Predicting disease-free survival following curative-intent resection of right-sided colon cancer using a pre- and post-operative nomogram: a prospective observational cohort study

James Lucocq, MBChB,PG, MRCS,RAMC\*,Tom Trinder, MBChB, BSc (Hons), Kate Homyer, MBChB, BSc (Hons), Hassan Baig, MBChB, MSc, MRCS, DOHNS, PradeepPatil,MBBS,MD,FRCS,Girivasan Muthukumarasamy,MBBS,FRCS

Introductio:Diseaseprognsticationcaneachived throughthederivationfbiolgicalyandclinicallyintgratedpredition modelsTh preent studyreports 1, 3, and 5-eardisease-free suival Dipatients udergoing rght hemiclctomf curative intet and bthderives and valiatsapre-andpst-peratie predictiontooforDFS forprognsticationand rsk stratificationpurposes.

Method:Cneutivpatntsergngrg-siddcuratiintentetionforclretalcance2interiacare were folweduppropectielyfreurrenceand sivalutcmuival analysesweeed tderive prandpost-oerati models predicting 1, 3, and 5-yearDF.Calbationwasreported and inteal valdation was perfomed using botstrapp. Results: A total of 822 patients underwent resection and 528 had 25 -year follow-up. The 1-, 3-, and 5-year DFS rates were 85.6% 72.5% and 57.6% , respectively. Variables associated with death/recurrence included: increasing age (H\mathsf{R}>1.95 P=0.037 ), male gender (HR 1.62, P{<}0.001 ,ASA ^{23} (HR 1.79, P{<}0.001 ), low albumin (HR 1.54, P{<}0.001 ),T4 stage (HR 2.35, P=0.023 , R1 status (HR 1.63, P=0.024 。， {>=}4 positive lymph nodes (\mathsf{H R}>1.74 ， P<0.001 ) and Clavien-Dindo ^{23} (HR 2.83, P{<}0.001 .The preand post-operativemodels contained9and 13demographic,linical, biochemical, operative andpatholgical variablrtilCnxad,rtilldindgrapilicaladratiarialsiniat reduced the C-index of the pre-(0.62) and post-operative models (0.70).

ConclusionThpreentdpredictiontolsfrwillhelclnicansstratiyrsoffearopriatadjuvant trtmt ad predict long-termDFS following curative-intent right-sided colon cancer resection.

Keywords: colon cancer,prediction,recurrence,right hemicolectomy,survival

Introduction

HIGHLIGHTS

Colorectal cancer is the second most common cause of cancer death in the United Kingdom and resection remains the mainstay of treatmentl12]. Disease-free survival (DFS) is increasingly being recognized as avaluableoutcomemetricincolorectal cancer and is commonly being reported in colorectal clinical trials.Five-year DFS rates approach 80% for elective right hemicolectomyl3-9]1

·Demographic, clinical, biochemical, operative and pathological factors influence the likelihood of diseasefree survival.
· The pre-and post-operative prediction tools estimate 1-, 3-, and 5-year disease-free survival with good accuracy following curative-intent right hemicolectomy.
The models have a role in risk stratification, offering appropriate adjuvant treatment and predicting long-term survival.

Colorectal adenocarcinomas are currently staged and managed based on the TNM (Tumor, Nodes, Metastasis) classification of solid tumorsinadults,developedby theUnion for InternationalCancer Control(UICC）andused bythe American Joint Committee on Cancer (AJCC). TNM staging is used to plan management and predict outcomes such as recurrence by grouping patients based on their stage and applying outcomes fromclinicaltrials specifictothe typeandsite of cancer.Predictingoutcomesbasedonmedianvaluesderived from clinical trials is not patient centric or individualized; patients with the same TNM stage cancer can have varying survival 0,1l The TNM staging relies only on the tumor stage and does not take other patient characteristics into consideration such as asymptomatic patients (screen-detected), age, comorbidities and both operative and perioperative outcomes, which may have a bearing on long-term DFS. In other words, individualized outcome prediction is lacking for cancer patients. This information,if availablewillbetterinformbothcliniciansandpatients while enabling shared decision making regarding appropriate modalities of care. There is currently no known method of predicting cancer recurrence or survival specific to right colon cancers[12-16]

