Pathologically examining a minimum of three lymph nodes could better determine node negativity in patients with non-metastatic chromophobe renal cell carcinoma

Objective: To evaluate the prognostic value of lymph node (LN) involvement for patients with chromophobe renal cell carcinoma (chRCC) and ascertain the minimum number of LNs that need to be pathologically examined to reliably diagnose a patient with node negative chRCC. Methods: From 2004 to 2014, non-metastatic chRCC patients receiving radical nephrectomy together with lymphadenectomy were identified from the Surveillance, Epidemiology and End Results (SEER) database. The primary outcome was overall survival (OS). Results: Two hundred and forty-six patients received lymph node dissection during the surgery. Of the patients, 24 (10%) had pathologically confirmed positive LN. Multivariate Cox regression model showed that positive LN was an independent unfavorable predictor for OS (HR = 2.83, 95% CI = 1.14 – 6.98, P = 0.024). More importantly, LN( − ) patients with at least three LNs dissected had significantly better OS compared with when 1 – 2 LNs were examined ( P = 0.048). Multivariate analysis confirmed that in LN( − ) patients, the examination of three or more LNs could independently predict better OS compared with patients with only 1 – 2 LNs dissected (HR ≥ 3LNs = 0.362, 95% CI = 0.135 – 0.972, P = 0.044). Additionally, the likelihood of finding at least one positive LN was significantly higher on dissection of ≥ 3 LNs compared with examination of 1 – 2 LNs (15% vs 5%, P = 0.018). Decision curve analysis found a better clinical validity of the ‘ 3 LNs examined ’ -based classification compared with the traditional LN( − )/LN( + ) classification. Conclusion: The proportion of positive LNs in chRCC was far from neglectable and LN metastasis could independently predict unfavorable OS. We recommended a minimum of three LNs should be pathologically examined in order to reliably determine node negative.


Introduction
Renal cell carcinoma (RCC) is one of the most common urinary malignancies with clear cell, papillary and chromophobe RCC being the three main histological types (1,2). Among them, chromophobe renal cell carcinoma (chRCC) accounts for~5% of all renal tumors (3), while its incidence could be up to over 10% (4). Compared with clear cell RCC, for many years chRCC did not attract much attention and the prognostic factors that can be useful in precisely guiding clinical decisions and accurately aiding counseling for patients with chRCC had not been fully discovered.
Lymph node (LN) status is embedded into the American Joint Committee on Cancer (AJCC)/International Union against Cancer (UICC) staging system for renal tumor and the pathologic node positivity was associated with a significantly worse overall survival (OS) compared to node-negative [LN(−)] clear cell RCC (5). However, the prognostic value of LN status on patients' prognosis in chRCC remained unclear. Besides, staging lymphadenectomy during radical nephrectomy (RN) for chRCC has not been defined, resulting in only selective lymphadenectomy being performed currently.
In fact, for many other sorts of cancers, studies have focused on identifying a minimum number of LNs that need to be pathologically examined for accurate staging evaluation. For example, a minimum of 12 and 4 LNs was recommended for recognition of reliable 'node negative' in colorectal and adrenocortical carcinoma for pathologists, respectively (6,7). Therefore, using a large cohort of chRCC patients receiving RN together with lymphadenectomy, the aim of the current study was to evaluate the prognostic value of positive LNs [LN(+)] for chRCC patients and to reveal whether the number of LNs examined could play a role in staging and survival of chRCC patients.

Study population
The chRCC patients from 2004 to 2014 were identified from Surveillance, Epidemiology, and End Results (SEER) database. The inclusion criteria were as follows: (1) diagnosed with chRCC (tumor site: kidney) by pathological confirmation (ICD-O-3 code 8317/3, renal cell carcinoma, chromophobe type); (2) non-metastatic (M0) patients; (3) chRCC as the primary and first malignancy; (4) receiving radical nephrectomy (surgery of primary site code 50) together with lymph node dissection; (5) complete or known information of pathologically T stage, pathological differentiation, race, tumor size, laterality, chemotherapy and radiation; (6) known status of lymph node metastases and number of dissected lymph nodes; (7) experienced active follow-up during the study period until the cutoff follow-up date or occurrence of endpoint event. Patients were excluded if (1) they received surgical procedure to distant sites/nodes or other regional sites; (2) the status of LN was assessed by only biopsy; (3) they were recorded to receive neoadjuvant chemotherapy or neoadjuvant radiation. Flow chart of the selection process is listed in Supplementary Figure S1. Our study was granted an exemption from Ethics Review Board in our institution because the SEER Program collects data from population-based cancer registries with anonymous information. Hence, no ethical approval is needed.

