Circulating tumor DNA analysis guiding adjuvant therapy in stage II colon cancer: 5-year outcomes of the randomized DYNAMIC trial

Abstract

Early data from the DYNAMIC study of circulating tumor DNA (ctDNA)-guided adjuvant chemotherapy (ACT) versus standard approach met its primary outcome demonstrating reduced ACT use without compromising 2-year recurrence-free survival (RFS) for stage II colon cancer. We report here other prespecified analyses of overall survival, ctDNA clearance and ctDNA level. At a median follow-up of 59.7 months, 5-year RFS was 88% and 87% with ctDNA-guided and standard management, respectively (difference 1.1%, 95% confidence interval −5.8% to 8.0%), and 5-year overall survival is similar (93.8% versus 93.3%, hazard ratio (HR) 1.05; P = 0.887). For treated ctDNA-positive patients, ctDNA clearance was observed at the end of ACT (EOT) in 35 out of 40 patients (87.5%). A higher than median postoperative tumor-derived mutant molecules per milliliter plasma was associated with worse 5-year RFS (HR 10.62; P = 0.005). For treated ctDNA-positive patients, post hoc analysis of ctDNA clearance at EOT assessed by a new assay that evaluated an average of 29 tumor-derived mutations per patient predicted for a favorable 5-year recurrence-free probability of 97% versus 0% for ctDNA persistence (P < 0.001). Mature DYNAMIC outcome data support a ctDNA-guided approach to ACT for stage II colon cancer, with potential to further risk stratify ctDNA-positive patients based on ctDNA burden and EOT results. Australian New Zealand Clinical Trials Registry Identifier: ACTRN12615000381583.

Circulating tumor DNA (ctDNA) is a promising marker of molecular residual disease across various solid tumor types^1–5, including curatively resected colorectal cancer, where initial observational studies demonstrated that ctDNA detection was associated with a recurrence risk approaching 100% in untreated patients⁶. The DYNAMIC study randomized 455 patients with stage II colon cancer to a ctDNA-guided approach to adjuvant chemotherapy (ACT), where ctDNA-positive patients received ACT and ctDNA-negative patients were observed, or to standard management based on conventional clinicopathologic criteria. The study’s primary analysis reported a reduced proportion of patients receiving chemotherapy when guided by postoperative ctDNA results (15%) compared to standard management (28%), without compromising 2-year recurrence-free survival (RFS)⁷.

Patient selection for ACT for stage II colon cancer has traditionally been guided by the presence or absence of high-risk clinicopathologic features. This approach has limitations, as high-risk features, such as T4 status, although associated with increased recurrence risk, are not clearly associated with treatment benefit^8–11. For the ~20% of patients with stage II colon cancer with deficient mismatch repair (dMMR), the prognosis is excellent, and current guidelines do not recommend adjuvant therapy in these patients^12,13. Although individual gene mutations within the tumor do not currently inform routine care, they are of potential interest in early-stage disease, with for example KRAS and BRAF mutations reported to have negative prognostic impact¹⁴. The relationship between these mutations and postsurgery ctDNA status has previously not been examined in early-stage colon cancer cohorts.

Beyond the binary read-out of the presence or absence of tumor-derived DNA, ctDNA can also be quantified as a potential measure of the bulk of minimal residual disease. Such analyses of molecular burden could further stratify post-surgery recurrence risk in those with a positive ctDNA result. Repeated measurement of ctDNA following adjuvant therapy may also represent a useful real-time marker of adjuvant treatment benefit or resistance, with data from multiple observational studies now indicating that persistently detectable ctDNA is a poor prognostic indicator^1,15,16. In the GALAXY study for example, patients with detectable postoperative ctDNA who achieved ctDNA clearance following ACT had a significantly better 12 months RFS than those with persistently detectable ctDNA (85.7% versus 20.7%, P < 0.0001).

Here we report updated data from the DYNAMIC trial, including mature RFS and overall survival (OS) data, the association between end-of-treatment (EOT) ctDNA status and outcomes, clinical outcomes in subgroups of interest, postoperative ctDNA molecular burden and a new assay.

Results

Patients and treatment

Between 10 August 2015 and 2 August 2019, 455 patients were randomized. Of these, 441 patients were included in the intention-to-treat population, 294 in the ctDNA-guided group and 147 in the standard management group (Fig. 1). Patient characteristics were generally well matched between the two groups (Table 1), with T4 disease being present in 15% and 14%, and dMMR tumors in 20% and 18% within the ctDNA-guided and standard management groups, respectively. At least one cycle of ACT was received by 45 (15%) patients in the ctDNA-guided arm and 41 (28%) patients in the standard management arm. In terms of chemotherapy regimen, 6% and 9% received single-agent fluoropyrimidine and oxaliplatin-based doublet chemotherapy in the ctDNA-guided group, respectively; 25% and 3% received single-agent fluoropyrimidine and oxaliplatin-based doublet chemotherapy in the standard management group, respectively (Table 1). Following relapse, 54% of patients (13/24) in the ctDNA-guided arm and 71% (10/14) in the standard management arm underwent curative intent surgical resection of oligo-metastases (n = 17) or local recurrence (n = 6).

Fig. 1 |. Patient disposition.

Flow diagram showing patient recruitment, randomization, follow-up and analysis populations. Of the 14 patients excluded from intention-to-treat (ITT) analysis, 6 did not have week 7 postoperative blood draw, 3 were found to be ineligible (2 due to metastatic disease, and 1 had a major cerebrovascular event and was deemed unfit for ACT) and 5 patients withdrew consent from study procedures and clinical follow-up.

Table 1 |.

Baseline characteristics and treatment delivered

Characteristics/treatment	Standard management (n = 147)	ctDNA-guided management (n = 294)	Overall (N = 441)
Male sex, n (%)	81 (55.1)	154 (52.4)	235 (53.3)
Age (years), median (range)	62 (28, 84)	65 (30, 94)	64 (28, 94)
Age group, n (%)
≤70 years	113 (76.9)	207 (70.4)	320 (72.6)
>70 years	34 (23.1)	87 (29.6)	121 (27.4)
Tumor stage, n (%)
T3	127 (86.4)	250 (85.0)	377 (85.5)
T4	20 (13.6)	44 (15.0)	64 (14.6)
Poor tumor differentiation, n (%)	17 (11.6)	43 (14.6)	60 (13.6)
Lymph node yield <12, n (%)	7 (4.8)	13 (4.4)	20 (4.5)
Tumor perforation, n (%)	7 (4.8)	7 (2.4)	14 (3.2)
Bowel obstructiona, n (%)	18 (12.2)	26 (8.9)	44 (10.0)
Lymphovascular invasion, n (%)	38 (25.9)	82 (27.9)	120 (27.2)
MMR status, deficient, n (%)	27 (18.4)	59 (20.1)	86 (19.5)
Adjuvant-chemo received, n (%)	41 (27.9)	45 (15.3)	86 (19.5)
Oxaliplatin-based chemo, n (%)	4 (2.7)	28 (9.5)	32 (7.3)
Single-agent fluoropyrimidine, n (%)	37 (25.2)	17 (5.8)	54 (12.2)
Curative intent resection of recurrence, n (%)	10/14 (71.4)	13/24 (54.2)	23/38 (60.5)

^aInformation on bowel obstruction was missing for three patients.