The aim of this study is to develop preoperative and postoperative prediction models for long term (5 years) survival for patients with right sided colon cancer. Patient-specific preoperative prediction models will help shared decision-making regarding treatment modalities. Postoperative model adding perioperative complications and tumor pathological characteristics will estimate risk of recurrence which can guide follow up intensity and need for adjuvant treatments.The models are intended to be simple, user friendly and applicable to all general hospitals throughout the world whilst demonstrating strong statistical ability to predict outcomes in individual patients.

Methods

Study design

We undertook an observational, cohort study of patients undergoing right-sided hemicolectomies for colorectal cancer from January 2010 to December 2020 in a tertiary care unit, reported according to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE)17 Data usage approval was granted by our institutional review board and the study was conducted in accordance with the Declaration of Helsinki. The manuscript has been reported in line with the STROCSS criterial18]

All patients aged over 18 years undergoing right hemicolectomy for colorectal cancer were identified from institutional electronic records. The following patients were excluded: procedures not performed for curative intent (n=92 ),patients who died in the immediate post-operative period( _{<30} days, n=24 ) those with metastasis (n=42) ,those without adenocarcinoma on pathology (n=29 and distal transverse tumors (n=70 0

Datavariables

Background patient characteristics were documented such as demographics,comorbidities, American Society of Anesthesiologists (ASA) score, metabolic equivalent (MET) score, smoking history, body mass index (BMI), colon cancer symptoms, presentation (e.g.emergency,elective,or screen-detected),and referral pathway.The type of right-sided colonic resection (e.g.total mesentery excision,extended)was performed at the discretion of the surgeon and based on the clinical suspicion and local invasion of the tumor. The surgical approach (laparoscopic versus open), the need for conversion, the urgency of the procedure and the needfora stoma were documented.

Histopathological data were extracted from postoperative pathology reports and malignant tumors were included. The histological type of cancer (e.g. mucinous adenocarcinoma) was noted, as well as the differentiation, lymphovascular invasion,T and N stage, number of lymph nodes in the specimen and the number of positive lymph nodes. The resection“R1” status wasdetermined on a cut-off distancefrom the tumor to the resection margin of <1mm .Contrast-enhanced computed tomography (CT)of the chest, abdomen and pelvis were used to determine the stage of each patient. The use of adjuvant chemotherapy was reported.

Outcomes

The diagnosis of recurrence was made by radiological or histological diagnosis and MDT review.The site of recurrence was classified as either loco-regional or systemic. The length of follow-up for each patient was recorded and the date or recurrence or death,if applicable,were recorded.The time from the date of surgery to the date of the first event (recurrence or death), was recorded. Data was censored for patients without recurrence and still alive at the end of follow-up. The rate of recurrence and DFS were reported using Kaplan-Meier (KM) curves.

KMcurveswereusedtodeterminevariablesassociatedwith DFS using the log-rank test accounting for the length of followup.In the log-rank analysis, patient age was categorized by decade, BMI was divided by 2\hat{5}~kg/m^{2} ,ASA was dichotomized by a cut-off of 1-2 versus \scriptstyle>=3 and METs score was divided by 1-4 versus >=5. LowAlbuminwasdefined as {<}35\ g/dL and a lowHb was {<}120\ g/dL for females and {<}130g/dL for males. The number of lymph nodes in the specimen were reported and the number of positive lymph nodes were grouped as follows: 0; 1-3; 4-5; 6- 10;{>}10 .For all statistical tests,patients with missing data were excluded from analyses.Cox-Proportional Hazards Models (CPHM) were used as a multivariate approach to determine variables associated with DFS.Factors considered forinclusion in the CPHM were determined from previous studies and by a P<0.1 demonstrated from the logrank analysis.

