Research On Prognostic Index PRIMA-PI Combined With Ki67 To Predict Survival Of Patients With Follicular Lymphoma

Objective:Follicular lymphoma(FL)is one of the common subtypes of non-Hodgkin’s lymphoma(NHL).The disease progresses slowly and has a long median survival but is difficult to treat and prone to recurrent relapse.Stratification of FL and screening of patients at high risk for FL for early intervention are the focus of therapeutic trials for follicular lymphoma.To this end,various prognostic risk models regarding FL have been proposed in recent years,among which the simplified PRIMA-PI risk index can even be used as a basis for more complex and comprehensive biomolecular prognostic models.Notably,Progression of disease within24 months after first-line treatment(POD24)was found to be a significant poor prognostic factor in follicular lymphoma and was used as an important predictor of prognosis for FL patients by various prognostic models.There is no optimal prognostic model to accurately predict patients with early disease progression,and given the complexity and cost of the technology,genetic bio-clinical prediction models are difficult to be scaled up in the clinical setting,so simple and efficient tumor microenvironment is a current research hotspot,especially the index detection of immune checkpoints.The aim of the full paper is to combine traditional prognostic models with new indicators to develop new prognostic models and to use the construction of Nomogram models to predict survival more accurately in high-risk patients with FL in order to provide clinical utility.Methods:Data on patients with primary follicular lymphoma with complete medical records from January 2015 to December 2020 in the Department of Hematology of Shanxi Cancer Hospital and from an independent group of patients from Tianjin Cancer Hospital were collected.(1)Single-factor 2 test and multifactor logistic regression were used to screen variables for the main risk factors of FL patients in order to construct new prediction models and calculate their correlation performance(sensitivity,specificity,accuracy and survival curves);(2)three methods were used to initially screen variables by single-factor COX regression,Best Subsets Regression,and LASSO regression with cross-validation.The variables screened by the three methods were incorporated into multi-factor COX regression for final screening.Models are constructed and compared using the variables finally selected by the three methods.The best model will be used as the construction of a predictive model(Nomogram),and subsequently using calibration curves,area under the ROC curve(AUC),and decision curves analysis(DCA)to validate the assessment performance.Results:1.predictive modeling: among 135 patients with primary FL,a total of 32(23.7%)patients developed PFS and 20(15%)patients developed POD24,among which 70 FL patients underwent immunohistochemical testing(IHC)and 12 patients developed POD24.by one-way 2 test on relevant variables for primary screening,screening of MUM-1positivity,Ki67 high expression,and PRIMA-PI high risk as risk factors for POD24(all P <0.05),and multifactorial logistic regression showed that patients with PRIMA-PI high risk and Ki67 high expression were risk factors for POD24(all P < 0.05),and a new prognostic model(PRIMA-PIC)was established by combining PRIMA-PI with Ki67.The sensitivity of PRIMA-PIC in predicting POD24 was calculated to be higher than that of PRIMA-PI and the results of log-rank test using Kaplan-Meier survival curve showed that PRIMA-PIC with high risk(P=0.016)was more predictive of FL patients than PRIMA-PI with high risk(P=0.034).2.Nomogram model building and evaluation: In the full paper,70 patients who all received IHC were used as the training set,the remaining 65 patients were the internal validation set,and the external validation set consisted of 74 patients from an independent group in Tianjin Cancer Hospital.The results of the initial screening variables showed: 4variables for the initial screening of one-factor COX;6 variables for the initial screening of Best Subsets Regression;and 4 variables for the initial screening of LASSO regression.Further inclusion of multi-factor COX regression results showed that the model was constructed based on single-factor COX regression with 2 variables,based on Best Subsets Regression and LASSO regression with 4 variables after the exclusion of variables.Using the ROC curve,which had the largest AUC value(83.8)and the smallest AIC value(101.63)as the evaluation criteria,the models constructed using the variables screened by Best Subsets Regression and LASSO regression were superior.Therefore,gender,histological grading,NK cell percentage and PRIMA-PIC risk group were finally selected as variables for the construction of Nomogram.The Nomogram model constructed in the article were better able to predict the value of survivability in FL patients as validated by calibration curves,AUC and DCA.Conclusions:1.MUM-1 positivity,high Ki67 expression,and high risk of PRIMA-PI are risk factors for POD24.And the ability of PRIMA-PIC established with PRIMA-PI combined with Ki67 to predict the survival of FL patients is better than the performance of PRIMA-PI.PRIMAPIC index has practical value in improving the sensitivity of PRIMA-PI to predict POD24 in patients while maintaining the characteristic high specificity advantage.2.The Nomogram model was constructed based on 3 methods of screening variables:one-way COX regression,Best Subsets Regression,and LASSO regression plus crossvalidation,and finally the Nomogram constructed with gender,histological grading,NK cell percentage,and PRIMA-PIC risk group as the prediction model was able to predict well the prognostic survival of patients with follicular lymphoma.