Establishment And Prognostic Analysis Of Preoperative Hematological Prognostic Score Model In Patients With Glioblastoma

ObjectiveGlioblastoma(GBM)is the most common primary malignant brain tumor.Recent studies have shown that preoperative hematological biomarkers have emerged as a powerful tool to predict the prognosis of patients with cancer.A single hematological marker has limited prognostic value,and is not an ideal prognostic indicator.Therefore,we aimed to establish a comprehensive prognostic scoring model to improve the prognostic prediction in patients with GBM.MethodsThis retrospective study included patients who were newly diagnosed with GBM at the First Affiliated Hospital of Zhengzhou University between June 2016 and January 2019,and patients were randomly divided into a training set and external validation set to develop and validate a hematological prognostic score(HPS).The following variables were obtained for each patient:age at diagnosis;sex;first presenting symptoms;clinical history;preoperative Eastern Cooperative Oncology Group performance status(ECOG PS)score;tumor size;tumor location;extent of resection;isocitrate dehydrogenase(IDH)mutation status;postoperative adjuvant therapy;and laboratory index values,which were collected from our hospital case documents.Receiver operating characteristic(ROC)curve was used to determine the optimal cutoff value of each laboratory index,and the cutoff values were used to adjust all indicators to binary variables.Survival rates were calculated using the Kaplan-Meier method,and the significance of differences between the survival curves was determined using the log-rank test.Cox regression models were used to identify independent prognostic factors in patients with GBM.The least absolute shrinkage and selection operator Cox proportional hazards regression analysis was used to determine the optimal covariates that constructed the prognostic score.Furthermore,a quantitative survival-predicting nomogram was constructed based on HPS combined with clinical index.The results of the nomogram were validated using bootstrap resampling and the external validation set.Finally,we further explored the relationship between the HPS and clinical prognostic factors.Results1.According to the ROC curve,the optimal cutoff value for HPS was 0.839 in the overall data.In the training cohort,patients were successfully classified into different prognostic groups based on their HPSs,that is,the high-HPS group and low-HPS group.According to the results of univariate and multivariate survival analysis,the overall survival(OS)of the low-HPS group was significantly longer than that of the high-HPS group(P<0.001).The areas under the curve(AUCs)of the HPS were 0.67,0.73,and 0.78 at 0.5,1,and 2years,respectively.The same cutoff value of HPS(0.839)was used for risk grouping in the external validation cohort.According to the results of univariate and multivariate survival analysis,patients in the low-risk group had significantly better outcomes than those in the high-risk group(P<0.001).The 0.5-y,1-y,and 2-y AUCs of the HPS were 0.51,0.70,and 0.79,respectively.2.In the training group,the C-index value of the nomogram based on HPS and clinical prognostic factors was 0.81;the calibration plot of the nomogram model showed that the predicted 0.5 and 1-year OS was in good agreement with the actual survival time shown by Kaplan-Meier analysis;the decision curve analysis results showed that the nomogram had clinical net benefit.In the validation group,a high C-index value of 0.82 could still be achieved;the calibration curve showed that the predicted and observed survival rates at 0.5 and 1 year were similar.3.In the overall data,violin plots were used to show the HPSs in different groups of patients which were divided by age,first symptoms,IDH mutation status,and tumor location,respectively.Patients aged>50 years(P<0.001),with seizures as the first presenting symptom(P<0.001),with IDH-wild-type GBMs(P<0.001),and with tumor growth in mixed sites(P<0.005)had significant higher HPSs.4.1n the subgroup of clinical factors,patients were divided into the low-HPS group and high-HPS group by the optimal cutoff value.There were significant differences in the survival rates between the low-HPS group and high-HPS group(P<0.05),except among the "ECOG PS-3 or 4"(P=0.36 vs.P=0.45),"tumor location-thalamus"(P=0.87);"surgical resection-partial"(P=0.84)and "therapy status-none"(P=0.57)subgroups.5.Among them,HPS and IDH mutation status could be subgrouped with each other,and there were statistically significant differences between each subgroup(P<0.05).The survival curves combining HPS with age,first presenting symptoms and IDH mutation status showed that HPS can carry out more accurate risk stratification based on those clinical characteristics.Among them,HPS and IDH mutation status could be grouped with each other,and there were statistically significant differences between each subgroup(P<0.05).Conclusion1.The HPS was an independent prognostic factor for OS,and had robust prognostic value for GBM patients.2.The nomogram model based on HPS had a high predictive ability for OS in GBM patients.3.The HPS showed specific correlations with age,first presenting symptoms,IDH mutation status and tumor location,which can synergize with those clinical factors to identify high-risk patients.4.The HPS can be widely suitable for the prediction of prognosis in most GBM patients,and was able to further stratify each clinical feature into different risk subgroups.