OS, DSS and RFS in all patients

After a median follow-up of 59.7 months (interquartile range (IQR) 55.0–61.5), the prespecified 5-year OS was 93.8% and 93.3% in the ctDNA-guided and standard management groups, respectively (hazard ratio (HR) 1.05; 95% confidence interval (CI) 0.47–2.37; P = 0.887; Fig. 2a). There were no treatment-related deaths. Overall, there were 17 deaths (5.8%) in the ctDNA-guided arm and 9 deaths (6.1%) in the standard management arm. Only 11 deaths (42%) were attributable to colorectal cancer recurrence, 6 deaths (23%) were due to another cancer diagnosed during follow-up and 9 deaths were unrelated to any cancer. A post hoc analysis of disease-specific survival (DSS) demonstrated an estimated DSS at 5 years of 97.9% and 97.2% in the ctDNA-guided and standard management groups, respectively (HR 1.19; 95% CI 0.35–4.09; P = 0.79; Fig. 2b). The median time from recurrence to death was 21 months with ctDNA guidance and 26 months with standard management.

Fig. 2 |. Kaplan-Meier estimates in the intention-to-treat population according to management arm.

a–c, OS (a), DSS (b) and RFS (c).

The number of patients who experienced a recurrence was 24 (8.2%), whereas 9 patients (3%) died without evidence of recurrence in the ctDNA-guided arm. The corresponding statistics were 14 (9.5%) and 3 patients (2%) for the standard management arm. The probabilities of survival without disease (RFS) at 5 years were 88.3% with ctDNA guidance and 87.2% with standard management (absolute difference 1.1%, 95% CI −5.8% to 8.0%; Fig. 2c). 73.7% (28 of 38) and 89.5% (34 of 38) of recurrences were observed within the first 2 and 3 years following randomization, respectively. The median time from randomization to recurrence was 13 months with ctDNA guidance and 18 months with standard management.

OS, DSS, and RFS for T4 and dMMR tumors

Given that T4 stage is the strongest adverse risk feature in stage II colon cancer and frequently influences adjuvant treatment recommendation, additional prespecified exploratory subgroup analyses were performed for T4 patients. Among patients with T4 disease (n = 44 in ctDNA-guided and n = 20 in standard management), the estimated OS at 5 years were 90.5% with ctDNA guidance and 80.0% with standard management (HR 2.16; 95% CI 0.53 to 8.79; P = 0.27; Extended Data Table 1 and Extended Data Fig. 1a). Five-year RFS was 81.2% and 70.0% (HR 1.79; 95% CI 0.62–5.15; P = 0.28; Extended Data Fig. 1b) for ctDNA-guided and standard management arms, respectively. A similar trend of more favorable OS and RFS in the ctDNA-informed group was observed when the analysis is restricted to proficient MMR T4 tumors only (5-year OS 87.1% versus 77.8%, P = 0.363; 5-year RFS 74.6% versus 66.7%, P = 0.495). Significantly fewer patients with a T4 tumor in the ctDNA-guided group received ACT (12 of 44, 27.3%) compared to the standard management group (14 of 20, 70.0%; P = 0.002). Of those who received chemotherapy, seven (58.3%) in the ctDNA-guided group and two (14.3%) in the standard management group were treated with oxaliplatin-based regimen.

Given that current guidelines do not recommend adjuvant therapy for patients with dMMR stage II colon cancers but disease will recur in some patients with such dMMR cancers, additional prespecified exploratory subgroup analyses were also performed for dMMR patients. The estimated DSS rates at 5 years were 98.3% and 100% in the ctDNA-guided and standard management groups, respectively (HR not evaluable due to zero event in one of the comparator arms; Extended Data Table 1). The estimated RFS and OS results are shown in Extended Data Table 1 and Extended Data Fig. 2. Postoperative ctDNA was detected in 5 of 59 patients (8.5%) with dMMR cancers (Extended Data Fig. 3). As prespecified in the protocol, all of these five patients received ACT, and none had recurred or died at the last follow-up. Of the 27 dMMR patients in the standard management arm, 3 (11.1%) received ACT and 1 recurred and died.

OS, DSS and RFS in the ctDNA-guided population

We performed prespecified RFS analysis and post hoc OS and DSS analyses for the ctDNA-guided cohort. The Schoenfeld residuals test did not indicate any violation of the proportional hazards assumption for all outcomes assessed. Among the 294 patients in the ctDNA-guided arm, postoperative ctDNA analysis was successful in 291 patients. Plasma samples were collected for ctDNA analysis from all patients at weeks 4 and 7 after surgery. The second week 7 testing was performed to see if this increased test sensitivity due to an anticipated increase in ctDNA detection rate with time after surgery and with more volume sampled. The ctDNA results for the two timepoints were as follows: 31 patients with week 4 timepoint positive and week 7 timepoint positive, 8 patients with week 4 timepoint positive and week 7 timepoint negative and 6 patients with week 4 timepoint negative and week 7 timepoint positive. Overall, 45 (15.5%) patients had at least one positive ctDNA result at week 4 or 7 after surgery. The ctDNA detection rates according to key prognostic subgroups are shown in Extended Data Fig. 3. Five-year OS was significantly worse in ctDNA-positive patients who received ACT versus ctDNA-negative patients who were observed (85.6% versus 95.3%, HR 3.30; 95% CI 1.20–9.05; P = 0.014; Fig. 3a). Similar to the overall population, only 7 of the 17 deaths (41.2%) were a direct consequence of colorectal cancer relapse. The 5-year DSS and RFS rates are shown in Extended Data Table 1 and Fig. 3b,c. For the 45 ctDNA-positive patients, recurrence rates for those with ctDNA positivity at both week 4 and week 7 timepoints and those with a positive result at only one postoperative timepoint were 20.0% and 13.3%, respectively (P = 0.699).

Fig. 3 |. Kaplan-Meier estimates in the ctDNA-guided population.

a–c, OS (a), DSS (b) and RFS (c) stratified by ctDNA status. d,e, OS (d) and RFS (e) in ctDNA-negative patients stratified by T stage.

We performed further post hoc exploratory analysis of the postoperative ctDNA-negative subgroup according to T stage. The probabilities of both OS and RFS at 5 years were numerically higher for T3 tumors than for T4 tumors (OS 96.0% versus 90.6%; HR 2.45; 95% CI 0.58 to 9.25; P = 0.171; Fig. 3d; RFS 91.7% versus 81.3%; HR 2.45; 95% CI 0.97 to 6.22; P = 0.051; Fig. 3e). Interestingly, 9 of 16 postoperative ctDNA-negative patients with cancer recurrence (56.3%) had relapse at locoregional sites only (peritoneum, anastomotic, pelvic or mesenteric node), including 4 of 5 recurrences in patients with T4 tumors. In contrast, of the postoperative ctDNA-positive patients who recurred, all eight relapses involved distant sites (four liver only, one lung only and three other combinations; P = 0.015).