Long-term(5 year) disease-free survival

To demonstrate the association of variables with long-term DFS (5-year) a subgroup analysis was performed. Of those with 5-year follow-up, variables associated with 5-year DFS were determined. Baseline categorical data were presented and analyzed using theindependent samples X^{2} test. Continuous data were assessed for normality and managed accordingly using a one-way ANOVA or Kruskal-Wallis test as appropriate.

Developingpredictivemodelfor1-,3-,and5-year disease-freesurvival

We adhered to the AJCC criteria for developing risk models during the development of our risk model81. Variables were included in the initial CPHM model based on significant variables (P<0.1) in the log-rank analysis. Two separate models were created: a model derived from all variables accessible pre-operatively and a further post-operative model derived from variables available once the final specimen pathology was available.

Resulting hazard ratios (HR's) and P valuesfor theCPHM were reported.Akaike's information criterion(AIC)was minimized to select the final prediction modelusing a backward elimination process.Variance inflation factor (VIF）was used to assess for multicollinearity between variables and a value greater than 10 indicated significant collinearity. The concordance index (C-index)was used to determine discriminative ability for the training and test cohort, using the “rms” package in R studio 2022.02.01. The model was trained using 1000 bootstrapped samples and then correction for model optimism was performed byinternal validation on out-of-sample test data. Calibration was assessed using calibration plots on bootstrap resampling data to simulate an out-of-sample validation.The models were displayed as nomograms.

Sensitivity analysiswas assessed byplottingreceiver operator characteristic curves and assessing the area under the curve. To demonstrate the impact of adding demographic,clinical and operative variables into the predictive model in addition to the pathological variables, the C-index was reported for models with and without the addition of demographic, clinical and operative variables.

Results

A total of 822 patients (median age,73;IQR,58-82）with malignant right-sided colon cancer undergoing resection with curative intent were included (Table 1). Figure 1 demonstrates the risk of recurrence (Fig. 1A) and DFS (Fig. 1B) over time after a median follow-up of 5 years and 7 months. The 1-, 3-, and 5-year recurrence rates were 7.2% ， 14.5% and 17.6% ,respectively (Table 2).The 1-, 3-,and 5-year DFS rates were 85.6% 72.5% and 57.6% ,respectively.

As demonstrated in the KM analysis, a multitude of variables were associated with recurrence and/or death (Table 1).Figure 1 displays KM analyses for DFS based on T stage (Fig. 1C), number of positive lymph nodes (Fig. 1D),operative approach (Fig. 1E), and urgency of procedure (Fig. 1F).

In CPHM, a number of variables were positively associated with recurrence/death:increasing age {HR}>1.95 ， P=0.037; ， male gender (HR 1.62, it{P}<0.001 ，ASA \scriptstyle>=3 (HR 1.79, P<0.001) , low albumin (HR 1.54, P<0.001 ),T4 stage (HR 2.35, P=0.023 ),R1 status (HR 1.63, P=0.024 0,4-5 positive lymph nodes (HR 1.74, P<0.001 ),6-10 positive lymph nodes (HR 3.65, \begin{array}{r l r}{P}&{{}<}&{0.001}\end{array} ）andClavien-Dindo \scriptstyle>=3 (HR 2.83, P<0.001 0.The following variables were negatively associated with recurrence and/or death:20-29 lymph nodes in the specimen (HR 0.54, P=0.032 and 230 lymph nodes in the specimen (HR 0.44, P=0.012 .No variables suffered from significant multicollinearity (VIF>10 0

Long-term(5year)disease-freesurvival

The subgroup of patients \begin{array}{r}{(n=528)}\end{array} who were followedupfor at least 5-years were split into two groups: those with (n=304 and without \begin{array}{r}{(n=224).}\end{array} 5-yearDFS.The clinicopathological characteristics were compared between the two groups and are reported in Supplementary Digital Content (Table 1, http://links.lww.com/ JS9/D966).Patients who survived 5 years without recurrence were younger, less likely to be male,had lower rates of diabetes mellitus (DM), chronic kidney disease (CKD) stage >=3 ,ischaemic heart disease (IHD), higher albumin levels and were more likely to be identified during screening but less likely to have undergone an emergency hemicolectomy (Supplementary Digital Content, Table 1,http://links.lww.com/JS9/D966).Further significant differences between the two groups include the 5-year DFS group having lower numbers of positive lymph nodes,greater number of resected lymph nodes and lower overall T and N stages.

