Postoperative Complications Are Not Elevated in well-compensated End-Stage Renal Disease Patients Undergoing Cardiac Surgery

Abstract

Background:

Patients with end-stage renal disease (ESRD) are at high risk for cardiac disease requiring surgery, and have been shown to have increased surgical risks. There have been significant improvements in ESRD management, surgical techniques, and patient selection over the past ten years. We evaluated rates of serious post-operative outcomes in stable, well-dialyzed patients with ESRD undergoing non-emergent cardiac surgery compared to the general cardiac surgery population.

Study Design:

In this propensity-score matched study, we evaluated 1,451 adult patients who underwent non-emergent cardiac surgery at the University of Colorado Hospital (UCH) between 2011 and 2016. Patients with ESRD were compared to non-ESRD patients. The primary outcome was a composite endpoint, including 30-day mortality, stroke, postoperative infection, and prolonged intensive care unit (ICU) stay.

Results:

A total of 35 had ESRD, and there were no statistically significant differences in the outcomes between ESRD and non-ESRD patients in either unadjusted or adjusted analysis. (OR for composite outcome 0.70, CI 0.29-1.72).

Conclusion:

Stable ESRD patients undergoing non-emergent surgery are not at increased risk of major postoperative complications when compared to those without ESRD. Overall, stable ESRD patients should not be excluded from surgical consideration.

Introduction

The number of patients with end-stage renal disease (ESRD) in the United States continues to rise, and over 500,000 are currently receiving either hemodialysis (HD) or peritoneal dialysis [1]. The primary cause of mortality within the ESRD population is cardiovascular disease, which occurs in 64.5% of patients > 65 years old and accounts for 39% of deaths in this patient group [1, 2]. With the rising prevalence of ESRD and high rate of cardiac disease within this population, cardiac surgeons are faced more than ever with decisions regarding whether or not to operate on patients with ESRD.

Previous studies of outcomes in patients with ESRD have demonstrated significant increases in perioperative morbidity and mortality compared to the non-ESRD population [3–9]. However, there have been encouraging trends among ESRD patients over the last 5–10 years. Mortality rates among ESRD patients have reached an all-time low at 16% annually, and the average life expectancy for patients initiating dialysis has reached a high of 6.8 years (even after excluding kidney transplants)[1]. There have also been increases in less invasive cardiac procedures including thoracic endovascular aneurysm repair (TEVAR), transcatheter aortic valve replacement (TAVR) and mini-sternotomy valve surgery, which have been shown to improve outcomes in ESRD patients compared to open surgical repair [5, 7, 10]. Even within coronary artery bypass grafting (CABG), there have been survival benefits shown in the ESRD population with the use of internal mammary artery grafting [11] and off-pump CABG [12, 13].

The largest studies showing adverse outcomes with cardiac surgery in the ESRD population predate many of these advances. It is possible that overall improvements in the outcomes of patients with ESRD, changes in surgical techniques, and more judicious patient selection have combined to reduce ESRD morbidity and mortality rates. The purpose of this study was to evaluate the rates of morbidity and mortality in stable patients with ESRD undergoing non-emergent cardiac procedures. We used multivariable regression and propensity-score matching to compare ESRD patients to those without ESRD.

We hypothesized that cardiac surgical outcomes in patients with stable ESRD would be similar to the non-ESRD patients in this selected population.

Methods

Study Population:

We conducted a retrospective chart review from January 2011 to May 2016 utilizing the University of Colorado Society of Thoracic Surgeons (STS) database. Institutional review board approval was obtained from the University of Colorado with a waiver of informed consent. Subjects were included if they were ≥18 years of age and underwent non-emergent surgery at the University of Colorado Hospital. Patients were excluded for the following reasons: 1) acute kidney injury (AKI) at the time of surgery (in the non-ESRD population), 2) infection treated with antibiotics within 2 weeks prior to surgery, 3) post-operative infection arising within 48 hours of surgery, 4) procedures performed for infective endocarditis, aortic dissection, or heart transplant, 5) ventricular assist device placement, 6) cardiogenic shock at the time of surgery, 7) ESRD patients not on a stable dialysis regimen for at least one month prior to surgery. These inclusion and exclusion criteria were chosen to select a population of stable patients undergoing lower-risk procedures without infection or other known modifiable contributors to poor outcomes at the time of surgery.

