Abstract
The consensus molecular subtype (CMS) classification divides colon tumors into four subtypes holding promise as a predictive biomarker. However, the effect of adjuvant chemotherapy on recurrence free survival (RFS) per CMS in stage III patients remains inadequately explored. With this intention, we selected stage III colon cancer (CC) patients from the MATCH cohort (n = 575) and RadboudUMC (n = 276) diagnosed between 2005 and 2018. Patients treated with and without adjuvant chemotherapy were matched based on tumor location, T‐ and N‐stage (n = 522). Tumor material was available for 464 patients, with successful RNA extraction and CMS subtyping achieved in 390 patients (surgery alone group: 192, adjuvant chemotherapy group: 198). In the overall cohort, CMS4 was associated with poorest prognosis (HR 1.55; p = .03). Multivariate analysis revealed favorable RFS for the adjuvant chemotherapy group in CMS1, CMS2, and CMS4 tumors (HR 0.19; p = .01, HR 0.27; p < .01, HR 0.19; p < .01, respectively), while no significant difference between treatment groups was observed within CMS3 (HR 0.68; p = .51). CMS subtyping in this non‐randomized cohort identified patients with poor prognosis and patients who may not benefit significantly from adjuvant chemotherapy.
Keywords: adjuvant chemotherapy, biomarker, CMS, colorectal cancer, stage III
What's new?
The consensus molecular subtype (CMS) classification divides colon tumors into four subtypes, whose promise as predictive biomarkers needs to be further clarified. This non‐randomized, observational study in stage III colon cancer patients found that patients with CMS1, 2, or 4 tumors, but not those with CMS3 tumors, benefited from treatment with adjuvant oxaliplatin‐based chemotherapy. CMS4 was also an independent predictor for poor prognosis. The findings emphasize the additional value of molecular subtyping and underscore its potential clinical utility in guiding the selection of stage III CC patients for adjuvant chemotherapy.
1. INTRODUCTION
Biomarkers that can predict response to chemotherapy are much needed to improve and tailor treatment strategies in non‐metastatic colon cancer (CC). Current treatment of CC is primarily based on the tumor‐node‐metastasis (TNM) classification 1 and the backbone of treatment for patients with stage I–III CC is surgical resection of the primary tumor. Adjuvant systemic chemotherapy, fluoropyrimidine either in combination with oxaliplatin or as monotherapy, is indicated in patients with high‐risk stage II and stage III CC.
Despite this intensive treatment, 25%–35% of the patients treated with adjuvant chemotherapy develop metastatic disease, while approximately 50%–60% of the high risk stage II and stage III patients would have remained disease‐free without adjuvant chemotherapy treatment. 2 , 3 Importantly, adjuvant chemotherapy is associated with significant side effects, ranging from gastrointestinal to neurological toxicity. Recent evidence indicates that reduction in the duration of adjuvant therapy from 6 to 3 months is not inferior in terms of clinical benefit. 4 As chemotherapy only benefits a subset of the CC patients, identifying the patients at risk of developing metastases, as well as those responding to chemotherapy is a clear unmet need in CC care and the development of new prognostic and predictive markers for chemotherapy response is therefore of utmost importance. Currently, a range of biomarkers, such as the Immunoscore, 5 , 6 tumor‐stroma ratio 7 , 8 and circulating tumor DNA, 9 , 10 have been identified, but their development toward clinical practice has proven difficult. In addition, these biomarkers are developed to identify patients at high risk for recurrence, while their role in predicting chemotherapy benefit remains uncertain.
The consensus molecular subtype (CMS) classification divides CC in four biologically distinct subtypes based on bulk mRNA expression data, and holds great promise as a biomarker. CMS1, the immunogenic subtype, is characterized by microsatellite instability (MSI), hypermutation, and BRAF mutations. CMS2, the most prevalent and so‐called canonical subtype, has epithelial features with marked WNT and MYC signaling. CMS3, the metabolic subtype, exhibits increased expression of multiple metabolic pathways and is enriched for KRAS mutations. Lastly, CMS4, the mesenchymal subtype, is associated with pronounced stromal invasion, angiogenesis and TGF‐β activation. Previous studies have shown that CMS might be clinically useful in predicting both prognosis and response to chemotherapy. 11 , 12 , 13 However, data on the predictive value of the CMSs for chemotherapy response in non‐metastatic CC are limited and warrant careful interpretation due to heterogeneous study designs with differences in sample size, treatment regimens and outcome measurements.
An important obstacle for clinical implementation of CMS is the current lack of robust and practical classification methods. The gold standard classification strategy relies on mRNA expression data from sufficient fresh frozen tissue (FF), while routine clinical specimens are usually prepared as formalin‐fixed, paraffin‐embedded (FFPE) tissue. This process leads to reduced RNA quality, which affects subsequent amplification steps required to generate RNAseq, microarray or PCR data. NanoString technology offers a solution for transcriptome‐profiling of FFPE tissues, as it can reliably profile partially fragmented RNA without the need for an amplification step. 14 , 15 , 16 To determine the impact of chemotherapy on the distinct molecular subtypes a novel, in‐house, NanoString‐based assay was developed, which allows for CMS stratification of both FF and FFPE material. 17
The biological differences between CMSs might contribute to the differences in patient outcome and response to therapy, even within the same TNM‐stage. In this study, we evaluated the added value of CMS in predicting both prognosis and benefit of adjuvant chemotherapy in stage III CC patients.
