Renal Function Following Retroperitoneal Sarcoma Resection with Nephrectomy: A Matched Analysis of the United States Sarcoma Collaborative Database

Abstract

Background:

Nephrectomy is often required during en bloc resection of a retroperitoneal sarcoma (RPS) to achieve an R0/R1 resection. The impact of nephrectomy on postoperative renal function in this patient population, who may also benefit from subsequent nephrotoxic systemic therapy, is not well described.

Methods:

The United States Sarcoma Collaborative (USSC) database was queried for patients undergoing RPS resection between 2000–2016. Patients missing pre- or postoperative measures of renal function were excluded. A matched cohort was created using coarsened exact matching. Weighted logistic regression was used to further control for differences in the nephrectomy vs non-nephrectomy cohorts. Primary outcomes included postoperative acute kidney injury (AKI), acute renal failure (ARF), and dialysis.

Results:

The initial cohort consisted of 858 patients, of which 3 patients (0.3%) required postoperative dialysis. The matched cohort consisted of 411 patients of which 108 patients (26%) underwent nephrectomy. Patients that underwent nephrectomy had higher rates of postoperative AKI (14.8% vs 4.3%, p<0.01) and ARF (4.6% vs 1.3%, p=0.04); however, no patients required dialysis postoperatively. Using logistic regression modeling, the risk of AKI (OR 5.16, p<0.01) and ARF (OR 5.04, p<0.01) following nephrectomy persisted despite controlling for age and preoperative renal function.

Conclusions:

Nephrectomy is associated with increased risk of postoperative AKI and ARF following RPS resection. This study is unable to statistically assess the impact of nephrectomy on postoperative dialysis, but the risk of post-operative dialysis is ≤0.5% regardless of nephrectomy status.

Introduction:

Retroperitoneal sarcomas (RPS) account for approximately 15% of all soft tissue sarcomas and comprise a heterogeneous group of many histologic tumor types, the most common of which are liposarcoma and leiomyosarcoma.¹ Surgical extirpation is the primary treatment modality of RPS with pre- or postoperative chemotherapy having an unproven impact on outcomes.² Compartmental or wide en bloc resection with macroscopically negative (R0 & R1) margins is the standard surgical approach to these tumors,^2,3 which often requires multivisceral resection of contiguously involved organs; such that multivisceral resections occur in nearly 40% of patients.⁴ Previous reports have suggested acceptable morbidity and mortality associated with these extended resections.^5–8

Anatomically, the kidney is one of the most frequently resected organs in a RPS multivisceral resection.⁴ Small, single institution retrospective cohorts (n~100) have examined the effect of concomitant kidney removal and have found limited adverse impact on postoperative renal function.^7,9–11 Following an initial nadir, kidney function improves and patients ultimately have mild progression of CKD, but progression to dialysis is rare.

Despite these findings, a large study (n>100,000) from the transplant donor nephrectomy population demonstrated that the risk of ESRD increased from a baseline of 3.9/10,000 patients to 30.8/10,000 patients (a nearly 10-fold increase) following donor nephrectomy, with a median follow-up of 7.6 years.¹² Given this disconnect, we wished to use a large, multi-institutional database of retroperitoneal sarcoma cases to determine if nephrectomy during RPS resection led to clinically significant adverse renal outcomes. We hypothesized that, following matched cohort analysis, renal outcomes would be worse for patients who underwent nephrectomy.

Methods

All patients within the United States Sarcoma Collaborative database (an 8-institution collaborative including University of Wisconsin, Emory University, Stanford University, Medical College of Wisconsin, Wake Forest University, The Ohio State University, University of Chicago Medicine, and the Siteman Cancer Center, Washington University in St. Louis) that underwent a primary or recurrent RPS resection between January 2000 and April 2016 were identified for inclusion in this study. Patients requiring dialysis at baseline (N=10) were excluded. We then elected to perform a complete case analysis (CCA). While missing covariate data did vary based on the site of the operation (e.g. some centers had better capture of specific variables, such as ASA Class, than others), there was no reason to suspect a correlation between the outcome variables relating to postoperative renal function and missingness (i.e., that patients with missing data had different postoperative renal outcomes than those without missing data), making a CCA appropriate.

