Practice-setting and surgeon characteristics heavily influence the decision to perform partial nephrectomy among American Urologic Association surgeons


Correspondence: Bradley C. Leibovich, Department of Urology, Mayo Clinic, 200 First Street SW Gonda 7, Rochester, MN 55905, USA.



What's known on the subject? and What does the study add?

  • There is great variability in the utilization of partial nephrectomy, but the causes of these variations are not well understood.
  • The present study underscores the already observed phenomenon of surgical volume influencing surgical planning and outcomes, but it gets at why this might be so. We observe that high-volume renal surgeons have different thresholds of ‘technical feasibility’.


  • To investigate why there continues to be wide variability in the application of partial nephrectomy (PN) for treating small renal masses despite guidelines in the US and Europe stating that a PN is a standard of care for a patient with a T1 renal mass.

Patients and Methods

  • In June 2009, 764 surgeon-members of the American Urologic Association (AUA) participated in a survey evaluating the management of renal masses.
  • Renal mass complexity was graded by nephrometry score (NS).
  • Multivariable logistic regression models with generalized estimating equations were constructed to evaluate how tumour, surgeon and practice-setting characteristics influence the use of PN.


  • The survey response rate was 19%. Each urological surgeon responded to eight scenarios, providing 6112 evaluable cases.
  • Tumour NS ranged from 4 to 10, and each unit increase in NS was associated with 59% increased likelihood of a surgeon offering RN on multivariable analysis (odds ratio [OR] = 1.59; 95% CI: 1.52–1.64).
  • When holding patient and tumour characteristics constant, the following surgeon and practice-setting characteristics significantly increased the odds of offering a PN: increasing renal case volume (OR = 1.57; 95% CI: 1.27–1.95), academic practice (OR = 1.80; 95% CI: 1.42–2.29), increasing PN % volume (OR = 3.7; 95% CI: 2.46–5.55) and younger surgeon age (≤40 vs >50 years) (OR = 1.64; 95% CI: 1.35–1.96).


  • The characteristics of a surgeon and the setting in which he or she practices influence the utilization of PN, the adherence to professional guidelines, and the threshold of tumour complexity at which a surgeon stops offering PN.

nephrometry score


odds ratio


partial nephrectomy


radical nephrectomy


Population-based data and single institutional reports have chronicled wide variability in the application of partial nephrectomy (PN) for small renal masses in recent years. Although the use of PN to treat renal masses has been slowly increasing in North America [1-4], recent population-based estimates suggest that overall PN rates are around 15–20% for potentially amenable tumours [5, 6]. These rates are much lower than those reported by single institutions which are as high as 70–90% for small renal masses [4, 7]. The reasons for these large disparities in practice pattern have been difficult to evaluate. Nearly all studies to date that compare radical nephrectomy (RN) with PN offer few anatomical details of the kidney tumours treated, apart from size, making it impossible to determine whether PN or RN is appropriate or even feasible. In addition, there are disparate reports in the literature of the benefit of PN, with some finding a survival advantage [5, 7-10] while one randomized trial did not [11]. Furthermore, it remains unclear whether these differences in PN application are due to patient and tumour characteristics or surgeon and practice-setting factors [12]. We surveyed members of the AUA to get an understanding of how surgeons approach patients with small renal masses and how tumour, surgeon and practice-setting characteristics influence the willingness to offer PN.

Subjects and Methods

In June 2009, all members of the AUA (n = 4513) with a listed email address were invited to participate in a survey evaluating the management of renal masses. The details of the survey have been published previously [13], but, briefly, 866 surgeons completed the survey (19%). After excluding non-urologists, residents and fellows from the analyses (n = 102), the responses of 764 staff urological surgeons remained available for analysis.

Participant surgeons were presented with eight clinical cases. For each case, a coronal drawing and an axial CT image of the renal tumour were provided (Fig. 1). Although all tumours were stage T1a, the tumours varied by location, depth of invasion and tumour size. Each of these renal masses was given a nephrometry score (NS) [14]; the eight tumours were given the scores of 4, 5, 6 and 8 in one tumour each, and the four remaining tumours all scored 10. In all cases, tumours were described as asymptomatic, solitary, sporadic, solid and enhancing renal masses, the kidneys had equal function, and there were no medical conditions that would preclude surgical intervention. In each case, respondents were asked to choose their first choice in management (RN, PN, thermal ablation or active surveillance). In the full survey, three standardized patients were presented for each of the eight cases. Since we were specifically interested in studying how surgeon traits and practice setting influence the determination of technical feasibility of PN, we limited our analysis to the ideal case for PN, a young healthy 45-year-old patient with a T1a tumour. The AUA and European Association of Urology (EAU) guidelines [15, 16] unequivocally state that PN should be offered as the standard of care in this case, reserving RN for cases where PN is not technically feasible.