Tumor mutation profile and ctDNA detection

All patient’s primary colon tumors were sequenced for the following commonly mutated genes in colorectal cancer: TP53, APC, KRAS, BRAF, PIK3CA, SMAD4, FBXW7, AKT1, CTNNB1, ERBB3, GNAS, NRAS, POLE, PPP2R1A and RNF43. In a prespecified exploratory analysis, the frequency of mutations found in the primary tumor and the ctDNA detection rate for each mutated gene are shown in Extended Data Fig. 4. The ctDNA detection rate for tumors harboring any of the top five most commonly mutated genes were TP53 (18.1%), APC (13.0%), KRAS (20.5%), PIK3CA (15.6%) and BRAFV600E (6.4%). Patients with a TP53 mutant or KRAS mutant tumors had a ctDNA detection rate that was numerically higher than that of patients with a wild-type tumor (TP53: 18.1% versus 9.3%, P = 0.075; KRAS: 20.5% versus 12.1%, P = 0.068). In contrast, tumors with BRAFV600E mutation had a numerically lower ctDNA detection rate compared with BRAF wild-type tumors (6.4% versus 17.0%, P = 0.077).

Molecular tumor burden and clinical outcome

A prespecified exploratory objective of the study was to correlate postoperative ctDNA level or molecular burden with ctDNA clearance and RFS. Among the postoperative ctDNA-positive patients, the molecular tumor burden was measured as the number of tumor-derived mutant molecules per milliliter plasma (TDMM; median of 0.38, IQR 0.135–3.8). Patients with a TDMM higher than the median had a lower ctDNA clearance rate (75% versus 100%, P = 0.047) and a worse RFS than patients with a TDMM less than the median, with a 5-year RFS of 58.9% versus 95.2% (HR 10.62; 95% CI 1.34–83.95; P = 0.005; Fig. 4a). The 5-year OS was also inferior for patients with a ctDNA molecular burden higher versus lower than the median (71.8% versus 100%; HR not evaluable due to zero event in one of the comparator arms; P < 0.001; Fig. 4b). We also observed a lower postoperative ctDNA molecular tumor burden in patients who had ctDNA clearance compared with those who did not (median 0.32 versus 5.35 TDMM, P = 0.016; Fig. 4c).

Fig. 4 |. Impact of postoperative ctDNA molecular tumor burden on clinical outcome following ACT in postoperative ctDNA-positive patients.

a,b, Kaplan-Meier estimates of RFS (a) and OS (b) stratified by the median postoperative ctDNA TDMM of 0.38. c, Box and whiskers plot showing the median, IQR and 5% to 95% percentile of postoperative ctDNA TDMM for patients with ctDNA clearance (n = 35) versus persistence (n = 5) following ACT; the Mann-Whitney test (two sided) was used to compare the distributions of the two groups.

ctDNA clearance after ACT

An EOT ctDNA sample, collected at 4 weeks after completion of ACT, was available in 40 of the 45 patients in the ctDNA-guided arm. ctDNA clearance (postoperative ctDNA positive to EOT ctDNA negative), analyzed with the same colorectal targeted panel as the postoperative ctDNA analysis, was observed in 35 (87.5%) patients. Clearance was achieved in 24 of 26 (92.3%) patients treated with oxaliplatin doublet chemotherapy and 11 of 14 (78.6%) patients treated with single-agent fluoropyrimidine chemotherapy (Fig. 5a). Notably, one patient who received single-agent fluoropyrimidine and failed to clear ctDNA received only 5 weeks of capecitabine treatment due to toxicity. The other 39 patients received at least 3 months of therapy. With respect to subsets of particular interest, no difference in ctDNA clearance rate was observed between T3 (28/32, 87.5%) and T4 tumors (7/8, 87.5%), whereas ctDNA clearance was observed in 3 of 4 (75%) and 32 of 36 (89%) dMMR and pMMR patients, respectively.

Fig. 5 |. ctDNA clearance and clinical outcome following ACT in postoperative ctDNA-positive patients.

a, Sankey plot of ctDNA clearance/persistence after oxaliplatin-based or single-agent fluoropyrimidine ACT and recurrence outcome for the 40 patients who had both postoperative and EOT ctDNA samples available for analysis using the colorectal cancer (CRC) targeted panel approach, b, Violin plot showing the distribution of the number of variants tracked in plasma as identified by tumor tissue sequencing with targeted panel versus WES with or without WGS in 38 patients with sufficient EOT plasma available. The Mann-Whitney test (two sided) was used to compare the distributions of the two groups; dashed line represent the median and dotted lines the quartiles, c, Sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) of each assay in predicting recurrence. d, Kaplan-Meier estimates of recurrence-free interval according to EOT ctDNA status using the tumor WES/WGS-guided ctDNA assay.

To evaluate the ability of EOT ctDNA to predict later relapse in these patients, we performed a post hoc analysis comparing two approaches in 38 patients with sufficient EOT plasma samples, both using a library preparation method that efficiently preserves template molecules and minimizes sequencing errors (Methods). In the first approach, we evaluated driver gene mutations identified in the primary tumor by targeted amplicon sequencing, as described previously¹⁷. This allowed us to evaluate a median of three driver gene mutations per patient. In the second approach, we performed whole-exome sequencing (WES) of the primary tumor from all patients to identify passenger gene mutations that were present in clonal fashion. Additional whole-genome sequencing (WGS) was performed in 3 of the 38 cases, because only a low number of suitable variants were detected with WES. A median of 29 variants (IQR 24–38) were assessed in the plasma with this tumor WES/WGS-informed approach (P < 0.001; Fig. 5b).

The evaluation of more variants through the WES/WGS-informed approach proved superior (Fig. 5c), and that approach was chosen for further analysis. In the seven patients with recurrence, EOT ctDNA was positive in six (sensitivity of 86%). For the 31 recurrence-free patients, EOT ctDNA was negative in all (specificity of 100%). The positive and negative predictive values were 100% and 97% (Fig. 5c). The number of variants, mutant molecules detected in each plasma sample with the two methods, and corresponding recurrence outcome are listed in Extended Data Table 2. Suitable variants identified by WES that were detected in the EOT ctDNA are listed in Extended Data Table 3. The probability of being recurrence-free at 5 years was 96.8% for patients without detectable EOT ctDNA and 0% for patients with detectable EOT ctDNA (HR not evaluable due to 100% event rate in one of the comparator arms; P < 0.001; Fig. 5d). The median time to recurrence from randomization for patients with EOT ctDNA positivity was 12.2 months (range 6.6–36.2) versus 47.5 months for the only patient with recurrence who was ctDNA negative following ACT.

Discussion

For patients with stage II colon cancer, a more personalized approach to ACT decision-making is an urgent need, given the uncertain benefit when patients are selected for treatment based on traditional criteria. The updated data from DYNAMIC, now with mature survival data, demonstrate that a ctDNA-informed approach to ACT versus standard management can reduce the proportion of patients receiving chemotherapy without compromising RFS or OS. These findings appear robust across patient subgroups, including with varying recurrence risk consistent with those expected for traditional high- and low-risk criteria such as T4 and dMMR tumors, respectively. The DYNAMIC data also suggest continued improvement in survival outcomes over time, with lower recurrence rates in this more contemporary stage II cohort when compared to historical practice-defining trials of adjuvant therapy. In further analyses, we report a potential role for serial ctDNA measurement as a dynamic marker of minimal residual disease (MRD), informing adjuvant therapy benefit and residual risk at treatment completion. Exploratory analyses also raise the potential for further risk stratification of postoperative ctDNA-positive patients by measurement of molecular burden (TDMM) and potential gains with a more sensitive testing methodology.

