Comparison of partial vs radical nephrectomy with regard to other-cause mortality in T1 renal cell carcinoma among patients aged ≥75 years with multiple comorbidities


  • M.S. and M.B. contributed equally to the work.

Correspondence: Maxine Sun, Cancer Prognostics and Health Outcomes Unit, University of Montreal Health Center, 1058, rue St-Denis, Montreal, QC, Canada H2X 3J4.



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

  • Surgical intervention is the established standard of care option in patients diagnosed with localized RCC.
  • The study found that better and rigorous selection of surgical candidates should be implemented in the context of localized RCC as some patients may not benefit from surgery.


  • To quantify the effect of partial nephrectomy (PN) vs radical nephrectomy (RN) on other-cause mortality (OCM) in elderly patients with localized renal cell carcinoma (RCC) and/or multiple comorbidities.


  • Using the Surveillance, Epidemiology, and End Results Medicare-linked database, patients with T1 RCC, aged ≥75 years, or who had ≥2 comorbidities, were identified (1988–2005).
  • To adjust for inherent differences between treatment types, propensity-based matched analyses were performed.
  • Competing-risks regression analyses for prediction of OCM were assessed according to treatment type.
  • The effect of PN and RN on OCM was examined in three sub-groups: patients aged ≥75 years; patients with ≥2 comorbidities; and patients aged ≥75 years with ≥2 comorbidities.


  • After propensity-based matched analyses and adjustment for all covariates, PN was found to exert a protective effect relative to RN with respect to OCM in all patients (hazard ratio [HR]: 0.84, P = 0.048).
  • In subanalyses, no difference was recorded between PN and RN in patients who were aged ≥75 years (HR: 0.83, P = 0.2), with ≥2 baseline comorbidities at diagnosis (HR: 0.83, P = 0.1), or in patients who were aged ≥75 years and who had ≥2 baseline comorbidities (HR: 0.77, P = 0.2).


  • Some elderly patients and/or those with multiple comorbidities at diagnosis may not benefit from PN with respect to OCM.
  • After rigorous patient selection, alternative treatment options could be considered.

other-cause mortality


partial nephrectomy


radical nephrectomy


Surveillance, Epidemiology, and End Results


Charlson comorbidity index


International Classification of Disease 9th revision


Current Procedural Terminology Coding System, 4th edition


socioeconomic status


hazard ratio


Owing to an increase in the incidental detection of small renal masses with non-invasive abdominal imaging in the last decade, the incidence of localized RCC has risen considerably [1, 2]. Nonetheless, mortality rates for localized RCC continue to rise, suggesting that the treatment management of such cases should be reconsidered. Indeed, only a minority of patients with localized RCC will have recurrence after surgical intervention and benign pathology is found in nearly 20% of cases [3]. As such, most patients diagnosed with localized RCC will not die from RCC, but of other-cause mortality (OCM).

In this context, the use of partial nephrectomy (PN) has increased tremendously in recent years [4, 5]. Previously, it has been shown that PN may provide improved renal function, a lower risk of cardiovascular events, and improved overall survival compared with radical nephrectomy (RN) [6-10]. Nonetheless, in the setting of localized RCC, several investigators have questioned the selection of elderly patients and/or patients with multiple comorbidities at diagnosis as viable candidates for surgical intervention [11]. Certainly, active treatment could be considered appropriate only in the event that such treatment could be curative and when cancer could potentially reduce a patient's life expectancy.

The aim of the present study was to investigate whether a survival benefit existed for individuals with advanced age and/or multiple comorbidities at diagnosis of localized RCC who were treated with nephron-sparing surgery (i.e. PN) compared with those who were treated with RN. To accomplish this, we examined the rates of OCM after nephrectomy in patients aged ≥75 years old and/or with multiple comorbidites (≥2) at baseline, who were diagnosed with non-metastatic T1 RCC (≤7 cm).


Data Source

The present study relied on the Surveillance, Epidemiology, and End Results (SEER) Medicare insurance programme-linked database. The SEER regions represent ∼14% of the US population before the year 2000, and 26% thereafter [12]. The Medicare-linked database is 98% complete for case ascertainment. It encompasses ∼97% of persons aged 65 years or older in the USA. Linkage to the SEER database is complete for ∼93% of the patients.

