Predictors of Mortality in Patients Treated with Continuous Hemodiafiltration for Acute Renal Failure in an Intensive Care Setting

Abstract

We performed this study to identify predictors of mortality in critically ill patients treated with continuous venovenous hemodiafiltration (CVVHDF) for acute renal failure in an intensive care setting. It was an uncontrolled, observational study that took place in a general intensive care unit in a university hospital. Forty-one patients undergoing CVVHDF for acute renal failure in a consecutive sample of 1,018 ICU treatments were studied. The underlying disease included 25 postsurgical cases and 16 medical cases. Between survivors (n = 23) and nonsurvivors (n = 18), the following factors were assessed: demographic data; the number and type of failed organs; Acute Physiology and Chronic Health Evaluation (APACHE) II scores; urine production; pH; base excess; serum creatinine levels; bilirubin levels; lactate levels; platelet counts; and hemodynamic variables, including cardiac index and central venous pressure. On univariate analyses, the number of failed organs (p < 0.01), presence of hepatic failure (p < 0.01), APACHE II scores (p < 0.01), pH (p < 0.01), base excess (p < 0.001), average urinary production before the initiation of CVVHDF (p < 0.05), and serum bilirubin (p < 0.01) and lactate levels (p < 0.001) were significantly different. Multiple regression analysis identified serum bilirubin (p < 0.01) and lactate levels (p < 0.01) as the predictors of hospital mortality. Presence of hepatic failure was also predictive of hospital mortality (p < 0.01) in the analysis of the type of organ failure. The cut-off value set at bilirubin levels > 10 mg/dl or arterial lactate levels > 3.5 mmol/L provided 83.3% sensitivity and 90.9% specificity in the prediction of hospital death. The crucial factors in predicting outcome of critically ill patients undergoing CVVHDF for renal failure are elevated serum bilirubin and lactate levels at the onset of CVVHDF. Presence of hepatic failure, defined as both jaundice and coagulopathy, may also worsen outcome of critically ill patients undergoing CVVHDF for renal failure. The cut-off value set at bilirubin levels > 10 mg/dl or arterial lactate levels > 3.5 mmol/L may serve as beneficial predictors of hospital mortality.

ontinuous renal replacement therapy (CRRT), such as continuous venovenous hemodiafiltration (CVVHDF) or hemofiltration (CVVHF), has proven beneficial for the treatment of critically ill patients with acute renal failure. ^1–3 The advantages of continuous over intermittent dialysis include more precise fluid and metabolic control, decreased hemodynamic instability, and the potential benefit of removing proinflammatory cytokines. ^4,5 However, survival rates of those with multiple organ dysfunction syndrome (MODS) rarely exceed 50%, despite the introduction of continuous renal replacement therapy. ^6,7 The unsatisfactory survival rate is likely due to the fact that the age of patients continues to rise, surgical treatment is now indicated for patients with severe preexistent complications, and that more severe cases may survive a longer period, which results in the increasing prevalence of acute renal and other organ failure. ^3,8 Identification of prognostic factors in these patients for hospital mortality may contribute to more effective use of limited medical resources. The purpose of this study was to identify predictors of mortality in critically ill patients treated with continuous venovenous hemodiafiltration (CVVHDF) for acute renal failure in an intensive care setting.

Patients and Methods

The study subjects were selected from patients treated in the Intensive Care Unit (ICU) at Hokkaido University Hospital from April of 1996 through January of 1999. Among 1,018 patients who entered the ICU during the study period, a total of 41 patients (4.0%) treated with CVVHDF for acute renal failure were enrolled in the study. Any patients undergoing other blood purification, such as plasma exchange, direct hemoperfusion, or intermittent hemodialysis were not included in the study population. In addition, patients who underwent CVVHDF for nonrenal indications were also excluded from the study. The study subjects included 21 men and 20 women, with the mean age of 56.4 ± 3.2 years (range, 5-79 years). Historical, clinical, and laboratory data were collected in the period 24 hours before and at least 48 hours after the introduction of CVVHDF treatments. Patients were divided into two groups, survivors or nonsurvivors, at the time of discharge from the hospital. For patients who were discharged from the ICU and returned to the general ward, the outcome was followed up until each patient was discharged from the hospital. If the patient died after being transferred to another hospital, the patient was assigned to the nonsurvivors group. Mortality in this study was determined by hospital mortality, not by ICU mortality.

