Microsatellite instability and survival after adjuvant chemotherapy among stage II and III colon cancer patients: results from a population‐based study

Previous studies have reported conflicting results regarding the benefit of administering 5‐FU‐based chemotherapy to colon cancer (CC) patients with microsatellite‐instable (MSI‐high) tumors, and results from stage‐specific analyses are scarce. Patients with stage II or III CC were recruited as part of a population‐based study between 2003 and 2015. The Cox regression models including propensity score weighting were used to calculate hazard ratios and confidence intervals for the association between chemotherapy and cancer‐specific (CSS), relapse‐free (RFS), and overall survival (OS) by stage of disease and MSI status of the tumor. Median follow‐up was 6.2 years. A total of 1010 CC patients were included in the analysis (54% stage II, 46% stage III, 20% MSI‐high). Adjuvant chemotherapy was administered to 48 (8.7%) stage II and 366 (79%) stage III patients. Overall, patients who received adjuvant chemotherapy had better CSS [HR = 0.65 (0.49–0.86)] than those who received surgery alone. Among stage II patients, only 64 (12%) cancer‐related deaths occurred, none of which in MSI‐high patients who received chemotherapy. Patients with MSI‐high tumors who received adjuvant treatment showed better CSS and a tendency toward better RFS compared to MSI‐high patients who did not receive chemotherapy [HRCSS = 0.36 (0.15–0.82), HRRFS = 0.49 (0.22–1.06)]. Patients with microsatellite‐stable (MSS) tumors receiving adjuvant chemotherapy also had significantly better survival [HRCSS = 0.65 (0.48–0.87) and HRRFS = 0.68 (0.52–0.88)]. In this population‐based study including stage II and III CC patients, we observed a survival benefit of adjuvant chemotherapy for both MSS and MSI‐high tumors. Adjuvant chemotherapy seemed to be beneficial among high‐risk stage II patients with MSI‐high tumors.

Previous studies have reported conflicting results regarding the benefit of administering 5-FU-based chemotherapy to colon cancer (CC) patients with microsatellite-instable (MSI-high) tumors, and results from stage-specific analyses are scarce. Patients with stage II or III CC were recruited as part of a population-based study between 2003 and 2015. The Cox regression models including propensity score weighting were used to calculate hazard ratios and confidence intervals for the association between chemotherapy and cancer-specific (CSS), relapse-free (RFS), and overall survival (OS) by stage of disease and MSI status of the tumor. Median follow-up was 6.2 years. A total of 1010 CC patients were included in the analysis (54% stage II, 46% stage III, 20% MSI-high). Adjuvant chemotherapy was administered to 48 (8.7%) stage II and 366 (79%) stage III patients. Overall, patients who received adjuvant chemotherapy had better CSS [HR = 0.65 (0.49-0.86)] than those who received surgery alone. Among stage II patients, only 64 (12%) cancer-related deaths occurred, none of which in MSI-high patients who received chemotherapy. Patients with MSI-high tumors who received adjuvant treatment showed better CSS and a tendency toward better RFS compared to MSI-high patients who did not receive chemotherapy [HR CSS = 0.36 (0.15-0.82), HR RFS = 0.49 (0.22-1.06)]. Patients with microsatellite-stable (MSS) tumors receiving adjuvant chemotherapy also had significantly better survival [HR CSS = 0.65 (0.48-0.87) and HR RFS = 0.68 (0.52-0.88)]. In this population-based study including stage II and III CC patients, we observed a survival benefit of adjuvant chemotherapy for both MSS and MSI-high tumors. Adjuvant chemotherapy seemed to be beneficial among high-risk stage II patients with MSI-high tumors.

