Clinical Characteristics And Development Of Prognostic Models For Predicting Tumor-related Death In Second Primary Colorectal Cancer Patients And Its Clinical

PurposeThis retrospective study was designed to explore the clinical characteristics of the second primary colorectal cancer(SPCRC).A competing-risk prognostic model was developed to identify prognostic risk factors,followed by the assessment of its clinical application value.MethodsThe data were extracted from Surveillance,Epidemiology,and End Results(SEER)database to include SPCRC patients between 1988 and 2013.Standardized incidence ratios(SIR)and annual percentage change(APC)were used to describe relative risks and incidence trend respectively.Prognostic analysis was carried out using competing-risk methodology,in which death from other cause was regarded as a competitive event,and the cumulative incidence function(CIF)of tumor-related death and non-tumor-related death were calculated in univariate analysis.Covariates with statistical differences were selected as candidate predictors and subsequently incorporated into multivariate analysis.Least absolute shrinkage and selection operator(Lasso)regression was used to filter covariates with statistical differences for the construction of the competing-risk nomogram.The nomogram was implemented to predict the probability of tumor-related death at 1-year,3-year and 5-year.Finally,the performance and clinical application of the model were evaluated internally and externally using the Receiver Operator Characteristic Curve(ROC curve)and the calibration plot.ResultsDataset of 4381 patients who met the inclusion criteria were eventually selected for analysis.The median follow-up was 96 months.A total of 2631 deaths were studied,including 1641 tumor-related deaths and 990 non-tumor-related deaths.The 5-year CIF of tumor-related death and non-tumor-related death were 38.692%(95%CI 38.686%-38.700%)and 13.650%(95%CI 13.648%-13.653%),respectively.Incidence of second primary colorectal cancer gradually increased.Interval time<60 months,the left semicolon tumor,female patients were at increasingly high risk of developing second primary colorectal cancer.Tumor site and TNM stage of initial colorectal cancer(ICRC),time interval between the initial and secondary primary CRC,sex,race,marital status,tumor site,grade,TNM stage,surgical status,CEA level of SPCRC were independent predictors of tumor-related deaths(P<0.05).In the end,TNM stage of ICRC,time interval between the initial and secondary primary CRC,sex,tumor site,grade,TNM stage,surgical status,and CEA level were included in the prognostic model after Lasso regression and presented in form of nomogram.The nomogram was shown to have good discrimination after internal validation with area under curve(AUC)of 0.792(95%CI 0.782-0.810),and calibration curve performed well.The external validation showed good discrimination,with an AUC of 0.787(95%CI 0.712-0.801)and satisfactory calibration curve.Both the internal and external validation calibration indicated proper discrimination and good concordance between the predicted and actual outcomes.ConclusionThe incidence of SPCRC gradually increased,specifically those who interval time<60 months,the left semicolon tumor,female patients had increasingly high risk of developing second primary colorectal cancer.In addition,patients with SPCRC had worse prognosis than those with unique colorectal cancer.Using the large sample database and competing-risk analysis,a tumor-related death prediction model was developed.The internal and external validation both showed the good performance of the model and evaluated the potential clinical application value by personalized predicting the tumor-related death of SPCRC.