Covariates and follow-up information
Baseline factors included demographics (age at diagnosis, race and gender), tumor characteristics (laterality, T stage, pathological differentiation and tumor size) and treatment (adjuvant radiation and adjuvant chemotherapy). Total number of LNs examined and number of positive LNs were used for analyses. Patients were naturally classified into LN(+) and LN(−) groups according to the presence or absence of any positive LNs found on histopathology. The follow-up information including survival months and survival status was collected in our study. The major endpoint of interest in our study was overall mortality. Duration of overall survival (OS) was defined as the time from pathologic diagnosis of chRCC to the date of all-cause death or last contact.

LN cutoff point analysis and evaluation
To ascertain the minimum number of LNs required for adequate sampling, we established binary variables comprising patients with number of LNs examined below and above each sequentially increasing value. Patients' survival was then compared within these binary variables for each cutoff value in the LN(−) group. The cutoff value that led to the maximum of Chi square value and significant difference between the survival curves (set this cutoff value as X) was selected for further assessment (8,9). Next, in the overall cohort, the patients were divided into two groups by that cutoff value X and the likelihood of identifying at least one positive LN was compared between these two groups. To further analyze the clinical applicability of this cutoff value, a new three classificationbased LN status was established: (1) LN(+), defined as patients with positive LNs found; (2) LN(−)&D<X, defined as patients with no positive LN found but less than X LNs were examined; (3) LN(−) &D≥X, defined as patients with at least X LNs examined but no positive LN found. Decision curve analysis (DCA) was introduced to compared the clinical validity of the new three classificationbased LN status and the traditional classification (LN(−) and LN(+)).

Statistical analysis
Continuous variables are reported as median with interquartile range (IQR), and categorical variables are reported as frequency and percentages. Patients characteristics between groups were compared using Student's t test for continuous variables and Chi square for categorical variables. Survival curves were compared using the Kaplan-Meier method with the log-rank test. Multivariate Cox proportional hazards regression analysis was employed to evaluate the prognostic factors, and hazard ratios (HR) along with 95% confidence interval (95% CI) were calculated. The likelihood of identifying positive LN was compared using a two-sided Chi square test.
Two-sided P values of less than 0.05 were considered statistically significant. All statistical analyses were performed using the SPSS version 23.0 (IBM SPSS Statistics, Chicago, IL, USA). Kaplan-Meier survival curves and were plotted by GraphPad Prism version 7.0 (GraphPad Software, San Diego, CA, USA). Decision curve analysis was conducted using R version 3.4.0.

Patient cohort
For all M0 chRCC patients identified from 2004 to 2014, a total of 1717 patients received radical nephrectomy while 1061 patients received partial nephrectomy. Among these patients, only 32 patients (3%) received partial nephrectomy together with lymphadenectomy while 246 patients (14%) received radical nephrectomy combined with lymphadenectomy. These 246 patients were defined as the study cohort of our study according to the inclusion/exclusion criteria and were taken into further analyses. Of the overall study cohort, the median follow-up time was 49 months. Among these, 133 (54%) were male and more than half (n = 139, 57%) were under 55 years old at the diagnosis with a median age of 53 (interquartile range [IQR] 23-91). Median tumor size was 10 (IQR [7][8][9][10][11][12][13][14] cm. The majority (n = 167, 68%) of the tumor were staged as T1/ T2 (local). Besides, the median number of LNs dissected was two (IQR 1-7) per tumor specimen. Only seven patients (3%) were administered postoperative chemotherapy.

