The Risk-Adjusted Impact of Intraoperative Hemofiltration on Real-World Outcomes of Patients Undergoing Cardiac Surgery

Methods: To address this we retrospectively followed 7620 patients who underwent CABG between April 2001 and March 2006. Logistic regression was used to risk adjust in-hospital outcomes. Cox proportional hazards analysis was used to risk adjust Kaplan-Meier freedom from death curves. Outcomes were adjusted for American Heart Association and American College of College of Cardiology recommended variables.


INTRODUCTION
Cardiopulmonary bypass is known to adversely impact upon perioperative renal function in particular for individuals with diabetes mellitus or an existing renal impairment [1]. Renal injury may occur resulting in acute kidney injury (AKI) which is the standard term for an abrupt and sustained decrease in renal function resulting in the retention of nitrogenous (urea and creatinine) and non-nitrogenous waste products. Depending on the severity and duration of the renal dysfunction, this accumulation is accompanied by metabolic disturbances, such as metabolic acidosis and hyperkalaemia, changes in body luid balance, and effects on many other organ systems. De initions of AKI have varied from the severe (i.e., that require dialysis) to the slight increases in serum creatinine concentration by 44·2 μmol/L [2]. Severe AKI, de ined as renal failure that requires renal replacement therapy is a major complication of cardiac surgery with an associated mortality of 60-100% when treated with standard hemodialysis [3]. More recently, continuous renal replacement techniques have been introduced, which circumvent the hemodialysis instability associated with intermittent hemodialysis and its limited ability to control the patient's volume state [4]. One such form of continuous renal replacement therapy is continuous venovenous hemodialysis (CVVH) that has been shown to substantially facilitate improved care of patients with severe AKI after cardiac operations when used early and intensively [5].
Hemo iltration was introduced in the 1980s for CPB management [6][7][8][9], and has been shown in pediatric cardiac surgery to be effective at regulating hemostasis, removal of cytokines and complement components [10]. Similarly, in adults zerobalance hemo iltration was shown to be effective at removing in lammatory mediators during CPB [10,11]. In addition, one recen.t prospective, randomized double-blind trial demonstrated that hemo iltration but not steroids results in earlier tracheal extubation following CPB [12]. Initially hemo iltration was intended to correct the accumulation of extravascular water during or immediately following the surgical procedure. Several new bene its have been identi ied, such as the reduction of postoperative blood loss and immediate improvement in hemodynamic [13,14].
In addition, recent evidence has shown that mild renal dysfunction is a predictor of outcomes in terms of in-hospital mortality, morbidity, and midterm survival in patients undergoing coronary artery bypass graft" (CABG) surgery [15]. Since hemo iltration is well tolerated in patients with compromised circulatory status this technique has become useful in the postoperative period for treating patients with acute renal failure. Indeed, early and aggressive continuous venovenous hemo iltration (CVVH) has been associated with better than predicted survival in severe postoperative AKI [16]. However it is still unknown as to whether real-world outcomes of cardiac operations incorporating intraoperative hemo iltration as applied during cardiopulmonary bypass surgery are markedly different from those with postoperative hemo iltration initiated in intensive care units or wards. The objective of this study was to analyse real-world data to assess whether outcomes differ for patients with preoperative kidney disease treated with hemo iltration modalities either during CPB or 1-2 days after coronary artery bypass graft (CABG)" to "CABG.

