What's known on the subject? and What does the study add?
Treatment options for small renal masses include radical nephrectomy (RN), and nephron sparing modalities (NSM) such as partial nephrectomy (PN), and thermal ablation (Cryo- and radiofrequency ablation, C/RFA). Prior studies had demonstrated gross underutilization of PN; however overall treatment trends for C/RFA had not been well studied using a population-based cohort. In this study, which examined management trends of localized renal masses in the USA, we identified an increased prevalence of RN, PN and C/RFA over the study period, with PN increasing the most rapidly, and with RN continuing to account for the vast majority of procedures.
This is the first study to examine surgical management of renal masses in patients with non-dialysis dependent chronic renal insufficiency. Although nephron sparing modalities were increasingly utilized over the study period, it is particularly concerning that patients with pre-existing non-dialysis dependent chronic renal insufficiency are receiving less nephron sparing approaches. Further investigations are required to confirm these findings and to identify impediments to the dissemination of nephron sparing modalities.
To evaluate the diffusion of nephron-sparing modalities (NSM) for the treatment of renal neoplasms in the USA over the last decade and to identify the factors associated with renal procedure selection.
Patients and Methods
The Nationwide Inpatient Sample was utlized to identify patients undergoing cryo/radiofrequency ablation (C/RFA), radical nephrectomy (RN) and partial nephrectomy (PN) from 1998 to 2008.
Annual trends in procedure prevalence were determined.
Multivariate analyses were performed to query the influence of age, race, sex and comorbid disease on surgery selection.
We identified 443 853 procedures performed during the study period: 25 599 C/RFA, 79 568 PN and 338 687 RN.
The prevalence per 100 000 hospital admissions in 1998 was 3.7 for C/RFA, nine for PN and 87.1 for RN. All procedures increased over the study period, by 1.05, 3.1 and 2.2/100 000 admissions per year, respectively (all P < 0.001).
Diabetes, urban, teaching and large capacity hospitals were associated with NSM (either C/RFA or PN) compared to RN (all P ≤ 0.011). Age ≥70 years, female, hypertension, diabetes, chronic kidney disease (CKD) and region outside the Northeast favoured C/RFA over PN (all P ≤ 0.026).
Compared to those without CKD, patients with CKD had an almost twofold higher probability of undergoing RN than NSM (odds ratio, 1.88; 95% confidence interval, 1.7–2.1). Despite increasing NSM utilization over the study period, most patients with CKD still received RN.
Although the prevalence of NSM is increasing, RN is more common.
The low utilization of NSM in patients with pre-existing CKD warrants further investigation.
A paradigm shift is underway with respect to the treatment approach employed for renal masses, leading the AUA to release guidelines for the management of clinical stage 1 renal masses in 2009 for the first time . On the one hand, the epidemiology of renal masses is changing . Not only is the incidence of RCC increasing overall , but also there is a stage migration towards the discovery of smaller, incidental lesions , probably as a result of an increased utilization of abdominal imaging . On the other hand, there is accumulating evidence that surgically-induced chronic kidney disease (CKD) [6, 7] may increase the risk of cardiovascular  and metabolic sequelae [9, 10], and that the risk of developing Stage IV CKD is significantly increased with radical nephrectomy (RN) compared to nephron-sparing modalities (NSM) [6-11]. Although partial nephrectomy shows oncological outcomes equivalent to RN , it does so at the cost of at least a higher procedure specific complication rate [13-15]. Ablative techniques have limited long-term oncological data supporting their use . Thus, the surgical management of a renal mass is an increasingly complex decision, factoring the risks of potential malignancy, as well as metabolic and surgical complications, in the context of an individual patient's co-morbid conditions.
The present study aimed to examine trends in the utilization of extirpative and ablative treatments for localized renal masses, and to analyze factors associated with their utilization in the USA from 1998 to 2008 using the Nationwide Inpatient Sample (NIS) .