The underlying diseases in the 41 patients included 14 cases of postcardiovascular surgery renal failure, 7 cases of postgastrointestinal tract surgery, 4 cases of post-liver transplantation, 4 cases of metabolic disorder or autoimmune disease, 5 cases of hematologic or malignant disease, 3 cases of infectious disease or sepsis, 1 case of hepatic failure unrelated to liver transplantation, 2 cases of post–bone marrow transplantation, and 1 case of cardiomyopathy (Table 1). The primary indication for CVVHDF was renal failure in all 41 cases. CVVHDF was initiated regardless of the underlying disease when the patient met at least one of the following criteria, which were modified from those reported by Schwilk and associates ⁹ : (1) oligouria or anuria, despite maximum diuretic treatment with furosemide, mannitol, or both, in the presence of sufficient hydration; (2) insufficient water clearance resulting in unacceptable positive fluid balance and increasing tissue edema or increasing central venous pressure, despite maximum diuretic treatment; (3) insufficient and unacceptable electrolyte clearance resulting in serum potassium concentration > 5.5 mEq/L, or decreasing pH values, despite maximum diuretic treatment and other conventional medical treatment; (4) decreasing urine production and increasing serum creatinine levels above 3.0 mg/dl or serum urea above 80 mg/dl BUN, despite maximum diuretic treatment. In the presence of single or multiple organ failure, CVVHDF may be started earlier with an increasing serum creatinine level above 2.0 mg/dl or serum urea above 60 mg/dl.

Table 1: Comparison of Underlying Diseases

Hemodiafiltration Technique

Vascular access for venovenous hemodiafiltration was obtained by inserting a flexible double-lumen catheter (Flexxicon Dual Lumen Catheter, 11 Fr, Vas-cath, Inc., Mississauga, Ontario, Canada) into either the femoral vein, internal jugular vein, or subclavian vein. A system for automated renal replacement therapy (JUN-600 or JUN-500, UBE Medical, Inc., Tokyo, Japan) was used. The dialyzer used for CVVHDF was a cellulose triacetate hollow fiber dialyzer with a surface area of 0.7, 1.1, or 1.5 m² (UT-700, UT-1,100, or FB-150U, Nipro, Inc., Osaka, Japan). The flow rate of the dialysate (Sublood-B; sodium 140, potassium 2, calcium 1.75, magnesium 0.5, chloride 111, bicarbonate 35, acetate 3.5, and glucose 5.51 mmol/L, FUSO Pharmacy, Inc., Osaka, Japan) was held at 500-1,500 ml/hr, depending upon clearance. Blood flow was maintained at 80-100 ml/min. Ultrafiltrate was obtained at 300-500 ml/hr to control fluid balance and was replaced with Sublood-B in the postdilutional mode. A continuous infusion of nafamostat mesilate (10-20 mg/hr) into the circuit before the hemofilter was used as the anticoagulant. There were no patients who were converted from intermittent to continuous therapies in this study. CVVHDF or CVVHF performed in the ICU is usually carried out by intensivists. Some patients were converted from continuous to intermittent therapies when the individual patient’s condition became hemodynamically stable. To prevent hemodynamic deterioration by the use of intermittent therapies, the first two or three intermittent dialyses for these patients were performed by intensivists in the ICU. There were no technical or methodologic differences between intensivists and nephrologists.