2. MATERIALS AND METHODS
2.1. Study design and population
We conducted a multicenter observational cohort study in stage III CC patients. Patients were selected from one academic hospital (RadboudUMC) in the Netherlands and the MATCH cohort. 18 The MATCH study is a multicenter cohort study including patients with stage I–III colorectal cancer (CRC) from 2007 until December 2017 in seven hospitals (one academic, six non‐academic teaching hospitals) in the region of Rotterdam, the Netherlands. The aim of the MATCH study was to collect fresh frozen (FF) tissue samples and matched clinical data to develop molecular markers for CRC patients. 19
Patients with stage III CC who did not undergo neoadjuvant therapy were selected from both cohorts, resulting in 575 patients in the MATCH study and 276 from the RadboudUMC (Figure 1). Sufficient follow‐up (≥2 years) data were available for 191 patients within the RadboudUMC cohort. Of these 191, patients treated with surgery plus adjuvant chemotherapy (adjuvant chemotherapy group) were case‐matched based on tumor location (right vs. left), T‐stage (T1‐3 vs. T4) and N‐stage (N1 vs. N2) with patients treated with surgery only (surgery alone group), resulting in 112 patients. FFPE blocks were collected and RNA extraction and CMS classification was successful for 83 patients (surgery alone group n = 42 and adjuvant chemotherapy group n = 41).
FIGURE 1.
Patient flowchart.
Since FF tissue was collected for a subset of patients within the MATCH study, we performed a parallel patient selection according to tissue availability. Out of the 575 MATCH patients, FF tissue was available for 304. After case‐matching of the treatment groups (on tumor location, T‐stage and N‐stage) out of these 304, 228 patients remained, of which 165 samples passed RNA quality control (RIN ≥1 and tumor percentage >30%) and were subjected to RNA analysis and subsequent CMS classification using our newly developed NanoString platform that classifies CRC with high fidelity into the CMSs. 17 Of these, 82 patients received adjuvant chemotherapy, while 83 were treated with surgery alone.
Parallel to this, we performed an additional patient selection with the remaining 410 patients, who were not represented in the final FF set. Patients who presented with a synchronous second primary colon tumor (n = 4) or when follow‐up was not available for ≥2 years (n = 78) were excluded. Subsequently this set was subjected to case‐matching for treatment groups as well, resulting in 182 patients divided over the two arms. From these FFPE tissue was available for 151 patients, RNA extraction and CMS classification was successful in 142 cases resulting in a surgery alone group of n = 67 and adjuvant chemotherapy group of n = 75.
These three sets were combined for further analysis, resulting in a total of 390 stage III CC patients that were CMS subtyped and included for further analysis, 192 in the surgery alone group and 198 in the adjuvant chemotherapy group.
2.2. CMS classification
CMS classification was performed using an in‐house developed NanoString‐based classifier. Details on the development and validation of this NanoClassifier (NanoCMSer) are described elsewhere. 17 Briefly, we selected a set of 55 genes with high discriminatory power based on differential expression analysis to develop a custom NanoString codeset. This codeset was consequently used to generate gene‐expression data of paired FF and FFPE samples, in order to develop a classifier that would work on both materials. The gold‐standard CMS labels were obtained for these samples by RNA sequencing and the single‐sample predictor function of the CMSclassifier package. 20
The accuracy of the developed NanoCMSer was 95% and 92% for FF and FFPE respectively, as compared to the gold‐standard RNA sequencing method. 17
2.3. RNA extraction
RNA was extracted from FF tissue using RNA‐Bee (Tel‐Test, Inc., USA) and Trizol (ABP Biosciences, Rockville, USA) according to manufacturer's protocol. Samples with a tumor percentage <30% were excluded. For FFPE tissue, three 10 μm‐slides with flanking H&Es were prepared and tissue deparaffinization was performed with Xylene. Subsequently, tumor tissue was macro‐dissected to enrich for tumor tissue. RNA was extracted using the RNeasy FFPE kit (QIAGEN GmbH, Hilden, Germany) according to manufacturer's instructions. RNA quantity and quality were measured using Tapestation (Agilent).
2.4. NanoString profiling
Gene expression levels of the included FF and FFPE samples were measured using the developed codeset and positive and negative probes were included to monitor the quality of the runs. Sample input was 50–150 ng of total RNA, depending on the RNA quality. Log2‐transformation and quantile normalization were performed as general method of normalization in this study.
2.5. Statistical analysis
Case‐matching between the two treatment groups, surgery plus adjuvant chemotherapy versus surgery alone was performed on tumor location, T‐stage and N‐stage, using the case–control function in SPSS. Differences in baseline characteristics between the two treatment groups, were tested with Pearson's Chi‐square analysis or Mann–Whitney U‐test as appropriate. Association of clinical variables and CMS subtypes were analyzed with Pearson's Chi‐square test or Mann–Whitney U‐test as appropriate. The primary endpoint was recurrence free survival (RFS), defined as the time from the date of surgery to the first documented recurrence (local or distant) or last moment of follow‐up. Patients who died before evidence of a recurrence were censored. Kaplan Meier method was used to describe the distribution of time‐to‐recurrence and log‐rank test was used to compare the treatment groups. Median follow‐up was calculated from time of surgery to last moment of follow‐up, using reverse Kaplan Meier method. After establishing that 7.9% of tumor differentiation data was missing at random, multiple imputation using the monotone method was applied. Prognostic significance of clinical variables and CMS was assessed with univariable and multivariable Cox proportional hazard models. Known predictors (age, tumor location, T‐stage, N‐stage, number of examined lymph nodes and differentiation grade) for clinical outcome in CC were used in the multivariable Cox model. The Schoenfield test was used to evaluate the proportional‐hazard assumption, and any variables found to violate this assumption were modeled as stratification factors. In order to evaluate the predictive value of CMS, we evaluated the effect of adjuvant chemotherapy in strata per CMS subtype. p‐values are two‐tailed and results <.05 were considered significant. Statistical analyses were performed in IBM SPSS software version 28 (SPSS, Inc. and IBM Company, Chicago, IL, USA) and R version 4.2.0.