Study Population

Data on patients with a primary diagnosis of non-metastatic T1 RCC (C64.0) between 1988 and 2005 were extracted. Patient age at diagnosis was obtained from the Medicare file. Patient comorbidity was quantified using the Klabunde modification [13] of the Charlson comorbidity index (CCI) [14], which gives a weighted score based on diagnosis claims. A higher score indicates a patient with a greater number of comorbid conditions (0, 1, 2, ≥3). For the purpose of our analyses, patients had to fulfill at least one of two inclusion criteria: age ≥75 years and ≥2 baseline comorbidities at diagnosis. Moreover, only patients with both Medicare Part A and Part B claims available 12 months before the first recorded diagnosis and 6 months after diagnosis, and who were not enrolled in a health maintenance organization throughout the duration of the study period, were included. This allowed the calculation of a non-cancer comorbidity score and ensured we had relevant treatment information after diagnosis. Patients with RCC diagnoses based on autopsy, or patients whose original or current reason for Medicare entitlement was listed as disability or who had a Medicare status code including disability, were removed from the study.

Treatment Type

Treatment type was identified by searching Part A and Part B Medicare files, and the outpatient claims file, using the International Classification of Disease 9th revision (ICD-9) or Current Procedural Terminology Coding System, 4th edition (CPT-4) codes within 6 months of primary diagnosis of RCC. Patients who underwent RN were identified using ICD-9 procedure code: 55.5x or CPT-4 codes: 50220, 50225, 50230, 50545 and 50546. Patients who underwent PN were identified via the presence of ICD-9 procedure code: 55.4 or CPT-4 codes: 50240, 50280, 50290, 50543. Patients who had a claim for PN followed by a claim for RN within 30 days were identified as part of the RN group, and vice versa.

Patient and Sociodemographic Characteristics

Information on sex and race (white, black or other) was obtained using the SEER demographic datafile. Population density of county of residence (urban or rural), and marital status (married, single, unmarried or unknown) was obtained from the Medicare file. Patients with a baseline diagnosis of hypercalcaemia (ICD-9: 275.42), hyperlipidaemia (ICD-9: 272.4, 272.0, 272.2) and anaemia (ICD-9: 280.9, 281.0, 281.9, 282.3) were also identified.

Socioeconomic status (SES) was defined according to three county-attribute variables: education level as a percentage of persons who did not complete high school; poverty as a percentage of persons living below the poverty line at the time; and the median family income. These county-attribute variables were measured using the census 2000 documentation files available at: We created a composite variable of SES, using the three variables mentioned above, based on a previous methodology [15]. Firstly, we recoded the variables individually to ensure that low values represented low SES, and vice versa. Secondly, we transformed all values into standardized scores. Thirdly, we took the sum of these scores and categorized the total scores into two groups according to the median, which resulted in our low and high SES.

Statistical Analyses

Frequencies and proportions were assessed for categorical variables, while means, medians and interquartile ranges were determined for continuously coded variables. The chi-squared test and independent Mann–Whitney test were used to test the statistical significance of proportions and medians, respectively.

Owing to inherent differences between patients undergoing PN and RN in terms of baseline patient and disease characteristics, we used 1:1 nearest-neighbour-propensity-score-matched analyses to adjust for these differences. The use of the propensity score method eliminates the customary bias associated with the conventional multivariable modelling approach [16, 17]. Propensity scores were calculated by modelling a logistic regression with the dependent variable as the odds of undergoing a PN, and the independent variables as patient age, sex, race, baseline CCI, year of surgery, marital status, SES, population density, baseline hypercalcaemia, baseline hyperlipidaemia, baseline anaemia, tumour size and tumour stage. Subsequently, covariate balance between the matched groups was examined.

Univariable and multivariable competing-risks regression analyses for prediction of OCM were performed to assess the effect of treatment type within the post-propensity matched group. Covariates comprised cancer-specific mortality, patient age, baseline CCI, sex, race, SES, marital status, population density, tumour size, baseline hypercalcaemia, baseline hyperlipidaemia, baseline anaemia, Fuhrman grade, and year of surgery. Competing-risks regression analyses were repeated exclusively in three sub-groups: (i) patients aged ≥75 years; (ii) patients with ≥2 baseline comorbidities; and (iii) patients aged ≥75 years with ≥2 baseline comorbidities.