Developing apre-operativeandpost-operativemodelfor1- and5-yeardisease-freesurvival

Based on the AIC,9 pre-operative variables were selected for the final pre-operative prediction model and were used to predict 1-, 3-, and 5-year DFS (Fig. 2). Using the model for an individual patient, the probability of DFS can be calculated by summing together the points derived from each variable. The C-index for the pre-operative model was 0.75 and once adjusted for internal validation,the C-index was 0.74 with a slope of 0.84.The calibration plots for 1-, 3-, and 5-year DFS outcomes are reported (Supplementary Digital Content, Figure 1,http:// links.lww.com/JS9/D899). Once excluding demographic, clinical and operative variables and only including the pre-operative pathological variables, the C-index for the internal validation was 0.62.

Based on the AIC,13 variables were selected for the final optimal post-operative prediction model and were used to predict 1-,3-,and 5-year DFS (Fig.3).The C-index for the model is 0.79 and once adjusted for internal validation, the C-index was 0.77.The slope of the model was 0.86. The calibration plots for 1-,3-,and 5-year DFS outcomes are reported for the postoperative model (Supplementary Digital Content, Figure 2, http://links.lww.com/JS9/D900). Once excluding demographic, clinical and operative variables and only including the postoperative pathological variables,the C-index for the internal validation model was 0.70.

Discussion

The present study reports oncological and survival outcomes for patients undergoing a right-sided colonic resection for malignancy with curative intent and identifies variables that are associated with long-term (5-year) DFs.We derive and validate preand post-operative prediction tools for 1-, 3-,and 5-year DFS that accounts for demographic,clinical,biochemical,operative and pathological factors. Incorporating all variables boosts the predictive ability for DFS compared to pathological factors alone.

The present prediction tool determines the factors that independently predict outcome, many of which are well known to be linked to recurrence or survival.For example, the number of lymph nodes resected has been demonstrated as a prognostic factor in right hmicolectmyl20 as areT stage,N stageand the presence of lymphovascular invasion[19]. Less well understood is how such factors culminate in an overall risk of recurrence or death. By aggregating these factors, we gain an appreciation of the numerous aspects that influence prognosis and a more accurate estimation of the long-term outcome. Whilst the post-operative model has the highest utility once the formal specimen pathology is reported, both the pre-operative and post-operative models demonstrate high predictive ability. The post-operative model has a role in counseling patients on prognosis and long-term follow-up. Utilizing the pre-operative model has a role in patient counseling on the peri-operative risk and can help clinicians decide between surgical candidates, particularly in high-risk patients[21].

Todate,the derivation ofpredictive modelsfor survival outcomes following the resection of right-sided colon cancer for curativeintenthasnot been conducted.Increasing evidence supports the disparate patient characteristics and outcomes of right and left colectomy, more specifically, the worse prognosis, higher rate of post-operative complications and longer length of hospital stay following right hemicolectomy as well as the differing response to adjuvant chemotherapyl12-16]. There are significant genetic differences between right and left sided cancers, such as an association of microsatellite instability (MSI) with right sided cancers, compared to the chromosomal instability pathway mutations on the leftl2231. The above is evidence that predictive models for right and left-sided cancers should be derived and implemented independently.