Differentiation between LN(+) and LN(−) patients' characteristics
In the study cohort, 222 (90%) and 24 (10%) patients were found pathologically node negative and node positive, respectively. The LN(+) group showed a higher median number of LNs evaluated compared with the LN(−) group (five versus two, P = 0.044) ( Table 1). The positive group was also associated with a higher proportion of advanced tumor (T3/T4) (62% versus 29%, P = 0.001) and poorly differentiated/undifferentiated cancer (71% versus 44%, P = 0.016). The two groups were comparable in terms of age, gender, race, laterality as well as tumor size.

Prognostic factors related to overall survival
In the univariate analysis, overall survival (OS) was significantly unfavorable in LN(+) patients, >55 years patients, patients with higher T stage or worse differentiation and patients receiving chemotherapy ( Table 2). The 5 year OS rate of LN(+) patients decreased to 47% when compared with 88% for the LN(−) group (P < 0.001), which revealed that positive lymph node could significantly lower chRCC patients' prognosis. Besides, the multivariate analysis found that a positive LN was an independent unfavorable prognostic indicator of OS for chRCC patients (Table 2).

Identifying cutoff point of LN
Due to the obviously negative impact of positive LNs on survival, it is important to identify patients that with LN involvement. Hence, we aimed to identify the cutoff point of LNs that were required to accurately determine LN negativity. When analyzing the serially increasing number of LNs dissected in LN(−) patient, we noticed that a threshold of three LNs had the capability to impact OS of patients (Table 3). Those patients defined as LN(−) on assessment of one to two LNs had a 5-year OS of 83%, whereas for those who had three or LNs dissected, the 5-year OS was significantly improved to 93% (1-2 LNs versus 3 or more LNs P = 0.048) (Fig. 1). Multivariate analysis also proved that in LN(−) patients, the assessment of three or more LNs was an independent favorable predictor for OS compared with patients with one to two LNs examined (HR ≥3LNs :0.362, 95% CI 0.135-0.972, P = 0.044, '1-2 LNs' as reference) after adjusted by confounders including age, gender, race, laterality, T stage, pathological differentiation, tumor size and adjuvant treatments. Moreover, in the study cohort, the likelihood of recognizing at least one positive lymph node was significantly higher for chRCC patients with 'at least three LNs examined'

Discussion
Assessment of LN status is crucial for accurate staging for patients with malignant tumors (7,10,11). For chRCC patients, we found that radical nephrectomy was combined with lymph node dissection in only 16% of the patients between 2004 and 2014. In these patients, we revealed that a positive LN status could independently predict unfavorable OS. More importantly, we suggested that the assessment of a minimum of three LNs was necessary to accurately determine node negative chRCC for pathologists. Due to the rarity of chRCC, the prognostic features of this disease are not well defined (12,13). In the present study, we revealed a relatively low risk of aggressive clinical behavior of chRCC, which was in line with previous studies (14,15). The postoperative 5-year survival rate was 83%, which was much higher than that of clear cell RCC (16). As for prognostic factors that were associated with chRCC, some clinicopathological variables such as age, tumor size and T stage were found to be relevant to patients' OS. Volpe et al. also pointed out that T stage and tumor size was associated with prognosis of chRCC. But they declared the potential prognostic value of gender (17), which was confirmed by neither our study nor any other previous study of chRCC (14,18). Besides, no recommendation of the administration of adjuvant chemotherapy or radiation for chRCC has been made in the new EAU guidelines for renal cell carcinoma (19). In the present study, based on the limited cases of  patients receiving postoperative adjuvant treatment, we did not find any survival benefit that radiation could bring to chRCC patient's prognosis. Interestingly, both the univariate and multivariate analyses suggested that adjuvant chemotherapy could actually shorten the survival time of chRCC patients, which is a very unexpected result. To be honest, the reliability of such result was decreased because of the rather small proportion (3%) of patients that received adjuvant chemotherapy. But it could let us reconsider whether it is worth using systemic chemotherapy in this kind of tumor with a low risk of progression, metastasis and death (17). LN involvement is one of the most important prognostic factors in RCC. As early as 1980s, Golimbu et al. had pointed out that lack of invasion of regional lymph nodes was associated with a better prognosis in RCC (20). But most of the previous studies were based on clear cell carcinoma, and the role that positive LN played in chRCC remained unknown. Using a cohort of 102 chRCC patients, Cheville et al. found that LN(+) were significantly associated with death in chomophobe RCC by univariate analysis. To our knowledge, our study was the first study ever to report that positive LN was an independent prognostic factor of OS for chRCC using multivariate analyses. More importantly, we reported a proportion of 10% patients with positive LNs in the chRCC patients that received RN with lymph node dissection. Previous study pointed out that lymph node involvement was rare in chRCC (21)(22)(23). Our study comprehensively proved that the lymph node metastasis of chRCC patients was far from neglectable and could significantly lower the survival rate of patients.
The above-mentioned importance of LN status let us to consider how to accurately differentiate LN(+) patients from LN(−) patients. To be strict, it's hard to define a 'true N0' because almost no patient would undergo a thorough pathologic examination of all regional lymph nodes in clinical practice. Therefore, it is important in any cancer to assess the number of LNs that minimizes the risk of false negatives while accurately capturing all true positives (7). In our study, we found that LN(−) patients who had one to two LNs assessed had a significantly worse OS than those who had at least three LNs examined. And in LN(−) patients, '1-2 LNs examined' could independently predict worse OS after controlling for other factors that may influence outcomes. The survival weakness of '1-2 LNs' group could be potentially explained by the inadequate examination of LNs. This could lead to the failure to harvest and recognize one or more relevant positive LNs. This phenomenon suggested that some of those tumors designated as pN0 were in reality pN+ and is in line with the concept of 'stage migration' (24). In practice, suboptimal LN dissection might be a nidus for future recurrences/ metastasis (6). Additionally, we found that the likelihood of