All tests were two-sided and a P value <0.05 was considered to indicate statistical significance. Analyses were conducted using the statistical package for R (the R foundation for Statistical Computing, version 2.13.1).


Baseline Descriptives

Overall 7524 patients with non-metastatic T1 RCC, who were either aged ≥75 years or had ≥2 comorbidities at baseline, were identified (Table 1). Of these patients, 4310 (57%) were aged ≥75 years, and 5804 (77%) had multiple comorbidities at baseline (≥2), while 2590 (34%) were aged ≥75 years and had ≥2 comorbidities at diagnosis. Of the total 7524 patients, 8% and 88% underwent PN and RN, respectively. PN-treated patients were younger (mean: 74 vs 76 years), more likely to be male (57 vs 52%) and to be married (62 vs 57%), to have smaller tumours (mean size: 2.7 vs 4.0 cm), and to have a baseline diagnosis of hyperlipidaemia (74 vs 58%) compared with RN-treated patients. After propensity-based matched analysis, 924 patients in the RN group were matched to 924 patients in the PN group. Inherent differences between PN and RN according to patient and disease characteristics were reduced to a minimum (all standardized mean differences ≤10.0%, Table 1). The use of propensity-matched analysis resulted in 819 patients aged ≥75 years old, 1231 patients with ≥2 baseline comorbidities, and 430 patients aged ≥75 years and who had ≥2 comorbidities at baseline.

Table 1. Baseline descriptive characteristics of 7524 patients from the SEER Medicare-linked database with T1 RCC, treated with PN or RN, 1988–2005.
VariableOverallPre-propensity matchingPost-propensity matching
PNRNStandardized mean differencesPNRNStandardized mean differences
No. of patients (%)7524924 (8.3)6600 (87.7)924 (50)924 (50)
Age, years   −0.370  0.035
Mean (median)75 (76)74 (74)76 (76)74 (74)73 (73)
Interquartile range71–7969–7871–7969–7869–77
Sex, n (%)   −0.101  0.000
Male3958530 (57)3428 (52)530 (57)530 (57)
Female3566394 (43)3172 (48)394 (43)394 (43)
Race, n (%)   0.018  −0.033
White6439782 (85)5657 (86)782 (85)765 (83)
Black58481 (9)503 (8)81 (9)98 (11)
Other50161 (7440 (7)61 (7)61 (7)
Marital status, n (%)   −0.017  −0.001
Married4363572 (62)3791 (57)572 (62)563 (61)
Single52170 (8)451 (7)70 (8)62 (7)
Previously married2422246 (27)2176 (33)246 (27)267 (29)
Unknown21836 (4)182 (3)36 (4)32 (4)
SES, n (%)   0.037  −0.021
High4064484 (52)3580 (54)484 (52)474 (51)
Low3460440 (48)3020 (46)440 (48)450 (49)
Population density, n (%)   −0.051  −0.003
Urban6485810 (88)5675 (86)810 (88)809 (88)
Rural1039114 (12)925 (14)114 (12)115 (12)
CCI, n (%)   0.151  −0.011
01476128 (14)1348 (20)128 (14)126 (14)
124418 (2)226 (3)18 (2)29 (3)
21203166 (18)1037 (16)166 (18)150 (16)
≥34601612 (66)3989 (60)612 (66)619 (67)
Tumour size, cm   −0.832  −0.016
Mean (median)3.9 (3.8)2.9 (2.7)4.0 (4.0)2.9 (2.7)2.9 (2.8)
Interquartile range2.7–5.02.0–3.53.0–5.02.0–3.52.0–3.6
Tumour stage, n (%)   −0.781  −0.027
T1a4536784 (85)3752 (57)784 (85)775 (84)
T1b2988140 (15)2848 (43)140 (15)149 (16)
Baseline hypercalcaemia, n (%)15220 (2)132 (2)0.01120 (2)19 (2)0.007
Baseline hyperlipidaemia, n (%)4492681 (74)3811 (58)0.362681 (74)676 (73)0.012
Baseline anaemia, n (%)2864342 (37)2522 (38)−0.025342 (37)343 (37)−0.002