Figure 1.

An example of a low (A), moderate (B) and highly (C) complex renal tumours. If a patient with tumour (C) presented to a 40-year-old high-volume, high-PN % volume academic surgeon, he or she would be offered a PN 81% of the time. If that same patient were to present to 55-year-old, low-PN % volume, low-volume surgeon in private practice, he or she would be offered a PN only 30% of the time.

Case volume was determined by respondent selection from several case volume ranges for each treatment type. We calculated a point estimate of the number of each type of operation in each individual surgeon's practice/year by fixing the estimate in the middle of the range, or 55 if ‘greater than 50’ was selected. PN % volume was the percentage of cases/year divided by the total number extirpative cases, i.e. [PN/(RN + PN)] × 100. The odds of offering PN were evaluated by NS and surgeon and practice-setting characteristics using multivariable logistic regression with generalized estimating equations to control for repeated measurement within each respondent. Case volumes of the respondents were divided into approximate tertiles, and were defined as follows: high-, intermediate- and low-volume surgeons were those who evaluate >20, 11–20 and ≤10 patients with renal masses per/year, respectively.

Differences in proportions were evaluated with Pearson's chi-squared test. Predicted probabilities were generated from the multivariable logistic regression equation. The c-index was used to evaluate the respective contributions of each of the individual components for the final model. Fellowship training status was removed from the final model because it did not improve the c-index and was not significant in the multivariable model. The RN threshold was defined as the most complex tumour (as defined by NS) at which the predicted probability of PN was greater than 50%. Since NS values of 11 and 12 were not in the measured range in this survey, they were replaced by ‘>10’ in the predicted thresholds. Statistical analyses were performed using the SAS software package (SAS Institute, Cary, NC, USA). All tests were two-sided and P < 0.05 was considered to indicate statistical significance.


The characteristics of the 764 responding staff surgeons are given in Table 1. The distribution of renal tumour management strategies did not appear to be substantially different according to practice setting (Fig. 2). On average, most surgeons use all the modalities, i.e. RN, PN, active surveillance and thermal ablation to treat renal masses, but academic surgeons manage more patients with each of the modalities. A similar pattern was noted with volume category and PN % volume.

Figure 2.

The distribution of renal tumour management strategies according to renal case volume (A) and practice setting (B).

Table 1. Characteristics of surgeons completing the survey on management of small renal masses
CharacteristicTotal (N = 764) [n (%)]
Surgeon age, years
<3479 (10.4)
34–40114 (14.9)
41–50240 (31.4)
51–60236 (30.9)
>6095 (12.4)
Practice setting
Missing3 (0.4)
Academic hospital309 (40.5)
Community hospital301 (39.5)
Community (individual practice)151 (19.8)
Time in practice, years
Missing3 (0.4)
≤5142 (18.7)
6–10114 (14.9)
11–15114 (14.9)
16–20106 (13.9)
>20285 (37.3)
Number of renal masses evaluated per year
05 (0.7)
1–5108 (14.1)
6–10177 (23.2)
11–20227 (29.7)
21–50174 (22.8)
51–10053 (6.9)
>10020 (2.6)
Tertiles of annual renal mass case volume
Low (0–10)290 (38.0)
Medium (11–20)227 (29.7)
High (>20)247 (32.3)
Post residency fellowship in either oncology or minimally invasive surgery
No537 (70.3)
Yes227 (29.7)
PN % volume (n = 685 due to missing data) (% of extirpative masses treated by PN annually)
0%76 (10.5)
1–25%146 (20.3)
25–50%215 (29.8)
>50%248 (39.3)

Annual renal case volume, practice setting and PN % volume are associated with significant differences in the surgical decision-making process (Fig. 3). High-volume, academic and high-PN % volume surgeons tend to focus primarily on factors that affect long-term postoperative renal function when deciding when to perform a PN (solitary kidney status, multiple renal tumours/genetic syndrome and preoperative renal function). By contrast, low-volume, private practice and low-PN % volume surgeons report that technical concerns such as tumour size and location are as important as, or more important than, preoperative renal function.

Figure 3.

Factors that surgeons report influence the decision to perform PN stratified by surgical volume (A), practice setting (B) and PN % volume (% of PN vs RN). Groups with an asterisk are statistically significant (p-value < 0.05).