Patients
We retrospectively followed 7620 patients who underwent on-pump irst-time isolated cardiac bypass graft surgery between April 2001 and March 2006. All data were collected (Table 1) during in-patient admission as part of the routine clinical practice as previously described [17]. In-hospital mortality was de ined as deaths within the same hospital admission regardless of cause or within 30 days of hospital discharge. Postoperative stroke was de ined as a new focal neurological de icit and comatose states' occurring postoperatively that persisted for 24 hours after its onset and was noted before discharge. Transient confused states and temporary intellectual impairment were not included. Postoperative myocardial infarction was de ined as unequivocal ECG changes, elevation of cardiac enzyme(s) to 3 times upper limit of creatinine kinase (CKMB) and above twice the upper limit of normal or elevated troponin (T/I), and chest pain typical of ischemia lasting for more than 20 minutes. The European Risk Strati ication Score System (EuroSCORE) was used to assess risk associated with cardiac surgery [18].
Patients with renal dysfunction were described as those with an estimated glomerular iltration rate (eGFR) <60 mL/minute/1.73 m 2 that has persisted for at least 3 months before the date of surgery. We strati ied patients into 3 groups: (1) patients treated with intraoperative hemo iltration (n=112) during cardiopulmonary bypass; (2) patients treated with postoperative hemo iltration (n=38); (3) a control group without hemo iltration use (n=7006). Patients with preoperative dialysis dependent renal insuf iciency were excluded from further analysis.
Intraoperative hemo iltration was applied for either secondary renal impairment or correction of luid retention due to cardiac failure/excessive hemodilution. Standard intraoperative hemo iltration involves connecting a parallel circuit to the cardiopulmonary bypass via hollow ibre ilters (Polysulfone PS, Hospal Lyon, France). Blood low rates were maintained at 300-500 ml/min (about 10% of the pump output) with a transmembrane pressure of about 300 mm Hg.
Cases of severe AKI were identi ied postoperatively and treated with continuous venovenous hemo iltration (CVVH). AKI was as previously de ined by the KIDGO consensus conference [19,20]. The technique of CVVH involved a double lumen catheter connected to a blood pump, through which blood was pumped into a module at low rates of 200 to 250 mL/m with an intravenous pump controlling the ultra iltration rate at 2 L/h. The module was also equipped with pressure alarms and an air bubble trap. Standard Polysulfone PS ilters were used in all patients. Anticoagulation of the circuit was performed in accordance with local clinical practice governed by duration of hemo iltration. Replacement luids were administered at the pre-ilter stage with rates chosen to suit the desired luid therapy goals.

Patient follow-up
The patients were followed up through a national tracing service using patient's name, National Health Service unique number and date of birth, gender, and postcode.

Study endpoints
The primary endpoint was the risk-adjusted freedom from death within the 5-year study period. Secondary outcomes include risk-adjusted in-hospital death, atrial ibrillation, myocardial infarction, stroke, re-operation for bleeding, and surgical wound infections.

Statistical analyses
Due to non-normality of data, continuous variables are shown as median with 25th and 75th percentiles. Categorical variables are shown as a percentage. Comparisons were made with Kruskal-Wallis tests and Chi-square tests as appropriate. Logistic regression was used to risk adjust in-hospital outcomes for differences in patient and disease characteristics [21]. Deaths occurring as a function of time were described using the product limit methodology of Kaplan and Meier [22]. Cox proportional hazards analysis was used to calculate adjusted hazard ratios (HR) and to risk adjust the Kaplan-Meier survival curves [23]. Logistic EuroSCORE [18] was used for risk adjustment. This approach aims to balance out patient preoperative characteristics by incorporating a commonly used cardiac surgery risk scoring method. In all cases a p value <0.05 was considered signi icant. All statistical analysis was performed with SAS for Windows Version 8.2.

In-hospital outcomes
A total of 113 patients had intraoperative HF, 38 had postoperative HF and 7006 had no HF. Tables 1 and 2, shows patient's preoperative characteristics and in-hospital outcomes. In the unmatched data (Table 2), we observed increased in-hospital death, and re-operation for bleeding outcomes in both the HF groups as compared to the control group without HF. More volume was removed from the postoperative haemo iltration group (Table 2). Interestingly, all the patients in the postoperative HF group had preoperative renal impairment and that 97% of these patients went on to developed acute kidney injury that necessitated hemo iltration in intensive care unit.
In contrast, 38% of the patients on intraoperative hemo iltration had preoperative renal impairment and 27% developed perioperative acute kidney injury. As shown on table 3, only 39.5% of patients with preoperative renal dysfunction that undergone intraoperative HF had AKI compared with 33.5% of those in the group that had no preoperative kidney impairment. This compares favourable with patients that had the conventional delayed postoperative CVVH where all patients who had preoperative kidney disease suffered AKI except one.
The adjusted odds ratios (OR) for in hospital outcomes (Table 4) indicated that in both intraoperative and postoperative HF groups there was no association between the use of HF with in hospital strokes, re-operation for bleeding, myocardial infarction and surgical wound infections. In contrast, a strong association was shown with inhospital death for intraoperative HF (OR 2.8, CI 1.5-5.1, P<0.001) and more so in the postoperative HF group (OR 12.1, CI 5.7-25.6, P<0.001). Atrial ibrillation was also signi icantly associated with postoperative HF group but not the intraoperative HF group.