3. RESULTS
3.1. Case‐matching of colon cancer patients
In total, 390 stage III CC patients were included in the analyses (Figure 1). The median age was 70 years (IQR 63–76) and 47.4% (n = 185) were female (Table 1). In the adjuvant chemotherapy group, most patients received capecitabine and oxaliplatin combination therapy (CAPOX) (n = 118, 59.6%), followed by capecitabine monotherapy (n = 41, 20.7%) and 5‐FU plus oxaliplatin (FOLFOX) (n = 33, 16.7%).
TABLE 1.
Demographic variables.
| All patients (n = 390) | Chemotherapy (n = 198) | No chemotherapy (n = 192) | p‐value | |
|---|---|---|---|---|
| Age | ||||
| Median (IQR) | 70 (63–76) | 65 (59–71) | 76 (69–80) | <.001 |
| Gender | ||||
| Male | 205 (52.6) | 106 (53.5) | 99 (51.6) | .696 |
| Female | 185 (47.4) | 92 (46.5) | 93 (48.4) | |
| ASA score | ||||
| I–II | 260 (77.2) | 162 (89.5) | 98 (62.8) | <.001 |
| III–IV | 77 (22.8) | 19 (10.5) | 58 (37.2) | |
| Missing | 53 | 17 | 36 | |
| Tumor location | ||||
| Right colon | 186 (47.7) | 95 (48.0) | 91 (47.4) | .908 |
| Left colon | 204 (52.3) | 103 (52.0) | 101 (52.6) | |
| pT‐stage | ||||
| 1 | 11 (2.8) | 5 (2.5) | 6 (3.1) | .888 |
| 2 | 63 (16.2) | 31 (15.7) | 32 (16.7) | |
| 3 | 243 (62.3) | 127 (64.1) | 116 (60.4) | |
| 4 | 73 (18.7) | 35 (17.7) | 38 (19.8) | |
| pN‐stage | ||||
| 1 | 267 (68.5) | 129 (65.2) | 138 (71.9) | .153 |
| 2 | 123 (31.5) | 69 (34.8) | 54 (28.1) | |
| Tumor differentiation | ||||
| Good | 3 (0.8) | 1 (0.6) | 2 (1.1) | .837 |
| Moderate | 293 (81.6) | 145 (81.5) | 148 (81.8) | |
| Poor | 63 (17.5) | 32 (18.0) | 31 (17.1) | |
| Missing | 31 | 20 | 11 | |
| Lymph nodes examined | ||||
| <12 | 111 (28.5) | 60 (30.3) | 51 (26.6) | .413 |
| ≥12 | 279 (71.5) | 138 (69.7) | 141 (73.4) | |
| CMS classification | ||||
| CMS1 | 77 (19.7) | 39 (19.7) | 38 (19.8) | .793 |
| CMS2 | 180 (46.2) | 87 (43.9) | 93 (48.4) | |
| CMS3 | 52 (13.3) | 28 (14.1) | 24 (12.5) | |
| CMS4 | 81 (20.8) | 44 (22.2) | 37 (19.3) |
Note: Data are presented as number (%) unless otherwise stated. p‐values calculated with Pearson's Chi‐square analysis or Mann–Whitney U‐test as appropriate.
Abbreviation: CMS, consensus molecular subtype.
Although we ensured that several key tumor characteristics were similar for both treatment groups, treatment assignment had an inherent bias. As expected, patients receiving adjuvant chemotherapy were on average younger (median age 65 vs. 76, p < .001) and had a lower ASA‐score (11% ASA I–II vs. 37% ASA III–IV, p < .001) (Table 1). Other known variables like gender, number of examined lymph nodes and tumor differentiation were equally distributed between the two groups.
3.2. Distribution of CMS
To determine the molecular subtype of all 390 tumors, RNA was isolated and profiled using the NanoString codeset and the classification algorithm. 17 Of the 390 patients, 77 (19.7%) were classified as CMS1, 180 (46.2%) as CMS2, 52 (13.3%) as CMS3 and 81 (20.8%) as CMS4. These results are in line with published proportions for stage III CC. 20 The distribution of patients who received surgery alone or surgery in combination with adjuvant therapy was balanced within each CMS.
Next, we evaluated the association between clinicopathological features and the CMS subtypes (Table 2). As expected, CMS1 was associated with poor tumor differentiation, female gender and right‐sided tumors (all p < .001). CMS2 tumors were more often located in the left colon and were associated with pT1‐3 status (both p < .001). No associations were seen for CMS3, while CMS4 was associated with pT4 tumors (p = .012). These results are consistent with previous studies 11 , 21 , 22 , 23 , 24 , 25 , 26 and further substantiate the performance of the developed NanoString classifier. Further analysis indicated that the distribution of CMS subtypes was not completely equal among the tissue types (FF vs. FFPE) used within this study (Table S1). However, further analysis of additional datasets revealed no significant differences. Specifically, when the PETACC3 FFPE dataset was compared directly with the TCGA FF dataset, no differences in CMS distribution were detected (Table S2), suggesting that the observed difference between tissue types in our study was not a result of classifier‐induced effects.
TABLE 2.