Univariate and multivariable logistic regression models showing the variables associated with a surgeon's recommendation for a PN are presented in Table 2. The c-index change was used to evaluate the independent individual contributions to PN prediction. As tumour complexity increased, the differences in practice style became more apparent (Fig. 4). The difference in the predicted PN frequency in the low-complexity tumours was 18% between groups 1 and 3, but in the highest-complexity tumours the PN frequency difference rose to 52%. The definition of ‘technical feasibility’, i.e. the threshold at which tumour complexity would push a surgeon over the RN threshold, appeared to vary by surgeon and practice-setting characteristics (Table 3).

Figure 4.

Predicted PN frequency in increasingly complex renal tumours when controlling for patient factors. Group 1, private practice, surgeon age > 50 years, low PN rate and annual renal tumour case volume; group 2, community base hospital practice, surgeon age 41–50 years, and moderate PN rate and annual renal tumour case volume; group 3, academic practice setting, surgeon age ≤ 40 years, and high PN rate and annual renal tumour case volume. Vertical lines mark the PN threshold; on average, any tumour above this complexity will be treated by RN for the respective surgeon groups.

Table 2. Univariate and multivariable logistic regression with generalized estimating equations to predict PN for all tumours with non-missing data (N = 5737)
FeatureUnivariate modelMultivariable modelc-index change*
OR (95% CI)P valueOR (95% CI)P value
  1. *The c-index change represents the difference between the c-index from the full multivariable model (0.785) and the c-index obtained by excluding the feature listed. For example, a model that contained all of the features listed except for age had a c-index of 0.780, resulting in a c-index change of 0.005. Fellowship training was removed to make the most parsimonious multivariable model. It was not significant (P = 0.09) and added no predictive improvement compared with the model without it.
Age (years)     
≤401.0 (reference)0.0021.0 (reference)0.0310.005
41–500.75 (0.62–0.90)<0.0010.81 (0.67–0.98)<0.001 
>500.53 (0.44–0.64) 0.61 (0.51–0.74)  
Private1.0 (reference)<0.0011.0 (reference)<0.0010.005
Academic2.05 (1.67–2.52)<0.0011.80 (1.42–2.29)0.023 
Community1.44 (1.18–1.75) 1.30 (1.04–1.62)  
No1.0 (reference)<0.001Not included*0.090.000
Yes1.64 (1.38–1.96)    
Surgical volume     
Low1.0 (reference)0.0211.0 (reference)0.490.003
Moderate1.22 (1.03–1.45)<0.0011.07 (0.88–1.31)<0.001 
High2.08 (1.73–2.49) 1.57 (1.27–1.95)  
% treated with PN     
01.0 (reference)0.0091.0 (reference)0.330.012
>0–251.43 (1.09–1.88)<0.0011.19 (0.84–1.69)<0.001 
>25–502.09 (1.63–2.67)<0.0011.91 (1.39–2.61)<0.001 
>504.36 (3.12–6.09) 3.70 (2.46–5.55)  
Nephrometry score0.65 (0.63–0.68)<0.0010.63 (0.61–0.66)<0.0010.145
Table 3. Predicted RN threshold NS in a healthy 45-year-old patient, stratified by surgeon and practice-setting characteristics. Threshold level is the most complex tumour or highest NS where the predicted majority of surgeons in each group would choose PN, i.e. a threshold of 7 means all tumours more complex (NS > 7) would, on average, be treated by RN in that particular surgeon group
Surgeon characteristicsRN threshold NS by practice setting
Annual renal case volumeSurgeon age, years% of cases where surgeon treats tumours with PNAcademic practiceCommunity hospital-basedPrivate practice

Most surgeons responding to the survey performed at least one PN a year (n = 688 [90%]), and were 3.9 times more likely to offer a PN in at least one of the presented scenarios than those who did not perform PN in the proceeding year (unadjusted OR = 3.9, 95% CI: 1.2–13.1).


Population-based data [5, 6] and single institutional data [4, 7] have chronicled wide variability in the application of PN for small renal masses. There is evidence that a large portion of the variability is due to surgeon and hospital characteristics and not to the patient characteristics [12]. We sought to determine if these discrepancies are due to differing views on the utility of PN, different practice compositions, or differences in the perceived ‘technical feasibility’ of PN.

Differences in PN application do not appear to be due to divergent beliefs in the utility of PN. PN was offered by 98.3% of all surgeons in at least one case scenario, underscoring a near-uniform belief that PN offers advantages over RN in certain settings. Only the subset of surgeons who do not perform PN in their own practice (n = 76) appeared to believe that PN offered little advantage over RN in any setting. They were 75% less likely to offer RN, regardless of the tumour characteristics. Likewise, the distribution of renal tumour management strategies did not appear to influence the use of PN; most surgeon groups report using a combination of PN, RN, active surveillance and thermal ablation in their own patients (Fig. 2).

