| Background:Patients with primary cancer of non-pancreatic origin are more likely to develop second primary pancreatic cancer(spPC).However,the survival and therapy to spPC have not been fully elucidated to date Methods:Characteristics,prognostic factors,and treatment of spPC were explored in this retrospective cohort study.All the data were collected from registered pancreatic cancer patients in the National Cancer Institute's SEER from Jan.1974 to Dec.2015.To understand the changes in incidence over time,we used the older SEER9 database during 1974-2015 for age-standardized incidence every 100,000 person-years.Cumulative number of cancer survivors was calculated by 20-year limited-duration prevalence,which was the number of surviving patients who had been diagnosed as the cancer in the last two decades at a specific time.Therefore,the number of cancer survivors before 1994 was uncalculated.JoinPoint Regression Analysis Software 4.6.0.0(NCI,Bethesda,Maryland)was used for determining the changing moment of the trend of incidence.6 An algorithm was used to mark the optimal counts of inflection points in which incidence altered in direction and magnitude.The number of joinpoints was controlled to be less than 5.Log-transformed curves were marked as annual percentage change(APC),representing the trends in malignancy rates over time.To explore independent factors in survival,SEER18 was used to recalculate the patients diagnosised during 1974-2013.Incomplete data were not excluded from the study.Both the log-rank test and the Kaplan-Meier method were used in univariate survival analysis;while Cox hazard regression modelling was applied in multivariate survival analysis with its significance examined in a likelihood ratio test and hazard ratio in propensity score-matching analysis Hazard ratios(HRs)along with 95%confidence intervals(CIs)were calculated.An interaction term was added to the model to determine the correlations.SPSS 25.0.0(IBM,Armonk,New York)and R 2.13.2(r-project.org)were used with P<0.05 indicating significant difference.A nomogram was built to forecast the 3-year overall survival(OS)rates on the basis of the predictive model with the known prognostic factors.A C-index ranging from 0.5 to 1 was used to test discrimination.A higher value in C-index meant a better prognostic model.Overfit bias was decreased by resampling 500 of them in Bootstrap.Calibration was conducted by comparing the mean actual survival rate with the mean predicted survival rate in a Kaplan-Meier analysis after arranging by tens the nomogram-predicted survival rate.Also,the receiver operating characteristic(ROC)curve AUC was used to test the accuracy of the 3-year survival forecasts.Results:The proportion in the annual incidence of spPC among pancreatic cancer(PC)as a whole raised 6.97-fold from 1974(2.13 percent)to 2015(14.85 percent),and the incidence per million of spPC from cancer survivors increased by 0.9%(95%Cl:0.8-1.0)(from 48.37 in 1997 to 61.75 in 2015).spPC's incidence will further rise according to the APC-based predictive model.By 2025,its proportion in the annual incidence among PC as a whole was projected to increase to 22.99%and its incidence was estimated to rise to 67.54 per million cancer survivors.Pancreatic cancer patients without previous cancer were younger than spPC patients(mean age,67.9 years-old versus 69.7 years-old;P<0.001).There were no significant differences between their tumor grade,surgery utilization and radiotherapy.Disease stage differed based on tumor's sequence,which can be seen through the lower proportion of metastatic tumors(48.63%vs.52.33%;P<0.001)and higher percentage of localized patients(11.57%vs.8.80%;P<0.001)in spPC patients.For women,pancreatic cancer is mainly secondary to breast(N=2108,39.3%),tuteri(N=1777,33.1%)and digestive system cancer(N=1019,19.0%).While the first cancer in men is mainly prostate(N=3311,48.3%),digestive system(N=1150,16.7%)and bladder cancer(N=546,7.9%).SpPC patients presented similar OS as 1st PC patients(6 months vs.6 months in median survival time).Pancreatic cancer replaced other diseases as the dominant determinant of survival after the diagnosis of spPC(the half-,1-,3-and 5-year cumulative cancer-specific mortality was 43.05%,59.37%,74.88%and 77.44%,respectively).A multivariate Cox proportional hazards model and competing risk analysis were performed to define the factors independently influencing survival.The coefficients from the two analysis models fitted very well for most factors.Surgery was the strongest predictors of determining survival(unperformed versus performed:HR,2.71,P<0.001;sHR,1.92,P<0.001).Not surprisingly,advanced stage(regional versus local:HR,1,77,P<0.001;sHR,1.74;distant versus local:HR,2.27,P<0.001;sHR,1.93,P<0.001),higher age at diagnosis(?80 versus<60:HR,1.88,P<0.001;sHR,1.93,P<0.001),and