EURURO Vol. 71 No. 4

671 692

4. Discussion
We report an update of the SSIGN score using the original
RN cohort with longer follow-up and in new cohorts of
contemporarily managed patients treated with RN and PN.
We found that the SSIGN score retained its predictive
accuracy in the originally characterized patients with longer
follow-up and after accounting for competing risks of non-
RCC death. We also confirmed that despite significant
differences in clinical features and an era-specific improve-
ment in survival, that for contemporary ccRCC patients, the
SSIGN score is a valid predictive tool for death from RCC.
Our results are similar to the prior experience with the
SSIGN score. In the original description, the C-index assessing
the ability of SSIGN score to predict death fromRCCwas 0.84,
similar to the findings for all three cohorts described here
despite differences in clinical features In addition, our
findings confirm those of the subsequent validations of the
SSIGN score Notably, the third and largest validation of
the SSIGN score was performed using routine pathology and
found similar findings as our results
Although our findings are corroborated by prior valida-
tions, several important differences should be considered.
First, the original cohort has matured such that 39% of
patients were alive at 10 yr of follow-up compared with only
26% in the original description and 14.4–29.3% in the
validation cohorts Second, we included a separate
analysis of patients managed with PN, as opposed to
considering them with RN patients, and confirmed the
predictive capacity of the SSIGN score for these patients.
Certainly the SSIGN score validation studies included
nephron-sparing surgery; however, they represented the
minority of cases, and the performance of the SSIGN score for
those patients was not considered separately . Our
subset assessment of the SSIGN score in PN patients is
underscored by the association of PN with superior survival
in two validation studies . Finally, we included an
analysis of the competing risk of non-RCC death, which given
the long postsurgical survival of most RCC patients is an
important analysis that has to our knowledge not been
reported. The results of these competing risk analysesmodify
the estimated CSS, adjusting for an overestimation of the risk
of RCC death, and therefore may serve to improve patient
counseling.
We found that after controlling for SSIGN score, survival
was improved in the contemporary RN patients compared
with the originally described cohort. Although there has been
an established stage migration in RCC which could
explain any apparent survival improvement for all patients
considered together, our finding of improved survival among
contemporary patients when controlling for SSIGN score and
among M0 and M1 patients suggests that stage migration
alonemay not explain this result. Our resultsmay be partially
explained by era of treatment, given the findings from
Surveillance, Epidemiology, and End Results analyses,
whereby treatment in the targeted-therapy era was inde-
pendently associated with improved survival after control-
ling for stage Notably, when PROG scores were
assigned to the M0 original and contemporary RN patients,
era of treatment was not associated with a reduced risk of
metastasis during follow-up (Supplementary Table 2). Taken
together with our finding that contemporary management
was associated with a reduction in the risk of death fromRCC
among M0 patients after controlling for SSIGN, this would
suggest that improvements in survival may be a function of
patient treatment after developing metastatic disease.
As previously mentioned, the SSIGN score is increasingly
being utilized to assess the effect of new biomarkers on
survival in RCC. Hakimi and colleagues evaluated the effect of
the MET rs11762213 single nucleotide polymorphism in
272 patients in The Cancer Genome Atlas, and after adjusting
for SSIGN score, they found an increased risk of death from
RCC . In addition, BAP1, a deubiquitinating enzyme
whose loss is associated with inferior CSS survival ,
was further characterized at our institution. Using the
SSIGN score to adjust for clinical features, BAP1 was found
only to be associated with inferior survival when restricted
to lower risk patients (SSIGN score 3) Conversely,
the BioScore, a composite of B7-H1, survivin, and Ki-67
expression, was found to be informative toward survival
only in patients with intermediate and high SSIGN scores
(>3) . Together, these findings demonstrate that the
SSIGN score is not only useful as a continuous adjustment,
but also for revealing the complex interaction between
biomarkers and the clinical features associated with
survival. Therefore, the continued evaluation and valida-
tion of the SSIGN score is of utmost importance in assuring
that these biomarker studies, and other ongoing research,
retain their clinical significance.
Despite the findings reported by these biomarker
studies, the predictive capacity of existing models such as
SSIGN must be considered. Specifically, the SSIGN score is
often grouped into three categories, which results in a loss
of granularity as demonstrated by a reduction in the
reported C-index for SSIGN alone, and thus an apparent
incremental gain in the C-index when combined with new
biomarkers , which may not exceed that of the SSIGN
when used as described.
As previously outlined, we noted an era-specific improve-
ment in survival. Despite this finding, the SSIGN score
retained the predictive capacity for RCC-specific death with
excellent discrimination (C-index 0.84 and 0.82 in contem-
porary RN and PN patients, respectively) as originally
described and using the originally assigned point values.
Taken together with our desire not to invalidate existing
reports utilizing the SSIGN score and the excellent perfor-
mance of the model as originally described, we did not
attempt to incorporate new variables or adjust the assigned
point values.
Certain limitations with this work warrant further
discussion. First, these results are obtained from a retrospec-
tive review of our Nephrectomy Registry and are subject to
the many inherent biases associated with this approach.
Although outcomes in our registry are collected prospec-
tively, it is likely that a small number of recurrences or
RCC-related deaths were not captured. In addition, this
investigation includes patients managed 40 yr ago,
and significant advances have been made in imaging,
E U R O P E A N U R O L O G Y 7 1 ( 2 0 1 7 ) 6 6 5 – 6 7 3 671