The impact of perioperative blood transfusion on survival after nephrectomy for non-metastatic renal cell carcinoma (RCC)

Authors


Abstract

Objective

  • To evaluate the association of perioperative blood transfusion (PBT) with survival after nephrectomy.

Patients and Methods

  • We identified 2318 patients who underwent partial or radical nephrectomy at Mayo Clinic between 1990 and 2006.
  • PBT was defined as transfusion of allogenic red blood cells during surgery or postoperative hospitalisation.
  • Survival was estimated using the Kaplan–Meier method and compared with the log-rank test. Cox proportional hazards regression models were used to evaluate the association of PBT with outcome.

Results

  • In all, 498 patients (21%) received a PBT. The median (interquartile range) number of units transfused was 3 (2, 5).
  • Patients receiving a PBT were significantly older at surgery (P < 0.001), more likely to be female (P < 0.001), with more frequent symptomatic presentation (P < 0.001), worse Eastern Cooperative Oncology Group performance status (P < 0.001), and more frequent adverse pathological features, such as high nuclear grade (P < 0.001), locally-advanced tumour stage (P < 0.001) and lymph node invasion (P < 0.001).
  • The median follow-up was 9.1 years. Receipt of a PBT was associated with adverse 5-year cancer-specific (68% vs 92%; P < 0.001) and overall (56% vs 82%; P < 0.001) survival.
  • On multivariate analyses, PBT remained associated with higher risk of death from any cause (hazard ratio [HR] 1.23; P = 0.02).
  • Among patients who received a PBT, an increasing number of units transfused was independently associated with increased all-cause mortality (HR 1.08; P = 0.001).

Conclusion

  • PBT is associated with a significantly increased risk of mortality after nephrectomy. While external validation is needed, continued efforts to minimise the use of blood products in these patients are warranted.

Introduction

In 2013, an estimated 65 150 new cases of kidney cancer will be diagnosed in the USA, with an estimated 13 680 patients dying from this disease [1]. For patients with non-metastatic tumours, surgical excision remains the mainstay of treatment. Clinicopathological features associated with outcome after nephrectomy for localised RCC have been well-defined, and include tumour size, pathological stage, nuclear grade, histological tumour necrosis, and the presence of sarcomatoid features [2, 3].

Notably, nephrectomy (including partial and radical) has been associated with a perioperative blood transfusion (PBT) rate ranging from 3.5% to 18.1% [4-8]. Previous reports have shown an increased risk of cancer recurrence in patients receiving a PBT after surgery for colon, hepatic and oesophageal malignancies [9-11]. It has been postulated that this association may be secondary to an immunosuppressive effect of blood transfusion with decreased host tumour surveillance, or may be caused by growth factor delivery from red blood cells; however, a definitive biological link has not been confirmed [12-14].

Meanwhile, for genitourinary malignancies, conflicting results have been reported, e.g. regarding the association of PBT with prostate cancer recurrence after radical prostatectomy, as well as for bladder cancer after radical cystectomy [15-19]. In fact, we recently reported that PBT was independently associated with mortality after cystectomy, and that this association was dose-dependent [18]. Interestingly, few reports exist about the impact of PBT on kidney cancer outcomes after nephrectomy to date, and these have reported conflicting findings [20-22]. Nevertheless, the studies of PBT and nephrectomy have been limited by small sample sizes and historical patient cohorts [20-22].

Thus, in the present study, in a large cohort of patients, with long-term follow-up, we sought to evaluate the independent association of PBT with survival after nephrectomy for localised RCC.

Patients and Methods

After Institutional Review Board approval, we reviewed the Mayo Clinic Nephrectomy Registry to identify 2318 patients treated with partial (825 patients) or radical (1493) nephrectomy at our institution between 1990 and 2006 for unilateral, non-metastatic clear cell, papillary, or chromophobe RCC. Both laparoscopic (n = 235) and open (n = 2083) procedures were included for this study. Patients were excluded from analysis if non-lymph node metastases were documented at surgery (M1 stage) and/or if PBT status was unknown.