With a median follow-up of nearly 5 years, the mature data from DYNAMIC demonstrate that similar OS outcomes being achieved for stage II colon cancer patients managed with a ctDNA-informed versus a standard approach. Notably, excellent outcomes were achieved for all patients, including a 5-year RFS of 88.0% and 5-year OS of 93.6%, with a minority of patients receiving adjuvant therapy. These outcomes do not appear to be driven by enrolling a select low-risk patient population, as the expected rate (40%) of patients had high-risk features. Rather, the results are consistent with trends for improved survival over time for early-stage colon cancer^18,19, gains that can in part be attributed to the now routine use of high-quality imaging at baseline that may detect previously unseen low volume metastatic disease that was previously missed, upstaging some stage II patients to stage IV. Other potential contributors to survival gains are more accurate nodal staging due to an increased number of lymph nodes being examined by pathologists, minimizing undiagnosed stage III disease¹⁹. Notably, only 4.5% of the DYNAMIC study cohort had an inadequate lymph node yield of <12 (ref. 20). Improved management of any recurrence may also be a major factor impacting OS outcomes, including better systemic therapy and more curative intent salvage surgery. The latter occurred in more than 50% of DYNAMIC patients who relapsed, consistent with the reported increased use of salvage therapy in community series²¹.

Although detectable ctDNA is a powerful prognostic marker, conventional clinicopathologic risk factors retain prognostic significance in a ctDNA stratified subset. This includes for T4 disease, the most adverse feature and anecdotally a major driver of ACT use in routine care. In DYNAMIC, the 5-year recurrence rate was 15% for the ctDNA-negative T4 tumors, similar to all clinical low-risk stage II patients combined, noting that there is no evidence supporting a survival benefit from adjuvant treatment in unselected T4 cases. In contrast, for patients with dMMR tumors where cancer guidelines recommend against treatment due to the low recurrence rate and possible fluoropyrimidine treatment resistance, some dMMR cases still do recur. In our previous observational studies in early-stage colon cancer, a recurrence rate of 66% was seen in a small number of untreated ctDNA-positive dMMR tumors^6,15. The DYNAMIC study provides suggestion of adjuvant therapy benefit in any patient with dMMR, with ctDNA clearance in three of four dMMR patients (75%) treated with chemotherapy (three receiving oxaliplatin-based treatment) as per protocol, and no recurrences with mature follow-up data. This finding needs to be further explored.

We observed a high ctDNA clearance rate at the completion of ACT, with an expected numerically higher clearance rate with oxaliplatin doublet chemotherapy than single-agent fluoropyrimidine (92.3% versus 78.6%). The relatively high ctDNA clearance rate for ctDNA-positive patients who received single-agent fluoropyrimidine is consistent with the relative contribution to disease-free survival of fluoropyrimidine and oxaliplatin in non-ctDNA selected clinical high-risk stage III colon cancer subgroup, where around 60–65% of the disease-free survival benefit from the combination regimen can be attributed to the fluoropyrimidine alone²². To our knowledge, the relative clearance rate of single-agent fluoropyrimidine and doublet oxaliplatin regimen has not been reported previously. Notably, we used a single timepoint at 4 weeks after completion of ACT for assessment of ctDNA clearance to minimize the known issue of transient clearance during chemotherapy observed in our previous observational study⁶. Even with this single timepoint, a reasonable sensitivity (86%) and very high (100%) specificity were achieved with the WES/WGS-informed approach. Serial ctDNA assessments during surveillance, which was not collected in this study, may provide further information to help define the optimal timing and metrics (for example, one landmark timepoint or multiple timepoints) for assessing ctDNA clearance.

The DYNAMIC data, along with previous observational data^1,15, indicate that patients with persistently detectable ctDNA despite ACT represent a particularly high-risk group (Fig. 5d). Further treatment of these patients before they develop clinically detectable disease could be beneficial, a strategy being explored in several current studies. Early data from the PEGASUS study, with early follow-up and small numbers, suggest the possibility of benefit from salvage treatment for patients with persistently detectable ctDNA, using FOLFIRI chemotherapy²³. Other studies are investigating the use of agents not previously trialed in the MRD context, including agents active in metastatic disease, such as trifluridine/tipiracil, various targeted therapies in molecularly defined subsets and low toxicity strategies such as personalized vaccines. ctDNA status at EOT may also usefully inform the intensity of surveillance strategies, which merits further exploration such as is being explored in the IMPROVE-IT2 trial²⁴.

Beyond the binary readout of a ctDNA result as positive or negative, when detectable, the amount of ctDNA may have prognostic significance, potentially reflecting the ‘burden’ of MRD. In an earlier combined analysis of three early-stage colorectal cancer observational cohorts, recurrence risk was found to be higher in patients with higher postoperative ctDNA mutant allele fraction in those treated with ACT²⁵. Data from DYNAMIC are consistent with this finding, where a higher postoperative ctDNA molecular burden was associated with a lower ctDNA clearance rate in patients treated with ACT and a markedly worse 5-year RFS (HR 10.62, P = 0.005). The GALAXY observational study has similarly demonstrated that patients with higher postoperative ctDNA molecular burden are less likely to achieve ctDNA clearance with chemotherapy¹. The immediate relevance of these data is that ctDNA molecular burden should be considered as a stratification factor for future studies of ctDNA-informed therapy, including the addition of novel therapies that could focus on patients with higher ctDNA levels. Such studies could conceivably enroll only patients with higher ctDNA levels, as standard treatment appears very effective in patients with a low ctDNA burden, with a 100% clearance rate in a small patient cohort in DYNAMIC. For this subset, future strategies could focus on reducing treatment duration or intensity.

To date, tumor-informed ctDNA MRD assays have shown high positive predictive value for recurrence, but clinical sensitivity at the postoperative timepoint was consistently around 50% with the first-generation MRD assays²⁶, indicating scope for improved sensitivity. Naturally, this ‘false-negative’ rate is a potential barrier to routine adoption of MRD testing for the purpose of informing adjuvant treatment de-escalation, especially in situations where adjuvant therapy is currently considered standard of care. Given this, we explored a new assay using residual EOT samples. Using an average of 29 tumor-derived mutations, this exploratory analysis found a very high specificity (100% of 31 patients) and suggested improved sensitivity (86% of 7 patients). The one patient who had a negative ctDNA test result at EOT recurred only after 4 years, much later than the patients with a positive ctDNA test result. These results suggest potential for an improved ctDNA test, using an assay that detects many mutations simultaneously and has an extremely high specificity, such as the one described herein. Data from prospective clinical trials are needed to substantiate this potential.