Patients were excluded when missing: preoperative creatinine, postoperative creatinine, postoperative dialysis requirement, or any other variable required to estimate glomerular filtration rate (GFR), including age, sex, or race (Figure 1). GFR was then calculated using the CKD-EPI 2009 equation from the nephro package in R, and divided into CKD stages as defined by the National Kidney Foundation: G1=GFR>89, G2=GFR 60–89, G3a=GFR 45–59, G3b=GFR 30–44, G4=GFR 15–29, G5=GFR<15.^13,14 Acute kidney injury (AKI) and acute renal failure (ARF) were calculated using the Risk, Injury, Failure, Loss, ESRD (RIFLE) criteria.¹⁵ AKI was defined as a postoperative creatinine >2x the preoperative creatinine, while ARF was defined as a postoperative creatinine >3x the preoperative creatinine or a postoperative creatinine > 4 mg/dL. Preoperative creatinine was defined as the preoperative creatinine value (mg/dL) obtained closest to the date of surgery (typically at a preoperative clinic visit), while the postoperative creatinine was the highest postoperative creatinine value obtained during the first 90 days after surgery. Our primary outcomes included postoperative: AKI, ARF, temporary need for dialysis, and permanent need for dialysis. The cutoff for all outcomes was 90 days postoperatively (i.e. temporary need for dialysis ceased before 90 days postop, any dialysis need that lasted beyond 90 days was considered permanent).

Figure 1:

Study Inclusion/Exclusion Criteria

For demographic data, metastatic disease was defined as metastatic disease at the time of presentation of the current tumor. Recurrent disease was defined as RPS presenting after a tumor of the same histologic type was treated.

Coarsened exact matching (CEM), a monotonic imbalance reducing method, was performed using the MatchIt package.¹⁶ CEM allows for researcher-guided coarsening of the data into clinically important subgroups pre-matching, followed by exact matching on the coarsened data. This allows for higher numbers of matches than exact matching while still minimizing clinically significant imbalance between treated and control groups. Patients that underwent a nephrectomy were matched to those that did not by sex, race, age (coarsened to 0,25,50,60,70, and >80), and GFR (coarsened to G1, G2, G3a/b, and G4/G5).

A data analysis strategy consisting of matching followed by weighted regression was deliberately chosen—matching methods are not designed to compete with regression, and the two methods have been shown to work best in combination.¹⁷ Matching reduces model dependence and bias, strengthening the conclusions that can be drawn from retrospective data. Then, just as in randomized experiments; regression adjustment is used to clean up small residual imbalance between matched groups.¹⁷ This methodology also creates a clear demarcation between the 1) design and 2) data analysis of retrospective studies.

After matching, weighted logistic regression was performed to determine the impact of nephrectomy on postoperative renal outcomes and reported as odds ratios. When performing a logistic regression, a CCA is asymptotically unbiased as long as the missingness is not dependent on the outcome variable.¹⁸ All statistical analysis was performed in R 3.6.1 (R Foundation for Statistical Computing, Vienna, Austria). A p-value < 0.05 was considered statistically significant unless otherwise indicated. Students t-tests and chi-squared tests were used for bivariate analysis. All institutions obtained Institutional Review Board approval with an approved waiver of consent before beginning any research efforts.

Results

Unmatched Cohort

A total of 858 patients underwent RPS resection; 3 (0.35%) required temporary or permanent postoperative dialysis (of these 3 patients, 1 patient had a nephrectomy, 2 did not). The rate of postoperative dialysis was 1/191 (0.52%) in the nephrectomy group and 2/667 (0.3%) in the non-nephrectomy group. Of the 507 patients out of the original 858 with sufficient data to calculate a preoperative GFR, 2 (0.4%) required dialysis (neither underwent nephrectomy, preoperative GFR categories were G3b and G4, respectively).

Matched Cohort

Following exclusions, coarsened exact matching based on age, sex, race, and GFR was used to create a cohort of 411 matched patients (of which 108 underwent nephrectomy with their RPS resection). Demographic data on this matched cohort is displayed in Table 1. Both groups had similar mean ages (59 v 62, p=0.05), sex (56% vs 53% female, p=0.55), BMI (27.7 vs 26.3, p=0.08), neoadjuvant chemotherapy rates (15% vs 10%, p=0.20), ASA Class (p=0.14), and GFR classifications (p=0.61). Despite matching, patients undergoing resection with nephrectomy had lower grade tumors (31% vs 22% Low-Grade, p=0.02), larger tumors (22.3 cm vs 12.5 cm, p<0.01), higher rates of liposarcoma (63% vs 33%, p<0.01), were less likely to have recurrent disease (8.3% vs 46.2%, p<0.01), longer operative times (349 min vs 263 min, p<0.01), and higher intraoperative transfusion rates (49% vs 31%, p<0.01) than those without nephrectomy.