Materials and Methods
The NIS is the largest database of inpatient care in the USA, encompassing almost eight million (20%) of all annual hospital admissions. It can be weighted to estimate the national incidence . Analyses using databases with de-identified patient data such as the NIS are exempted from individual review by our institutional review board under an umbrella provision. We queried the NIS, identifying all hospital admissions in patients aged >18 years with a primary or secondary procedure code for cryoablation/radiofrequency ablation (C/RFA) (International Classification of Diseases, ICD-9 55.3, 55.32, 55.33, 55.34, 55.35 and 55.39), partial nephrectomy (PN) (55.4) or RN (55.5, 55.51, 55.52 and 55.54) between 1998 and 2008. The ICD-9 procedure code for nephrectomy does not distinguish between radical and simple operations. Consequently, we excluded all patients with a concomitant ICD-9 diagnosis code for kidney donor (V59.4), as well as infectious aetiologies including acute and chronic pyelonephritides (590.0, 590.00, 590.01, 590.1, 590.10 and 590.11) and renal/perinephric abscess (590.2), because these diagnoses may represent the indication for performing the procedure. We also excluded those patients with a diagnosis of renal pelvic or ureteral tumour (189.1 or 189.2) because nephrectomy and nephroureterectomy utilize the same ICD-9 procedure code. The extent of disease affects surgical selection; patients with locally advanced, locoregional or metastatic disease had a higher probability of undergoing RN. The NIS does not contain disease-specific staging, and thus we excluded patients undergoing concomitant surgeries suggestive of advanced disease, including splenectomy (41.4, 41.42, 41.43), liver resection (50.2, 50.21–26, 50.29, 50.3), pancreas resection (52.5, 52.51–53, 52.59, 52.6, 52.7), bowel or colon resection (45.5, 45.50–52, 45.61–62, 45.7, 45.71, 45.73–76, 45.79, 45.8, 45.81–83), or thrombectomy with vascular reconstruction (37.10, 38.05, 38.07, 38.45, 38.47, 38.65, 38.67, 38.75, 38.77, 38.87, 39.6, 39.61, 39.63, 39.66), aiming to reduce the potential for bias by burden of disease. We also excluded autosomal dominant polycystic kidney disease (753.12, 753.13, 753.14) because the diagnosis may affect preoperative renal function, as well as surgical indication and selection. Co-morbid conditions assessed included hypertension (Dx CCS 98, 99), diabetes (Dx CCS 49, 50) and obesity (ICD-9 278.0, 278.01, 278.00). Because the prevalence of ICD-9 coded obesity was far below the national prevalence of the condition (6.4% in our cohort compared to 30.9% of adults aged 20–74 years in the National Health and Nutrition Examination Survey 1999–2000) , we excluded obesity from the analysis.
A subgroup analysis of patients with a discharge ICD-9 diagnosis of chronic renal insufficiency (585x) was performed. Beginning in 2005, specific ICD-9 codes classified CKD based on stage. We divided patients with a specified CKD stage into those with end-stage renal failure (ESRF) requiring dialysis (CKD stage 5+, ICD-9 585.5 and 585.6) and those with milder renal failure (CKD stages 1–4, ICD-9 585.1–585.4).
The 20% NIS sample was weighted to estimate all national inpatient stays and used for all calculations. Rates of each treatment over time were calculated using the total number of inpatients in each year, and simple linear regression models approximated the change in procedure incidence over time. Demographics, as well as clinical and hospital characteristics, were compared between groups using chi-squared (Rao–Scott second-order correction), ANOVA and Student's t-tests (using Bonferroni correction for intergroup comparisons). We performed multivariate analysis using binary logistic regression models to examine the influence of age, race, sex, co-morbid disease and hospital characteristics on surgery selection; only variables that reached statistical significance in the multivariate model were ultimately included in the final models. SVY coding in STATA, version 11.1 (StataCorp, College Station, TX, USA) was used to account for the NIS sampling methodology, and the probability of a type I error was defined a priori as α = 0.05.
