2.1 Data source and study cohort

We followed the STROBE statement and this study is fully compliant with the STROCSS criteria [11]. Our registration unique identifying number is researchregistry7016.

Following institutional review board approval, we retrospectively reviewed our database (Peking University Third Hospital Thrombus Database, PUTH-TD) containing medical records of all patients with renal vein thrombus or inferior vena cava (IVC) thrombus from January 2014 to January 2021 (n = 350). The database has been constructed prospectively since January 2014. Two full-time clinical data managers had all access to the data through electric case report forms (eCRF) and were responsible for data entry, verification and quality control. The demographic, perioperative, postoperative and follow-up data were accurately recorded. The inclusive criteria were as follows: (1) pathologically confirmed tumor of renal origin; (2) detailed information on surgical procedures; (3) no comorbidity of hemorrhagic disease; (4) minimum follow-up of 6 mon. The patients who were treated by robot-assisted LRN-VT (RALRN-VT) or who didn’t receive surgical treatment were excluded.

2.2 Surgical procedures

The techniques of LRN-VT and ORN-VT at our institution have been described previously [8, 12, 13, 14]. For LRN-VT, both transperitoneal approach and retroperitoneal approach were used in our center depending on the thrombus level, tumor side, surgery history and invading the IVC wall or not. For level 0 thrombus, we didn’t block the IVC and we cut off the renal vein from the segment without tumor thrombus. For level I thrombus, the IVC was partially clamped using a Satinsky clamp. We incised the IVC and removed the thrombus and then irrigated the IVC with heparinized saline before suturing. For level II thrombus, we clamped the IVC below the thrombus, the contralateral renal vein and the IVC above the thrombus sequentially after we cut off the renal artery on the tumor side. The contralateral renal artery was blocked for the left tumor, while the left renal artery was not blocked for the right tumor. Segmental resection of the IVC would be considered if the thrombus invaded the IVC wall extensively. If minor involvement of the IVC wall occurred, we performed resection of the invaded vessel wall rather than segmental resection. For level III or IV thrombus, the first porta hepatis should be blocked and the Foley catheter-assisted six-step thrombectomy technique [12] was used in our center. If the right atrium was involved, ORN-VT is usually needed. We usually incised the thoracoabdominal midline and the “milking” technique is performed to squeeze the intra-atrial thrombus back into the IVC.

2.3 Follow-up

We provided the same follow-up plan to all patients and follow-up data were regularly collected (symptoms and signs, laboratory tests, imaging examination of the chest, abdomen and pelvis). The laboratory tests included routine blood test and blood biochemical test. The imaging examination included computed tomography (CT), magnetic resonance imaging (MRI) and X ray. Patients were followed up every 3 mon after surgery in the first year, then 6 mon to the third year, then annually thereafter. We determined the death reason according to the death certificate issued by the local medical institution if a patient died during the interval of follow-up visit or we determined that by the treating physicians.

2.4 Outcomes and definitions

Primary outcomes contained surgical outcomes and oncologic outcomes. The surgical outcomes were represented by perioperative and postoperative outcomes (operative time, blood loss and complication rate, et al) between the two groups and the oncologic outcomes were represented by the overall survival (OS), tumor-specific survival (TSS), metastasis-free survival (MFS) and local recurrence-free survival (LRFS). Complications were evaluated within 30 days after surgery. OS meant the length of time from surgery to death from any cause. TSS was defined as the time from surgery to death due to tumor. LRFS was defined as the time from surgery to local recurrence (tumor recurrence in or abutting the previous surgical bed) based on CT or MRI. MFS was defined as the time from surgery to the development of metastatic disease (new lesions in other organs, brain, lung, liver, bone, et al) based on CT or MRI. MFS was only assessed in patients with M0 diseases.

The presence of local symptoms was defined as a palpable mass, pain, gross hematuria. Patients with edema, fever, swelling, fatigue and weight loss, et al were thought to have systemic symptoms. The American Society of Anesthesiologists Physical Status classification system (ASA level) [15] was introduced to estimate the operative risk of patients. Thrombus level was classified using the Mayo Clinic classification of tumor thrombus level [1]. Postoperative complications were graded according to the Clavien-Dindo grading system [16].

The histological diagnosis of renal tumors was based on the World Health Organization (WHO) classification (2004 and 2016 version) [17, 18]. The Fuhrman system was applied to RCC nuclear grading [19]. A sarcomatoid differentiation was defined as RCC accompanied by histological appearance of spindle-cell sarcoma. The 2017 version of the tumor-node-metastasis (TNM) classification was used for clinical staging based on postoperative pathological specimen.

2.5 Data quality and bias control

We designed a precise eCRF containing all the necessary medical information and standardized the definition, naming method and data type of each variable. We set up mandatory fields to reduce data missing and two trained data managers performed data entry to reduce logic errors. The data managers were blinded to the study design. A third party checked the database regularly to keep the authenticity and accuracy.

We set strict criteria for inclusion and exclusion to reduce the selection bias. We performed a propensity-matched comparative analysis to minimize the influence of confounding bias and to identify a cohort of patients with comparable baseline characteristics. The propensity score was estimated by a multivariable logistic regression model and the 1:1 matching without replacement was performed with a caliper width equal to 0.02. Patients were matched regarding age, body mass index (BMI), ASA level, comorbidity, tumor diameter, thrombus level and metastasis at diagnosis. Matching was performed by a researcher who didn’t know the outcomes.

An independent researcher who was blinded to the matched-cohort analyzed the outcomes to control the information bias. Except for the routine review after surgery, the data managers conducted telephone interviews every 3–6 mon and collected the follow-up information to reduce the withdraw bias.

