with A. ctDNA positivity at 4-weeks after surgery; HR was adjusted by sex, and performance status. B. ctDNA negativity at 4-weeks after surgery; HR was adjusted by age, p-stage, MSI and performance status. A landmark analysis at 8 weeks was implemented to address the immortal time bias, whereby DFS was measured starting from day 60. Hazard ratio (HR) and 95% CI were calculated by the Cox proportional hazard model. P-value was calculated using the two-sided log-rank test.

Extended Data Fig. 7. ctDNA-based MRD testing is Predictive of Response to ACT in Oligometastatic stage IV ctDNA-positive patients who did not receive neoadjuvant chemotherapy.

Kaplan-Meier estimates for disease-free survival stratified by observation (red) and adjuvant chemotherapy (blue) in oligometastatic stage IV patients with A. ctDNA positive patients who received neoadjuvant therapy; HR was adjusted for gender. B. ctDNA-positive patients who did not receive neoadjuvant chemotherapy; HR was adjusted for gender and performance status. A landmark analysis at 8 weeks was implemented to address the immortal time bias, whereby DFS was measured starting from day 60. HR and 95% CI were calculated by the Cox proportional hazard model. P-value was calculated using the two-sided log-rank test.