Propensity-Adjusted OCM Rates

Within the propensity-based matched population (n = 1848), the 2- and 5-year OCM rates were 7.1 and 20.5% for PN vs 7.9 and 24.1% for RN (hazard ratio [HR]: 0.84, 95% CI: 0.70–0.99, P = 0.043, Fig. 1). Among individuals aged ≥75 years (n = 819), the 2- and 5-year OCM rates were 8.9 and 26.0% for PN vs 10.1 and 28.5% for RN (HR: 0.79, 95% CI: 0.62–1.01, P = 0.05, Fig. 2A). Among patients with ≥2 baseline comorbidities (n = 1231), the 2- and 5-year OCM rates were 7.9 and 22.2% for PN vs 8.6 and 26.1% for RN (HR: 0.84, 95% CI: 0.68–1.04, P = 0.1, Fig. 2B). Finally, 430 patients aged ≥75 years had ≥2 baseline comorbidities. For PN and RN, respectively, the 2- and 5-year OCM rates were 10.0 and 28.4% vs 12.4 and 32.6% (HR: 0.71, 95% CI: 0.52–0.97, P = 0.03).

Figure 1.

Cumulative incidence plots depicting OCM according to treatment type in the matched cohort (n = 1848).

Figure 2.

Cumulative incidence plots depicting OCM according to treatment type in A, patients aged ≥75 years old (n = 819) and B, patients with ≥2 baseline comorbidities (n = 1231).

Propensity-Adjusted Competing-Risks Regression Analyses

Table 2 shows competing-risks regression analyses that accounted for cancer-specific mortality, patient age, year of surgery, race, sex, marital status, SES, population density, baseline comorbidities, tumour stage, tumour size, baseline diagnoses of hyperlipidaemia, hypercalcaemia and anaemia. Within the entire matched population (n = 1018), PN-treated patients had a lower risk of OCM compared with RN-treated patients (HR: 0.84, 95% CI: 0.70–0.98, P = 0.048). In subgroup analyses that considered only patients aged ≥75 years (HR: 0.83, 95% CI: 0.65–1.07, P = 0.2), patients with ≥2 baseline comorbidities (HR: 0.83, 95% CI: 0.67–1.03, P = 0.1), and patients ≥75 years with ≥2 baseline comorbidities (HR: 0.77, 95% CI: 0.54–1.10, P = 0.2), PN failed to achieve independent predictor status of OCM relative to RN.

Table 2. Association between treatment type (PN vs RN) and OCM in the post-propensity matched cohort and subgroupsa.
 PN vs RN HR (95% CI)P
  1. aAll models adjusted for cancer-specific mortality, patient age, year at surgery, race, sex, marital status, SES, population density, baseline comorbidities, tumour stage, tumour size, baseline hyperlipidaemia, baseline hypercalcaemia, and baseline anaemia.
Entire population0.84 (0.70–0.98)0.048
≥75 years old0.83 (0.65–1.07)0.16
≥2 baseline comorbidities0.83 (0.67–1.03)0.10
≥75 years and ≥2 baseline comorbidities0.77 (0.54–1.10)0.15


In the last two decades, the incidence of localized RCC has risen dramatically, with a concurrent rise in mortality rates [2]. In a parallel context, life expectancy continues to rise in North America [18]. Consequently, an increasing number of older individuals are being diagnosed and seen for incidentally detected renal masses, suspected to be RCC. While a number of such patients will be selected for minimally invasive interventions or active surveillance [4], active treatment options, such as PN or RN remain long-standing practices, even in older patients and patients with multiple comorbidities. Provided that it is technically feasible, PN has emerged as an established standard of care in patients with T1a RCC, and has also taken prominence in patients with T1b tumours [5, 19]. Nephron-sparing is associated with several benefits, namely the lower risk of cardiovascular events, preservation of renal function, and equal cancer control outcomes over RN [6, 7, 10, 20].

In the present study, after propensity-matching, inherent differences between PN- and RN-treated patients were reduced to a minimum, which rendered the two groups comparable. In general, OCM rates were higher in RN-treated patients. For example, the 5-year OCM rates were 21 vs 24% for PN and RN, respectively (P = 0.04). However, when analyses were performed within older patients (≥75 years) and/or patients with ≥2 comorbidities, OCM rates were equivalent for PN and RN. Specifically, among patients aged ≥75 years, PN failed to exert a significant protective effect relative to RN (26 vs 29%, P = 0.05). The same observation can be made for patients with ≥2 baseline comorbidities (22 vs 26%, P = 0.1). In multivariable analyses that adjusted for all examined covariates, including that of cancer-specific mortality, PN achieved independent predictor status over RN for a lower risk of OCM in the matched cohort. However, when analyses were restricted to patients aged ≥75 years and/or with ≥2 baseline comorbidities, PNwas no longer protective with respect to OCM, relative to RN.