Current predictive models in patients undergoing colectomy have aimed to predict recurrence or cancer-specific survival following colectomyl242sAlthoughpreviousmodels incorporate pathological variables (e.g. T stage, number of positive lymph nodes,adjuvant chemotherapy） similar to the present study, there are fundamental differences.37241.The MSKCC clinical calculator aims to predict recurrence but censors for patients who have died without recurrencel24]. As a model orientated towards pathologicalfactors thatmaybe associated with recurrence it does not account for survival or acknowledge clinical or operative factors. Furthermore, a pre-operative prediction of outcome is not provided. Zheng et al have created a prediction tool to predict cancer-specific survival25]. This model does not predict recurrence or mortality that is not directly attributed to cancer. After accounting for a multitude of aspects such as demographic factors, significant comorbidities,operative factors,biochemical factors and pathological factors the present study uses DFS as an overall outcome metric that incorporates both recurrence and survival whilst providing pre- and post-operative estimates.

With the emergence of molecular phenotypes and biomarkers future studies should consider how these could be incorporated into models to help improve prognostication further, particularly in conjunction with targeted therapiesl26]. The presence of MSI and both KRAS and BRAF genes help inform prognosis but also response to adjuvant treatment.For example,MSI Stage II CRC patients do not appear to benefit from 5-FU chemotherapy, but may respond to checkpoint inhibitorsl27-291. Circulating tumor DNA may also offer prognostic information pre- and post-operatively and may be monitored in response to treatmentl30,31l. Of course, there are a number of barriers to overcome before universal use by all healthcare providers can be achieved, such as lack of technical standardization betweenlaboratories,low numbers of validating studies and inadequately-sized patient cohorts.

There were a number of variables which were not explored in the present study.Pre-operative CEA has been shown to be predictive of recurrence in univariate analysis and could hold predictive value. Post-operative CEA reportedly offers even more value in predicting recurrence and could help improve the prediction of DFS during the post-operative periodl32l. Other factorswhichwerenot considered includeperineuralinvasion and tumorinfiltratinglymphocytes.Althoughperineuralinvasion was not considered, it is unlikely to have influenced the results significantly due to its high correlation with lymphovascular invasion.Whilst molecular phenotypes and biomarkers were not used in the present model, the included variables should be available to the vast majority of pathology and surgical departments which improves the generalizability of the models133,34]

Figure 2.Preoperative nomogram to predict likelihood of 1-,3-,and 5-year DFS.

Figure 3.Postoperative nomogram to predict likelihood of 1-,3-,and 5-yearDFS.

Therewere a number ofconsiderations when constructing the present modelsl35]. Including too many variables can result in overfitting of the model to our data cohort.Secondly,with numerous variables particularly in the final model, it was important that we assessed formulticollinearity to ensure appropriate model specification.Multicollinearity could falsely declare a variable as significantly associated with DFS due to confounding. As demonstrated by the slopes of the models and the VIF, both overfitting and multicollinearity were not significant concerns.Finally, we opted to use bootstrapping for internal validationin preferenceto an alternativecohortwhichmaximised the number of patients that could be included in the model.

There are limitations to the current study.Firstly, the present studyisasinglecenterthatrelies oninternalvalidation.External validation is required to ensure that the model functions accurately in different populations and to ensure that the model does not suffer from overfitting.Furthermore, although our center had a well-defined post-operative surveillance pathway, it is possible that patients had recurrence that was not identified,out with the surveillance period. Lastly, the type of resection (e.g. complete mesocolic excision) was left to the discretion of the surgeon and accounting for such details may have improved the predictive accuracy further.

In conclusion, the present study identifies variables that are associated with long-term（ 25 year)DFS in patients whohave underwent a resection for a right sided colon cancer with curative intent.The pre-and post-operative prediction tools successfully estimate the 1-, 3-,and 5-year DFS using demographic, clinical, biochemical, operative and pathological factors.

Ethical approval

Data was collected as a regional audit and included anonymised data.A reference number is not routinely provided.

Consent

As per our regional ethical committee, no patient consent was required.

Sources of funding

The authors received no financial support for this research and/ or authorship of this article

Authorcontributions

J.L.: study concept, design, analysis, writing, interpretation; T.T.: writing draft; P.P.: supervision, writing; G.M.: data collection, study concept.