Predictor Variable:

ESRD patients were defined based on a previous diagnosis of ESRD in the patient chart. To ensure that ESRD patients were stable and well-dialyzed, patients were excluded if not on dialysis for at least one month, and if they were not deemed to be compliant with dialysis at the time of surgery.

Outcomes:

The primary outcome was a combined endpoint consisting of 30-day mortality, post-operative stroke, post-operative infection, and prolonged ICU length of stay. Secondary outcomes were individual evaluations of 30-day mortality, post-operative stroke, post-operative infection, and prolonged ICU length of stay (LOS). Post-operative infection was defined as a new surgical site infection (SSI), either deep or superficial, positive blood, urine, or tracheal secretions culture, or imaging suggestive of new-onset pneumonia.

Matching Variables:

Variables used for adjustment were age, gender, hematocrit before surgery, cardiopulmonary bypass duration, intraoperative blood transfusions, previous cardiovascular interventions, and Charlson comorbidity index [14].

Statistical Analysis:

When comparing ESRD to non-ESRD patients, we risk-adjusted for covariates by two different methods: 1) propensity-matched cohorts as the primary analysis and 2) multivariable logistic regression as a sensitivity analysis [15]. For the propensity model the β-coefficients were combined with the patient’s values for each covariate to generate individual propensity scores. The patient-level propensity scores were then used to match patients with ESRD to similar patients without ESRD to produce the propensity-matched cohort using the 1-to-1 greedy matching method [16, 17]. The matching algorithm first matches pairs within eight decimal places of the propensity score. If there are no matches at eight decimals then the matching algorithm moves to seven decimal places and so on down to one decimal place. Generalized estimating equation (GEE) models were used to test the propensity-matched cohort and to account for correlation within each matched pair. For all outcomes and risk-adjusted results, statistical significance and confidence intervals were adjusted for multiple comparisons (i.e., four outcomes and two risk-adjustment methods) using the Bonferroni method by multiplying the p values or dividing the alpha levels by the number of comparisons. All statistical tests were considered to be significant at a 2-sided p < 0.05. All analyses were performed using SAS software version 9.4 (SAS Inc., Cary, NC).

Results

Baseline Data.

A total of 2,362 patients underwent cardiac surgery at the University of Colorado during the study period, of which 1,451 met inclusion criteria (Figure 1). Baseline characteristics of patients with and without ESRD are summarized in Table 1. Patients with ESRD were less often Caucasian, were more likely to have hypertension and diabetes, and had lower pre-operative hematocrit. ESRD patients were also much more likely to undergo coronary artery bypass compared to the non-ESRD population (Table 2).

Figure 1.

Cohort selection flow diagram depicting patient screening and exclusions.

Table 1.

Demographics, preoperative, and operative characteristics of patients with and without end-stage renal disease.

Characteristics	No ESRD (n=1,416) N (%)*	ESRD (n=35) N (%)*	P value†
Age, years, mean (SD)	62.6 (14.3)	58.8 (13.0)	.10
Female	421 (29.7)	10 (28.6)	.88
Caucasian race	1,199 (84.7)	23 (65.7)	.002
Body mass index, mean (SD)	28.4 (6.2)	27.3 (5.2)	.21
Hypertension	1,027 (72.5)	32 (91.4)	.01
Hyperlipidemia	846 (59.8)	25 (71.4)	.16
Diabetes	434 (30.7)	22 (62.9)	<.0001
Peripheral vascular disease	102 (7.2)	4 (11.4)	.32
Cerebrovascular disease	165 (11.7)	2 (5.7)	.42
Previous cardiovascular intervention	471 (33.3)	10 (28.6)	.56
Ejection fraction, mean (SD)	53.6 (14.8)	50.7 (13.2)	.24
Sleep apnea	277 (19.6)	8 (22.9)	.63
Chronic lung disease	327 (23.1)	8 (22.9)	.97
Chronic liver disease	41 (2.9)	1 (2.9)	.99
Cancer	128 (9.0)	4 (11.4)	.55
Hematocrit, mean (SD)	41.0 (5.7)	31.9 (4.3)	<.0001
White blood cell count, mean (SD)	7.3 (2.4)	6.7 (1.9)	.13
Charlson comorbidity index, mean (SD)	3.2 (1.9)	3.2 (1.9)	.95

Abbreviations: ESRD, end stage renal disease; SD, standard deviation.

^*Values are frequency and column percent unless otherwise specified.

^†P values were from chi-square or Fisher’s exact for categorical variables and Student t-test for continuous variables. All significant P values are bolded.