The present study is the first to attempt to fine-tune the ideal patient characteristics for PN, relative to RN. Among previous studies that have examined the effect of PN and RN on OCM or overall survival [6, 10, 20, 21], none has focused exclusively on older and/or patients with multiple comorbidities. Specifically, previous studies that compared PN and RN focused on all patients, regardless of age or comorbidities. In consequence, although previous analyses corroborate the protective effect of nephron-sparing on OCM relative to RN, the present study indicates an absence of OCM benefit in older and/or patients with multiple comorbidities.

The present findings are clinically important. Nonetheless, given the retrospective nature of the study, the interpretation of our results warrants consideration. While PN was not associated with a significant protective effect against OCM relative to RN in older patients and those with multiple comorbidities, it may reflect an inherent selection bias that we could not adjust for. Simply put, it is possible that older patients and patients with multiple comorbidities, with limited life expectancy, selected to undergo nephrectomy would not have benefitted from the intervention, regardless of the type of nephrectomy the patient received. Clearly, surgical intervention can only be appropriate provided that it can be curative and if the cancer would reduce the life expectancy of a given individual. In this context, previous investigators have voiced their concerns with respect to overtreatment in patients with localized RCC. Indeed, a recent institutional report has shown that patients ≥75 years are unable to benefit from either PN or RN relative to non-surgical management [22]. It is therefore possible that only in properly selected stage I patients, PN could emerge as a better treatment option over RN, when OCM is considered as an endpoint.

Consequently, it cannot be recommended that all older patients and those with multiple comorbidities should not receive active treatment. In a recent study, Sun et al. [23] showed that in the setting of low-stage and grade RCC, older individuals are at a higher risk of RCC-specific mortality relative to their younger counterparts. In consequence, it can be stated that at least in some individuals who are older or who have multiple comorbidities at diagnosis, a surgically-derived benefit should not be denied these patients. From this perspective, a rigorous patient selection process should be implemented during counselling for treatment decision-making. These include existing preoperative nomograms [11, 24, 25], as well as extensive questionnaires developed by the geriatric society for the assessment of life expectancy in such individuals [18]. In patients with a limited life expectancy, other minimally invasive approaches and active surveillance may be more appropriate. Nonetheless, surgical intervention may still remain beneficial in some select patients. A rigorous patient selection process is therefore essential during treatment decision-making.

The present study has some limitations. First, only patients aged ≥65 years old are captured in the SEER-Medicare database. It is possible that a considerable proportion of younger patients who may have had multiple comorbidities at baseline were not considered in the present analyses. In addition, comorbidities were coded based on administrative claims, which may be associated with an unavoidable misattribution bias. It may be postulated that some important comorbidities were missed, which would then result in an inherent bias, but the coding of comorbidities using administrative data has been previously validated [26], and represents an established methodology. Similarly, misattribution and coding errors pertaining to cause-of-death information may have occurred; however, this limitation is present in all retrospective data and cause of death in the SEER database has been shown to be highly reliable [27]. Additionally, the present study fails to account for baseline renal function and postoperative renal insufficiency, and lack of central pathology represents another limitation. Finally, only age and comorbidities were considered in the present study. Ideally, age or comorbidities alone cannot be considered as primary deterrents for surgical omission. For example, some octogenarians may have an up to 10-year life expectancy with an average health status, and up to 20 years with an excellent health status. A rigorous patient selection process involves both the assessment of comorbid status and age simultaneously in order to make viable decisions with respect to cancer treatment and tolerability. In addition, other considerations should not be omitted in the process (i.e. functional measures, demographic measures, social support, nutritional status, etc.) [18].

In conclusion, older patients and patients who have multiple comorbidities at diagnosis of localized RCC should undergo a rigorous patient selection if surgical intervention is an option. Other minimally invasive options and active surveillance should be considered.


Pierre I. Karakiewicz is partially supported by the University of Montreal Health Center, Urology Specialists, Fonds de la Recherche en Sante du Quebec, the University of Montreal Department of Surgery and the University of Montreal Health Center (CHUM) Foundation.

Conflict of Interest

None declared.