As an external validation dataset, we first validated the studied four-tiered grading system in a cohort of Asian patients with ChRCC. In the present study, we found that grade according to the four-tiered ChRCC grading system proposed by Avulova et al. [15] was significantly associated with cancer-specific death and distant metastasis, and that its predictive ability was stronger than that of any previous grading system. Nevertheless, separating Grades 2 and 3 did not add further prognostic value. Although grade according to the exploratory four-tiered system, which classifies tumors with necrosis as Grade 3, is associated with survival outcomes, its predictive ability is low. Similarly, Grades 2 and 3 have the same predictive power.

It has been shown that Fuhrman tumor grade is unable to predict the prognosis of ChRCC [13, 14, 19]. In the absence of a pathological grading system that accurately predicts the prognosis of ChRCC, several grading systems aimed at identifying patients with adverse outcomes have been proposed. Paner et al.’s system [19], in which 124 patients were studied, and the median follow-up time was 37 months, is reportedly able to accurately stratify patients with ChRCC according to prognosis. This three-tiered ChRCC grading system separates Grades 1 and 2 according to nuclear crowding and atypical cells, whereas Grade 3 tumors have sarcomatoid differentiation. Four recent studies have validated this grading system. Some studies [16,17,18, 24] have raised questions and dissenting opinions on the three-tiered grade system proposed by Paner et al., Avulova et al.’s study being an exception [25]. In a study of 266 US–American patients, these authors confirmed the prognostic value of chromophobe tumor grade as proposed by Paner et al. for cancer-specific death and distant metastasis [19], which is similar to our results. Cheville et al. studied 124 patients with ChRCC and concluded that, after adjustment for TNM stage, there was no difference in cancer-specific survival between Grades 2 and 1 in Paner et al.’s grading system. In patients with non-sarcomatoid ChRCC, grade in the three-tiered grading system is not associated with cancer-related death rate [16]. The concordance indexes of the studies by Cheville et al. and Avulova et al. for CSS are reportedly 0.766 and 0.84, respectively. According to Finley et al., exclusion of sarcomatoid ChRCC results in grade in the three-tiered grading system no longer being an independent predictor of recurrence [17]. The AUC of the ROC curve for recurrence-free survival in all patients was 0.822. Ohashi et al. reported similar overall survivals as determined by Kaplan–Meier curves for Grades 1 and 2 in the three-tiered chromophobe grading system [18].

In 2019, Ohashi et al. [18] reported a novel two-tiered grading system with a concordance index of 0.79 that was based on findings in 382 patients from multiple medical centers. In this system, the presence of sarcomatoid differentiation and/or necrosis results in a classification of high grade. High grade, advanced age, lymph node and/or distant metastasis, and advanced T stage were associated with overall survival according to multivariate regression analysis. It has been pointed out that overall survival does not reflect cancer-specific survival and that it is inappropriate to include sarcomatoid changes and necrosis in a single grade. Five-year CSS rates reportedly differ significantly between sarcomatoid (44%) and non-sarcomatoid ChRCC (61%) with tumor necrosis [25]. In our study, 5 year CSS rates in the above groups were 37.5% and 90.3%, respectively.

Many studies have found that tumor necrosis is a predictor of prognosis of ChRCC [5, 7, 21, 26, 27]. Avulova et al. [15] accordingly sought to improve Paner et al.’s three-tiered grading system [19] by proposing a four-tiered grading system. As far as we know, this four-tiered grading system has not yet been externally validated. This grading system incorporates necrosis with a high c-index of 0.85, which is why we performed external verification on an Asian cohort.

Avulova et al.’s study included 266 patients; the 5 year CSS rate for Grades 1, 2, 3, and 4 was 98.4%, 90.6%, 61.2%, and 44.5%, respectively. They reported that higher grade in their grading system, larger tumor, higher pT stage, pN1 stage, necrosis, and sarcomatoid change were all significantly associated with poor survival rates according to univariate analysis. Additionally, associations between grade according to the four-tiered system and CSS and DMFS outcomes remained significant after adjustment for other variables, which is consistent with our findings. In the subgroup of 247 patients with Grade 1 or 2 tumors, tumor size, pN stage, and necrosis were associated with CSS, whereas pT stage was not. This apparent discrepancy may be attributable to the association between high pT stage and sarcomatoid change. Avulova et al.’s grading system was not evaluated in this subset. In contrast, Ohashi et al. [28] reported that Grade 3 (P = 0.871) and Grade 2 (P = 0.182) in the four-tiered grading system were not associated with CSS in a cohort of 245 patients.

In our study, LVI was found to be an independent predictor of metastasis after nephrectomy for ChRCC with or without sarcomatoid change. To the best of our knowledge, this is the first such report. Many studies have found that LVI is a predictor of distant metastasis in patients with clear cell renal cell carcinoma [20, 29]. When our analysis was restricted to 67 patients with Grades 2 and 3 in the four-tiered grading system, the risk of death from RCC and metastasis was approximately twice as high for Grade 3 as for Grade 2, and grade tended to be associated with cancer-specific death and metastasis; however, this difference was not statistically significant. Avulova et al. did not analyze or discuss the difference in predictive capability between Grades 2 and 3. We speculate that the association between tumor necrosis and survival may be attributable to the presence of tumor necrosis in sarcomatoid ChRCCs. After adjustment for sarcomatoid differentiation, coagulative tumor necrosis was no longer associated with survival. Despite the c-indexes (0.875 and 0.877) for the three-tiered grading system for CSS and DMFS being lower than those (0.892 and 0.897) for the four-tiered grading system, the increased stratification did not result in more predictors of prognosis. Thus, we consider that the three-tiered grading system more accurately predicts the risk of death and metastasis of ChRCC.

A grading system that includes a separate grade for the presence of tumor necrosis has not previously been explored. In the absence of tumor necrosis, Grades 1 and 2 are separated on the basis of nuclear crowding and cellular atypia. Once tumor necrosis and sarcomatoid differentiation or frank anaplasia have developed, tumors are classified as Grades 3 and 4, respectively. There are statistically significant associations between grade according to this system and CSS and DMFS. However, it has the same problem as the grading system proposed by Avulova et al., in that there is no difference in predictive capability between Grades 2 and 3.

The present study has the following strengths: To the best of our knowledge, this is the first validation of a chromophobe grading system in an Asian cohort. This study was large, being conducted at a large regional medical center. The study patients had all undergone laparoscopic surgery, unlike in the previous studies [15], in which open surgery was performed on most patients. Laparoscopic surgery has fewer complications and is, therefore, currently the preferred surgical approach for most patients. Because this study more closely matches current treatment protocols, the survival times likely more accurately reflect current clinical outcomes. We acknowledge that this study had some limitations. First, it was retrospective and conducted in a single center; a large multicenter study is needed. Like in other studies of ChRCC, the mortality and metastasis rates were low. When there are so few target events, adjustment of multivariable Cox regression analysis by too many variables is prone to result in over-fitting [30]. We, therefore, adjusted only one variable at a time to maximize the reliability of our findings. Nevertheless, our multivariate analysis of multiple variables may not be accurate.