Introduction
Colorectal cancer is a leading cause of cancer incidence and mortality worldwide (Bray et al., 2018). For stage III colon cancer (CC) patients, the standard of treatment includes curative resection and adjuvant chemotherapy (National Comprehensive Cancer Network, 2019). For stage II CC patients, administration of adjuvant chemotherapy is recommended only when high-risk features are present: T4 tumors; inadequate lymph node sampling (< 12); poorly differentiated histology [for microsatellite-instable (MSI)-H]; lymphovascular or perineural invasion; bowel obstruction; localized perforation; or close, indeterminate, or positive margins (National Comprehensive Cancer Network, 2019).
Current guidelines recommend that stage II patients with MSI-high tumors should not receive adjuvant treatment with 5-FU chemotherapy (National Comprehensive Cancer Network, 2019). This recommendation was based on evidence from two clinical trials that found no survival benefit in MSI-high tumors receiving adjuvant 5-FU (Ribic et al., 2003;Sargent et al., 2010). However, other studies found no evidence for MSI status to be predictive of either beneficial or harmful effects of chemotherapy (Bertagnolli et al., 2011;Hutchins et al., 2011), and evidence from two trials (Klingbiel et al., 2015;Quasar Collaborative Group, 2007) and a large population-based analysis (Casadaban et al., 2016) suggested that adjuvant treatment in this group of patients was beneficial. In stage III patients, the role of the MSI-high status as a predictive biomarker is also not completely clear (Klingbiel et al., 2015;Sinicrope et al., 2013). A recent pooled analysis of data from the N0147 and PETACC-8 trials showed a small but significant decrease in the risk of recurrence for MSI-high tumors among stage III CC patients receiving adjuvant chemotherapy (Zaanan et al., 2018).
Various studies have attempted to clarify these associations, but controversies still exist about the usefulness of different molecular markers in predicting survival in the context of adjuvant chemotherapy for CC. In this study, we describe the response to adjuvant chemotherapy in stage II and III CC patients and provide analyses according to the MSI status of the tumor.

Study population
The DACHS population-based study was conducted in 22 hospitals of the Rhine-Neckar region in Germany. Details of the study have been previously described (Blaker et al., 2019). Between 2003 and 2015, patients older than 30 years of age, with a histologically confirmed diagnosis of CRC, and able to participate in an interview were recruited. Long-term follow-up was performed at 3, 5, and 10 years after diagnosis, including information on type of therapy, comorbidities, and recurrence of disease. Vital status, and date and cause of death were determined from population registries and death certificates issued by the health authorities. Only stage II and III CC patients were included in this analysis. Patients with rectal cancer, and stage I or IV tumors, missing information on stage or MSI status, who died in the first month after surgery, and who received neoadjuvant chemotherapy were excluded from this analysis. Figure S1 presents the selection process and corresponding number of excluded patients. The study was approved by the ethics committees of the University of Heidelberg and of the Medical Chambers of Baden-W€ urttemberg and Rhineland-Palatinate, and all participants signed an informed consent.

Marker characterization
Tumor tissue analyses were performed on formalinfixed, paraffin-embedded samples. MSI status was determined using a mononucleotide marker panel (BAT25, BAT26, CAT25; Findeisen et al., 2005) or as reported from the patient's medical records for 93% and 7% of patients, respectively. Determination of MSI using the triple-marker panel has been shown to correctly classify 100% of MSI-high cases when compared to the traditional five-marker panel recommended by the National Cancer Institute (BAT25, BAT26, D2S123, D5S346, and D17S250; Findeisen et al., 2005). The MSI status extracted from pathology reports (7% of patients) was mostly determined using immunohistochemistry for MLH1, MSH2, or MSH6, or by genetic testing.

Adjuvant chemotherapy
All patients underwent surgical resection of the tumor. The information on adjuvant chemotherapy and administered scheme was reported by the treating physician during follow-up. The type of chemotherapy administered was categorized as 'FU-based' when the reported treatment included either 5-FU or capecitabine alone or 5-FU+ leucovorin, and as 'oxaliplatinbased' when it included oxaliplatin in addition to 5-FU/leucovorin (FOLFOX), or in addition to capecitabine (CAPOX, XELOX). Due to low sample sizes in some of the specific chemotherapy schemes, the main exposure variable used in the analyses remained as having received adjuvant chemotherapy vs surgery alone.

