Prognostic Value of Different Laboratory Measures of Renal Function for Long‐Term Mortality After Contrast Media‐Associated Renal Impairment

Abstract

Background: Contrast media‐induced nephropathy (CIN) is associated with markedly increased morbidity and mortality. Although creatinine is at present routinely used to characterize renal function, many studies and guidelines recommend using the estimated glomerular filtration rate (eGFR) since it was found to be much more accurate.

Hypothesis: To assess whether the eGFR or creatinine alone provided a better predictive value for long‐term mortality after contrast media‐associated renal impairment.

Methods: From a prospective trial with 412 patients undergoing heart catheterization, creatinine and eGFR before and after 24 h, 48–72 h, and 30 d after contrast‐media exposure were assessed as well as long‐term mortality.

Results: Univariate Cox regression models identified increases in creatinine after 48 h (hazard rate ratio [HRR] 1.754, 95% confidence interval [CI] 1.134–2.712) and 30 d (HRR 3.157, 95% CI 1.968–5.064) as well as decreases in eGFR after 30 d (HRR 0.962, 95% CI 0.939–0.986) to be significant predictors of long‐term mortality. However, by multivariable Cox regression, only increases in creatinine after 48 h (HRR 1.608, 95% CI 1.002–2.581) and after 30 d (HRR 2.685, 95% CI 1.598–4.511) turned out to be significant and independent predictors of mortality. With regard to a possibly critical threshold of creatinine increase, our data confirmed the historically grown increase in creatinine of 0.5 mg/dl or more during the first 48 h as being associated with increased mortality (p = 0.016, log rank test).

Conclusions: Serum creatinine, but not eGFR, was predictive for long‐term mortality, with a threshold of 0.5 mg/dl or more indicating worse prognosis. Copyright © 2010 Wiley Periodicals, Inc.

Supported by an unrestricted research grant from Schering AG, Berlin, Germany.

Introduction

Contrast‐media induced nephropathy (CIN) is a worldwide problem in patients undergoing heart catheterization. The number of patients with pre‐existing chronic kidney disease (CKD) increases due to a variety of synergistic trends.1 The number of patients with pre‐existing CKD who are exposed to contrast media, is raising, too.1, 2, 3 About 5%–15% of these patients will develop CIN4, 5 which has a fundamental impact on their future outcome. Patients with CIN suffer from a high rate of complications, including up to a 20‐fold increase in mortality during their in‐hospital stay.4, 5, 6, 7 Furthermore, the cumulative long‐term mortality rates of patients with CIN were dramatically high, ranging from 30% to 40% after 1 year for those not requiring dialysis4, 5 and from 45% at 1 year4 to 80% at 2 years7 for those who developed dialysis‐dependent renal failure. Thus, CIN is one of the most powerful independent predictors of 1 year mortality in patients with pre‐existing CKD.8 Since CIN has such an unfavorable prognosis, many efforts have been made to evaluate which changes of renal deterioration after heart catheterization predict worse outcome.

Serum creatinine is the most widely used measure to identify deterioration in renal function, especially because it is easy to determine. Historically, an increase by 0.5 mg/dl or more in serum creatinine during the first 48 h after contrast‐media administration was chosen to define CIN.5 However, the glomerular filtration rate (GFR) is still considered to give a better overall index of renal function in health and disease than serum creatinine, but it is difficult to measure.2, 3 Therefore, many national and international societies recommend using the estimated GFR (eGFR) calculated by the Modification of Diet in Renal Disease (MDRD) formula because it seems to be, at present, the most accurate method to assess renal function.2, 3, 9

Since the eGFR is now considered to be more suitable and accurate than serum creatinine, we evaluated whether changes in the eGFR may have a better value for predicting long‐term survival after CIN than changes in creatinine. Therefore, the impact of changes in creatinine and in eGFR on acute and long‐term mortality was compared with data from a large prospective trial.10

Patients

Data from the prospective Dialysis‐versus‐Diuresis (DVD)‐trial were used.10 In this trial, all patients (n>8,000) who were admitted for elective left heart catheterization between January 1, 2001 and July 7, 2004 were screened for inclusion. Finally, 412 patients were included. Inclusion criterion was a serum creatinine level (determined by the Jaffe method) >1.3 mg/dl and <3.5 mg/dl. Exclusion criteria were acute or recent (within 7 d) myocardial infarction (MI), congestive heart failure (New York Heart Association class IV), recipient of transplanted organs, monoclonal gammopathy, and contrast dye exposure within the last 7 d.