Clinicopathological variables recorded included age, gender, constitutional symptoms at diagnosis (including rash, sweats, weight loss, fatigue, early satiety, and anorexia), Eastern Cooperative Oncology Group (ECOG) performance status, body mass index (BMI), preoperative haemoglobin, type of nephrectomy, estimated surgical blood loss, receipt of PBT and number of units transfused, pathological tumour stage and lymph node status, nuclear grade, coagulative tumour necrosis and the presence sarcomatoid differentiation. One urological pathologist (J.C.C.) re-reviewed all nephrectomy pathology specimens without knowledge of patient outcomes. Tumour staging followed the 2009 American Joint Committee on Cancer/Union Internationale Contre le Cancer 7th edition TNM classification [23].

PBT was defined as transfusion of allogenic red blood cells intraoperatively or during the postoperative hospitalisation. Notably, the decision to administer PBT was at the discretion of the treating physicians, with no institutional standardised transfusion threshold used. The retrospective nature of the present study precluded a standardised follow-up protocol in all patients. However, follow-up after nephrectomy at our institution has generally been recommended semi-annually for the first 2 years after surgery, to include serum electrolyte panel as well as imaging of the chest, abdomen and pelvis, and annually thereafter. Recurrence was defined as distant metastases. Vital status was identified from physician correspondence or death certificates. For patients followed elsewhere, the Mayo Clinic Nephrectomy Registry monitors outcomes annually by correspondence with the patient and treating physician.

Outcomes measured included recurrence-free (RFS), cancer-specific (CSS) and overall survival (OS). Survival was estimated as the time from nephrectomy to event using the Kaplan–Meier method, and was compared between patients who did vs did not receive a PBT with the log-rank test. Cox proportional hazard regression models were used to evaluate the association of PBT with outcomes, controlling for clinicopathological variables. In these models year of surgery (1-year increase), age at surgery (10-year increase), preoperative haemoglobin (1-g/dL increase), BMI (5-kg/m2 increase), tumour size (1-cm increase), and units transfused (1-unit increase) were analysed as continuous variables, and the 2009 primary tumour classification (1-stage increase) was evaluated as an ordinal variable. The remaining variables in the multivariable models, including gender, constitutional symptoms, ECOG performance status, histological subtype, the 2009 regional lymph node classification, nuclear grade, sarcomatoid differentiation, and PBT were considered categorical variables.

Statistical analyses were performed using the SAS software package (SAS Institute, Cary, NC, USA). Continuous features were summarised with medians and interquartile ranges (IQRs); categorical features were summarised with frequency counts and percentages. Comparisons of features between patients with and without a PBT were evaluated using Wilcoxon, chi-square, Fisher's exact, and Cochran-Armitage trend tests. Logistic regression models were used to evaluate the association of PBT with short-term outcomes of 30-day mortality. All statistical tests were two-sided, with a P < 0.05 considered to indicate statistical significance.

Results

Overall, 498/2318 patients (21%) received a PBT, with a median (IQR) number of units transfused of 3 (2, 5). Comparisons of the clinical and pathological features between patients who did and did not receive a PBT are shown in Table 1. As can be seen, patients who received a PBT were significantly older (P < 0.001), more likely to be female (P < 0.001), with symptomatic presentation (P < 0.001), worse ECOG performance status (P < 0.001), and a higher rate of adverse pathological features, such as high nuclear grade (P < 0.001), more frequent locally advanced disease (P < 0.001) and more frequent lymph node invasion (P < 0.001). In addition, patients who received a PBT had a significantly higher median intraoperative estimated blood loss at nephrectomy (878 mL) than patients who did not receive a PBT (200 mL) (P < 0.001).