higher grade were correlated with quicker death(G3G4 versus G1G2:HR,1.53,P<0.001;sHR,1.27,P<0.001).Compared with metachronous cancer,patients with simultaneous pancreatic cancer tended to live longer(interval>6 versus interval<6:HR,1.33,P<0.001;sHR,1.45,P<0.001).Ethnicity and radiotherapy were not proved to be independent predictors of survival in both two analysis models.Patients who refused chemotherapy(HR,1.32;95%CI:1.20-145;P<0.001)and male patients(HR,1.16;95%Cl:1.03-1.30;P<0.001)had a disadvantage in overall survival.Although chemotherapy(sHR,1.07;P=.255)and gender(sHR,1.11;P=.142)did not meet statistical significance in competing risk analysis,similar trends in survival risk were seen when compared with multivariate analysis.When it comes to the previous cancer,prostate cancer had a better prognosis than other types of cancer(HR,0.82;95%CI:0.72-0.95;P=0.01)in overall survival.It's worth noting that differentiation,disease stage and surgery of the first tumor did not influence overall survival.Treatments for 1st PC and spPC victims were assessed with a multivariate analysis on OS,which was further stratified by the differentiation and staging of the neoplasm.Its sequence and the advantages of treatments were also evaluated.In local-staged PC subjects,no significant difference between sequence and therapy benefits was found in their interactions.The only clear optimal treatment was surgical resection.Radiotherapy could not prolong the life of local PC.Chemotherapy was correlated with survival advantage in local G3/G4 PC patients:although not statistically significant for spPC patients,interaction analysis demonstrated that chemotherapy had no difference in the efficacy between the two groups When looking only at PC with regional disease,performing surgery was associated with longer survival time in all subgroups and the curative effect remained in line between sequences Radiotherapy was not regarded as an independent prognostic factor,except in G3/G4 1st PC group(P for interaction=0.041).Chemotherapy was found to be correlated with OS in all grade,sequence combination groups and the curative profit was more significant in G1/G2 in spPCs(HR,1.36 vs.1.11,P for interaction=0.019).In patients with G1/G2 stage ?,implementing chemotherapy provided a better outcome for both spPC and 1st PC.In patients with G3/G4 stage IV,implementing chemotherapy and radiotherapy provided a better outcome for both spPC and 1st PC,with the therapeutic result of radiotherapy more effective for spPC(HR,1.76 vs.1.14,P for interaction=0.050).Although no decreased survival was seen in spPC in contrast to 1st PC,an inferior OS was found on the regional stage stratification(24 months vs.46 months in median survival time,P<0.001).When classified by sequence and tumor stage,spPC patients had lower odds of receiving surgery in local stage classification(35.40%vs.43.30%;P<0.001).The survival curves between sequence of resected local PC patients(107 months vs.75 months in median survival time,P=0.118)and unresected(8 months vs.7 months in median survival time,P=0.552)overlapped.In other words,if resection rates are consistent,the survival rates of the spPC and 1st PC will be almost the same.Therefore,we could deduce that the decrease in survival in local spPC can be explained by the underutilization of surgery.We further studied the features and prognostic factors of spPC secondary to metastatic initial cancer.Cumulative incidence function curve revealed that in patients with metastatic first cancer,the rates of cumulative incidence competing risk of dying of pancreatic cancer at 20-month and 5-year follow-up were 37.48%and 42.25%.Multivariate analysis revealed surgical therapy of spPC retained strong statistical significance(unreceived versus received:HR,2.2 7;95%Cl:1.40-3.69;P<0.001).Surgical therapy of the first cancer was not an independent prognostic factor(HR,0.78;95%CI:0.56-1.09;P=0.160).The nomogram based on the Cox PHs model was established to foretell the 3-year OS rates The C-index was 0.708(SD=0.011)in the study population,embodying moderate discrimination ability of established nomogram.The AUC in the nomogram forecasting the 3-year OS rates was 0.853,exhibiting excellent specificity and sensitivity.The predicted 3-year OS rates were within 10%margin of error(dotted)of the actual survival rateConclusions:The incidence of spPC is increasing.History of previous cancer did not reduce the efficacy of PC.As the dominant survival factor of spPC patients,not enough surgery was performed in locally-staged patients in contrast to 1st PCs.Curative surgery,when possible,and chemotherapy for spPC be conducted similarly to that for 1st PC.Increasing screening rates and surgical utilization are ways to improve survival of spPC patients. |
PDF Full Text Request