Freedom from death analysis
Freedom from death curves are shown in igure 1. After adjusting for differences in a number of confounding factors (Figure 2), there was a signi icantly higher mortality during the follow-up period for the postoperative HF patients as compared with the intraoperative HF group.

DISCUSSION
The application of hemo iltration before and after CABG surgery is very common for a number of indications including luid overload and preoperative kidney impairment. Studies [24] have shown that up to 10% of patients sustain renal dysfunction due to kidney injury induced by CPB during CABG surgery [25,26]. Karkouti et al. [27], also demonstrated that CPB was associated with a 33.6% prevalence of postoperative acute kidney injury (AKI), which was associated independently with a greater than 4-fold increase in the odds of death. The indings are consistent with our observations of 6.4% postoperative AKI rates for patients with no history of preoperative kidney disease and that these rates increase signi icantly (3.5-15 fold) for patients with a history of preoperative kidney disease irrespective of the type of renal replacement intervention used. These indings are in agreement with previous studies [15] that also demonstrated that preoperative mild kidney disease was highly predictive of both AKI and in-hospital deaths [15].
Mechanisms related to the occurrence of AKI in these patients is however still uncertain. Several studies [28][29][30] have previously demonstrated that renal injury sustained during cardiac surgery may be mechanistically related to pre-existing renal dysfunction, diabetes mellitus, ventricular dysfunction, older age, hypertension, microembolic and macroembolic processes, in lammatory mediators, prolonged CPB time, sensitivity to sympathetic stimulation, and perturbation in renovascular resistance and low. Indeed indings from our study suggest that when data are corrected for pre-existing renal dysfunction, diabetes mellitus, ventricular dysfunction, older age, hypertension, the risk of AKI is signi icantly reduced for patients on intraoperative HF, suggesting that multiple comorbidities contribute to incidents of AKI. Interestingly, the incidents of AKI in patients with preoperative kidney injury were comparable to those without when intraoperative HF is given. This may suggest that intraoperative HF may protect patients from CPB-induced injury.
Other authors [31] have also demonstrated that the occurrence of postoperative mild to moderate acute kidney injury (AKI) was associated with a mortality rate of 10% to 20%, indings that are consistent with results from our study. Interestingly, AKI incidents in this study were associated with even much higher in-hospital mortality rates (50%) despite the application of postoperative CVVH. When the results are corrected for preoperative chronic kidney disease and other risk factors, the evidence also suggests that preoperative kidney disease is the single most signi icant risk factor for in-hospital mortality irrespective of renal support treatment modalities.
Currently, there is a lack of consensus on indications for and timing of initiation of renal support for patients with preexisting renal impairment or those at risk of AKI in the perioperative period. Some studies have shown that postoperative HF improves outcomes if initiated early [32][33][34][35][36]. In contrast, others such as Bouman and co-workers in 2003 [37] studied the effects of the initiation time of continuous venovenous hemo iltration on patients with circulatory and respiratory insuf iciency leading to oliguric AKI. The authors observed that recovery of renal function was not improved by using high ultra iltrate volumes or by early initiation of hemo iltration. In addition, other randomised studies failed to demonstrate that intensity/duration of renal replacement therapy could improve outcomes [38,39]. Issues remain regarding the proper de inition of AKI [19,20]. While there is an unanimous agreement that renal complication severely impact patient outcome, speci ic aspects of management, including the method of renal replacement therapy, i.e. dialysis based on the diffusion principle or hemo iltration (based on convection); the dose of renal replacement therapy (Urea KT/V for dialysis vs. hemo iltration low (ml/kg/hour)), and the timing (early vs. late) of renal replacement therapy are all controversial. Even after a consensus  de inition was put forward [the R.I.F.L.E classi ication: 2nd Consensus Conference, ADQI Group [40][41][42], there is still great uncertainty as what is the best course of action when treatment decisions have to be made speci ically aimed at patients with preexisting renal impairment. Existing evidence would suggest that even slightly elevated serum creatinine predicts AKI requiring hemo iltration after cardiac surgery [43]. Despite the limitations, our study is the irst to show that in high risk category of patients with pre-existing renal impairment, hemo iltration during cardiopulmonary bypass can minimise incidences of postoperative AKI or has survival bene its.