Table 2.

Type of procedure performed in patients with and without end-stage renal disease.

Procedure Type	No ESRD (n=1,416) N (%)	ESRD (n=35) N (%)	P value*
Aneurysm Repair	145 (10.2)	1 (2.9)	.15
Valve replacement or repair	415 (29.3)	7 (20.0)	.23
Coronary artery bypass grafting	417 (29.5)	18 (51.4)	.005
Off-pump	10 (2.4)	0 (0)	.35
Other	226 (16.0)	5 (14.3)	.79
Combined/Multicomponent†	213 (15.0)	4 (11.4)	.55
Aortic Valve Types
Mechanical	90 (6.4)	2 (5.7)	.66
Bioprosthetic	362 (25.6)	4 (11.4)
Intraoperative Variables
Cardiopulmonary bypass time, mean (SD)	125.2 (78.7)	130.9 (84.4)	.69
Intraoperative blood products	481 (34.0)	17 (48.6)	.07
Circulatory arrest	150 (10.6)	1 (2.9)	.14

^*P values were from chi-square or Fisher’s exact and was calculated as each surgery type compared to all other surgery types. All significant P values are bolded.

^†Multicomponent procedures including combined valve replacement or repair, coronary artery bypass grafting and or aneurysm repair.

Propensity Matching

Propensity score matching between ESRD and non-ESRD patients was successful as defined by an average standard error <0.1 on bivariate analysis following matching (Supplementary Table 1).

ESRD was not associated with increased postoperative complications.

Patients with ESRD at the time of surgery had rates of the composite outcome that were no different than the non-ESRD population in both the unadjusted (Table 3) and adjusted analyses (Table 4). The odds ratio in the propensity-matched model was 0.70 favoring the ESRD population, although this did not reach statistical significance. There were also no differences in unadjusted or adjusted analyses in the secondary outcomes of 30-day mortality, stroke, infection, prolonged ICU stay, or total hospital days in patients with ESRD, although there was a notable trend towards decreased infection in the propensity-matched ESRD group (OR 0.27, CI 0.07-1.07, p=0.06). Table 3 gives the absolute event rate in each group, and Table 4 provides to results of statistical analysis. Of note, because there were no mortality events and only one stroke event in the 35 patients with ESRD, these secondary outcomes could not be adjusted using multivariable or propensity-score matched analysis.

Table 3.

Bivariate Comparison of Early postoperative outcomes in patients with and without end-stage renal disease

Outcomes*	No ESRD (n=1,416) N (%)*	ESRD (n=35) N (%)*	P value†
	Composite Outcome	572 (40.4)		17 (48.6)	.33
	30-day mortality	27 (1.9)		0 (0)	.99
Post-operative infection	138 (9.8)	3 (8.9)	.82
Deep Sternal Wound Infection	17 (1.2)	0 (0)	.52
Postoperative sepsis	55 (3.9)	2 (5.7)	.65
Pneumonia	51 (3.6)	1 2.9)	.81
ICU length of stay > 72 hours	507 (35.8)	17 (48.6)	.12
Stroke	29 (2.1)	1 (2.9)	.52
Hospital length of stay, median (IQR)	8 (6-11)	8 (7-13)	.06
Disposition Location			<.001
Extended Care/Rehab/SNF	62 (4.4)	7 (20.0)
Other Acute Care Hospital	51 (3.6)	3 (8.6)
Home/Unknown	1276 (92.0)	25 (71.4)

Table 4.

Unadjusted and adjusted outcomes comparing the patients with and without end-stage renal disease (ESRD).

Outcome*	ESRD versus no ESRD
	Odds Ratio (95% CI)	P value
Composite Outcome⁂
Unadjusted†	1.39 (0.71-2.73)	.33
Multivariable adjusted†	0.86 (0.41-1.80)	.75
Propensity matched‡	0.70 (0.29-1.72)	.44
Post-operative Infection
Unadjusted†	0.87 (0.26-2.87)	.82
Multivariable adjusted†	0.55 (0.16-1.90)	.13
Propensity matched‡	0.27 (0.07-1.07)	.06
ICU LOS > 72 hours
Unadjusted†	1.69 (0.87-3.32)	.12
Multivariable adjusted†	1.14 (0.55-2.39)	.42
Propensity matched‡	1.13 (0.43-2.91)	.80
Hospital LOS§
Unadjusted†	1.19 (0.99-1.44)	.07
Multivariable adjusted†	0.94 (0.79-1.13)	.52
Propensity matched‡	0.89 (0.62-1.28)	.53

Abbreviation: ESRD, end-stage renal disease; CI, confidence interval; ICU, intensive care unit; LOS, length of stay.