We identified 613 723 procedures performed during the study period of which 443 853 met the inclusion criteria: 25 599 C/RFA, 79 568 PN and 338 687 RN. The incidence per 100 000 hospital admissions in 1998 was 3.7 for C/RFA, 9.0 for PN and 87.1 for RN. All procedures increased over the study period, by 1.05, 3.1 and 2.2/100 000 admissions per year, respectively (all P < 0.001). Figure 1 shows the change in procedure incidence annually over the study period. All procedures increased significantly (all P < 0.001), with NSM (either C/RFA or PN) increasing at a faster rate than RN (1424 cases per year compared to 1142 per year). RN still represents most of the surgical cases performed (76% overall; 67% in 2008).
Table 1 details the clinical and demographic factors associated with the three procedures. On multivariate analysis, the patient factors most associated with the utilization of NSM were younger age (<70 years; odds ratio, OR, 1.08; P < 0.001), diabetes (OR, 1.13; P = 0.006) and normal renal function (OR, 0.53; P < 0.001) (Table 2). There was no statistically significant difference in the probabilty of undergoing NSM as a result of race, sex or hypertension (P > 0.05) and, as such, these factors were not included in the final model. Patient factors favouring C/RFA over PN included age ≥70 years (OR, 2.15; P < 0.001), CKD (OR, 1.28; P = 0.012), diabetes (OR, 1.26; P = 0.007), female sex (OR, 1.12; P = 0.003) and hypertension (OR, 1.12; P = 0.003) (Table 3). Age, female sex, hypertension, diabetes mellitus and CKD also impacted the utilization of C/RFA over RN (P < 0.004).
Table 1. Demographics, clinical and hospitalization characteristics
Cryotherapy/RFA (n = 25 599)
Partial nephrectomy (n = 79 568)
Radical nephrectomy (n = 338 687)
P value by chi-squared, anova (Bonferroni correction). *Statistically significant (two-tailed). IQR, interquartile range; RFA, radiofrequency ablation.
The variable most strongly associated with RN vs NSM was a diagnosis of CKD (OR, 1.88; P < 0.001). In total, 19 333 (4.4% of the total cohort) patients carried this diagnosis; of these, 10 034 (51.9%) had milder renal failure (CKD stages 1–4) and 6185 (32%) had ESRF requiring dialysis (CKD stage 5+). The remaining 3114 patients had CKD of an unclassified stage. A series of subset analyses using binary logistic regression were performed to test the association of CKD with the use of RN in different subpopulations. The multivariate model comparing RN use (dependent variable) in CKD vs non-CKD patients was adjusted for other (independent) demographic, clinical and hospital variables, including age, diabetes mellitus, hypertension, hospital region, hospital location, teaching/non-teaching, urban/rural and size of hospital (bedsize) (Fig. 2). In all subgroups, patients with CKD had a higher probability of undergoing RN. Figure 3 shows the evolution in proportion of each surgical procedure performed over the course of the study period in patients with CKD. Those patients with ESRF had a significantly higher probability of undergoing RN than those without renal insufficiency (OR, 9.1; 95% CI, 6.8–12.3; P < 0.001). The cohort of patients with milder renal insufficiency had a probability of undergoing RN similar to that for patients with a normal glomerular filtration rate (OR, 1.02; 95% CI, 0.9–1.13; P = 0.775). On multivariate analysis, no factor was predictive of undergoing NSM in dialysis independent CKD patients compared to patients with normal renal function (all P > 0.05).
Hospital and regional factors also impacted procedure selection. The largest fraction of all renal surgeries occurred in the South. NSM was more common in the Northeast compared with all regions (OR 1.18–1.21; P < 0.022), in teaching compared to non-teaching hospitals (OR 1.52; P < 0.001), in large hospitals compared to small hospitals (OR 1.24; P < 0.011) and in urban rather than rural areas (OR 1.3; P < 0.001).