946
Three LNs examined for pN0 diagnosing in chRCC identifying at least one positive LN significantly enhanced when three or more LNs were examined, which further verified our theory. Besides, we should notice that high staging accuracy does not necessarily mean better usefulness in clinical practice. Therefore, we introduced DCA, which was widely used by previous studies (22,25), to assess the clinical validity. The results showed that the new three classification-based LN status (LN(+), LN(−)&D<3, LN (−)&D≥3) showed enhanced clinical validity because enhanced net benefits were observed when threshold probabilities was less than 0.3, with net benefits of other region similar.
In fact, the role of lymphadenectomy in renal cell carcinoma has always been debatable, regardless of the pathological types. According to a recent study which included over 1400 patients with renal clear cell carcinoma (ccRCC), lymphadenectomy could not bring survival benefit (26), while in another study with over 3500 ccRCC patients, lymphadenectomy was shown to improve patient's cancer specific survival (27). For chRCC, however, few study focused on the therapeutic role of lymphadenectomy. Also, as for the extent of lymph node dissection, although controversial, great progress has been made in the studies on ccRCC and the LNs from the ipsilateral great vessel and the interaortocaval region was recommended to be removed from the crus of the diaphragm to the common iliac artery for high-risk ccRCC patients (28). Again, similar study in chRCC lagged far behind that in ccRCC. To our knowledge, our study is the first on to recommend an extent or number of lymph node dissection for chRCC patients, which we hope to narrow the gap between studies on ccRCC and chRCC Although this is the first study to analyze the impact of a minimum number of lymph node examined on chRCC staging, some inevitable limitations should be acknowledged. First, the SEER registry did not offer information of recurrence status, therefore diseasefree survival could not be calculated. Second, the recently updated SEER dataset used 'No/Unknow' as one inseparable group in terms of chemotherapy and radiation, which referred to patients with no evidence of chemotherapy or radiation in the medical records. Hence, we used group named 'no evidence' rather than 'no' as the counterpart of 'Administered'. Third, information about which LN basins were included in each nodal assessment is not captured by the SEER dataset. Fourth, there might be some bias difficult to avoid due to the nature of retrospective studies, thus prospective studies are needed for further verification of our conclusions.

Conclusion
This study proved that the proportion of patients with LNs involvement could not be neglected in chRCC patients received RN. Among the patients who had RN, pathologically positive LN was an independent unfavorable predictor of OS. Above all, at least three LNs should be examined per specimen to accurately diagnose a patient with pathologically node negative chRCC.

Supplementary data
Supplementary data are available at Japanese Journal of Clinical Oncology online.