The determination of whether to offer PN or RN appears to hinge primarily on ‘technical feasibility’ and the need to preserve renal function. Among all surgeons in this survey, tumour size was the fifth most important variable in deciding whether to perform a PN. In descending order of importance, the most important factors were: solitary kidney (98.6%), tumour location (88%), genetic syndrome/multiple tumours (87.6%), preoperative renal function (86.9%), tumour size (77.3%), patient comorbidity (62.5%) and patient age (45.4%). Unfortunately, the only information given for most studies comparing the use of PN with RN in single institution series and population-based studies is tumour size. This lack of detail in the current literature has weakened the strength of the conclusions these studies can offer. This shortcoming, in part, has led to the development of several standardized reporting systems of anatomical tumour information [14, 17, 18]. Future studies that compare PN with RN should take into account all these variables, to limit confounding and to make the comparison more meaningful.

Not surprisingly, increasing complexity of renal tumours, as scored by the NS, was significantly associated with decreased PN use among all surgeons (OR = 0.63; 95% CI: 0.61–0.66 per each unit increase in NS). As NS rises, a point will be reached where it is no longer ‘technically feasible’ to perform a PN. But the definition of ‘technical feasibility’ is imprecise and greatly influenced by experience and practice setting.

The threshold for performing a PN also varies with surgeon and practice-setting characteristics. For example the RN threshold for a 55-year-old surgeon based at a community hospital who performs around one PN and five RNs annually would be a NS of 7. Or, in other words, any tumour with a NS > 7 would generally be treated by RN by this surgeon group. This is particularly striking since a NS > 7 is a very common presentation and often ‘technically feasible’ to other surgeons. In fact, a recently published series at a high-volume academic medical centre found that 72% of tumours treated by PN were NS > 7 [19].

These data also offer some insight into the volume-outcome debate. Increasing hospital and surgeon case volume have been significantly associated with meaningful outcomes such as mortality and complication rates in a wide variety of urological [20, 21] and non-urological [22, 23] procedures. Our findings suggest that there could be differences in the way high- and low-volume surgeons perceive renal tumours. Furthermore, it appears that high- and low-volume surgeons prioritize patient and tumour characteristics differently when deciding to perform PN. These differences could translate into long-term patient benefits, such as those observed in patients with preserved renal function [8, 9, 24].

We acknowledge that the present study is limited by the relatively low response rate (19%). The physician response rate to most surveys is poor [25]. However, we hypothesize that the non-response bias has actually moved our estimates towards the null hypothesis, because those who responded to the survey appear to be the surgeons most familiar with the management of renal masses and more likely to adhere to professional guidelines. This conclusion is based on three differences between our cohort and other published cohorts. First, the definition of high volume in other cohorts was three cases/year [12], whereas our low-volume cohort was 10 cases/year or less. Secondly, 40.5% of our respondents work in an academic setting, which is nearly double the percentage quoted by the AUA [26]. Finally, the PN rate in the present cohort was much higher than current population estimates. Respondents to our survey report that, on average, they remove 36.5% (95% CI: 35.0–37.9) of all renal tumours by PN, whereas recent population estimates are that only 11% of tumours are removed by this fashion in the US [6]. We also note that these responses might be subject to recall bias and that the ‘operative planning’ could be affected by social desirability bias. All these differences, however, would probably move our estimates towards the null, and the true differences in practice could therefore be even greater.

Finally, we are aware that surgeons are individuals and are not defined solely by the groups designated in Table 3. There might be high-volume, young surgeons in academic settings who have a low RN threshold and there might be low-volume surgeons over 50 in private practices who have a very high RN threshold. Nevertheless, it is noteworthy to discover that, by knowing the surgeon and practice-setting characteristics, we can correctly predict the operation offered to a patient with a T1a renal tumour 64% of the time, without knowing a single detail about the patient's health or the tumour size or location (Table 3).

In conclusion, increasingly complex tumours (as measured by NS) are more likely to be removed by RN. The experience of a surgeon and the setting in which he or she practices appear to heavily influence the application of PN. The definition of technical feasibility of a PN and the adherence to AUA and EAU guidelines are influenced by renal tumour case volume, PN % volume and practice setting.


We thank the AUA for allowing access to their member registry.

Conflict of Interest

None declared.