Treatment

Eligible patients were randomly assigned to receive 1 of the following treatments: hydration only (500 ml glucose 5% and 500 ml NaCl 0.9% over 12 h before and after heart catheterization); hydration and hemodialysis (500 ml glucose 5% and 500 ml NaCl 0.9% over 12 h before heart catheterization; hemodialysis within 20 min after catheterization and subsequently 500 ml glucose 5% and 500 ml NaCl 0.9% over 12 h); hydration and N‐ACC (500 ml glucose 5% and 500 ml NaCl 0.9% over 12 h together with 2 oral doses of 600 mg N‐ACC before catheterization; 500 ml glucose 5% and 500 ml NaCl 0.9% over 12 h after catheterization together with 2 oral doses of 600 mg N‐ACC).

Heart Catheterization and Hemodialysis

Coronary angiography and percutaneous coronary intervention (PCI) was performed using arterial access from the femoral or brachial arteries. In all patients, angiography was performed with iopromide (Ultravist 370^™ ), a nonionic iso‐osmolar contrast dye (from Schering AG, Berlin, Germany, who was also the sponsor of this trial). Left ventricular ejection fraction (EF) was determined from 30‐degree right anterior oblique projections of left ventricular cine angiograms. The number of diseased vessels and left ventricular EF were categorized in accordance to the revised classification of the American Heart Association and the American College of Cardiology.11

Hemodialysis was performed for 120 min using volumetrically controlled devices (MTS 4008C, Fresenius Medical Care, Bad Homburg, Germany) at a blood flow rate of 180 ml/min, with a net ultrafiltration rate of zero. All patients received unfractionated heparin for anticoagulation.

Determination of Renal Function

Renal function at baseline was determined from serum creatinine determined by the Jaffe method because it was rapidly available for the large number of patients being screened (>8,000 in the study period). If patients were included based on the Jaffe method, baseline creatinine was determined again enzymatically. Furthermore, enzymatically determined creatinine values were taken at 24 h and 48–72 h after catheterization to control the course of the creatinine values and to identify possible CIN.

To estimate the GFR we used the simplified Modification of Diet in Renal Disease (MDRD) formula:3, 9

(1)

Cardiovascular Risk Factors

Cardiovascular risk factors were assessed at presentation and were defined as follows: history of smoking (if the patient had smoked within the last 10 y); hypertension (if blood pressure >140/90 mm Hg had been documented); family history of cardiovascular disease (stroke, myocardial infarction, or coronary intervention had occurred in a first degree relative); diabetes was assumed if a patient was taking oral antidiabetic medication or insulin.

Follow‐up

From July to October 2004, a questionnaire asking for adverse events and repeat interventions was sent to all patients. If the patient did not return her/his questionnaire, follow‐up was performed by a telephone call to the patient, to relatives, or to referring physicians. For patients who died during follow‐up, the treating physicians were contacted to obtain information about the cause of death.

Ethical Committee

The study protocol had been approved by the Ethical Committee of the Landesärztekammer Westfalen‐Lippe and the Medical Faculty of the University of Münster, Münster, Germany. Written informed consent was obtained from all patients including their consent for long‐term follow‐up.

All patients were covered by an insurance contract made with Ecclesia, Sampo Industrial Insurance N.V., Köln, Germany.

Statistics

Differences in basic clinical characteristics were analyzed for continuous variables with normal distribution by Student t test, in continuous variables with skewed distribution by the Mann‐Whitney U test, and by the chi‐square test for categorical variables.