Nutritional support for all patients was provided according to the principles described in the textbook by Ayres and associates. ¹⁰ Resting energy expenditure (REE) was calculated by the Harris-Benedict equation in all patients every day, and total estimation of energy requirement was calculated by Calvin Long’s activity factor, injury factor, and REE. In addition, indirect calorimetry was used to calculate energy expenditure with a Puritan Bennett 7,250 metabolic monitor (Puritan Bennett, Inc., Carlsbad, CA) for adult patients. Daily administered calories varied with patient body weight and activity and injury factors, but, in principle, the caloric requirement was fully administered unless the blood sugar level was abnormally high. Because the fluid balance is easily regulated with the use of CVVHDF, administered fluid volume was not restricted after the institution of CVVHDF. Nutritional support was usually performed by the parenteral route, but support in patients where the gut was available was done by means of the enteral route if possible. For assessment of nutritional support, routine laboratory tests were performed in the central laboratory every 8–12 hours, and fluid balance was checked every 2 hours in all patients. These laboratory tests included blood cell counts, serum biochemical assay, and urine analysis (except for those anuric patients). In addition to the routine laboratory tests, arterial blood gas analysis was performed 4 to 6 times a day using a Radiometer ABL 520 and EML 100 (Radiometer, Inc., Copenhagen, Denmark). Blood sugar levels, lactate levels, and electrolyte levels, including sodium, potassium, chloride, and calcium, were simultaneously measured with every measurement of arterial blood gas tension. These data were used as references in the supplementation of calories, electrolytes, and nitrogen. All patients in the ICU received equal nutritional support and assessment; thus, there was no evident bias or difference of the intensity of management between survivors and nonsurvivors.

Between survivors (n = 23) and nonsurvivors (n = 18), the following factors were compared by univariate and multivariate analyses: routine demographic data such as age or gender, the number and type of failed organs, Acute Physiology and Chronic Health Evaluation (APACHE) II scores, urine production, pH, base excess, serum creatinine levels, bilirubin levels, lactate levels, blood sugar levels, platelet counts, and hemodynamic variables, including cardiac index and central venous pressure. Total time on CVVHDF and the length of ICU stay were also compared. In principle, data were at least acquired and assessed at the onset of CVVHDF. APACHE II scores were calculated at the time of admission to the ICU, because this score is originally designed to be used as a predictive tool at the time of admission to the ICU. Presence of organ failure was determined according to the definition of Knaus and Wagner, ¹¹ and systemic inflammatory response syndrome (SIRS) according to the report by the American College of Chest Physicians/Society of Critical Care Medicine Consensus Conference. ¹²Table 2 shows the definition of organ failure and SIRS. Infection is defined as a microbial phenomenon characterized by an inflammatory response to the presence of microorganisms or the invasion of normally sterile host tissue by those organisms. ¹² Sepsis, listed as an underlying disease in this study, is defined as the systemic inflammatory response to infection. ¹² The average urine production during 6 hours before the initiation of CVVHDF was recorded as one prognostic value that might be indicative of the level of residual renal function.

Table 2: Definitions of Organ Failure and Systemic Inflammatory Response Syndrome

Statistical Analysis.

All continuous variables are presented as the mean ± standard error (SEM). Univariate analysis was conducted with Student’s t-test for comparisons of continuous variables, Mann-Whitney U-test for nonparametric comparisons, and chi-square for dichotomous variables. Covariates were entered into the logistic regression analysis to determine predictors for hospital mortality, at a p value of < 0.05 in univariate analysis. All analyses were performed with SPSS software (SPSS, Inc., Chicago, IL). Difference was considered significant at p < 0.05.

Results

Table 1 summarizes the list of underlying diseases in both survivors and nonsurvivors. Survival rate in cases of postcardiovascular surgery seems to be relatively higher than that in other groups. However, difference in underlying disease between survivors and nonsurvivors was not significant (p = 0.180). According to the definitions of infection and sepsis by the American College of Chest Physicians/Society of Critical Care Medicine Consensus Conference, the presence of infection, sepsis, or both, as an underlying disease were found in three cases: two in survivors and one in a nonsurvivor. Table 3 shows the results of univariate analysis of risk factors. The number of patients suffering from each type of organ failure and the comparison of failed organs between survivors and nonsurvivors are also shown. The following factors were significantly different between survivors and nonsurvivors: number of failed organs (1.57 ± 0.11 vs. 2.11 ± 0.14;p = 0.003), presence of hepatic failure (3 of 23 vs. 11 of 18;p = 0.002), APACHE II scores (21.1 ± 1.5 vs. 28.1 ± 1.6;p = 0.004), pH (7.42 ± 0.01 vs. 7.35 ± 0.02;p = 0.003), base excess (2.14 ± 0.97 vs. −4.08 ± 1.12;p < 0.001), the average urinary production during the 6 hours before initiation of CVVHDF (1.52 ± 0.26 vs. 0.90 ± 0.34 (ml/kg per hr);p = 0.047), serum bilirubin levels (2.8 ± 0.7 vs. 7.8 ± 2.2 (mg/dl);p = 0.006) and lactate levels (1.69 ± 0.16 vs. 7.41 ± 1.55 (mmol/L);p < 0.001). No significant difference was found in blood sugar levels or platelet counts. Hemodynamic variables at the onset of CVVHDF did not show any significant differences in terms of central venous pressure or cardiac index.