Statistical analyses
Clinical and pathological characteristics were described for the entire study population and by the treatment group (adjuvant chemotherapy vs no chemotherapy). The Kaplan-Meier plots were computed to estimate survival curves stratified by stage and MSI status and to calculate 5-year survival rates and log-rank tests. Cancer-specific survival (CSS) was defined as time from diagnosis until death from CC, relapse-free survival (RFS) from diagnosis until reappearance of disease, metastases, cancer death or death from other causes, and overall survival (OS) until death from any cause.
To account for the observational nature of the study, propensity scores (PSs) were calculated and used to correct for the covariate imbalance between the treated and untreated groups (Brookhart et al., 2013;Sturmer et al., 2014). PSs were estimated using logistic regression by modeling the treatment status as a function of potential confounders including age, sex, proximal or distal location, MSI status, histological grade, stage, number of affected and examined lymph nodes, and presence of comorbidities. The Charlson comorbidity index was used to calculate an overall comorbidity score to group patients into four categories from score 0 (no comorbidities) to score 3 (severe comorbidities) (Charlson et al., 1987;Quan et al., 2005).
Propensity scores were then used to weight the study population to estimate the average treatment effect on the treated (ATT: 1 for the treated group and PS/[1-PS] for the control group; Austin, 2014). Balance of potential confounders between treated and untreated groups was assessed (Austin and Stuart, 2015): (a) by graphical inspection of the overlapping distributions of PS before and after adjustment; and (b) by calculating absolute mean differences for each covariate, where a threshold of 0.10 was considered to indicate residual imbalance between exposure groups (Nguyen et al., 2017). PS models were refitted independently in subgroup analyses for stage II, stage III, microsatellite stability (MSS), and MSI-high patients to improve residual balance, to exclude the variable used for stratification when calculating the PS (i.e., stage or MSI status), and to account for the different characteristics considered by clinicians in the decision of whether to administer chemotherapy to stage II and III patients. PS-weighted Cox proportional hazards models were used to estimate hazard ratios (HRs) and confidence intervals (95% CIs) for the overall study population and for each subgroup. Due to the small sample size in subgroups of combined stage and MSI status (i.e., stage II MSS, stage II MSI), balance after PS weighting was not achieved. Therefore, only descriptive statistics are presented for these subgroups. Also, because of the limited number of patients in the MSI subgroup, no test for interaction of MSI status and chemotherapy was performed with respect to survival. All statistical analyses were performed using SAS version 9.4 (SAS Institute Inc., Cary, NC, USA) and R version 3.5.1 software (R Core Team, 2018).

Meta-analysis of results
Given the low number of patients with stage II MSIhigh cancers who received chemotherapy, a meta-analysis was performed to compare the results obtained from the survival analysis in the subgroup of stage II MSI-high patients with previously reported literature. Results from studies reporting survival for stage II colon or colorectal cancer patients stratified by MSI status were included. These studies were identified using a combination of search terms (see Data S1) in PubMed and by manual search of the bibliographies from relevant studies. The search strategy aimed to identify studies reporting HR and 95% CI for survival outcomes [i.e., disease-free survival (DFS) or OS] for adjuvant chemotherapy compared to no chemotherapy (surgery alone). Clinical trials comparing the addition of a chemotherapy agent to an already existing treatment were not included in this meta-analysis.

Overall characteristics
One thousand and ten CC patients were included in this analysis, of which 54% (n = 549) and 46% (n = 461) had stage II and III disease, respectively. Overall, 41% (n = 414) received adjuvant chemotherapy, 88% (n = 366) of whom were stage III. Table 1 presents the baseline characteristics of the study population overall and by adjuvant chemotherapy status. Patients who received chemotherapy were younger and had a lower comorbidity index compared to those not receiving adjuvant treatment. Overall, 20% (n = 206) of patients had MSI-high tumors, a third (n = 69) of whom received adjuvant chemotherapy.
FU-based or oxaliplatin-based treatment was administered to 200 (31 stage II, 169 stage III) and 203 (14 stage II, 189 stage III) patients, respectively. The distribution of patients by chemotherapy scheme is presented in Table 2.