^*30-day mortality and stroke could not be calculated due to no events in the ESRD group.

^†No ESRD N=1,416 vs. ESRD N=35

^‡No ESRD N=34 vs. ESRD N=34

^§A negative binomial model was used for this outcome.

^⁂Composite of 30-day mortality, post-operative stroke, post-operative infection, and prolonged ICU length of stay

Discussion

In this retrospective study of a selected group of stable ESRD patients undergoing non-emergent cardiac surgery, there were no statistical differences in outcomes compared to patients without ESRD. Previous studies generally have, though not universally, demonstrated worse outcomes for ESRD patients undergoing cardiac surgery. Parikh et al. demonstrated in patients undergoing CABG that mortality rates in ESRD patients were 3-fold higher (5.4% ESRD vs. 1.8% non-ESRD) when compared to non-ESRD patients[8]. Thourani et al. demonstrated that valve surgery for ESRD patients on pre-operative dialysis had significantly worse mortality both in terms of in-hospital mortality (5.2% vs. 18.3%) as well as overall survival (median survival 12.3 vs. 1.8 years) [18]. However, Powell et al. showed in a prospective, nested case-control study that ESRD patients undergoing CABG had no worse mortality than those without ESRD[19]. Similar results have been observed in non-cardiac surgeries as well[20–23].

Over the past decade, there have been meaningful improvements in overall ESRD survival [1], advances in surgical techniques, perioperative care, less-invasive surgical procedures [5, 7, 10], and more judicious selection of ESRD patients for surgery. We have evaluated a selected population of ESRD patients who were well-dialyzed and undergoing non-emergent surgery. In our study, the ESRD patients had an average age of 58.8, more than three years younger than the non-ESRD population. Also, there were no statistical differences in Charlson comorbidity score, suggesting that despite their renal disease this selected population had low rates of additional comorbidities. The highly selected ESRD population distinguishes our findings from a recently published study of 174 dialysis-dependent patients undergoing isolated CABG that found elevated operative risk. These ESRD patients had significantly higher rates of comorbidities, as evidenced by the fact that calculated pre-surgical mortality risk was nearly four times higher in the ESRD group (8.4% vs 2.3%, p<0.001)[24].

Under the circumstances in our study wherein ESRD patients were selected to be younger, with low rates of comorbidities and undergoing non-emergent surgery, ESRD mortality rates were not higher than in non-ESRD patients. There were no cases of 30-day mortality in our ESRD cohort. Similarly, there were no differences in rates of post-operative stroke, post-operative infection, prolonged ICU stay, or overall hospital length of stay. ESRD patients were less likely to be discharged home, however (Tables 3 and 4). In relatively young, healthy, and stable ESRD patients, surgical risks should not be considered elevated compared to the general cardiac surgery population. While more extensive studies are needed, it is possible that the pendulum has swung too far in the past ten years, and that ESRD patients are too often excluded from consideration when they might benefit from surgical intervention.

Recent studies suggest that ESRD patients benefit from advances in surgical techniques. For instance, off-pump CABG has been shown to provide survival benefits within the ESRD population[12, 13]. Less invasive cardiac procedures including thoracic endovascular aneurysm repair (TEVAR), transcatheter aortic valve replacement (TAVR) and mini-sternotomy valve surgery have also been shown to improve outcomes in ESRD patients compared to open surgical repair [5, 7, 10]. Notably, percutaneous interventions, when compared to CABG, have not been shown to improve long-term survival[24]. Finally, the use and type of circulatory arrest used during aneurysm repair can be a factor influencing kidney outcomes[25].In our report, there was no statistical difference in the utilization of circulatory arrest. We recently reported our outcomes on 295 patients undergoing open aortic arch surgery requiring hypothermic circulatory arrest. On the multivariable analysis, significant predictors of AKI were the history of hypertension (P = 0.03) and CPB time (P = 0.02). Interestingly, hypothermic circulatory arrest time was not associated with the AKI[26].

