In this retrospective study, we collected files of patients diagnosed with non-metastatic ccRCC with VTT in Jinling Hospital, Drum Tower Hospital, and Changhai Hospital between June 2011 and November 2021. In total, 402 RCC patients with VTT were selected among patients with RCC. The exclusion criteria were: a) incomplete clinical data; b) unknown survival time; c) more than one primary tumor; d) not clear cell renal cell carcinoma; e) lymph node or distant metastasis. After exclusion criteria, 328 patients were adopted in this study. Patients were then randomly divided into training (n =213) and validation cohorts (n = 115) with a ratio of 7:3.

The following clinical data were collected: the clinicopathological parameters (such as age, gender, symptoms, comorbidity, ECOG Performance Status Scale, TNM stage, Mayo classification, tumor size, WHO/ISUP grade, etc.) and several preoperative blood routine examination results (absolute counts of leukocytes, neutrophils, lymphocytes). All laboratory data were performed within 1 week before surgery. The SII was calculated through the following equation: SII = platelet count × neutrophil count/lymphocyte count. All patients had complete preoperative investigations, including clinical, computed tomography (CT) scan and/or magnetic resonance imaging (MRI) examinations to confirmed the diagnosis. Adjuvant therapy included postoperative adjuvant tyrosine kinase inhibitors and immunotherapy. All patients were regularly followed up after surgery. Overall survival (OS) was defined as the period between the date of surgery and the date of death for all causes or the most recent follow-up. Disease-free survival (DFS) was defined as the time between the date of surgery and the date of recurrence of tumor or death, or last follow-up.

Continuous variables are presented as median (IQR) and categorical variables are presented as percentages. Data were tested for normality using the Shapiro-Wilk test. The Student’s t-test was performed to compare continuous variables, and Pearson’s Chi-square test was employed for categorical variables. In case the normal distribution of continuous variables was not assumed, the Mann Whitney U test was performed. Receiver operating characteristic (ROC) curves were plotted for SII to evaluate their sensitivity and specificity. And to determine the optimal cut-off value of inflammatory indices, Youden’s index was used. To evaluate the independent prognostic factors, cox proportional hazards models were utilized for univariate and multivariate analyses. Kaplan-Meier survival curve was plotted to compare the OS and DFS between different groups. Harrell’s C-index was used to measure the predictive accuracy of each model. To further explore clinical utility, we performed a decision curve analysis. To predict patients’ OS, nomogram was created using the rms package in R. Multivariable Cox regression models were trained via the least absolute shrinkage and selection operator (Lasso) to determine which variables to include in the nomogram. X-tile (17) was used for survival and relapse risk stratification. All p values were two-tailed, and p<0.05 was considered statistically significant. All statistical analyses were performed by R software (Version 4.1.3).

There were no significant differences in the clinical and pathological characteristics of the training and the validation cohorts. The clinical characteristics of the patients in the training and validation cohorts were summarized in Table 1 . Among 328 participants, 210 (64.0%) were male and 118 (36.0%) were female. The median age was 61years (IQR 53-68) and 59 (18.0%) of them were older than 70 years. Those who presented with no symptoms were 141 (43.0%) patients. Pathological T stage was T3a in 163 cases (50.0%), T3b in 127 (38.7%), T3c in 26 (7.9%), and T4 in 11 (3.4%). Mayo classification was 0 in 171 patients (52.1%), 56 (17.1%) in I, 58 (17.7%) in II, 22 (6.7%) in III, and 21 (6.4%) in IV. All patients underwent radical nephrectomy with tumor thrombectomy (RNTT), including open RNTT in 168 (51.2%) patients, laparoscopic RNTT in 123 (37.5%), and robot-assisted RNTT in 37 (11.3%) patients. Nearly half (48.2%) of all patients had WHO/ISUP grade III tumors, and 23.8% had grade IV tumors. The maximum diameter of the tumor ranged from 2 to 20 cm with a median size of 7.2 cm (IQR 5.5-9.0). The median follow-up was 24 months (IQR 8-49).

According to the optimal cut-off analyzed by using the maximum Youden index, 912 were applied as the optimal cut-off value for SII. The value of the area under the curve (AUC) is calculated to validate it as depicted in Supplementary Figure 1 .

To analyze the association between the SII with clinicopathological parameters, the optimal cut-off of SII was used to dichotomize the 328 patients into high-SII (SII≥912) and low-SII (SII<912) groups. There were 125 (38.2%) patients in the high-SII group and 203 (61.9%) patients in the low-SII group. Supplementary Table 1 shows that higher SII was significantly correlated with more symptoms (Fever or overt weight loss, p<0.001), longer surgical time (p=0.001), higher WHO/ISUP grade (III-IV)(p<0.001), and longer tumor diameter (p<0.001).