Table 3: Univariate Analysis of Risk Factors

Multivariate analysis to determine predictors of hospital mortality was performed among covariates with p < 0.05 on univariate analysis. Number of failed organs, APACHE II scores, pH, base excess, urinary production just before the onset of CVVHDF, serum bilirubin levels, and lactate levels were entered into logistic regression analysis. Table 4 summarizes the standardized regression coefficient with a 95% confidence interval and p value for each factor. Serum bilirubin (p = 0.006) and lactate levels (p = 0.006) were identified as the significant predictors of hospital mortality. Table 5 also shows the result of multivariate analysis to identify that organ failure associated with significant risk for hospital mortality. Among the organ systems investigated, presence of hepatic failure was identified as the only significant predictor (p = 0.002) for hospital mortality.

Table 4: Multivariate Analysis to Identify Incremental Risk Factors for Hospital Mortality

Table 5: Multivariate Analysis to Identify That Organ Failure at Significant Risk for Hospital Mortality

Figures 1 and 2 show the changes in serum bilirubin and lactate levels for the first 48 hours after the onset of CVVHDF; these variables demonstrated significant differences between the two groups. From these variables, an attempt was made to determine the cut-off value beneficial for the prediction of mortality. When the cut-off value for predicting hospital death was set at bilirubin levels > 10 mg/dl or arterial lactate levels > 3.5 mmol/L, it provided 83.3% sensitivity and 90.9% specificity.

Figure 1: Changes in serum bilirubin levels after the onset of continuous venovenous hemodiafiltration.

Figure 2: Changes in lactate levels after the onset of continuous venovenous hemodiafiltration.

Discussion

In critically ill patients, continuous forms of renal replacement therapy are preferred for their improved cardiovascular tolerance. They have proved to be beneficial modalities for treatment of acute renal failure. ^1,3,13 This technique is now widely used in the ICU in various situations. In the ICU in our institute, the first choice in renal replacement therapy for patients with acute renal failure is CVVHDF or CVVHF, because the majority of patients in the ICU are hemodynamically unstable at the time of admission. Of course, the use of conventional hemodialysis is considered if the hemodynamic conditions are very stable. In our institute, determination of hemodynamic stability or instability is usually made depending on the use of inotropes and the need for pulmonary artery catheter monitoring. Study subjects may have been identified as very sick at the time of referral to the ICU, but there is no bias in the selection of renal replacement therapy after admission. In a few patients undergoing cardiovascular surgery, the introduction of CVVHDF was postponed until postoperative bleeding was controlled. However, the outcome in these patients was not poor, and the few hours’ delay did not affect the prognosis.

In patients treated with CVVHDF or CVVHF, the following factors have been reported as predictors of mortality: age, need for artificial ventilation, use of inotropes, urine volume, serum bilirubin, arterial base deficit, and number of failed organs. ^6,9,14 The present study identified serum bilirubin and lactate levels as significant predictors of hospital mortality. Impact of hepatic failure on the outcome of the patients studied was also demonstrated in this study. It has been well documented that elevated serum bilirubin levels predict a worse outcome in patients treated in the ICU. ^6,15,16 The prognosis of patients with single organ failure is generally acceptable, but it becomes worse if associated with liver dysfunction. ^3,14 This is likely due to the fact that renal replacement therapy is a beneficial substitute for filtrative function of the kidneys, but that there is no complete substitute for the complex functions of the liver.

Age was not identified as a significant predictor of outcome in this study. As a feature of the general ICU population, some patients with blood diseases are referred from the pediatric department for the treatment of acute renal failure after chemotherapy or bone marrow transplantation. Due to the use of agents that might impair renal function, recovery of renal function is usually delayed. In addition, these patients are at higher risk for the development of infection or sepsis due to immunosuppression. We have five cases of hematologic disease and two cases of post–bone marrow transplantation in this study, and prognosis for these patients was generally poor. The poor prognosis for patients from pediatrics was the main reason that age was not identified as a predictor of outcome in this study.