Table 1. Clinicopathological features of patients undergoing nephrectomy for RCC who did vs did not receive a PBT
 No PBTPBTP
Number of patients1820498 
Year of surgery, n (%):  0.08
1990–1994372 (20)124 (25)
1995–1999436 (24)117 (23)
2000–2004690 (38)171 (34)
2005–2006322 (18)86 (17)
Median (IQR) age at surgery, years63 (53, 71)67 (58, 75)<0.001
Gender, n (%):  <0.001
female558 (31)199 (40)
male1262 (69)299 (60)
Constitutional symptoms at presentation (n = 2315), n (%)229 (13)168 (34)<0.001
ECOG performance status (n = 2315), n (%)  <0.001
01632 (90)408 (82)
1151 (8)68 (14)
219 (1)9 (2)
314 (1)9 (2)
43 (<1)2 (<1)
Median (IQR) preoperative haemoglobin, g/dL14.0 (13.0, 15.1)11.8 (10.3, 13.3)<0.001
Median (IQR) BMI (n = 2061), kg/m228.6 (25.8, 32.1)27.7 (25.8, 32.0)0.006
Pathological tumour stage (n = 2295), n (%)  <0.001
pT11325 (74)190 (38)
pT2238 (13)78 (16)
pT3233 (13)214 (43)
pT43 (<1)14 (3)
Median (IQR) tumour size (n = 2305), cm4.3 (2.7, 6.7)7.8 (4.5, 10.2)<0.001
Histological subtype, n (%)  <0.001
clear cell1355 (74)423 (85)
papillary350 (19)56 (11)
chromophobe115 (6)19 (4)
pN1, n (%)27 (1)47 (9)<0.001
Nuclear grade: n (%)  <0.001
1155 (9)28 (6)
21011 (56)148 (30)
3600 (33)249 (50)
454 (3)73 (15)
Coagulative tumour necrosis, n (%)375 (21)218 (44)<0.001
Sarcomatoid differentiation, n (%)20 (1)26 (5)<0.001

The median (IQR) follow-up after surgery in those alive at last follow-up was 9.1 (6.9, 12.8) years. During this time, 491 patients had disease recurrence and 1065 died, including 401 who died from kidney cancer. Of note, receipt of a PBT was associated with significantly decreased 5-year RFS (63% vs 88%; P < 0.001) and CSS (68% vs 92%; P < 0.001; Fig. 1). Similarly, patients who received a PBT were likewise found to have worse 5-year OS vs patients who did not (56% vs 82%; P < 0.001; Fig. 2). Of the 498 patients with a PBT, 10 (2.0%) died at ≤30 days of surgery, compared with four of 1820 (0.2%) patients who did not receive a PBT (univariate odds ratio 9.30; 95% CI 2.91, 29.79; P < 0.001).

Figure 1.

CSS after nephrectomy for RCC, stratified by receipt of PBT.

Figure 2.

OS after nephrectomy for RCC, stratified by receipt of PBT.

We next assessed the independent association of PBT with patient's risk of disease recurrence, death from kidney cancer, and all-cause mortality on multivariate analyses, controlling for various clinicopathological features. We found that receipt of a PBT remained associated with all-cause mortality (HR 1.23; P = 0.02; Table 2), while the association of this variable with death from RCC (HR 1.15; P = 0.31) and tumour recurrence (HR 1.04; P = 0.77; data not shown) did not reach statistical significance. Meanwhile, established pathologicalprognostic features including tumour and nodal stage, nuclear grade and the presence of tumour necrosis were associated with increased risks of recurrence, death from any cause, and kidney cancer death. Older patient age and worse ECOG performance status were also associated with significantly increased risks of both cancer-specific and all-cause mortality.

Table 2. Multivariate analysis of factors associated with death from kidney cancer and all-cause mortality after nephrectomy for RCC
VariableDeath from any causeDeath from kidney cancer
HR95% CIPHR95% CIP
Year of surgery0.980.97, 1.000.030.950.93, 0.98<0.001
Age at surgery1.631.53, 1.75<0.0011.141.03, 1.270.02
Gender (ref. male)1.271.09, 1.480.0020.850.66, 1.090.2
Symptomatic presentation1.170.98, 1.390.081.240.95, 1.620.12
ECOG performance status (ref. 0)2.261.89, 2.71<0.0011.961.40, 2.73<0.001
BMI1.020.95, 1.090.581.010.91, 1.130.82
Preoperative haemoglobin0.960.93, 1.000.071.040.98, 1.110.16
Receipt of PBT1.231.04, 1.460.021.150.87, 1.530.31
Pathological tumour stage1.071.01, 1.130.021.321.22, 1.42<0.001
Tumour size1.021.00, 1.050.061.041.00, 1.070.03
Histologic subtype (ref. papillary/chromophobe)1.411.18, 1.68<0.0012.251.53, 3.29<0.001
pN11.911.39, 2.62<0.0012.131.48, 3.05<0.001
Nuclear grade (ref. 1–2):      
31.301.11, 1.530.0022.992.13, 4.19<0.001
42.211.58, 3.10<0.0014.802.89, 7.95<0.001
Coagulative tumour necrosis1.461.23, 1.74<0.0012.261.73, 2.95<0.001
Sarcomatoid differentiation1.691.08, 2.650.021.590.96, 2.640.08