At present, differences in sample preparation methods, molecular recovery, genomic DNA contamination and methods of quantification have hindered the harmonization of ctDNA assays for widespread implementation as an MRD test. The SaferSeqS-based assay used in this study potentially overcomes these challenges in several ways. First, all steps can be automated and streamlined for consistently high molecular recovery ( > 70%)¹⁷. Second, mutant quantification does not require an external standard but relies on molecular barcodes that are introduced at the beginning of the workflow to uniquely tag every initial DNA molecule, allowing for precise quantification of every mutant molecule. Third, error-reduction strategies, including duplex sequencing and molecular barcoding, can exponentially decrease error rates associated with sample preparation and sequencing, highlighting the perfect specificity obtained with the WES/WGS-informed approach. Last, the assay can be used on a variety of sequencing platforms. We believe that this approach establishes a foundation for the next generation of MRD assays. Additionally, the current cost of goods for this WES/WGS assay is less than US$1,000 and is expected to further decrease with lowering costs of existing sequencing platforms and the development of new sequencing technologies that are more cost-effective. Once WES/WGS is completed and the personalized panel has been built, each subsequent assay is less expensive, costing <US$200 per test. This is substantially less expensive than the cost of imaging and has the added advantage of being more convenient, requiring only a sample of blood.

There are limitations to the DYNAMIC study, which enrolled a relatively small patient number compared to traditional adjuvant therapy trials. As such, the DYNAMIC study was not powered to show noninferiority in OS outcomes. Also, a subset of the analyses, due to the small sample size, are exploratory in nature and need to be confirmed in future studies. There was an imbalance in the use of oxaliplatin-based chemotherapy in the two arms, with an additional 7% of patients receiving oxaliplatin-based treatment in the ctDNA arm, as the choice of adjuvant therapy was at clinician discretion. This imbalance may have contributed to a similar outcome being achieved in both study arms despite lower rates of chemotherapy use, but we would note that in pivotal randomized studies the addition of oxaliplatin to 5-fluorouracil versus 5-fluorouracil alone had minimal impact on survival outcomes^11,27. Whether the addition of oxaliplatin provides benefit in the ctDNA-positive subset of stage II colon cancer, but not in all patients, needs to be further explored. Further, given oxaliplatin is associated with potential long-term neurotoxicity, data on adverse events and quality of life, which were not specifically collected in the study, would have been helpful to fully understand the consequences of treatment de-escalation.

The DYNAMIC study represents an important step toward validating a biomarker-driven, more personalized approach to adjuvant therapy selection in patients with stage II colon cancer. The new data reported herein, now with mature follow-up, confirm survival outcomes are not compromised with a ctDNA-informed approach. We also demonstrate the potential for gains in future studies, including moving beyond a binary qualitative result, repeating analyses at EOT (with testing on treatment and during surveillance also worth exploring) and exploring refined methodology. The outcomes of further studies underway for stage II colon cancer and for other postcurative resection scenarios, including stage III colon cancer and resected metastatic disease, are eagerly awaited.

Online content

Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41591-025-03579-w.

Methods

Study design and participants

The DYNAMIC study was conducted in accordance with the International Ethical Guidelines for Biomedical Research Involving Human Subjects, Good Clinical Practice guidelines, and the principles of the Declaration of Helsinki. The study protocol and amendments were approved by the ethics committees of the following hospitals or institutions: Barwon Health, Bendigo Health, Border Medical Oncology, Cabrini Hospital, Calvary Mater Newcastle Hospital, Canberra Hospital, Chris O’Brien Lifehouse, Eastern Health, Epworth Freemasons, Fiona Stanley Hospital, Flinders Medical Centre, Johns Hopkins Medicine, Melbourne Health, Monash Health, Nepean Cancer Centre, Newcastle Private Hospital, Northern Health, Peter MacCallum Cancer Centre, Royal Brisbane and Women’s Hospital, Royal Hobart Hospital, South West Health Care, St Vincent’s Hospital Melbourne and Western Health. All patients provided written informed consent.

DYNAMIC (ACTRN12615000381583) is a multicenter, randomized, controlled, phase 2 biomarker-driven adjuvant therapy trial. The trial was registered on 27 April 2015 at https://www.anzctr.org.au/Trial/Registration/TrialReview.aspx?ACTRN=12615000381583. Details of the trial design and study protocol have been published⁷.

Eligible patients included those with curatively resected histological confirmed stage II (T3-4, N0, M0) colon or rectal adenocarcinoma with negative resection margins. Patients with rectal cancer were eligible unless they have had neoadjuvant combined chemoradiation or were scheduled for postoperative chemoradiation. A representative paraffin-embedded tumor sample is available for molecular testing within 28 days after surgery. Patients had to be medically fit for adjuvant oxaliplatin- or fluoropyrimidine-based chemotherapy, with Eastern Cooperative Oncology Group performance status 0 to 2, accessible for follow-up and no evidence of macroscopic metastatic disease on computed tomography of the chest, abdomen and pelvis performed within 8 weeks of enrollment. Exclusion criteria included the presence of synchronous primary colorectal cancer, history of another primary cancer within the last 3 years, with the exception of nonmelanomatous skin cancer and carcinoma in situ of the cervix, treatment with neoadjuvant chemoradiotherapy or medical or psychiatric condition or occupational responsibilities that may preclude compliance with the protocol.

Patients were enrolled within 3 weeks of surgery, and adequate tumor sample from the resection specimen were to be provided for mutation analysis by 4 weeks after surgery. Following enrollment, patients were randomized in a 2:1 ratio to be managed according to ctDNA results (ctDNA-guided management) or by the treating clinician according to standard clinicopathologic criteria (standard management). Individual patients were assigned to study arms using block randomization stratified by the participating center location (regional or metropolitan) and T stage (T3 or T4). Patients were allocated to the study arms by block randomization with block size of 3. The randomization table was generated using R, and REDCap was used to allocate each participant based on the randomization table. For patients randomized to ctDNA-guided management, week 4 and week 7 blood samples were analyzed concurrently for the presence of ctDNA and ctDNA results were made available to the treating clinician 8 to 10 weeks after surgery. Patients with a positive ctDNA result at either week 4 or week 7 received adjuvant single-agent fluoropyrimidine or oxaliplatin-based chemotherapy, with the treatment regimen at clinician’s discretion. Patients with negative ctDNA results at both week 4 and week 7 were not treated with ACT. In the standard management arm, all treatment decisions were based on conventional clinicopathologic criteria. Acceptable chemotherapy regimens for patients in either arm included single-agent fluoropyrimidine (modified De Gramont, QUASAR, Roswell Park or capecitabine) or oxaliplatin-based doublet (modified FOLFOX6 or CAPOX/XELOX). A protocol amendment in February 2018 allowed clinicians the flexibility of administering 3 months of ACT in light of emerging data from the IDEA meta-analysis suggesting 3 months of ACT is likely noninferior to 6 months of treatment. Dose modifications to chemotherapy were as per local standard practice.