Conflicts of interest disclosure

The authors declare that they have no competing interests.

Researchregistration uniqueidentifying number (UIN)

Research registry (researchregistry10432).

Guarantor

James Lucocq, Girivasan Muthukumarasamy.

Provenance and peer review

Not commissioned,externally peer-reviewed.

Dataavailabilitystatement

Data will be made available upon reasonable request

References

[1] Bertelsen CA, Neuenschwander AU, Jansen JE,et al. Danish Colorectal Cancer Group.Disease-free survival after complete mesocolic excision compared with conventional colon cancer surgery: a retrospective, population-based study. Lancet Oncol 2015;16:161-68.
[2]MitryE,BouvierAM,EsteveJFaivreJ.mprovement ncolorectal cancer survival: a population-based study. Eur J Cancer 2005;41:2297-303.
[3] Yin J, Salem ME,Dixon JG, et al. Reevaluating disease-free survival as an endpoint vs overall survival in stage III adjuvant colon cancer trials. J Nat Cancer Inst 2022;114:60-67.
[4] Morris JS. Disease-free survival is a promising surrogate for overall survival in colorectal cancer studies. J Natl Cancer Inst 2022;114:5-6.
[5]Auer RC,Ott M,Karanicolas P,et al.PERIOP-01 investigators.efficacy and safety of extended duration toperioperative thromboprophylaxis with low molecular weight heparin on disease-free survival after surgical resection of colorectal cancer (PERIOP-01): multicentre, open label, randomised controlled trial. Bmj 2022;378.e071375.
[6] Yu L, Liu Z, Chen Z, et al. Pathways of lymph node metastasis and prognosis after right hemicolectomy for cecal cancer: results from a retrospective single center.World J Surg Onc 2023;21:281.
[7] Gasser E, Braunwarth E, Riedmann M, et al. Primary tumour location affects survival after resection of colorectal liver metastases:a two-institutionalcohort study withinternationalvalidation,systematic meta-analysis and a clinical risk score. PLoS One 2019;14.e0217411.
[8] Sargent DJ, Wieand HS, Haller DG,et al. Disease-free survival versus overall survival as a primary end point for adjuvant colon cancer studies: individual patient data from 20,898 patients on 18 randomized trials. J Clin Oncol 2005;23.8664-70.
[9] Birgisson H, Wallin U, Holmberg L, et al. Survival endpoints in colorectal cancer and the effect of second primary other cancer on disease free survival. BMC Cancer 2011;11:438.
10] Gold JS, Gonen M, Gutierrez A,et al.Development and validation of a prognostic nomogram for recurrence-free survival after complete surgical resection of localised primary gastrointestinal stromal tumour: a retrospective analysis.Lancet Oncol 2009;10:1045-52.
11] Balachandran VP, Gonen M, Smith JJ, DeMatteo RP. Nomograms in oncology:more than meets the eye.Lancet Oncol 2015;16:e173-80.
12] Yang CY, Yen MH, Kiu KT, Chen YT, Chang TC. Outcomes of right-sided and left-sided colon cancer after curative resection.Sci Rep 2022;12:11323.
13] Grass F,Lovely JK,CrippaJ,et al.Comparison of recovery and outcome after left and right colectomy.ColorectalDis 2019;21.481-86.
14] Masoomi H, Buchberg B, Dang P, et al. Outcomes of right vs. left colectomy for colon cancer.J Gastrointest Surg 2011;15:2023-28.
[15] Miron Fernandez I, Mera Velasco S, Turino Luque JD, Gonzalez Poveda I, Ruiz Lopez M, Santoyo Santoyo J. Right and left colorectal cancer: differences in post-surgical-care outcomes and survival in elderly patients. Cancers (Basel) 2021;13:2647.