After a median follow-up of 24 months, 121 (36.9%) patients had progressed and 72 (22.0%) patients had died. The median OS was not reached and the median DFS was 70 months (IQR: 38.0-NR). The 1, 3, and 5-year OS was 92.4%, 73.4%, and 66.7% and the 1, 3, and 5-year DFS was 77.7%, 57.5%, and 51.5%. Figure 1 reflects Kaplan-Meier curves for OS (A) and DFS (B) stratified by SII. The log-rank test estimated a significant difference between SII<912 and ≥912 in OS (p < 0.001) and DFS (p < 0.001). In the high-SII group, the 1, 3, 5-year OS were 86.2%, 60.0% and 52.7% and the 1, 3, 5-year DFS were 68.8%, 40.2% and 35.9%. In the low-SII group, the 1, 3, 5-year OS were 96.2%, 82.2% and 75.7% and the 1, 3, 5-year DFS were 83.3%, 69.3% and 61.9%. Table 2 shows that pT stage (pT3b: p=0.002; pT3c: p<0.001; pT4: p<0.007), Mayo classification (Level I: p=0.01, Level II: p=0.016; Level III: p=0.005; IV: p<0.001), WHO/ISUP grade 3 (p<0.001), grade 4 (p<0.001), adjuvant therapy (p=0.043), and SII (p<0.001) were significant predictors for OS in univariate analysis. And multivariate analysis revealed that Mayo classification (Level I: p=0.021, Level II: p=0.009; Level III: p=0.043; IV: p<0.001), WHO/ISUP grade 3 (p=0.002), grade 4 (p<0.001), adjuvant therapy (p=0.01), and SII (p=0.002) were independent prognostic factors for OS. And higher SII ≥912 increased the risk of death 2.22 times compared with those in low-SII group. The result of the univariable analysis shown in Table 3 revealed that BMI (p=0.01), ECOG (P=0.007), surgical approach (Laparoscopic RNTT: p<0.001), pT stage (pT3b: p=0.012; pT3c: p<0.001; pT4: p<0.001), Mayo classification (Level I: p=0.026, Level II: p=0.024; Level III: p=0.07; IV: p<0.001), WHO/ISUP grade 3 (p<0.001), grade 4 (p<0.001), tumor diameter (p=0.016), and SII (p<0.001) were correlated with DFS. Factors that remained significant in multivariable analysis were Mayo IV (p=0.003) thrombus, WHO/ISUP grade 3 (p<0.001), grade 4 (p<0.001), and SII (p=0.001). And patients with higher SII increased the risk of recurrence 1.886 times.

Comparisons of the c-index between dichotomized SII (0.630 for OS and 0.595 for DFS) and other indicators including ECOG (0.583 for OS and 0.539 for DFS), tumor diameter (0.586 for OS and 0.565 for DFS), adjuvant therapy (0.561 for OS and 0.523 for DFS), and tumor necrosis (0.551 for OS and 0.524 for DFS) demonstrated the superiority of SII in predicting the outcomes ( Table 4 ). The pT stage (0.644 for OS and 0.606 for DFS) and mayo classification (0.652 for OS and 0.602 for DFS) showed no significant difference with SII. The WHO/ISUP grade (0.640 for OS and 0.623 for DFS) were the only indicator better than SII.

We integrated dichotomous SII into existing models including UISS risk group (18), GRANT score (19), Leibovich model (20), and SSIGN score (21) to see whether it can improve their predictive accuracy. As anticipated, the addition of SII improved models’ performance for predicting OS (0.684 for UISS versus 0.732 for UISS+SII, 0.677 for GRANT versus 0.723 for GRANT+SII, 0.747 for Leibovich versus 0.77 for Leibovich +SII, 0.735 for SSIGN versus 0.765 for SSIGN+SII) and DFS (0.658 for UISS versus 0.692 for UISS+SII, 0.685 for GRANT versus 0.702 for GRANT+SII, 0.692 for Leibovich versus 0.709 for Leibovich +SII, 0.693 for SSIGN versus 0.717 for SSIGN+SII). All the models achieved higher c-indexes than the original models ( Table 5 ). It was also confirmed by ROC analysis in Supplementary Figure 2 . A decision-curve analysis (DCA) shown in Supplementary Figure 3 demonstrates that the use of the SII in prognostic models contributes added clinical value.

Given the excellent predictive power of SII for OS, to predict survival probability of patients, we constructed a nomogram incorporating the clinical predictive factors and SII. Based on the results of LASSO Cox regression analysis shown in Figure 2 , five predictive features including ECOG, Mayo classification, WHO/ISUP grade, adjuvant therapy, and SII were selected in the nomogram for OS. Then we developed a nomogram to predict OS in training cohort. Figure 3 showed the 1-, 3-, and 5-year OS nomogram developed by the Cox proportional hazards results. Time-independent ROC analysis ( Supplementary Figure 4 ), DCA ( Supplementary Figure 5 ) and calibration curves ( Supplementary Figure 6 ) revealed well-powered superiority in prognostic accuracy and clinical value in the training and validation cohorts.

By using X-tile, risk stratifications based on the total points of nomogram were made. The cohorts were divided into three risk groups, with patients in each group having high (total points≥158), middle(108<total points ≤ 158), and low (total points ≤ 108) risk of death, respectively. The Kaplan-Meier OS curves displayed great differences in three groups ( Figure 4 ).