Follow-up

All patients in both arms were evaluated with serum CEA every 3 months for 2 years, then every 6 months for another 3 years. Contrast-enhanced computed tomography of the chest, abdomen and pelvis were performed every 6 months for 2 years and then at 3 years. All patients were to be followed for up to 5 years from randomization. Assessments were made for relapse, second cancers and death. Because only standard-of-care treatments were used in this study, adverse events were not assessed.

ctDNA analysis method

We used a tumor-informed personalized approach for ctDNA analysis, where somatic mutations were first identified by sequencing of each patient’s tumor tissue, and the presence of the same mutation was then assessed in the plasma samples. For the primary aim of the DYNAMIC study, which was to select patients for adjuvant therapy, targeted sequencing of commonly mutated colorectal cancer genes was performed to identify somatic variants for plasma analysis. For a post hoc exploratory aim to assess MRD after adjuvant therapy, we tracked more variants. In particular, WES (n = 35) or WES plus WGS (n = 3) of tumor tissue was performed in the 38 cases where residual EOT plasma samples were available. All tumor tissue mutation and ctDNA analyses were performed by the study scientists (J.D.C., K.L., Y.W., J.P., N.S., L.D., M.P. and B.V.).

Targeted tumor sequencing.

Formalin-fixed paraffin-embedded tumor tissue from the primary tumor was analyzed for somatic mutations in 15 genes recurrently mutated in colorectal cancer (SMAD4, TP53, AKT1, APC, BRAF, CTNNB1, ERBB3, FBXW7, HRAS, KRAS, NRAS, PIK3CA, PPP2R1A, RNF43 and POLE). Tumor sections were macro-dissected under a dissecting microscope to enrich neoplastic cell content. DNA was purified with a Qiagen FFPE Kit (Qiagen, #56494). Primers were designed and sequencing results analyzed as previously described⁶.

Whole-exome/genome tumor sequencing.

WES of the tumor and matched white blood cell was performed to an average depth of >200× (Ashion/EXACT Sciences). WGS was done on an Illumina NovaSeq instrument to an average depth of >20×. Tumor-specific somatic mutations were selected after eliminating mutations that were present in the matched whole blood cells. In addition, mutations in repetitive regions, regions with difficult alignment or amplification, and common polymorphisms were excluded.

Plasma sample collection.

Blood samples (30–60 ml) were collected in K2-EDTA tubes and processed within 3 h by double centrifugation; buffy coat was collected after the first centrifugation. All samples were stored at −80 °C before extraction and analysis. At least 10 ml plasma was purified from each patient using the QIAamp Circulating Nucleic Acid kit (Qiagen, #55114).

ctDNA analysis with mutations identified by targeting sequencing of tumor.

For each patient, at least one mutation identified from targeted sequencing of the tumor tissue was assessed in cell-free DNA from the plasma. The detection and quantitation of ctDNA were performed using the Safe-Sequencing System error-reduction technology for the detection of low frequency mutations^6,28,29 with plasma DNA divided into 12, 24 or 95 wells per sample. Leukocyte DNA was used to exclude constitutional polymorphisms.

For plasma DNA samples partitioned into 12 or 24 wells, ctDNA was classified as detectable (ctDNA-positive) or undetectable (ctDNA-negative). This classification was based on exact permutation tests as described previously^6,28,29 that compared the difference between the average mutant allele frequency across the wells containing the sample of interest with that of the wells containing the control sample for each mutation. One-sided P values were calculated using the permTS function of the R perm package (version 3.5.1). A sample was classified as ctDNA positive if the P value was <0.1. For plasma samples divided into 95 wells, the mutant allele frequencies of all observed mutations were used to model the amplicon-specific distribution of assay noise. The P values corresponding to the mutations of interest (that is, those detected in the primary tumor tissue) were combined using Fisher’s method to calculate a final P value for the patient. A patient was classified as ctDNA positive using the same P-value threshold ( < 0.1) described above.

The number oftumor molecules in plasma was estimated based on the number of distinct molecular barcodes or unique identifiers that tagged each DNA fragment (PMID: 21586637) in the plasma sample, relative to the experiment-specific normal plasma control with a known number of haploid genome equivalents.

ctDNA analysis with mutations identified by WES or WGS of tumor.

SaferSeqS assays¹⁷ were designed for each mutation chosen from WES or WGS. All mutations per patient were tested in one multiplex PCR using the KAPA HiFi HotStart ReadyMix (Roche). Mutation analysis was done as previously described¹⁷. Only mutant positions that could be uniquely mapped to the reference genome, at all positions except the mutant one, were evaluated. For evaluating plasma, a mutation found in the tumor through WES or WGS had first to be validated to be present in the tumor using the multiplex PCR test described above. The mutation also had to be absent in both DNA from the matched normal white blood cells from the same patient and in an unmatched plasma sample from a healthy control. In addition, any mutation with detected numbers of mutant molecules in the plasma that was highly discordant (Z-score >5) with other mutations in the same plasma sample was discarded.

Trial endpoints

The primary study endpoint of RFS rate at 2 years and key secondary endpoint of proportion of patients in each study arm treated with ACT has previously been reported⁷. Other key secondary endpoints included time-to-recurrence and 5-year OS. Preplanned secondary endpoints included RFS in ctDNA-positive and -negative patients in the ctDNA-guided group and EOT ctDNA clearance rate. Prespecified exploratory efficacy analyses were conducted on limited subgroups of interest such as T4 stage and dMMR tumors. Preplanned exploratory analysis was conducted for postoperative ctDNA level and tumor mutation profile. Given the substantial number of nonrecurrence related death events, a post hoc analysis of DSS by study arm and ctDNA status was also performed.

Statistical analysis

The overall sample size was developed to ensure a minimum of 30 patients would yield a ctDNA-positive result in the ctDNA-guided arm and an acceptable noninferiority margin for 2-year RFS of 8.5%, to exclude the largest absolute benefit that could be derived from adjuvant oxaliplatin-based chemotherapy for high-risk stage II patients. A total sample size of 450 patients provides 80% power to demonstrate noninferiority between the two study arms, assuming a type-I error of 5%, 2-year RFS with standard management of 84% and ctDNA-guided management of 85%, and allowing 10% dropout rate.

The primary analysis for 2-year RFS was performed when all participants have completed the 24-month follow-up visit or dropped out prior to the 24-month follow-up visit (data cut-off 15 October 2021) and has been reported previously⁷. The primary outcome analysis of the 2-year RFS rate was based on the intention-to-treat population. An interim analysis was conducted to demonstrate the feasibility of returning ctDNA results to clinician within the protocol specified timeframe. The interim analysis was planned after the first 20 patients have been randomized to ensure that the postoperative ctDNA results can be fed back to the treating clinicians within 10 weeks after surgery. Secondary efficacy analysis for the key secondary endpoint of OS was conducted after a minimum follow-up period of 5 years. Other secondary outcomes including reduction in chemotherapy use and changes in ctDNA status during ACT will also be reported. The primary endpoint and the key secondary endpoints will be assessed on a limited number of subgroups. The subgroup analyses will remain exploratory and hypothesis-generating given the study is not specifically powered to test any treatment effect. Association of ctDNA levels and tumor mutation status with recurrence outcome will be explored.

