Preoperative Decision Making for Nephron-Sparing Procedure in the Renal Mass: Time for Using Standard Tools?

Introduction

Nephron-sparing surgery (NSS) has become the standard of care for cT1a and when feasible for cT1b renal masses. ^{1 –4} Convincing data show equivalence of partial nephrectomy (PN) and radical nephrectomy (RN) for cancer control and the superiority of PN in terms of functional outcomes, cardiac morbidity, mortality rate, and overall survival (OS). ^1,2,5 Currently, nephron-sparing procedures (NSPs) include PN, enucleation, enucleoresection, and ablation. ³ Once the discussion about the feasibility of sparing renal parenchyma is settled, the next step is to tailor the type of NSPs to the individual patient with a given renal mass. Typically, decision-making on interventional options has been an individual process, likely driven by patient's age and comorbidity, tumor characteristics, and physician preferences. The literature is rich in observational and case–control studies comparing NSP approaches. ^6,7 Although anatomic differences are inferred between ablated and excised renal masses, standard tools for anatomic characterization of renal tumors have not been described until recently. ^8,9 Furthermore, comorbidity has been erratically assessed without standard indexes allowing for comparison among studies. ¹⁰

Because of its recent incorporation, tumor complexity scores have been mostly applied retrospectively and strong recommendations on which comorbidity classifications should be used are still lacking. Waiting for contemporary series incorporating prospectively these standard tools, it seems logical to test their influence in the choice of NSP.

The objective of the present work is to assess the application of standard measures in the decision-making process between two different NSP options in the treatment of renal masses (cryoablation [CA] and NSS).

Patients and Methods

Results

During the study period, 251 patients underwent a NSP (261 tumors). In total, 124 cases underwent NSS (81 cases open and 43 cases laparoscopic) and 137 cases underwent CA (121 cases laparoscopic and 16 cases percutaneous). Table 1 shows baseline patient and tumor characteristics. Mean (standard deviation [SD]) and median follow-up time was 44 (28.8) and 43 (0–108) months, respectively, for patients who underwent CA. For patients who underwent NSS, the mean (SD) and median follow-up time was 33 (29.9) and 25 (0–125) months. A significant difference was seen between the two groups for age, ASA score, solitary kidney, CCI (median), CACI, CCI groups, cMTD, cTNM stage, PADUA and RENAL scores, and PADUA and RENAL complexity groups. In the NSS group, 21 (16.9%) masses were benign. In the CA group, 22 (16.1%) showed benign tumor pathology following biopsy, while 32 (23.4%) were nondiagnostic, and in 4 initial cases, biopsy was not performed (2.9%).

Univariate regression analysis showed age (p < 0.001), ASA score (p = 0.027), CACI (p < 0.001), CCI groups (p = 0.04), cMTD (p < 0.001), PADUA (p < 0.001) and RENAL scores (p = 0.001), and PADUA (p < 0.001) and RENAL (p = 0.001) high-complexity groups to be predictors of the type of NSP choice (data not shown).

Different multivariate models were constructed avoiding variables that were correlated or included within the scoring systems, for example, cMTD and anatomic scores or age and CACI (Table 2). CCI was nonsignificant when included in models 4 and 7 (p 0.532 and 0.509, respectively), likely overpowered by age and thus was not included in the final models to avoid confounding. Multivariate regression analysis demonstrated the odds of undergoing NSS vs CA for the different models (Table 2).

Model 1 with our two primary variables of age and cMTD exhibited the highest discriminative ability (AUC = 0.79), showing that for each cm increase in cMTD there was a 2.11 higher odds of receiving NSS. Conversely, per each year of increase in age, the odds decreased 0.92-fold. Four models incorporating standardized measures of tumor complexity or comorbidity or both showed similar and nonstatistically different discriminative ability from model 1. Model 8, incorporating the quantitative assessment of comorbidity weighted by age, had an AUC of 0.75 with similar increasing cMTD odds of receiving NSS and a more realistic odds decrease for each point increase of CACI (0.69). The discriminative ability of models 4, 7, and 9 incorporating anatomic scores was 0.73 and nonstatistically significantly different from models 1 and 8. Figure 1 depicts the discriminative ability of those models with the best cutoff combination on treatment prediction with AUC >0.72 (models 1, 4, 7, 8, and 9). The calibration slope for these models (shrinkage factor) ranged from 0.95 to 0.99, indicating a little overoptimism.

OPEN IN VIEWER

FIG. 1.

Comparison of the performance and discrimination of different models using receiver operating curve (ROC). Only models with area under the curve (AUC) >0.72 are depicted (not bootstrap corrected). Model 1 includes clinical maximal tumor diameter (cMTD) and age; Model 4 includes RENAL score, age; Model 7 includes PADUA score, age; Model 8 includes cMTD and CACI; Model 9 includes PADUA groups and CACI.

The predictive ability of models 3, 5, and 6 was statistically significantly lower (p = 0.001, p = 0.038, p = 0.001) compared to model 1.

The 5-year OS probability for both NSS and CA groups was not significantly different (91% for NSS and 79% for CA, p = 0.058) (Fig. 2). The 5-year CSS, DFS, and RFS were not significantly different between the two groups (Fig. 2).

OPEN IN VIEWER

FIG. 2.

