Data
We identified 21411 mRCC patients with ICD-0-3 primary site code C64.9 and M1 stage. Among them, 8478 patients were recommended for surgery. Only 7515 patients meted the code of RX Summ--Surg Prim Site 0 (NS), 40 (CN) or 50 (CN). Then 449 pitients with TX and 428 patients with NX were excluded. After that, 106 cases with unknown, abnormal or inaccurate tumor size were excluded (CS tumor size code 0, 600, 700, 800, 900,920,930,950,980,989,990-999). Finally, a total of 6532 cases which met our inclusion criteria were retrieved, including 6043 that underwent CN and 489 cases that received NS. Table 1 showed patient demographics and clinical characteristics of these patients.
Multivariate cox regression analysis and nomogram plot
Through X-Tile, we figured out the best cut-off points for age at diagnosis and tumor size as 76-years-old (maximum chi-square 31.3, Miller-Seigmund P<0.0001) and 82 mm (maximum chi-square 45.5, Miller-Seigmund P<0.0001), respectively. According to the latest guideline and for the convenience of analysis, we unified N2 and N1 in derived AJCC T, 6th ed. (2004-2015) as N1. Nomogram plot in Figure 1 showed the weight of every variable in the result of multivariate cox analysis. It demonstrated that sex(P=0.01), age(P=0), race(P=0.02), T stage(P=0), N stage(P=0) and tumor size(p=0) were independent risk factors for CSS of total mRCC patients. Therefore, in the next subgroup analysis, the PSM was performed for balancing sex, race, age, tumor size, T stage, and N stage in CN- or NS-treated mRCC patients.
Subgroup analysis
First, we compared the CSS of mRCC patients treated by CN or NS for subgrouping under each variable (age, sex, race, T stage, N stage, and tumor size). Other variables were balanced by PSM to reduce potential confounding effects and treatment selection bias when comparing subgroup of one variable. Survival benefits were found in patients treated by CN in almost all subgroups except T4 subgroup. Although there were 8 months benefit of median CSS time in T4 subgroup patients, the HR (95% CI) was 0.801 to 1.816, crossing 1 (Figure 2). This was consistent with the result of KM survival analysis, showing that in T4 subgroup the CSS between CN (95 cases) and NS (50 cases) had no statistical significance (Figure 4A, p=0.29).
In order to further explore which factors may prevent patients of T4 subgroup from benefiting from CN, we matched T4 stage with age, N stage and tumor size respectively to group them in more detail (Figure 3). Table 2 showed the result of PSM in T4 stage patients. The result showed that the patients in T4&N0 subgroup and T4&age<76yr subgroup could still obtain CSS benefit from CN while T4&N1 subgroup and T4&age≥76yr could not, meaning N1 and age≥76yr were high risk factors which influenced prognosis (Figure 4B, C). In addition, there was no significant difference in CSS between CN- and NS-treated mRCC patients, whether in T4&tumor-size<82mm subgroup or T4&tumor-size≥82mm subgroup. It suggested that although multivariate COX results show that tumor size was an independent prognostic factor, tumor size had little effect on CSS of T4 stage mRCC. At last, we compared OS between patients who received CN or NS in T4, T4&N1 and T4&age≥76yr subgroups. The result showed the patients of these subgroups also cannot acquire OS benefit from CN (Figure 4D, E, F).