This analysis focuses on the long-term survival outcomes according to management arm, postoperative ctDNA status, T stage or MMR status. The analyzed outcomes included recurrence-free interval (RFI), RFS, DSS and OS. RFI time was calculated from the date of randomization to the date of recurrence confirmation or the last date at which the patient was known to be free of disease (censoring time). RFS was calculated from the date of randomization to the date of recurrence confirmation or death from any cause (whichever occurred earlier), or the last date at which the patient was known to be free of disease (censoring time). OS time was calculated as the duration from the date of randomization to the date of death from any cause. Patients alive at the cut-off date for analyses were censored at their last date of contact. DSS time was calculated as the duration from the date of randomization to the date of death due to colorectal cancer. Patients who died from other causes are censored at their date of death, and those alive at the cutoff date for analyses are censored at their last date of contact. Each outcome was described using the Kaplan-Meier method stratified by group. Landmark points at 3 and 5 years were derived along with the associated 95% CI. Survival difference between groups was assessed using the log-rank test.

The relative management effect across outcome was estimated using Cox proportional hazard regression stratified by tumor stage (T3 or T4) and participating center location (regional or metropolitan). HRs and associated 95% CIs were also reported after evaluating the proportional hazard assumption using the Schoenfeld residuals test. Association between categorical parameters were analyzed using the Fisher’s Exact test. This analysis was conducted when 75% patients have completed their 5-year follow-up. The data cut-off date for this analysis was 17 January 2024. All statistical analyses were performed using R version 3.6.1 (R Core Team) and SAS (SAS/STAT User’s Guide, Version 9.4; SAS Institute).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Extended Data

Extended Data Fig. 1 |. Kaplan-Meier estimates for T4N0 disease according to management arm.

(a) overall survival, and (b) recurrence-free survival.

Extended Data Fig. 2 |. Kaplan-Meier estimates for deficient MMR tumor according to management arm.

(a) overall survival, and (b) recurrence-free survival.

Extended Data Fig. 3 |. Postoperative ctDNA detection rates according to key prognostic subgroups.

Histogram showing post-op ctDNA detection rates for known prognostic factors for stage II colon cancer: T stage, presence or absence of lymphovascular invasion (LVI), mismatch repair (MMR) status, and clinical risk status.

Extended Data Fig. 4 |. Distribution of primary tumor mutation and ctDNA detection in ctDNA-guided cohort.

Patient’s primary colon tumors were sequenced for 15 commonly mutated genes in colorectal cancer as shown below. The frequency of mutations (MUT) for each gene are shown in percentage; the number of cases with ctDNA detection for each mutated and wild-type (WT) gene are shown.

Extended Data Table 1 |.

Recurrence-free interval (RFI), recurrence-free survival (RFS), disease-specific survival (DSS) and overall survival (OS) according to management arm, postoperative ctDNA status, T stage and MMR status

Population	No. of Patients	5-year RFI	5-year RFS	5-year DSS	5-year OS
Overall
Standard management	147	89.8%	87.2%	97.2%	93.3%
ctDNA-guided	294	91.7%	88.3%	97.9%	93.8%
		HR 1.16 (0.60, 2.25) P = 0.65	HR 1.01 (0.56, 1.81) P = 0.93	HR 1.19 (0.35, 4.09) P = 0.79	HR 1.05 (0.47, 2.37) P = 0.887
T4N0
Standard management	20	70.0%	70.0%	84.2%	80.0%
ctDNA-guided	44	83.5%	81.2%	92.7%	90.5%
		HR 2.04 (0.68, 6.06) P = 0.20	HR 1.79 (0.62, 5.15) P = 0.28	HR 2.32 (0.47-11.5) P = 0.30	HR 2.16 (0.53-8.79) P = 0.27
MMR Deficient
Standard management	27	96.2%	96.2%	100%	96.2%
ctDNA-guided	59	98.3%	93.2%	98.3%	93.2%
		HR 2.21 (0.14, 35.3) P = 0.56	HR 0.48 (0.05, 4.27) P = 0.51	HR NE*, P = NE*	HR 0.55 (0.06, 4.92) P = 0.59
ctDNA-guided
ctDNA-negative	246	93.4%	90.4%	98.3%	95.3%
ctDNA-positive	45	81.7%	76.7%	93.0%	85.6%
		HR 2.86 (1.22, 6.68) P = 0.02	HR 2.31 (1.08, 4.96) P = 0.02	HR 4.25 (0.91, 19.9) P = 0.05	HR 3.30 (1.20, 9.05) P = 0.01
ctDNA-negative
T3 tumor	213	94.8%	91.7%	99.0%	96.0%
T4 tumor	33	84.5%	81.3%	93.6%	90.6%
		HR 3.15 (1.09, 9.07) P = 0.03	HR 2.45 (0.97, 6.22) P = 0.05	HR 6.81 (0.96, 48.4) P = 0.05	HR 2.45 (0.58, 9.25) P = 0.17

^*NE = not evaluable due to zero event in one comparator arm

Extended Data Table 2 |.

WES/WGS-guided and CRC panel-guided end-of-treatment (EOT) ctDNA assay results and recurrence outcome for the 38 patients with positive postoperative ctDNA treated with adjuvant chemotherapy

Patient	WES/WGS-guided ctDNA assay					CRC Targeted Panel-guided ctDNA assay				Recurrence	Time to recurrence or last follow-up (months)
	WES or WGS	# Mutations Assessed in plasma	# Mutations positive in EOT plasma	# Mutant Molecules in EOT plasma	ctDNA status	# Mutations Assessed in plasma	# Mutations positive in EOT plasma	# Mutant Molecules in EOT plasma	ctDNA status
LCRA 1038	WES	16	0	0	Negative	4	0	0	Negative	No	61.25
LCRA 1051	WGS	47	0	0	Negative	2	0	0	Negative	No	61.22
LCRA 1053	WES	34	0	0	Negative	3	0	0	Negative	No	59.38
LCRA 1078	WES	33	0	0	Negative	1	0	0	Negative	No	59.57
LCRA 1126	WES	34	0	0	Negative	2	0	0	Negative	No	58.95
LCRA 1135	WES	35	0	0	Negative	3	0	0	Negative	No	56.71
LCRA 1139	WES	17	0	0	Negative	6	0	0	Negative	No	61.18
LCRA 1182	WES	27	0	0	Negative	3	0	0	Negative	No	60.56
LCRA 1185	WES	38	0	0	Negative	5	0	0	Negative	No	61.18
LCRA 1188	WES	16	0	0	Negative	2	0	0	Negative	Yes	47.5
LCRA 1194	WES	21	0	0	Negative	2	0	0	Negative	No	60.79
LCRA 1218	WES	28	0	0	Negative	4	0	0	Negative	No	62.53
LCRA 1220	WES	43	0	0	Negative	3	0	0	Negative	No	60.39
LCRA 1235	WES	25	0	0	Negative	2	0	0	Negative	No	61.25
LCRA 1265	WGS	60	0	0	Negative	2	0	0	Negative	No	58.32
LCRA 1275	WES	29	0	0	Negative	4	0	0	Negative	No	60.10
LCRA 1298	WES	8	0	0	Negative	2	0	0	Negative	No	58.92
LCRA 1309	WES	24	0	0	Negative	5	0	0	Negative	No	59.77
LCRA 1315	WES	33	0	0	Negative	5	0	0	Negative	No	59.64
LCRA 1320	WES	24	0	0	Negative	2	0	0	Negative	No	52.54
LCRA 1327	WES	24	0	0	Negative	2	0	0	Negative	No	46.26
LCRA 1331	WES	26	0	0	Negative	3	0	0	Negative	No	52.67
LCRA 1336	WES	36	1	1	Positive	1	0	0	Negative	Yes	21.7
LCRA 1338	WES	28	0	0	Negative	3	0	0	Negative	No	60.66
LCRA 1347	WES	31	1	1	Positive	1	0	0	Negative	Yes	36.2
LCRA 1348	WES	27	0	0	Negative	4	0	0	Negative	No	58.82
LCRA 1375	WES	41	0	0	Negative	4	0	0	Negative	No	59.80
LCRA 1381	WGS	30	0	0	Negative	3	0	0	Negative	No	61.22
LCRA 1388	WES	38	0	0	Negative	4	0	0	Negative	No	57.80
LCRA 1397	WES	27	0	0	Negative	3	0	0	Negative	No	60.56
LCRA 1486	WES	15	13	401	Positive	4	2	31	Positive	Yes	6.6
LCRA 1488	WES	26	0	0	Negative	11	1	1	Positive	No	60.79
LCRA 1535	WES	41	0	0	Negative	3	0	0	Negative	No	54.58
LCRA 1573	WES	38	0	0	Negative	3	0	1	Negative	No	55.23
LCRA 1596	WES	6	4	37	Positive	1	1	14	Positive	Yes	7.5
LCRA 1609	WES	39	3	3	Positive	3	1	1	Positive	Yes	14.3
LCRA 1636	WES	38	0	0	Negative	1	0	0	Negative	No	54.64
LCRA 1774	WES	28	12	14588	Positive	2	2	4490	Positive	Yes	10.2