Clinicopathologic features per CMS.
| CMS1 | CMS2 | CMS3 | CMS4 | p‐value | |
|---|---|---|---|---|---|
| Age | |||||
| Median (IQR) | 77 (63–78) | 69 (63–77) | 71 (64–76) | 70 (61–77) | .805 |
| Gender | |||||
| Male | 21 (27.3) | 101 (56.1) | 32 (61.5) | 51 (63.0) | <.001 |
| Female | 56 (72.7) | 79 (43.9) | 20 (38.5) | 30 (37.0) | |
| ASA score | |||||
| I–II | 48 (71.6) | 121 (77.1) | 36 (83.7) | 55 (78.6) | .515 |
| III–IV | 19 (28.4) | 36 (22.9) | 7 (16.3) | 15 (21.4) | |
| Missing | 10 | 23 | 9 | 11 | |
| Tumor location | |||||
| Right colon | 63 (81.8) | 50 (27.8) | 27 (51.9) | 46 (56.8) | <.001 |
| Left colon | 14 (18.2) | 130 (72.2) | 25 (48.1) | 35 (43.2) | |
| pT‐stage | |||||
| 1 | 2 (2.6) | 5 (2.8) | 4 (7.7) | 0 (0.0) | <.001 |
| 2 | 8 (10.4) | 38 (21.1) | 12 (23.1) | 5 (6.2) | |
| 3 | 47 (61.0) | 116 (64.4) | 27 (51.9) | 53 (65.4) | |
| 4 | 20 (26.0) | 21 (11.7) | 9 (17.3) | 23 (28.4) | |
| pN‐stage | |||||
| 1 | 51 (66.2) | 129 (71.7) | 34 (65.4) | 53 (65.4) | .658 |
| 2 | 26 (33.8) | 51 (28.3) | 18 (34.6) | 28 (34.6) | |
| Tumor differentiation | |||||
| Good | 0 (0.0) | 2 (1.2) | 1 (2.3) | 0 (0.0) | <.001 |
| Moderate | 36 (52.9) | 159 (91.9) | 38 (86.4) | 60 (81.1) | |
| Poor | 32 (47.1) | 12 (6.9) | 5 (11.4) | 14 (18.9) | |
| Missing | 9 | 7 | 8 | 7 | |
| Lymph nodes examined | |||||
| <12 | 19 (24.7) | 58 (32.2) | 13 (25.0) | 21 (25.9) | .502 |
| ≥12 | 58 (75.3) | 122 (67.8) | 39 (75.0) | 60 (74.1) |
Note: Data are presented as number (%) unless otherwise stated. p‐values calculated with Pearson's Chi‐square analysis or Mann–Whitney U‐test as appropriate.
Abbreviation: CMS, consensus molecular subtype.
3.3. Prognostic markers
Although we performed case matching on several key tumor characteristics, treatment assignment had an inherent bias, as shown from the baseline differences in age and ASA score between the two groups. In order to minimize the effect of these factors on our analysis we used recurrence free survival (RFS) instead of overall or disease free survival as main endpoint. The median follow‐up time for all patients was 61 months (IQR 59–62). Patients receiving adjuvant chemotherapy had a median follow‐up of 67 months (IQR 60–73), while this was 47 months (IQR 42–51) for patients treated with surgery alone.
The Kaplan–Meier analysis showed an impaired RFS for the surgery alone group as compared to the adjuvant chemotherapy group (HR = 0.30, 95% CI 0.20–0.46, p < .001) (Figure 2A). In the overall cohort, with both treatment groups combined, CMS4 was significantly associated with an impaired RFS as compared to CMS 1–3 (HR = 1.66, 95% CI 1.13–2.43, p = .01) (Figure 2B) and this result was maintained after adjustment for clinicopathologic features (HR = 1.55, 95% CI 1.05–2.30, p = .03). The poor prognosis of CMS4 is consistent with previous studies, 22 , 24 , 25 , 27 , 28 , 29 and these results illustrate the potential added value of CMS over the classical clinicopathologic features in estimating prognosis.
FIGURE 2.
Recurrence free survival, stratified by (A) adjuvant chemotherapy treatment and (B) consensus molecular subtype classification. (C) Effect of adjuvant chemotherapy on recurrence free survival within each consensus molecular subtype. (D) Multivariate cox regression analysis on recurrence free survival per consensus molecular subtype. Numbers represent hazard ratios. CMS, consensus molecular subtype; CT, chemotherapy; LNN, lymph nodes.
3.4. Predictive value of CMS subtypes
The main objective of this study was to evaluate the predictive value of the CMSs for chemotherapy response. To this end, we constructed Kaplan–Meier curves for each CMS comparing patients treated with adjuvant chemotherapy or surgery alone (Figure 2C). The RFS was significantly better in the adjuvant chemotherapy group for CMS1 (HR 0.24, 95% CI 0.09–0.63, p < .01), CMS2 (HR 0.42, 95% CI 0.23–0.77, p < .01) and CMS4 (HR 0.40, 95% CI 0.20–0.78, p < .01), while this was not the case for CMS3 (HR 0.86, 95% CI 0.34–2.16, p = .75). Notably, CMS1 and CMS4 show a sharp decline in the first year post‐surgery, which then levels off. This trend was similarly observed in our previous research on recurrences in stages II and III CRC. 30 , 31 Interestingly, previous literature suggests that MSI tumors can be categorized into two prognostic groups: the majority demonstrate relatively good prognosis, while a minority exhibit early recurrences, particularly within stage III. 32 , 33 This stage dependent dichotomy might explain the observed sharp decline in our study.