(a) Five-year overall survival (OS) of patients who underwent partial nephrectomy/enucleoresection (nephron-sparing surgery [NSS]) vs cryoablation (CA) (p = 0.058). (b) Five-year cancer-specific survival (CSS) of patients who underwent partial nephrectomy/enucleoresection (NSS) vs CA (p = 0.092). (c) Five-year disease-free survival (DFS) of patients who underwent partial nephrectomy/enucleoresection (NSS) vs CA (p = 0.718). (d) Five-year recurrence-free survival (RFS) of patients who underwent partial nephrectomy/enucleoresection (NSS) vs CA (p = 0.072). Color images available online at www.libertpub.com/end

Discussion

In our cohort, multivariate regression models show that incorporation of standard tools such as PADUA and RENAL anatomic characterization and CACI could predict the type of NSP with similar accuracy as a simplistic model based on age and cMTD. Especially a model combining cMTD and age-adjusted comorbidity offers more realistic odds of receiving NSS over CA than when age is used as a variable.

Despite the abundant literature on the subject, the choice between ablation and NSS is still ill defined, but for tumor size limits, and subject to multiple influences such as physician and patient preferences ¹⁴ or to a certain extent of comorbidity subjective assessment. To minimize the influence of surgeon's technical preferences, we chose a time frame when NSS and ablation were available at our center and when indications for one or another NSP were discussed in a structured multidisciplinary manner. CCIs were chosen to standardize comorbidity due to their widespread use and because they have been validated for surgical procedures. ¹⁵ Anatomic scores for tumor characterization were chosen based on their functionality, external validations, and reproducibility, all of them justifying its clinical use. ^{8,16 –23}

As expected, patient's and tumor's characteristics treated by NSS or CA differed significantly reflecting our primary selection process. The differences persisted when standard criteria were applied to patient and tumor (anatomic) characteristics. Our data support the current literature, indicating that less complex anatomic tumors and patients with a higher comorbidity load are selected for ablation. ²⁴

Quantitative comorbidity was significantly different only when adjusted for age (CACI) or when CCI was categorically divided. The differences in comorbidity observed between the two treatment groups when standard metrics are applied support our primary qualitative assessment and the known correlation between age and comorbidity. Age and comorbidity scores are strongly correlated with patient mortality rate and, therefore, predictive of nonkidney cancer death. ^{25 –27} The lack of significant difference in 5-year OS between the two subcohorts supports a choice based on age and cMTD, otherwise fully replicated by a combination of cMTD and CACI and RENAL or PADUA scores combined with age.

Once proven that standard metric tools reproduce the classic selection criteria, the logical step is to test the ability of different models combining standardized classification tools. To answer this question, the significant variables were combined in different predictive models and their performance compared with the model that incorporated the classic selection variables along the study period (model 1).

Age, CACI, tumor size, PADUA and RENAL score, and complexity groups appeared as the most powerful variables in univariate analysis. When creating the different predictive models, it was taken into account to avoid variables that were already part of a scoring system (e.g., cMTD and anatomic scores or age and CACI). When comparing performance and discriminative ability of the different models, models 1, 4, 7, 8, and 9 showed the best cutoff combination on treatment prediction with AUC >0.72 (Fig. 1). These five models that included the most intuitive one (model 1, combining two clean nonstandardized variables as cMTD and age) to a model combining two scores (PADUA groups with CACI) exhibited a similar discriminative ability, showing that our intuitive choice is fully reproducible when applying standard tools with a nonsignificant decrease in accuracy. When incorporating anatomic complexity groups to the models (9 and 10), no differences were found in the NSS predictive odds for lower and intermediate groups. Consequently, the low morphometric complexity group was taken as the reference. In these models, the balance was clearly in favor of NSS for those complex renal masses.

An important question is the subjective appraisal of the comorbidity. Overall crude rates of different comorbid conditions are different between patients treated by NSS or by ablation. ²⁸ Whether they reflect the general condition depends on the degree of severity of the comorbid conditions and this is probably more weighted by means of a score than of a percentage. The reason why CACI had a predictive value by opposition to CCI might be due to the obvious difference in the age of both groups or overlapping scores, however, when groups were assigned, this difference became obvious (Table 1). Comorbidity increases with age, the latter being per se an important predictor of risk of death. Especially when follow-up periods longer than 1 year are expected, the weighted age comorbidity index might reflect better the real situation of a given patient than the CCI. ¹⁵

Our study is not devoid of limitations. It can be argued on the size and composition of the cohort. The study period was chosen based on the possibility of choice between both NSP modalities, and the relative abundance of CA in the cohort responds to the fact that we are center of reference for this technology. Although clinical tumor size and age played an important role in our treatment decision-making, the other tumor (e.g., location, exophytic percentage, relationship with collecting system, or hilar component) or patient factors influenced our treatment decision. The lack of 5-year differences in OS and oncologic outcomes in our cohort supports the appropriateness of the treatment choice, although a proxy of 5-year estimation may be insufficient to establish definitive conclusions in terms of survival outcomes in T1 mass. AUCs <0.8 for our primary and standardized models reflect that variations in treatment choice exists without compromise survival outcomes.

Although in our setting the application of a standardized model has proven to be of the same value as cMTD and age, a work remains to determine if the models will be suitable to establish a threshold in tumor complexity and patient comorbidity that could be included in a treatment choice algorithm. This will be better tested in prospective studies, where perioperative outcomes are reliably collected. Finally, we should not generalize assuming that our standardized models fit other practices and settings. Therefore, validation in different centers is recommended.

Conclusions

Standardized models reproduce the intuitive choice of NSP type for localized renal masses. In our setting, the models 4, 7, 8, and 9 using standard tools such as RENAL, PADUA, and CACI are of value for future standardization on the choice between NSS and CA. Although using validated tools might be a consuming task, it overcomes selection biases, provides the basis for comparison, and if positive external validation, they should be used in the decision-making process.