Extended Data Table 3 |.

Suitable mutations detected in end of treatment (EOT) plasma samples

Patient ID	Chromosome	Position	Base from	Base to	Mutation Type	Mutation	Total # Molecules assessed	Mutant Allele Fraction in EOT Plasma
LCRA 1336	chr19	42509893	C	T	SBS	GRIK5 p.G749S,c.2245G>A	12914	0.008%
LCRA 1347	chr5	34863137	G	A	SBS	TTC23L p.V172M,c.514G>A	36239	0.003%
LCRA 1486	chr10	121558099	C	-	Indel	INPP5F p.Q365Nfs*2,c.1093del	26114	0.011%
LCRA 1486	chr19	38924505	C	G	SBS	RYR1 p.F12L,c.36C>G	39041	0.126%
LCRA 1486	chr10	114910832	A	-	Indel	TCF7L2 p.T318Qfs*33,c.952del	49418	0.113%
LCRA 1486	chr5	112175474	AGTTTTG	-	Indel	APC p.S1395Rfs*18,c.4185_4191del	40254	0.084%
LCRA 1486	chr6	32627938	C	A	SBS	HLA-DQB1 c.*75G>T	41808	0.098%
LCRA 1486	chr3	37067402	C	A	SBS	MLH1 p.A438D,c.1313C>A	43190	0.088%
LCRA 1486	chr7	127014711	-	A	Indel	ZNF800 p.G227Wfs*10,c.678dup	39423	0.066%
LCRA 1486	chr7	71801667	G	C	SBS	CALN1 c.120-57872C>G	28211	0.057%
LCRA 1486	chr17	71232977	C	A	SBS	C17orf80 p.G452=,c.1356C>A	42270	0.069%
LCRA 1486	chr5	150563127	C	A	SBS	CCDC69 p.E254D,c.762G>T	53618	0.069%
LCRA 1486	chr11	59189858	C	A	SBS	OR5A2 p.C190F,c.569G>T	38927	0.059%
LCRA 1486	chr6	46661713	G	C	SBS	TDRD6 p.D1950H,c.5848G>C	45561	0.061%
LCRA 1486	chr13	98674028	C	G	SBS	IPO5 p.L1100=,c.3300C>G	50708	0.041%
LCRA 1596	chr10	98092311	C	T	SBS	DNTT p.Y439=,c.1317C>T	58398	0.010%
LCRA 1596	chr19	14270014	T	C	SBS	LPHN1 p.T684=,c.2052A>G	36122	0.050%
LCRA 1596	chr4	36160475	G	C	SBS	ARAP2 p.H877D,c.2629C>G	40767	0.022%
LCRA 1596	chr10	102739049	G	A	SBS	MRPL43 g.chr10: 102739049G>A	42224	0.009%
LCRA 1609	chr18	34853926	A	G	SBS	CELF4 c.801+348T>C	13951	0.007%
LCRA 1609	chr3	46506285	C	T	SBS	LTF c.43+30G>A	11933	0.008%
LCRA 1609	chr2	241710435	G	A	SBS	KIF1A p.R432C,c.1294C>T	13862	0.007%
LCRA 1774	chr17	7578380	-	T	Indel	TP53 p.D184Rfs*2,c.549dup	9490	6.934%
LCRA 1774	chr2	43452300	T	A	SBS	ZFP36L2 p.I215F,c.643A>T	24267	5.534%
LCRA 1774	chr2	127816673	C	A	SBS	BIN1 p.A306S,c.916G>T	50927	5.066%
LCRA 1774	chr5	138724218	C	A	SBS	MZB1 p.G78=,c.234G>T	13844	4.731%
LCRA 1774	chr11	108382851	C	G	SBS	EXPH5 p.G1128A,c.3383G>C	1236	3.641%
LCRA 1774	chr19	30935469	C	A	SBS	ZNF536 p.P334T,c.1000C>A	54933	4.935%
LCRA 1774	chr11	128680826	C	G	SBS	FLI1 p.P434=,c.1302C>G	41225	4.381%
LCRA 1774	chr19	50094944	G	C	SBS	PRR12 p.G11=,c.33G>C	24599	4.191%
LCRA 1774	chr6	27806445	C	G	SBS	HIST1H2BN p.P2=,c.6C>G	32924	3.845%
LCRA 1774	chr8	144399930	G	C	SBS	TOP1MT p.S431=,c.1293C>G	44142	2.041%
LCRA 1774	chr16	48381402	TTC	-	Indel	LONP2 c.1939-8_1939-6del	25067	4.536%
LCRA 1774	chr13	109507842	C	G	SBS	MYO16 p.L412V,c.1234C>G	43896	1.04%

Chromosomal coordinates refer to genome version hg38.

Supplementary Material

The online version contains supplementary material available at https://doi.org/10.1038/s41591-025-03579-w.

Data availability

Qualified external researchers may request access to anonymized individual patient-level clinical data based on submitted curriculum vitae and reflecting non-conflict of interest. The request proposal must also include a statistician. Data requests should be sent to the Walter and Eliza Hall Institute of Medical Research (the sponsor) study chair at tie.j@wehi.edu.au. Approval of such requests is at the Walter and Eliza Hall Institute of Medical Research and the trial steering committee’s discretion and is dependent on the nature of the request, the merit of the research proposed, the availability of the data and the intended use of the data. We will attempt to respond to data requests within 2 months, but this timeframe may vary depending on the requester’s availability to respond to comments. Once approved, a data transfer agreement will be required before any data transfer.