Next, we employed multivariate Cox proportional hazard models within each CMS stratum (Figure 2D). Importantly, after adjusting for clinicopathologic features and tissue type, adjuvant chemotherapy was beneficial in terms of RFS for CMS1 (HR 0.19, 95% CI 0.06–0.68, p = .01), CMS2 (HR 0.27, 95% CI 0.14–0.51, p < .01) and CMS4 (HR 0.19, 95% CI 0.08–0.44, p < .01) cancers, while no significant chemotherapy benefit was seen within CMS3 (HR 0.68, 95% CI 0.22–2.11, p = .51). Beside age, there were no significant differences in baseline characteristics between treatment groups within CMS3 (Tables 3 and S3). A trend was seen in tumor location (p = .05).
TABLE 3.
Baseline characteristics within CMS3 comparing treatment groups.
| Chemotherapy (n = 28) | No chemotherapy (n = 24) | p‐value | |
|---|---|---|---|
| Age | |||
| Median (IQR) | 69 (59–73) | 75 (71–79) | <.001 |
| Gender | |||
| Male | 17 (60.7) | 15 (62.5) | .895 |
| Female | 11 (39.3) | 9 (37.5) | |
| ASA score | |||
| I–II | 20 (87.0) | 16 (80.0) | .538 |
| III–IV | 3 (13.0) | 4 (20.0) | |
| Missing | 5 | 4 | |
| Tumor location | |||
| Right colon | 18 (64.3) | 9 (37.5) | .054 |
| Left colon | 10 (35.7) | 15 (62.5) | .088 |
| pT‐stage | |||
| 1 | 1 (3.6) | 3 (12.5) | |
| 2 | 5 (17.9) | 7 (29.2) | |
| 3 | 19 (67.9) | 8 (33.3) | |
| 4 | 3 (10.7) | 6 (25.0) | |
| pN‐stage | |||
| 1 | 18 (64.3) | 16 (66.7) | .857 |
| 2 | 10 (35.7) | 8 (33.3) | |
| Tumor differentiation | |||
| Good | 0 (0) | 1 (5.0) | .530 |
| Moderate | 21 (87.5) | 17 (85.0) | |
| Poor | 3 (12.5) | 2 (10.0) | |
| Missing | 4 | 4 | |
| Lymph nodes examined | |||
| <12 | 8 (28.6) | 5 (20.8) | .521 |
| ≥12 | 20 (71.4) | 19 (79.2) |
Note: Data are presented as number (%) unless otherwise stated. p‐values calculated with Pearson's Chi‐square analysis or Mann–Whitney U‐test as appropriate.
Abbreviation: CMS, consensus molecular subtype.
4. DISCUSSION
In this non‐randomized, observational study, consensus molecular subtyping identified a subgroup of patients with poor prognosis and a subgroup of patients that may not derive benefit from adjuvant chemotherapy. The adjuvant chemotherapy group showed a more favorable prognosis in CMS 1, 2 and 4 patients, while the we did not find a difference in RFS for CMS3 patients treated with and without adjuvant chemotherapy. In addition, CMS4 was an independent predictor for poor prognosis. These findings emphasize the additional value of molecular subtyping and underscore the potential clinical utility of CMS in guiding the selection of stage III CC patients for adjuvant chemotherapy. However, further validation in independent cohorts is necessary to corroborate these findings.
Importantly, our findings indicate that the benefit of standard adjuvant chemotherapy may be minimal for patients with CMS3 CC, which was unexpected based on current knowledge. The epithelial subtypes, CMS2 and 3, were previously linked to chemotherapy benefit by several studies. 11 , 12 , 13 , 34 However, data on the predictive value of CMSs for chemotherapy response are limited in non‐metastatic CC and caution should be exercised in interpreting the data due to differences in study design, sample size, treatment regimen and outcome measurements. 35
One study 11 reported that adjuvant chemotherapy improved overall survival (OS) in IHC‐based CMS2/3 patients, but its small sample size, non‐consecutive patient selection, lack of case‐matching, and missing data on chemotherapy regimen limited its conclusions. Another retrospective study 34 reported chemotherapy benefit only for the epithelial subtype, but no case‐matching was performed, and the chemotherapy applied consisted of 5‐FU monotherapy. A study by Allen et al. 12 showed significant improvement in OS for CMS3 patients receiving adjuvant chemotherapy in stage III CC, but no case‐matching was performed, resulting in highly imbalanced treatment groups. Finally, Song et al. 13 performed a prospectively designed, retrospectively tested study on patients from the C‐07 trial, a randomized controlled trial compared 5‐FU monotherapy with 5‐FU plus oxaliplatin in stage II–III CC patients. 36 They showed that only patients with CMS2 tumors gained benefit from the addition of oxaliplatin on RFS. Interestingly, CMS3 favored 5‐FU monotherapy over 5‐FU plus oxaliplatin, but this result was not significant.
The mechanism behind the observed chemotherapy resistance in CMS3 tumors in our study remains unclear, but several insights are important to note. First, the clinical outcome of CMS3 appears to be related to the TNM stage at diagnosis. Analyses on the prognostic value of the CMSs stratified by stage within the TCGA 37 and Marisa (GSE39582) dataset, revealed that within stage II disease, CMS3 is an “early” subtype with favorable prognosis compared to CMS1, 2 and 4. However, within stage III, CMS3 is associated with worse RFS compared to CMS1 and 2 and resembles the poor‐prognosis subtype CMS4. It appears that CMS3 at early stage of disease where the tumor is still local, may represent a different biological entity than at later stage of disease, where the tumor has spread. This could be due to a progression of CMS3 to a more aggressive state or due to a subtype change toward a more aggressive CMS3 type.