The data reported in the present study show an increase in the prevalence of surgical procedures for renal masses, consistent with previous observations suggesting that the incidence of RCC is increasing . Changes in the proportion of RN compared to other procedures also suggest an increasing awareness of the importance of nephron preservation in the management of renal masses. The higher OR with respect to undergoing NSM in patients with hypertension and diabetes, which are the two most common causes of CKD in adults , implies the need for a considered application of these approaches in those patients who are at risk of glomerular filtration rate deterioration.
Nonetheless, the data obtained in the present study supplement previous studies suggesting that nephron-sparing approaches are still relatively underutilized [20-23]. Kane et al.  analyzed changes in RCC tumour stage in the National Cancer Database that substantially overlapped the period employed in the present study . Stage 1 tumours increased from 43% to 57.1%, and the mean tumour size in stage 1 tumours decreased from 4.1 to 3.6 cm between 1993 and 2004 . Additionally, because the probability of a diagnosis of RCC is directly proportional to the size of a mass on imaging , the present study underestimates the true proportion of small renal masses overall because it excludes benign histology more common at smaller sizes. Most of the renal masses currently identified therefore fall into the category of T1a lesions, many of which are amenable to a nephron-sparing approach. Recent analyses of the Surveillance, Epidemiology and End Results programme indicate that only 35.2% of T1a renal masses had a PN between 1999 and 2006 , and only 50% of renal masses <2 cm underwent PN and 48% of renal masses 2–4 cm underwent RN in 2008 . Furthermore, there is increasing evidence that appropriately selected T1b and even T2 tumours can be efficaciously treated with PN . In 2008, however, only 33% of patients underwent NSM in the present study despite attempts to eliminate competing diagnoses and advanced stage disease in our analysis.
Not all renal lesions are amenable to NSM. There are multiple tumour characteristics other than size that factor in the technical challenges of NSM, including location, relation to hilar and collecting system structures, and the degree of endophycity. Patient factors such as performance status, medical co-morbidity, previous surgery, social support, patient risk aversion and preference are also important considerations in treatment selection. The recent AUA guidelines on T1 masses are not dogmatic, and they recognize the complexity of all these factors in clinical decision-making in any particular case . It is precisely for this reason that large population-based studies such as the present one are important for determining whether we are applying renal surgery judiciously, and why some propose a standardized classification of renal lesions that can be used to measure prospective outcomes, such as the R.E.N.A.L. nephrometry score . Unfortunately, these tumour characteristics that are critical for surgical decision-making are not captured by the NIS.
The finding in the present study that patients with CKD had a higher probability of undergoing RN is troubling in this regard. We expected to see the preferential selection of NSM for patients with CKD because, by definition, they have a limited renal reserve; however, the factor with the strongest association with RN was CKD (OR, 1.88; P < 0.001). A recent analysis confirmed inferior renal functional outcomes in patients with CKD undergoing extirpative renal surgery that was directly proportional to decreasing GFR . Patients with an enhanced GFR of 45–59 had an adjusted hazard ratio of 4.20 and those with an enhanced GFR of 0–44 had an adjusted hazard ratio of 8.85 with respect to adverse renal outcome defined as a decrement in enhanced GFR to below 30, a decrement in enhanced GFR of ≥4 mL/min per 1.73 m2 per year or the initiation of dialysis . It has not been conclusively shown that medical and surgically-induced CKD share the same pathophysiology, nor whether they confer the same medical risks, although there is increasing evidence to support this [6-8, 11]. The degree of CKD cannot be assessed in many datapoints in the NIS during the period of the present study because the ICD-9 coding system used did not discriminate by stage until 2005. There was a difference in the prevalence of CKD coding by year, with most patients with reported CKD ICD-9 codes occurring at the end of the present study. The temporal trends in the application of renal surgery reported in this population should therefore be approached with caution. A statistically significant association of CKD with RN, however, was found in each year over the study period (all P ≤ 0.001) (Fig. 3). The finding that RN is more prevalent than PN is unexpected, although there are a number of reasons why it should not be considered as a statistical aberration. CKD (renal insufficiency lasting >3 months) has a set of ICD-9 diagnosis codes separate from those for acute kidney injury (ICD-9 code 584). Consequently, the CKD diagnoses in the NIS reflect preoperative renal function, with the exception of coding errors and the rare patient with an extended hospital stay over the course of months. The rate of CKD in patients undergoing renal surgery in the NIS was 4.4%, which is a similar rate to the 4.3% prevalence of recognized biochemical CKD (stages I–V) reported in Medicare patients in 2008 . This similarity suggests that there is no coding bias favouring the reporting of dialysis-dependent CKD over CKD stages 1–4. (Patients with dialysis-dependent CKD obtain less benefit from NSM than those without and so more frequently undergo RN; a finding corroborated in the present study). Most concerning, however, is the subset of patients with specific ICD-9 codes for dialysis-independent CKD, who showed a similar rate of NSM compared to controls despite their increased risk of renal deterioration (OR, 1.02; P = 0.775). There are inherent limitations to the database, which prevent further delineation of the complex medical, social or tumour-related subtleties that may represent hidden confounding factors or biases. Further investigations are necessary to verify this finding and identify patient, practitioner and system factors contributing to this apparent deviation from our guidelines for care .