Univariate predictors of mortality during follow‐up were analyzed by Cox regression models and calculation of hazard rate ratios (HRR) with 95% confidence intervals (95% CI). Apart from creatinine and eGFR, additional changes in the eGFR were evaluated as percentages of the baseline eGFR to refer to the fact that a decrease of 10 ml/min starting from a baseline eGFR of 20 ml/min displays a completely different clinical status than starting from 80 ml/min. Multivariable analyses of mortality was performed by Cox regression analyses using a forward conditional algorithm including creatinine, eGFR, and changes in the eGFR in percentages. Differences in 1 year mortality rates and cumulative survival were compared by the log rank test.

For all tests, p values <0.05 were taken as significant. All statistical analyses were performed with SPSS 11.5.1 for Windows.

Results

A total of 412 patients were included in this evaluation. At first, the baseline clinical characteristics and their association with long‐term mortality were analyzed. There were 338 survivors and 74 non‐survivors (Table 1). Among the non‐survivors, there were significantly more patients with diabetes, an EF <35%, a higher serum‐urea, and a lower hemoglobin. Serum creatinine was significantly higher in the non‐survivors than in survivors, whereas the eGFR was significantly lower in non‐survivors. Only the proportion of patients with preexisting hypertension was higher in the survivors group than in the non‐survivors group.

Table 1.

Clinical characteristics and their association with long‐term mortality

	Survivors (n = 338)	Non‐Survivors (n = 74)	All patients (n = 412)	p
Age, mean±SD, years	66.7±10.1	69.1±10.7	67.1±10.2	0.065
Men, n (%)	281 (83.1)	63 (85.1)	344 (83.5)	0.675
Diabetes, n (% of column)	90 (26.6)	30 (40.5)	120 (29.1)	0.017
Smoking, n (%)	65 (19.2)	15 (20.3)	80 (19.4)	0.838
Hypertension, n (%)	261 (77.2)	46 (62.2)	307 (74.5)	0.007
No CHD, n (%)	80 (23.7)	18 (24.3)	98 (23.8)	−
CHD with 1 diseased vessel, n (%)	54 (16.0)	5 (6.8)	59 (14.3)	0.181
2 diseased vessels, n (%)	79 (23.4)	17 (23.0)	96 (23.3)
3 diseased vessels, n (%)	125 (37.0)	34 (45.9)	159 (38.6)
EF <35%, n (%)	42 (12.4)	24 (32.4)	66 (16.0)	<0.001
Previous myocardial infarction, n (%)	147 (43.5)	30 (40.5)	177 (43.0)	0.642
Previous bypass grafting, n (%)	88 (26.0)	29 (39.2)	117 (28.4)	0.023
Creatinine, mean±SD, mg/dl	1.55±0.35	1.72±0.43	1.58±0.37	0.001
eGFR, mean±SD, ml/min	47.8±10.5	43.4±11.4	47.1±10.8	0.001
Serum Urea, mean±SD, mg/dl	7.10±1.95	7.87±2.62	7.23±2.10	0.005
Hemoglobin, mean±SD, g/dl	13.7±1.62	13.0±2.1	13.6±1.7	0.001
LDL cholesterol, mean±SD, mg/dl	107±40	102±36	106±39	0.452
HDL cholesterol, mean±SD, mg/dl	48.5±15.0	47.2±16.4	48.3±15.3	0.565
Triglycerides, median (range), mg/dl	128 (11–694)	127 (34–502)	128(11–694)	0.201*
Lipoproteina, median (range), mg/dl	27 (3–249)	30 (4–309)	28 (3–309)	0.552*
ACE‐inhibitors, n (%)	186 (55.0)	50 (67.6)	235 (57.0)	0.048
Loop diuretics, n (%)	122 (36.1)	37 (50.7)	159 (38.7)	0.020

Abbreviations: ACE = angiotensin converting enzyme; CHD = coronary heart disease; EF = ejection fraction. Differences between the patients who survived or died during long‐term follow‐up were analyzed in continuous variables with normal distribution by Student t‐test, in continuous variables with skewed distribution by the Mann‐Whitney‐test (marked by *), and by the chi‐square test for categorical variables