Second, recent findings have shown that MUC2, SPINK4 and REG4, which are known CMS3 markers, 38 are associated with non‐responsiveness to neoadjuvant chemoradiation in rectal cancer patients. 39
Our data also demonstrated that CMS classification can effectively identify patients with poor prognosis. Consistent with previous studies 11 , 13 , 23 , 24 , 27 , 28 , 29 , 34 and a recent meta‐analysis, 35 we observed that CMS4 is linked to dismal prognosis. The prognostic value of CMS4 remained after adjustment for clinicopathologic features such as T‐ and N‐stage, indicating that stratification by CMS can improve prognostication beyond the TNM classification.
Unfortunately, data on microsatellite instability status (MSI) are lacking in this study. MSI status is currently the only biomarker with predictive value for adjuvant chemotherapy benefit that is used in clinical decision making. MSI high‐risk stage II patients should not receive any adjuvant chemotherapy, while 5‐FU monotherapy should not be given to MSI stage III patients, since these treatment strategies have been proven ineffective. 33 MSI patients would mainly constitute the CMS1 subtype, 20 which showed significant benefit from adjuvant chemotherapy in our study. The majority (69%) of these patients received oxaliplatin‐based chemotherapy, which accordingly has been shown to be beneficial in MSI stage III CC patients. 40 , 41
This study has several limitations that should be acknowledged. First, the lack of randomization raises concerns about potential confounding factors that may have influenced treatment assignment and outcomes. In order to minimize the effect of these factors on our analysis we performed case‐matching in three important prognostic factors (tumor location, T‐stage and N‐stage). The reasons behind refraining from adjuvant chemotherapy were unknown for the majority of patients. Furthermore, information regarding the duration of chemotherapy and comorbidities was not available. These factors play a significant role in treatment response and overall prognosis, and their absence hinders a comprehensive understanding of the observed outcomes. Additionally, the follow‐up in this study adhered to standard guidelines, which introduces potential variability in monitoring and surveillance practices over time and across different treatment groups. These variations can impact the accuracy of recurrence detection and potentially bias the assessment of chemotherapy effectiveness. Future studies with expanded patient cohorts are warranted to further elucidate the potential variations in treatment response among CMS subtypes. Despite the acknowledged limitations, we compiled a well‐defined stage III CC cohort with careful consideration given to balance treatment groups based on several key tumor characteristics. We, therefore, believe that our findings highlight the clinical potential of the CMS classification in guiding treatment decisions for CC patients and underscore the importance of incorporating molecular subtyping in future trial designs.
In conclusion, while recognizing the limitations inherent in our study, we believe that our findings shed light on the clinical significance of CMS classification in stage III CC. The insights gained from this research emphasize the need for continued efforts to validate and refine CMS‐based approaches, enabling more targeted and effective therapeutic interventions. By embracing molecular subtyping strategies like CMS, we can enhance the precision and impact of treatment decisions, ultimately leading to improved outcomes and patient care in the management of CC.
AUTHOR CONTRIBUTIONS
Simone van de Weerd: Conceptualization; data curation; formal analysis; visualization; writing – original draft; writing – review and editing. Arezo Torang: Conceptualization; formal analysis; visualization; writing – original draft; writing – review and editing. Inge van den Berg: Data curation; writing – original draft; writing – review and editing. Veerle Lammers: Data curation; writing – original draft; writing – review and editing. Saskia van den Bergh: Data curation; writing – original draft; writing – review and editing. Nelleke Brouwer: Writing – original draft; writing – review and editing. Iris D. Nagtegaal: Writing – original draft; writing – review and editing. Miriam Koopman: Writing – original draft; writing – review and editing. Geraldine R. Vink: Writing – original draft; writing – review and editing. Frederieke H. van der Baan: Writing – original draft; writing – review and editing. Han van Krieken: Writing – original draft; writing – review and editing. Jan Koster: Writing – original draft; writing – review and editing. Jan N. Ijzermans: Writing – original draft; writing – review and editing. Jeanine M. L. Roodhart: Conceptualization; formal analysis; visualization; writing – original draft; writing – review and editing. Jan Paul Medema: Conceptualization; formal analysis; visualization; writing – original draft; writing – review and editing.
CONFLICT OF INTEREST STATEMENT
The authors declare no conflicts of interest.
ETHICS STATEMENT
The MATCH study was approved by the Medical Ethical Board of the Erasmus University Medical Center, Rotterdam, the Netherlands (MEC‐2007‐088). Written informed consent was obtained from all participants in the MATCH study for the use of their data. Similarly, the RadboudUMC study was approved by the Medical Ethical Board of the Radboud University Medical Center, Nijmegen, The Netherlands (MEC‐2014‐174). A waiver of consent was granted for this study and patients were given the option to refuse to participate by opting out. Patients who objected to the use of their data for scientific research were excluded from participation.
Supporting information
Data S1. Supporting information.
ACKNOWLEDGEMENTS
We would like to thank the pathology departments of the following hospitals for providing tumor tissue: Erasmus MC, Reinier de Graaf Gasthuis, Maasstad Ziekenhuis, Ikazia ziekenhuis, IJsselland ziekenhuis, Franciscus Gasthuis & Vlietland, Albert Schweitzer ziekenhuis, Radboud UMC. We would like to thank Kirsten Ruigrok‐Ritstier, Vanja de Weerd and Saskia Wilting for their work on the collection and RNA extraction of fresh‐frozen tissue samples from the MATCH cohort.
van de Weerd S, Torang A, van den Berg I, et al. Benefit of adjuvant chemotherapy on recurrence free survival per consensus molecular subtype in stage III colon cancer. Int J Cancer. 2025;156(2):456‐466. doi: 10.1002/ijc.35120
Simone van de Weerd and Arezo Torang contributed equally to this work.