Propensity score balance
The balance plots and calculated mean differences for the adjusted populations showed improved balance between the exposure groups for all variables, both in the overall population and in the subgroup analyses (Fig. S2).
Results from the Cox regression models using propensity score weighting showed significantly better CSS and RFS for patients who received chemotherapy [HR = 0.65 (0.49-0.86) and HR = 0.67 (0.52-0.87), respectively] compared to those who were treated with surgery alone (Table 3). Among stage II patients receiving adjuvant treatment, only four cancer deaths (8%) and eight relapse events (17%) occurred; significantly better CSS and RFS were observed for this subgroup. Also among stage III patients, better CSS

Overall survival
Death from any cause occurred in 38% (n = 207) and 48% (n = 222) of stage II and III patients, respectively. In propensity score-weighted Cox regression models, patients who received adjuvant chemotherapy had significantly better OS compared to those treated with surgery alone. In subgroup analyses, the survival benefit was maintained for both stage II and III patients and patients with MSS and MSI tumors (Table 3). MSI patients who received chemotherapy were younger than those who did not receive adjuvant treatment, and despite the inclusion of age in the PS model, this may explain the OS benefit.

Relapse-free survival by combinations of stage and MSI status
The number of treated patients and deaths or relapse events by subgroups combining stage and MSI status is presented in Table 4. Among stage II patients with MSI-high tumors who received adjuvant treatment (n = 13), no cancer deaths and two relapse events were observed. Multivariable Cox models in this subgroup of patients were used only to compare results with other studies in a meta-analysis; however, the low number of events precludes these results from generating any conclusions. Among stage II MSS patients who received adjuvant treatment, also few deaths and relapse events were observed. Similar to what was observed in the stage-specific analyses, stage III patients with MSS tumors who received adjuvant chemotherapy had better survival compared to those treated with surgery alone. Among stage III patients with MSI-high tumors (n = 73), the proportion of deaths and relapse events was lower for those who received adjuvant chemotherapy.

Meta-analysis of results for stage II MSI-high patients
Characteristics of the studies included in the metaanalysis are presented in Table S1 ( Bertagnolli et al., 2011;Hutchins et al., 2011;Kim et al., 2015;Sargent et al., 2010;Tougeron et al., 2016). The meta-analysis of results including five studies for MSI-high and four