We then performed a separate analysis among those patients who received a PBT to evaluate a potential dose-dependent association of PBT with outcomes (Table 3). We determined that an increasing number of units transfused was independently associated with an increased risk of all-cause mortality (HR 1.08; P = 0.001), with each unit transfused associated with an 8% increased risk of death. An increased number of units of PBT received tended to increase the risk of death from kidney cancer as well, although this association did not reach statistical significance in this analysis (HR 1.06; P = 0.08). Meanwhile, adverse pathological factors were significantly associated with patients' risk of mortality after surgery in this cohort as well.

Table 3. Multivariate analysis of factors associated with death from kidney cancer and all-cause mortality after nephrectomy for RCC among patients receiving PBT
VariableDeath from any causeDeath from kidney cancer
HR95% CIPHR95% CIP
Year of surgery0.980.95, 1.000.090.950.92, 0.990.01
Age at surgery1.371.21, 1.55<0.0011.010.85, 1.180.99
Gender (ref male)1.020.79, 1.330.860.760.53, 1.110.15
Symptomatic presentation1.020.77, 1.360.881.120.77, 1.650.55
ECOG performance status (ref. 0)2.461.83, 3.32<0.0012.051.31, 3.230.002
BMI0.960.87, 1.070.490.980.84, 1.140.77
Pre-operative haemoglobin0.980.92, 1.050.611.020.93, 1.120.69
No. units transfused1.081.03, 1.140.0011.060.99, 1.130.08
Pathologic tumour stage1.020.94, 1.110.681.191.06, 1.340.003
Tumour size1.030.99, 1.070.111.020.98, 1.070.31
Histologic subtype (ref. papillary/chromophobe)1.691.15, 2.470.0082.531.26, 5.070.009
pN11.541.02, 2.330.041.801.13, 2.860.01
Nuclear grade (ref. 1–2):      
31.100.80, 1.510.563.031.54, 5.970.001
41.671.00, 2.770.053.681.62, 8.370.002
Coagulative tumour necrosis1.531.12, 2.090.0082.281.49, 3.50<0.001
Sarcomatoid differentiation1.961.12, 3.430.021.931.00, 3.710.05

Discussion

In the present study, we found in a large cohort of patients treated with nephrectomy for RCC, that on long-term follow-up, receipt of a PBT was associated with a significantly increased risk of all-cause mortality. Additionally, we noted that among those patients who received a PBT, that increasing the number of units transfused was associated with a significant progressively increased risk of all-cause mortality. These associations remained when controlling for multiple clinical and pathological features, suggesting an independent association of receipt of a PBT with all-cause mortality after nephrectomy.

Previous studies investigating the impact of PBT on survival after nephrectomy have, in small cohorts, met with conflicting results. That is, Moffat et al. [21], in 126 patients undergoing nephrectomy from 1973 to 1985, with a transfusion rate of 63%, found no difference in survival between those patients who did vs did not receive a PBT. Meanwhile, Jakobsen et al. [20] noted an adverse 5-year CSS among the 208/258 of their patients who received a PBT with nephrectomy, although this association did not persist on multivariate analysis. Conversely, Edna et al. [23] evaluated the impact of PBT in a cohort of 201 patients undergoing nephrectomy, of whom 154 (77%) received a PBT, and on multivariate analysis found that transfusion with >4 units was independently associated with death from kidney cancer. It should be noted that the aforementioned studies included patients managed in the 1970s and 1980s, and given the significant changes in transfusion protocols and surgical management over time; we surmise that a contemporary evaluation is warranted. The present study thus augments the existing literature by presenting data from a large contemporary cohort of patients with RCC, with long-term postoperative follow-up. As such, the differences in the present findings from those of prior studies may be due to disparate sample sizes, lengths of follow-up, patient cohorts, and varied application of blood product administration.