Table 1:

Matched Cohort Demographics

	No Nephrectomy (N=303)	Nephrectomy (N=108)	p value
Age	58.7 (13.6)	61.6 (12.9)	0.05
Sex			0.55
- Female	170 (56.1%)	57 (52.8%)
Race			0.37
- Asian	4 (1.3%)	3 (2.8%)
- Black	18 (5.9%)	6 (5.6%)
- Latino	18 (5.9%)	9 (8.3%)
- White	263 (86.8%)	90 (83.3%)
BMI	27.7 (6.1)	26.3 (5.5)	0.08
ASA Class			0.14
- 1	5 (2.3%)	6 (7.1%)
- 2	82 (37.3%)	24 (28.2%)
- 3	123 (55.9%)	51 (60.0%)
- 4	10 (4.5%)	4 (4.7%)
Neoadjuvant Chemotherapy	46 (15.3%)	11 (10.3%)	0.20
Neoadjuvant Radiotherapy	34 (11.3%)	16 (15.0%)	0.32
Recurrent Disease	140 (46.2%)	9 (8.3%)	< 0.01
Metastatic Disease	87 (28.9%)	5 (4.7%)	< 0.01
Operative Time	262.7 (172.7)	348.5 (198.1)	< 0.01
Intraoperative Transfusion	89 (30.5%)	51 (48.6%)	< 0.01
Tumor Size	12.5 (8.9)	22.3 (13.1)	< 0.01
Tumor Histology			<0.01
- Leiomyosarcoma	73 (24.1%)	26 (24.1%)
- Liposarcoma	101 (33.3%)	68 (63.0%)
- Other Sarcoma	113 (37.3%)	11 (10.2%)
- Sarcoma NOS	16 (5.3%)	3 (2.8%)
Grade			0.02
- Grade cannot be assessed	31 (11.7%)	4 (3.7%)
- High-Grade	175 (66.0%)	70 (65.4%)
- Low-Grade	59 (22.3%)	33 (30.8%)
Preop Creatinine	0.9 (0.3)	0.9 (0.3)	0.40
GFR Stage*			0.61
- G1	136 (44.9%)	40 (37.0%)
- G2	125 (41.3%)	47 (43.5%)
- G3a	28 (9.2%)	14 (13.0%)
- G3b	12 (4.0%)	5 (4.6%)
- G4	1 (0.3%)	1 (0.9%)
- G5	1 (0.3%)	1 (0.9%)

^*Stages of chronic kidney disease as defined by the Kidney Disease Outcomes Quality Initiative Clinical Practice Guidelines

Patients that underwent a nephrectomy as part of their RPS resection had a higher postoperative peak creatinine (1.5 vs 1.0, p<0.01), as well as higher rates of postoperative AKI (15% vs 4%, p<0.01) and postoperative ARF (5% vs 1%, p=0.04) (Table 2). However, no patients in either group required dialysis postoperatively. Due to the significantly higher rate of recurrent disease in the no nephrectomy group (and the potential confounding of variables related to disease recurrence, including tumor size, tumor histology, and use of neoadjuvant therapies), a sensitivity analysis was performed using a subset of patients presenting with primary disease. This confirmed the worse renal outcomes seen in the nephrectomy group (postoperative creatinine 1.5 vs 1.0, p<0.01, postoperative AKI 14% vs 5%, p<0.01). The exception was ARF, which was no longer statistically significant (3.3% vs 2.5%, p=0.7), however this is likely related to the significantly reduced sample size (subset of 250 out of 411 patients) for an outcome with low prevalence in the dataset rather than a true difference in outcomes.

Table 2:

Postoperative Renal Outcomes

	No Nephrectomy (N=303)	Nephrectomy (N=108)	P value
Peak Postop Creatinine	1.0 (0.5)	1.5 (1.0)	< 0.01
Temporary Dialysis	0 (0.0%)	0 (0.0%)	NA
Permanent Dialysis	0 (0.0%)	0 (0.0%)	NA
Postop AKI	13 (4.3%)	16 (14.8%)	< 0.01
Postop ARF	4 (1.3%)	5 (4.6%)	0.04

The results of weighted logistic regression models to predict postoperative AKI and ARF are reported as log-odds ratios (odds ratios reported on the log scale for confidence interval symmetry) in Figure 2. After controlling for age, sex, BMI, GFR and surgical nephrectomy, AKI was associated with age (OR 1.05, 95%CI 1.01–1.10), GFR (OR 1.03, 95%CI 1.01–1.06) and nephrectomy (OR 5.16, 95%CI 2.17–12.63). After controlling for age, GFR and nephrectomy, ARF was associated with nephrectomy only (OR 5.04, 95%CI 1.20–25.86). A sensitivity analysis of only patients presenting with primary disease demonstrated similar results for AKI, however in this smaller sample the effect seen from nephrectomy in ARF did not reach statistical significance (Figure 3).