P-value
Overall (

Discussion
In this patient cohort, we investigated the benefit of adjuvant chemotherapy in patients with CC according to disease stage and MSI status of the tumor. Overall, survival was better among patients who received adjuvant treatment compared to those who were treated with surgery alone. This reflects the known beneficial effects of adjuvant chemotherapy that have been reported in several clinical trials (Andre et al., 2009;Yothers et al., 2011). In subgroup analyses of stage II and III patients and patients with MSS or MSI-high tumors, better CSS and RFS were observed. Recent results from large clinical trials suggest that MSI-high patients who received either 5-FU alone or combination therapy with FOLFOX maintained their survival advantage compared to MSS patients  (Klingbiel et al., 2015;Zaanan et al., 2018), confirming the generally better survival among MSI-high patients and suggesting that they retain this benefit when treated with oxaliplatin-based regimens (Andre et al., 2015). A recent systematic review (Webber et al., 2015) found no difference in the survival benefit following adjuvant 5-FU chemotherapy according to the MSI status, after including treated and untreated colorectal cancer patients. Including estimates from six studies, the pooled HR for DFS in MSS and MSI-high cancers was 0.62 (0.54-0.71) and 0.84 (0.53-1.32), respectively; the difference between these two estimates was not statistically significant (P = 0.11; Webber et al., 2015). However, this review did not differentiate between stage II and III and colon or rectal cancer. Our results for MSS patients are in line with those reported by the systematic review [HR = 0.68 (0.5-0.9)]; however, our results for RFS in MSI-high patients suggest better survival [HR = 0.49 (0.2-1.1)]. This difference may be explained by the inclusion of studies on rectal cancer, whereas our study is limited to CC patients.
Few studies have reported survival results specifically for stage II MSI-high patients, given the small number of patients who have received treatment and the resulting lack of power to detect a benefit of adjuvant treatment. Our results reflect the good prognosis that both early-stage and MSI-high patients have. The analysis in this small subgroup in our cohort and the comparison with other reported results suggest better survival for stage II MSI-high patients who received adjuvant treatment compared to those treated with surgery alone. Thus, the presence of MSI-high tumor might not be decisive for the use of chemotherapy in stage II patients with high-risk features, especially if the option of added oxaliplatin is also considered.
Patients included in this study were recruited after 2003, when the recommendation to not administer FU-based chemotherapy to MSI-high patients was introduced in the international literature (Benson et al., 2004;Ribic et al., 2003). In Germany, this recommendation was not officially introduced until 2017; however, it is possible that clinicians in the study region modified their treatment decisions in regard to the American recommendations. Since then, the number of stage II patients with MSI-high tumors who received chemotherapy has decreased, resulting in a low number of cases available for analyses. Nevertheless, the few deaths occurring among those patients who did receive adjuvant treatment point to a potential survival benefit. With the increasing use of combined chemotherapy, adding oxaliplatin to 5-FU-based regimens (i.e., FOLFOX) seems to also have a benefit in the survival of patients with MSI-high tumors (Zaanan et al., 2018). Future clinical trials will help elucidate the predictive utility of MSI and other molecular markers in the adjuvant setting for stage II and III CC patients.
The main limitations of this analysis include its observational nature and the small sample size in the stage II subgroup, since the study was not powered to detect effects of chemotherapy in small subgroups of patients. Because of the observational design of the study, the patient populations who received chemotherapy or surgery alone were not comparable at baseline. We corrected for this imbalance using propensity score weighting, which led to well-balanced covariate distributions between patients who received chemotherapy and those treated with surgery alone, both in the overall and in stage and MSI status subgroup analyses. The possibility of residual confounding due to unmeasured factors remains and may affect the PS model assumptions. The observed results for OS were generally stronger than those observed for cancer-specific and RFS. This reflects a real clinical practice scenario, where healthier patients are more likely to receive treatment and could be an indication thatalthough advanced statistical methods were employed to correct for the potential study design issues-such differences between treated and untreated patients cannot be completely removed. Information on adverse events was not taken into account in these analyses. Because the addition of oxaliplatin carries risk of adverse events, such as chronic neurological deficits (Meyers et al., 2017), the decision of administering adjuvant chemotherapy to patients with MSI-high cancers needs to be carefully considered by the clinician. Strengths of the analysis include the balance of baseline characteristics of the population to emulate a clinical trial design and the long-term follow-up. Additionally, our results reflect the effects of adjuvant chemotherapy in a real-world scenario and capture changes in treatment patterns as a result of the recommendations to not administer 5-FU to MSI patients and with the addition of oxaliplatin to the treatment schemes.

Conclusions
In conclusion, we found that adjuvant chemotherapy is beneficial for stage II and III CC patients with either MSS or MSI-high tumors. Although the result for stage II MSI-high patients was not statistically significant, the small number and even lack of events in these patients suggested a potential benefit from receiving adjuvant treatment. Results from larger studies and new clinical trials will contribute to elucidate the usefulness of MSI as a predictive marker in earlystage patients.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Fig. S1. Selection of patients for inclusion in analysis. Data S1. Search terms for meta-analysis. Fig. S2. Balance diagnostics for overall population after adjustment for propensity score weighting. Table S1. Characteristics of studies included in metaanalysis.