Jeanine M. L. Roodhart and Jan Paul Medema share last authorship.
DATA AVAILABILITY STATEMENT
The datasets generated and analyzed during the current study are submitted to the GEO repository with accession number of GSE250057.
REFERENCES
- 1. Amin MB, Greene FL, Edge SB, et al. The Eighth Edition AJCC Cancer Staging Manual: continuing to build a bridge from a population‐based to a more “personalized” approach to cancer staging. CA Cancer J Clin. 2017;67(2):93‐99. [DOI] [PubMed] [Google Scholar]
- 2. Wille‐Jørgensen P, Syk I, Smedh K, et al. Effect of more vs less frequent follow‐up testing on overall and colorectal cancer‐specific mortality in patients with stage II or III colorectal cancer: the COLOFOL randomized clinical trial. Jama. 2018;319(20):2095‐2103. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Primrose JN, Perera R, Gray A, et al. Effect of 3 to 5 years of scheduled CEA and CT follow‐up to detect recurrence of colorectal cancer: the FACS randomized clinical trial. Jama. 2014;311(3):263‐270. [DOI] [PubMed] [Google Scholar]
- 4. Grothey A, Sobrero AF, Shields AF, et al. Duration of adjuvant chemotherapy for stage III colon cancer. N Engl J Med. 2018;378(13):1177‐1188. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. El Sissy C, Kirilovsky A, Zeitoun G, et al. Therapeutic implications of the immunoscore in patients with colorectal cancer. Cancer. 2021;13(6):1281. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Pagès F, André T, Taieb J, et al. Prognostic and predictive value of the Immunoscore in stage III colon cancer patients treated with oxaliplatin in the prospective IDEA France PRODIGE‐GERCOR cohort study. Ann Oncol. 2020;31(7):921‐929. [DOI] [PubMed] [Google Scholar]
- 7. van Pelt GW, Sandberg TP, Morreau H, et al. The tumour‐stroma ratio in colon cancer: the biological role and its prognostic impact. Histopathology. 2018;73(2):197‐206. [DOI] [PubMed] [Google Scholar]
- 8. Mesker WE, Junggeburt JM, Szuhai K, et al. The carcinoma‐stromal ratio of colon carcinoma is an independent factor for survival compared to lymph node status and tumor stage. Cell Oncol. 2007;29(5):387‐398. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Tie J, Cohen JD, Lo SN, et al. Prognostic significance of postsurgery circulating tumor DNA in nonmetastatic colorectal cancer: individual patient pooled analysis of three cohort studies. Int J Cancer. 2021;148(4):1014‐1026. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Tie J, Cohen JD, Wang Y, et al. Circulating tumor DNA analyses as markers of recurrence risk and benefit of adjuvant therapy for stage III colon cancer. JAMA Oncol. 2019;5(12):1710‐1717. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Li Y, Yao Q, Zhang L, et al. IHC‐based consensus molecular subtypes as a prognostic and predictive biomarker for adjuvant chemotherapy in patients with stage II colorectal cancer. Oncologist. 2020;25:e1968‐e1979. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Allen WL, Dunne PD, McDade S, et al. Transcriptional subtyping and CD8 immunohistochemistry identifies poor prognosis stage II/III colorectal cancer patients who benefit from adjuvant chemotherapy. JCO Precis Oncologia. 2018;2018:1‐15. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Song N, Pogue‐Geile KL, Gavin PG, et al. Clinical outcome from oxaliplatin treatment in stage II/III colon cancer according to intrinsic subtypes: secondary analysis of NSABP C‐07/NRG oncology randomized clinical trial. JAMA Oncol. 2016;2(9):1162‐1169. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Veldman‐Jones MH, Brant R, Rooney C, et al. Evaluating robustness and sensitivity of the NanoString technologies nCounter platform to enable multiplexed gene expression analysis of clinical samples. Cancer Res. 2015;75(13):2587‐2593. [DOI] [PubMed] [Google Scholar]
- 15. Nielsen T, Wallden B, Schaper C, et al. Analytical validation of the PAM50‐based Prosigna breast cancer prognostic gene signature assay and nCounter analysis system using formalin‐fixed paraffin‐embedded breast tumor specimens. BMC Cancer. 2014;14:177. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Lee J, Sohn I, Do IG, et al. Nanostring‐based multigene assay to predict recurrence for gastric cancer patients after surgery. PLoS One. 2014;9(3):e90133. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Torang A, Weerd S, Lammers V, Hooff S, Koster J, Medema JP. Development of a CMS classifier for clinical CRC samples. bioRxiv. 2023;2023.06.28.546869. [Google Scholar]
- 18. van Vugt JLA, Coebergh van den Braak RRJ, Lalmahomed ZS, et al. Impact of low skeletal muscle mass and density on short and long‐term outcome after resection of stage I–III colorectal cancer. Eur J Surg Oncol. 2018;44(9):1354‐1360. [DOI] [PubMed] [Google Scholar]
- 19. Lalmahomed ZS, Coebergh van den Braak RRJ, Oomen MHA, Arshad SP, Riegman PHJ, IJzermans JNM. Multicenter fresh frozen tissue sampling in colorectal cancer: does the quality meet the standards for state of the art biomarker research? Cell Tissue Bank. 2017;18(3):425‐431. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Guinney J, Dienstmann R, Wang X, et al. The consensus molecular subtypes of colorectal cancer. Nat Med. 2015;21(11):1350‐1356. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Dunne PD, O'Reilly PG, Coleman HG, et al. Stratified analysis reveals chemokine‐like factor (CKLF) as a potential prognostic marker in the MSI‐immune consensus molecular subtype CMS1 of colorectal cancer. Oncotarget. 2016;7(24):36632‐36644. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Haasnoot KJC, Backes Y, Moons LMG, et al. Associations of non‐pedunculated T1 colorectal adenocarcinoma outcome with consensus molecular subtypes, immunoscore, and microsatellite status: a multicenter case‐cohort study. Mod Pathol. 2020;33(12):2626‐2636. [DOI] [PubMed] [Google Scholar]