Interestingly, in this study ESRD patients were not more likely to undergo off-pump bypass or less invasive techniques. In fact, none of the ESRD patients underwent off-pump CABG. The rates of TAVR and TEVAR (listed under “other” in Table 2) were also not higher in ESRD patients. In our study; therefore, the improved mortality rates compared to other ESRD studies seem to result from careful patient selection rather than the use of less invasive surgical techniques.

Another potential reason for the equivalent results observed between ESRD and non-ESRD in our study may be the relatively high rates of CKD in the non-ESRD population. About a quarter of the surgical population had an eGFR < 60 (CKD 3 or greater) at the time of surgery, and just over half had an EGFR < 90 (CKD stage 2). Previous studies have shown that operative mortality is increased in patients with CKD 3, and is greatly increased in patients with CKD 4 or 5 not treated with dialysis[27]. Notably, the well-dialyzed ESRD patients had lower rates of mortality that did un-dialyzed CKD 5 patients in prior studies. The number of CKD 4 and 5 patients was too low in this dataset for statistical analysis, but it is possible that a period of dialysis before surgery may be beneficial in those with high-grade CKD, and should be a topic of future research.

A final potential reason for higher rates of ESRD mortality in prior studies may be the conflation of ESRD and AKI requiring dialysis. Large surgical databases including STS and ACS-NSQIP do not specifically record ESRD at the time of surgery, but rather have variables indicating that a patient was on dialysis at the time of surgery without distinguishing between dialysis-dependence due to ESRD vs AKI. Rates of mortality in critically ill ICU patients requiring dialysis for severe AKI are >50%, making this condition one of the deadliest conditions commonly seen in US hospitals [28]. Even a small number of AKI patients needing dialysis before surgery mixed with ESRD will dramatically increase rates of mortality. For instance, in our STS dataset approximately 20% of patients undergoing non-emergent surgery and marked as pre-operative dialysis in the STS database had AKI needing dialysis rather than ESRD. If we had used the category of “dialysis at the time of surgery”, rather than confirming ESRD status, mortality would have been 11% rather than 0%. Gajdos et al. evaluated mortality rates in ESRD patients undergoing elective surgery using ACS-NSQIP and reported mortality rates of 12.7% compared to 1.5% in non-ESRD patients [20]. However, dialysis at initiation was used to determine ESRD status.

Similarly, Thourani et al. demonstrated that valve surgery for ESRD patients on pre-operative dialysis had significantly worse mortality both in-hospital (5.2% vs. 18.3%), but used a pre-operative dialysis variable rather than an ESRD specific variable [18]. Invited commentary on the Gajdos study remarked, “Results from this study represent a more realistic estimate of operative risk in dialysis patients than the 1% to 6% operative mortality rate reported in single-center series [29].” It is possible that in actuality these studies were measuring different patient populations. Because patients with dialysis-dependent AKI tend to be sicker than ESRD patients, the use of Predicted Risk of Mortality (PROM)[30, 31] scores in the STS database going forward may mitigate some of these differences; nonetheless, we would advocate that STS and other large surgical databases begin categorizing patients with pre-operative ESRD and patients with pre-operative AKI requiring dialysis separately.

There are several limitations to this study. First is the small sample size of our ESRD patients, and the low event rates within the ESRD population, necessitating the use of a composite primary outcome. It is possible that our study was underpowered to show statistically significant differences; however, the OR for the composite outcome as well as for post-operative infection and hospital ORs in the propensity-score model were less than 1, making it less likely that significantly worse ESRD outcomes were masked by power issues. It should also be noted that while these patients represented the actual decision-making process of our surgeons, this group was a younger, stable subset of ESRD patients, limiting generalizability. While we were able to show using the Charlson comorobidity index that our ESRD patient population was relatively healthy, STS PROM scores, which have become more frequent in the literature, were not available. Finally, this study focuses on the early post-operative outcomes (within 30 days), so we cannot comment on the long-term prognosis of these patients after cardiac surgery.

Conclusion

Our study demonstrates that appropriately selected, stable ESRD patients undergoing non-emergent surgery may not be at increased risk of post-operative complications including mortality, infection, stroke, and prolonged ICU stay compared to general non-ESRD patients. Although larger studies are needed to validate our results, it is possible that improvements in ESRD care, surgical techniques, and patient selection have led to meaningful but underrecognized improvements in patients with ESRD undergoing cardiac surgery. Surgeons should consider risks and benefits on a case-by-case basis and should not exclude patients based on the presence of ESRD.

Figure 2.

Forest Plot showing propensity-matched outcomes in ESRD vs. no ESRD patients.