Figure 2:

The odds of postoperative acute kidney injury and acute renal failure using logistic regression models

*Odds ratios are graphed using a logarithmic scale for symmetry, while the odds ratio values reported in the manuscript use the standard exponentiated scale (e.g. e⁽⁰⁾ = 1, e^(1.64) = 5.16)

**Models control for different risk factors due to the lower event rates for acute renal failure limiting the number of variables that could be included in the model

Figure 3:

The odds of postoperative acute kidney injury and acute renal failure using logistic regression models in a subset of patients presenting with primary disease

*Odds ratios are graphed using a logarithmic scale for symmetry, while the odds ratio values reported in the manuscript use the standard exponentiated scale (e.g. e⁽⁰⁾ = 1, e^(1.64) = 5.16)

**Models control for different risk factors due to the lower event rates for acute renal failure limiting the number of variables that could be included in the model

Discussion

This study of a multi-institutional cohort of 411 matched patients is the largest analysis to date of the impact of nephrectomy on post-RPS resection renal outcomes, and the first to utilize matching to minimize covariate imbalance between cohorts. Even after controlling for confounders including age and renal function, the odds of postoperative AKI or ARF were 5 times greater for patients that underwent nephrectomy (Figure 2).

While nephrectomy increased the odds of postoperative AKI or ARF, the data in this study were not sufficient to statistically estimate the impact of nephrectomy on postoperative dialysis. However, the rate of postoperative dialysis among the entire (unmatched) cohort was very low (≤0.5%), regardless of nephrectomy status. Additionally, patients that required postoperative dialysis had low preoperative renal function, with GFR stages greater than G3b (eGFR ≤44). Thus, in all patients undergoing RPS resection, the risk of postoperative dialysis is low and likely related to poor preoperative renal function. These findings are consistent with previous literature on this topic as described by Cho et al. and Kim et al., which suggest that while nephrectomy is associated with worse short-term renal outcomes (such as AKI or ARF), after RPS resection the rates of long-term renal dysfunction (dialysis) are low.^9,10

The evolving use of nephrotoxic systemic chemotherapy agents, such as ifosfamide, in the treatment of sarcoma is often considered when making decisions that may impact postoperative renal function.¹⁹ While this study cannot provide definitive recommendations on the impact of nephrectomy on the safety of subsequent nephrotoxic systemic therapy, neoadjuvant chemotherapy was used in both the nephrectomy and control groups (10.3% vs. 15.3%) without causing a high rate of postoperative dialysis—suggesting that nephrectomy is safe following nephrotoxic chemotherapy. Future research will be needed to fully elucidate the impact of the combination of nephrectomy followed by adjuvant nephrotoxic chemotherapy on renal function.

This study has several limitations. While confounders between groups were minimized using coarsened exact matching followed by logistic regression, the retrospective nature of the cohort still limits the strength of the findings. Patients that underwent nephrectomies had larger tumors, longer operating room times, and higher rates of intraoperative transfusions than those that did not (Table 1). These more physiologically taxing surgeries can contribute to AKI and ARF along with nephrectomy in a multifactorial fashion. Unfortunately, it is unlikely for any prospective, randomized trials to accrue to differentiate between the effect of nephrectomy vs. physiologic demand of longer operations given the rarity of RPS and the difficulty in determining the need for nephrectomy preoperatively. Finally, despite the size of our initial cohort the low frequency of postoperative renal dysfunction limited the depth of our analysis. Low rates of AKI/ARF limited the number of patient comorbidities we could control for in our weighted logistic regression.^20,21 Despite these shortcomings, this study represents the largest and only matched study addressing whether nephrectomy worsens renal outcomes following RPS resection. Continued accrual of patient data in multi-institutional collaboratives will undoubtedly help surgeons more conclusively answer these questions in the future.

Nephrectomy is associated with an increased risk of postoperative AKI and ARF following RPS resection. Data on the impact of nephrectomy on new postoperative dialysis are limited, but the risk of any patient undergoing RPS resection and subsequently requiring dialysis is ≤0.5% regardless of nephrectomy status.

Synopsis:

Nephrectomy for retroperitoneal sarcoma resection is associated with worse short-term renal function, but progression to dialysis is rare.