The reported rates of PBT with nephrectomy have varied considerably [4-8] and probably also reflect disparate populations studied, as well as a lack of standardised criteria for transfusion. Notably, a recent population-based study of 10 902 nephrectomies reported a transfusion rate of 18.1% [4], similar to our experience here (21%). Contributing to our rate of PBT, may also be the tertiary referral nature of our practice, as 46% of those who received a PBT had pT3/4 disease.

Regarding the present data that PBT was not significantly associated with the risks of tumour recurrence or death from RCC but did predict all-cause mortality, we submit that several potential factors may have contributed to these findings. For one, most patients in the present cohort (79%) did not receive a PBT, which, together with the fact that 62% of deaths in our series were not due to kidney cancer, may have limited the statistical power to detect such an association, despite the large size of our overall patient population. Alternatively, as the means by which PBT impacts survival remains to be established, it is possible that deleterious non-cancer effects of transfusion may be responsible for the present findings. For example, a randomised study of intensive care unit patients showed a significantly increased rate of complications including myocardial infarction and pulmonary oedema, among patients who received a transfusion to maintain a haemoglobin threshold of 10 g/dL vs a threshold of 7 g/dL [24], highlighting potential adverse impacts of PBT on non-oncological outcomes.

Importantly, multiple potential methods to limit use of red blood cell transfusion have been proposed, ranging from preoperative optimisation (erythropoietin administration and iron supplementation) [25], alternative blood products aside from allogenic red blood cells (autologous transfusion and intraoperative cell saver) [25], as well as the application of restrictive transfusion criteria [24, 26]. Additionally, with specific regard to patients with kidney cancer, preoperative arterial embolization has been used, albeit with conflicting reports on resulting transfusion rates [27-29]. Of these strategies, perhaps the most fertile area going forward for limiting the use of PBT will be the application of restrictive transfusion strategies. Indeed, several randomised trials have evaluated potential threshold levels for transfusion in patients in the intensive care unit [24], as well in postoperative orthopaedic patients with coronary artery disease [26]. These studies support the use of restrictive transfusion strategies (transfusion haemoglobin threshold of 7 g/dL in intensive care unit patients [24] and 8 g/dL in orthopaedic patients with coronary artery disease [26]). Extrapolation of these results for patients undergoing nephrectomy may limit the use of PBT. This represents an important area for future study in our field, ideally in the setting of a prospective randomised trial.

We recognise that the present study is limited by its retrospective non-randomised design. Importantly as such, significant clinicopathological differences existed between the cohort of patients who received a PBT vs those who did not (i.e. as depicted in Table 1). As such, considering the retrospective nature of our study design, even despite adjustment on multivariate analyses, these discrepancies may not have been entirely accounted for, and may therefore have impacted our reported findings. Additionally, the decision for PBT was based on the discretion of the treating physicians, without adherence to specific criteria thresholds for transfusion. Moreover, we could not evaluate the impact of transfusion of other blood products, such as fresh frozen plasma or platelets, which may have affected patient outcomes as well, and did not capture receipt of PBT before surgery. Furthermore, given the prolonged timeframe of the study, changes in both RCC presentation as well as handling of blood products have occurred. To account for these time trends, we included year of surgery into our multivariate models. Lastly, we acknowledge that these results are from a single, tertiary referral institution, and as such require external validation.

In conclusion, we found that receipt of a PBT was associated with a significantly increased risk of all-cause mortality among patients undergoing nephrectomy for non-metastatic RCC. This association remained after accounting for multiple potential confounding clinicopathological features. While these results represent outcomes from a single tertiary care centre, and external validation is needed, continued efforts to limit the use of blood products in these patients are needed.

Conflict of Interest

None declared.

Abbreviations
BMI

body mass index

CSS

cancer-specific survival

ECOG

Eastern Cooperative Oncology Group

IQR

interquartile range

OS

overall survival

PBT

perioperative blood transfusion

RFS

recurrence-free survival

Ancillary