2.6 Statistical analysis

Baseline characteristics were shown for categorical variables and continuous variables. We reported the medians and interquartile ranges for non-normally distributed continuous variables and the means and standard deviations for normally distributed continuous variables. Categorical variables were reported as frequencies and proportions. The chained multiple imputation was used to resolve the missing data. We compared differences between non-normally distributed continuous data and ordered categorical data with the Wilcoxon sum rank test and compared differences between unordered categorical data with the Chi-square test and Fisher exact test. In the matched cohort, the paired t-test, the Wilcoxon signed rank or the McNemar test was applied to compare differences. We used the Kaplan-Meier method to perform survival analysis and assessed the differences between the propensity-matched groups with log-rank test. Cox proportional hazards regression model was used to estimate the relative oncologic outcomes after adjusting the variables that satisfied the proportional hazards assumptions (thrombus level, lymph node metastasis and distant metastasis, perinephric fat invasion, adjuvant therapy, et al). A hazard ratio (HR) less than 1.0 favored LRN-VT. All statistical tests were performed by SPSS version 25.0 (IBM, Armonk, NY, USA) and the R statistics package version 3.6.1 (R Project for Statistical Computing, www.r-project.org). All tests were two-sided, and the significance level was set at p < 0.05.

3.1 Patient characteristics

Of consecutive 350 patients identified, 32 patients were excluded for pathologically confirmed tumors that were not renal origin, 6 patients were excluded for RALRN-VT, 6 patients were excluded because of conservative treatment and 4 patients were excluded for shorter than 6 mon of follow-up. 302 patients were matched, including 148 patients treated with LRN-VT and 154 patients treated with ORN-VT. Before propensity matching, significant differences existed in several baseline variables (Table 1). After matching, 94 patients were matched in each group and no significant differences were found in baseline characteristics.

3.2 Surgical outcomes

Surgical outcomes were showed in Table 2. Twenty-four patients (25.5%) in the LRN-VT group converted to open surgery. Forty-one patients (43.6%) received adrenalectomy in the LRN-VT group and 44 patients (46.8%) received adrenalectomy in the ORN-VT group (P = 0.66). Nine patients (9.6%) in the LRN-VT group and 23 patients (24.5%) in the ORN-VT group underwent segmental resection of IVC (P = 0.01). The median operative time of ORN-VT was longer than that of LRN-VT (326min vs 292min, P = 0.002). The median blood loss of ORN-VT was significantly greater than that of LRN-VT (1000 ml vs 500 ml, P < 0.001). Sixty-three patients (67%) in the ORN-VT group received intraoperative or postoperative blood transfusion, while only 38 patients (40.4%) in the LRN-VT group needed blood transfusion (P < 0.001). The median packed red blood cells transfusion and fresh frozen plasma transfusion in the ORN-VT group were significantly greater than those in the LRN-VT group (1200 ml vs 800 ml, 600 ml vs 400 ml, all P < 0.001). The complication rate of ORN-VT was higher than that of LRN-VT (39.4% vs 21.3%, P = 0.007) and the Clavien grade was higher in the ORN-VT group than that in the LRN-VT group (P = 0.005). The median postoperative hospital stay of LRN-VT group was shorter than that of ORN-VT group (8d vs 10d, P < 0.001).

3.3 Oncologic outcomes

After a median follow-up of 31mon (IQR 19–44 mon) in the LRN-VT group and 32 mon (IQR 17–40 mon) in the ORN-VT group, 28 deaths occurred in each group (29.8% vs 29.8%, P = 1.0). Twenty-four deaths (25.5%) in the LRN-VT group contributed to tumor and all deaths in the ORN-VT group contributed to tumor. The median time to death in the LRN-VT group and ORN-VT group were 17 mon (IQR 9–26 mon) and 12 mon (IQR 9–20.5 mon), respectively (P = 0.29). Among patients with M0 disease (LRN-VT group, n = 73 vs ORN-VT group, n = 65), 33 patients (45.2%) in the LRN-VT group and 36 patients (55.4%) in the ORN-VT group developed distant metastases (P = 0.23). The most common new metastatic sites were lung (n = 44), bone (n = 24) and liver (n = 18). Local recurrence occurred in 1 patient in the LRN-VT group and 5 patients in the ORN-VT group (P = 0.28). The median time to local recurrence in the LRN-VT group and ORN-VT group were 36 mon and 8 mon (IQR 6–15 mon), respectively (P = 0.007). Table 3 summarized the data on follow-up in each group.

Figure 1 depicted the Kaplan-Meier curves of oncologic outcomes. We couldn’t observe a statistically significant difference in either OS (Adjusted HR 0.96, 95%CI 0.67–1.40; P = 0.85), TSS (Adjusted HR 1.03, 95%CI 0.69–1.52; P = 0.90) or MFS (Adjusted HR 0.89, 95%CI 0.58–1.38; P = 0.61) between the two matched groups (Figure 1A, B and C). Patients who underwent LRN-VT had a lower risk of local recurrence compared to patients who underwent ORN-VT (Adjusted HR 0.18, 95% CI 0.05–0.62, P = 0.007) (Figure 1D).

3.4 Pathological outcomes

Table 4 showed the pathological results of the renal tumors between the two groups. We couldn’t observe a statistically significant difference in either T stage (P = 0.16), lymph node metastasis (0.09), perinephric fat invasion (0.39), histological type (P = 0.37), adrenal metastasis (0.25), venous wall involvement (0.61), sarcomatoid differentiation (P = 0.49) or Fuhrman grade (P = 0.12) between the two matched groups.