[16] Campana JP, Pellgrini PA,Rossi GL,Ojea Quintana G, Mentz RE, Vaccaro CA.Right versus left laparoscopic colectomy for colon cancer: does side make any difference? Int J Colorectal Dis 2017;32: 907-12.
[17] Vandenbroucke JP, von Elm E, Altman DG, et al. STROBE Initiative. Strengthening theReportingofObservationalStudiesinEpidemiology (STROBE): explanation and elaboration.PLoS Med 2007;4.e297.
[18] Mathew G,Agha R,for the STROCSS Group. STROCSS 2021: strengthening the Reporting of cohort, cross-sectional and case-control studies in Surgery. Int J Surg 2021;96:106165.
[19] Benz SR,Feder IS,Vollmer S,et al.Complete mesocolic excision for right colonic cancer: prospective multicentre study. Br J Surg 2022;110.98-105.
[20] De Lange G, Davies J,Toso C, Meurete G, Ris F, Meyer J. Complete mesocolic excision for right hemicolectomy: an updated systematic review and meta-analysis.Tech Coloproctol 2023;27:979-93.
[21] Kobayashi H, Miyata H, Gotoh M, et al. Risk model for right hemicolectomy based on 19,070 Japanese patients in the National Clinical Database.J Gastroenterol 2014;49.1047-55.
[22] Arakawa K, Hata K, Kawai K, et al. Predictors for high microsatellite instability in patients with colorectal cancer fulfilling the revised Bethesda guidelines. Anticancer Res 2018;38.4871-76.
[23] Baran B, Mert Ozupek N, Yerli Tetik N, Acar E, Bekcioglu O, Baskin Y. Difference between left-sided and right-sided colorectal cancer: a focused review of literature. Gastroenterology Res 2018;11: 264-73.
[24] Weiser MR, Hsu M, Bauer PS, et al. Clinical calculator based on molecular and clinicopathologic characteristics predictsrecurrence following resection of stage I-III colon cancer. J Clin Oncol 2021;39: 911-19.
[25] Zheng P, Lai C, Yang W, Guo J, Xiao S, Chen Z.Nomogram predicting cancer-specific survival in elderly patients with stages I-ll colon cancer. Scand J Gastroenterol 2020;55:202-08.
[26] Koncina E, Haan S, Rauh S, Letellier E.Prognostic and predictive molecular biomarkers for colorectal cancer: updates and challenges. Cancers 2020;12:319.
[27] Dj S, 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;28:3219-26.
[28] Basile D, Garattini SK, Bonotto M, et al. Immunotherapy for colorectal cancer: where are we heading?. Expert Opin Biol 2017;17:709-72.
[29] Bilgin B, Sendur MAN, Bilent Akinci M. Targeting the PD-1 pathway: a new hope for gastrointestinal cancers. Curr Med Res Opin 2017;33:749-59.
[30] Mas L, Bachet JB, Taly V, et al. BRAF mutation status in Circulating tumorDNAfrom patients withmetastatic colorectal cancer:extended mutation analysis from the AGEO RASANC study. Cancers Basel 2019;11:998.
[31] Reece M,Saluja H, HollingtonP,et al.The Use of Circulating Tumor DNA to Monitor and Predict Response to Treatment in Colorectal Cancer.Front Genet 2019;10:1118.
[32] Konishi T, Shimada Y, Hsu M, et al. Association of preoperative and postoperative serum carcinoembryonic antigen and colon cancer outcome. JAMA Oncol 2018;4:309-15.
[33] Reinert T, Henriksen TV, Christensen E,et al. Analysis of plasma cell-free DNA by ultradeep sequencing in patients with stages I to I colorectal cancer. JAMA Oncol 2019;5:1124-31.
[34] Kotani D, Oki E, Nakamura Y, et al. Molecular residual disease and efficacy of adjuvant chemotherapy in patients with colorectal cancer. Nat Med 2023;29:127-34.
[35]Steyerberg EW,VergouweY.Towards better clinical prediction models: seven steps for development and an ABCD for validation.Eur Heart J 2014;35:1925-31.