- 23. Williams DS, Mouradov D, Jorissen RN, et al. Lymphocytic response to tumour and deficient DNA mismatch repair identify subtypes of stage II/III colorectal cancer associated with patient outcomes. Gut. 2019;68(3):465‐474. [DOI] [PubMed] [Google Scholar]
- 24. Piskol R, Huw LY, Sergin I, et al. A clinical applicable gene expression classifier reveals intrinsic and extrinsic contributions to consensus molecular subtypes in primary and metastatic colon cancer. Clin Cancer Res. 2019;25:4431‐4442. [DOI] [PubMed] [Google Scholar]
- 25. Sveen A, Bruun J, Eide PW, et al. Colorectal cancer consensus molecular subtypes translated to preclinical models uncover potentially targetable cancer cell dependencies. Clin Cancer Res. 2018;24(4):794‐806. [DOI] [PubMed] [Google Scholar]
- 26. Kwon Y, Park M, Jang M, et al. Prognosis of stage III colorectal carcinomas with FOLFOX adjuvant chemotherapy can be predicted by molecular subtype. Oncotarget. 2017;8(24):39367‐39381. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Shinto E, Oki E, Shimokawa M, et al. A validation study for recurrence risk stratification of stage II colon cancer using the 55‐gene classifier. Oncology. 2020;1‐8:534‐541. [DOI] [PubMed] [Google Scholar]
- 28. Jary M, Hasanova R, Vienot A, et al. Molecular description of ANGPT2 associated colorectal carcinoma. Int J Cancer. 2020;147(7):2007‐2018. [DOI] [PubMed] [Google Scholar]
- 29. Dienstmann R, Villacampa G, Sveen A, et al. Relative contribution of clinicopathological variables, genomic markers, transcriptomic subtyping and microenvironment features for outcome prediction in stage II/III colorectal cancer. Ann Oncol. 2019;30(10):1622‐1629. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Coebergh van den Braak RRJ, Ten Hoorn S, Sieuwerts AM, et al. Interconnectivity between molecular subtypes and tumor stage in colorectal cancer. BMC Cancer. 2020;20(1):850. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. De Sousa EMF, Wang X, Jansen M, et al. Poor‐prognosis colon cancer is defined by a molecularly distinct subtype and develops from serrated precursor lesions. Nat Med. 2013;19(5):614‐618. [DOI] [PubMed] [Google Scholar]
- 32. Zaanan A, Cuilliere‐Dartigues P, Guilloux A, et al. Impact of p53 expression and microsatellite instability on stage III colon cancer disease‐free survival in patients treated by 5‐fluorouracil and leucovorin with or without oxaliplatin. Ann Oncol. 2010;21(4):772‐780. [DOI] [PubMed] [Google Scholar]
- 33. Sargent DJ, Marsoni S, Monges G, et al. Defective mismatch repair as a predictive marker for lack of efficacy of fluorouracil‐based adjuvant therapy in colon cancer. J Clin Oncol. 2010;28(20):3219‐3226. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34. Roepman P, Schlicker A, Tabernero J, et al. Colorectal cancer intrinsic subtypes predict chemotherapy benefit, deficient mismatch repair and epithelial‐to‐mesenchymal transition. Int J Cancer. 2014;134(3):552‐562. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Ten Hoorn S, de Back TR, Sommeijer DW, Vermeulen L. Clinical value of consensus molecular subtypes in colorectal cancer: a systematic review and meta‐analysis. J Natl Cancer Inst. 2021;114:503‐516. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Kuebler JP, Wieand HS, O'Connell MJ, et al. Oxaliplatin combined with weekly bolus fluorouracil and leucovorin as surgical adjuvant chemotherapy for stage II and III colon cancer: results from NSABP C‐07. J Clin Oncol. 2007;25(16):2198‐2204. [DOI] [PubMed] [Google Scholar]
- 37. Cancer Genome Atlas N . Comprehensive molecular characterization of human colon and rectal cancer. Nature. 2012;487(7407):330‐337. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38. Joanito I, Wirapati P, Zhao N, et al. Single‐cell and bulk transcriptome sequencing identifies two epithelial tumor cell states and refines the consensus molecular classification of colorectal cancer. Nat Genet. 2022;54(7):963‐975. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39. Chen TJ, Tian YF, Chou CL, et al. High SPINK4 expression predicts poor outcomes among rectal cancer patients receiving CCRT. Curr Oncol. 2021;28(4):2373‐2384. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40. Cohen R, Taieb J, Fiskum J, et al. Microsatellite instability in patients with stage III colon cancer receiving fluoropyrimidine with or without oxaliplatin: an ACCENT pooled analysis of 12 adjuvant trials. J Clin Oncol. 2021;39(6):642‐651. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41. André T, de Gramont A, Vernerey D, et al. Adjuvant fluorouracil, leucovorin, and oxaliplatin in stage II to III colon cancer: updated 10‐year survival and outcomes according to BRAF mutation and mismatch repair status of the MOSAIC study. J Clin Oncol. 2015;33(35):4176‐4187. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data S1. Supporting information.
Data Availability Statement
The datasets generated and analyzed during the current study are submitted to the GEO repository with accession number of GSE250057.