Measurements of lactate levels have been available during the past decade, but recent advances in blood gas analyzer systems allowed for routine measurement of arterial lactate levels at the bedside simultaneous with measurements of blood gas tensions. Lactate levels generally reflect the adequacy of peripheral circulation ^17,18 and have been identified as a predictor of outcome in critically ill patients. ^19–21 Our study also identified lactate levels as the second most important predictor of mortality. The elevation of lactate levels generally occurs in the presence of diminished peripheral circulation, resulting in impaired tissue oxygen delivery. ^17,18,22 In addition, a marked elevation often occurs in the presence of liver dysfunction or failure. ²³ Therefore, it may be suggested that the value of lactate levels is well correlated with outcome of critically ill patients. ^17,18,22 However, there is a pitfall in the interpretation of elevated lactate levels in the general or surgical ICU. There are two types of lactic acidosis based on the adequacy of tissue oxygen delivery. ^22,24 In patients after cardiopulmonary bypass, lactate levels may rise regardless of the patient’s condition or the presence of peripheral circulation insufficiency (type B). ²² In these patients, the cardiac index or presence of a low output syndrome will be a better predictor than the lactate levels.

Although the APACHE II scoring system seems to be beneficial for predicting hospital mortality in a variety of settings, outcome prediction of patients treated for CHDF has been reported as inaccurate. ^9,25 Barton and associates described a different scoring system based on the age, need for artificial ventilation, use of inotropes, urine volume, and serum bilirubin, that had a specificity of predicting death of 67% and a sensitivity of 76%. ⁶ The Liano score has been reported to identify a group of patients on dialysis for acute renal failure with a near 100% chance of mortality, but did not show discrimination between those patients who died in hospital and those who did not. ²⁶ As is shown in the report by Douma and associates, the use of both the APACHE and Liano scores would be better in predicting hospital mortality. The cut-off value, bilirubin levels > 10 mg/dl or lactate levels > 3.5 mmol/L, seems to be simpler and more useful for prediction of hospital mortality, which provides a specificity of predicting death of 90.9% and a sensitivity of 83.3%.

As a prognostic value, we measured serum creatinine levels every 12 or 24 hours in all cases and recorded the hourly volume of urine production during the stay in the ICU. The average urine production during the 6 hours before initiation of CVVHDF, along with serum creatinine levels, might be indicative of the level of residual renal function in this study. Serum creatinine levels were not significantly different between survivors and nonsurvivors, but the average urine production in the survivors’ group was higher than that in the nonsurvivors’ group. This finding suggests that nonoligouric versus oligouric renal function is likely to be one determinant of prognosis in patients requiring CVVHDF in an intensive care setting.

We investigated predictors of outcome in the general ICU population in this study. Of course, predictors may vary with different subjects, depending upon the medical or surgical ICU. However, identification of predictors of outcome in the general ICU would be required to develop standards in this field, as patients in all departments may enter and receive treatment. It is frequently pointed out that mortality due to acute renal failure remains high, despite recent progress in therapeutic modalities and strategy. ^1,6,7 This finding is probably because a large-scale, randomly assigned trial comparing CVVHDF and intermittent hemodialysis has not been reported. However, such a prospective, randomly assigned trial would be very difficult to carry out, because the application of intermittent hemodialysis in patients with hemodynamic instability is potentially hazardous. The severity of illness in patients treated in the ICU has apparently changed during the past decade; it, therefore, seems illogical to compare mortality regardless of changes in demography and severity of patients illness. ^9,27

In conclusion, the crucial factors in predicting outcome of critically ill patients treated with continuous venovenous hemodiafiltration for renal failure were elevated serum bilirubin and lactate levels at the onset of hemodiafiltration in our study subjects. Presence of hepatic failure, defined as both jaundice and coagulopathy, may also worsen outcome of critically ill patients undergoing CVVHDF for renal failure. The cut-off values set at bilirubin levels > 10 mg/dl or arterial lactate levels > 3.5 mmol/L may serve as beneficial predictors of hospital mortality.