Other studies have reported inequities in the application of nephron-sparing approaches. For example, Patel et al.  reported a similar analysis based on the NIS, suggesting that patients from zip codes with a lower median income received less PN. They also showed significant regional variation in the application of PN with greater utilization in the Northeast . However, they did not analyze ablative techniques. The regional disparity in PN that they noted is partially offset by the increased utilization of C/RFA identified in other areas of the country within the present study (OR of C/RFA compared to PN, 1.38–1.59; P ≤ 0.011 comparing other regions to the Northeast).
As a longitudinal view into the nationwide utilization of renal procedures based on ICD-9 diagnosis and procedure codes, the present study has the advantage of reflecting population-based trends, although there are also well-established limitations to studies based on medical coding [29, 30]. The large patient numbers in the database mean that even very small absolute differences are statistically significant (Table 1). Analysis of overall survival, oncological outcomes and surgical complications are beyond the scope of the dataset, which only captures patient data coded within a particular hospital admission. For similar reasons, the NIS does not include information on histological diagnosis, nephrometry score or stage, which prevents any subgroup analysis that may be possible through other cancer registries. For this reason, we used extremely strict inclusion criteria to tailor the present study cohort as closely as possible to patients with clinically localized, solid renal masses. As a result of these strict criteria, 27.7% of identified cases were excluded from the analyses, many of whom may have had small renal masses. Of these excluded cases, however, 95.2% underwent RN, and so the overapplication of RN shown by the present study is probably an underestimate.
The database has the advantage of capturing all renal surgeries, however, and therefore may reflect the landscape of operating practice for suspected renal malignancy more accurately than cancer-based registries. These datasets fail to include patients with a renal mass suspicious on imaging that ultimately prove to represent benign disease on surgical pathology and patients with a non-diagnostic or no biopsy before ablation. Because we derived the data in the present study from inpatient procedure and diagnosis coding, we cannot comment on the use of surveillance for renal masses. This strategy is an outpatient process. Attempts to do so using cancer registries or population-based databases are fraught with difficulty in any event; many patients do not have a histological diagnosis, and it is difficult to identify those patients who were consciously selected for surveillance. There is no specific diagnosis or procedure code for surveillance, and the practices used to implement this management strategy commonly occur in other scenarios. For example, the reason for a delay between diagnosis and treatment cannot be retrospectively defined in any particular case; it could reflect a conscious decision to pursue a surveillance programme just as easily as medical or social factors that result in delayed care.
In conclusion, in the present study, we identified an increased prevalence of RN, PN and C/RFA over the study period, with PN increasing the most rapidly. Although NSM was increasingly utilized over the study period, it is particularly concerning that the data obtained indicate that patients with pre-existing CKD are receiving less nephron-sparing approaches. Further investigations are required to confirm these findings and to identify impediments to the dissemination of NSM.
Conflict of Interest
The authors declare that there are no conflicts of interest.