Based on the observation from our study that patients with high preoperative risk scores (EuroSCORE >16.1) are at increased risk of further renal impairment we hypothesize that it is these types of patients who could derive signi icant bene its from intraoperative hemo iltration. Indeed our most recent randomised controlled trial on intraoperative hemo iltration during cardiac surgery for patients with preoperative chronic kidney disease demonstrated superior outcomes as compared with treatment as usual. This is in terms of signi icant reduction in rates of pulmonary infections and in-hospital deaths compared with a control usual care group without the use of intraoperative hemo iltration [44].
The randomised trial [44] appear to support the indings from this real-world study setting that after adjusting for case mixes intraoperative hemo iltration is associated with more favourable short term outcomes such as in-hospital death and reduced risk of atrial ibrillation as compared with equivalent patients that received the traditional delayed postoperative hemo iltration care.
The underlying mechanisms for these bene its have yet to be elucidated. It is possible that the bene its of intraoperative hemo iltration (conventional or modi ied ultra iltration) can be explained by the elimination of in lammatory mediators during the metabolic stressful period of coronary bypass, as demonstrated in studies before [6][7][8][9][10]. In addition, there is also evidence to suggest that intraoperative hemo iltration reduces levels of S100b release and cognitive dysfunction [11], resulting in an earlier tracheal extubation following cardiopulmonary bypass [12]. In other studies it has also been shown that intraoperative hemo iltration may reduce the requirement for post-operative blood transfusion, also increase BP and cardiac index immediately after hemo iltration [45].
In conclusion, indings from this study support the hypothesis that the use of intraoperative hemo iltration may offer short term clinical bene its and improve longer-term survival for patients with preoperative kidney disease. In contrast, the delayed use of postoperative hemo iltration on patients with known preoperative kidney impairment does not signi icantly impact occurrence of AKI and the impact of AKI on survival. It is with this reasoning and evidence from other studies [46][47][48] that we hypothesize that for patients with preoperative kidney disease routine application of intraoperative haemo iltration or early CVVH rather than the traditional delayed CVVH (which is commonly started 1-3 days after CABG in response to mostly oliguria symptoms and electrolytes imbalance) may improve outcomes of these vulnerable patients. The recent introduction of hemodial iltration techniques incorporating citrate-based regional anticoagulation systems soon after CPB may be a game changer for patients with preoperative kidney disease in routine clinical practice [49].

STUDY LIMITATIONS
Although our study included a large sample of patients undergoing coronary artery bypass surgery, it was not a prospective randomised trial, and therefore had a number of limitations. For example, patients who received HF on bypass were likely to have had less severe renal impairment and a lower EuroSCORE, and therefore this might impact upon the results giving the Intra-Op HF Group a better outcome because they were going to do well regardless. Our study, however, considered three unselected study groups of patients on the basis of use and timing of hemo iltration and therefore provides an insight into real practice. In addition, the follow-up data obtained via the National Institute of Statistics only provided us with knowledge of the survival status without the cause of death or differentiation between cardiovascular or other cause of death. We also do not have data on the underlying pathophysiology of renal dysfunction in the patients included in the study. The indings however suggest that a well-designed randomised study would offer a better evaluation of the treatment strategies for this group of patients.