Different Measures of Renal Function and Long‐Term Survival

Univariate Cox regression analyses of the impact of different measures of renal function and of basic clinical parameters on long‐term mortality are presented in Table 2. Changes in creatinine at 48 h (p = 0.011) and 30 d (p<0.001) as well as for the eGFR at 30 d (p = 0.002), and the changes in the eGFR at 30 d expressed in percentages (p = 0.017) had significant impact on long‐term mortality. In contrast, neither the absolute nor relative changes in eGFR at 48 h were significantly associated with survival. Furthermore, from the basic clinical parameters, age, diabetes, ejection fraction <35%, previous bypass grafting, and ACE‐inhibitor usage were significantly associated with long‐term mortality.

Table 2.

Univariate Cox regression analysis of long‐term mortality by laboratory measures of renal function and clinical parameters

Factor	HRR	95% CI	p
Laboratory measurements
Baseline creatinine	2.637	1.675–4.150	<0.001
Change in creatinine at 48 h	1.754	1.134–2.712	0.011
Change in creatinine at 30 d	3.157	1.968–5.064	<0.001
Baseline eGFR	0.960	0.940–0.979	<0.001
Change in eGFR at 48 h	0.995	0.968–1.024	0.738
Change in eGFR at 30 d	0.962	0.939–0.986	0.002
Changes (%) in eGFR at 48 h	0.994	0.982–1.006	0.339
Changes (%) in eGFR at 30 d	0.986	0.975–0.998	0.017
Baseline clinical parameters
Age	1.031	1.005–1.058	0.021
Diabetes	1.874	1.170–2.999	0.009
Hypertension	0.676	0.420–1.088	0.107
Ejection fraction<35%	0.333	0.204–0.543	<0.001
Number of diseased coronary vessels	1.162	0.943–1.431	0.160
Previous bypass grafting	1.930	1.200–3.103	0.007
ACE‐inhibitor usage	1.702	1.038–2.790	0.035

Survival rates were analyzed by unadjusted Cox regression models, with hazard rate ratios (HRR), 95% confidence intervals (CI) and p values

One year mortality as a function of an increase of creatinine or a decrease in eGFR at 48 h and 30 d are presented in Figure 1. There was a marked and significant increase in the 1 year mortality rates of patients who showed an increase in creatinine over 1.0 mg/dl at 30 d. Similar results were found for an eGFR decrease of >20 ml/min or of >25% of the baseline eGFR at 30 d.

Figure 1.

One year mortality rates were calculated by Cox regression models and are shown for distinct subgroups of patients with increases in serum creatinine (upper panels), decreases in eGFR (middle panels) and decreases in the eGFR expressed in percentages of baseline eGFR (lower panels). For each of these measures of renal function, results were analyzed with regard to changes at 48 h (left side, grey bars) and at 30 d after contrast media administration (right side, black bars). Significant differences in 1 year mortality rates were found depending on increases in creatinine at 48 h, creatinine at 30 d, decreases in eGFR at 30 d, and decreases in eGFR at 30 d in percentages of baseline (p values given in each panel; calculated by log rank test)

Those parameters which revealed a significant difference in mortality rates were subsequently analyzed regarding cumulative survival (Figure 2). Long‐term survival deteriorated with an increase in creatinine at 48 h and at 30 d (Figure 2A,B). An increase in creatinine at 48 h between 0 to 1.0 mg/dl showed a progressive deterioration in long‐term survival, with the best long‐term survival if an increase in creatinine of 0 to 0.25 mg/dl occurred. However, an increase >1.0 mg/dl was associated with a marked deterioration in long‐term survival with more than 70% of patients being deceased after 1,250 d (Figure 2A). With regard to the impact of increases in creatinine at 30 d on long‐term survival, increases of 0.25 to 0.5 mg/dl, 0.5 to 1.0 mg/dl and >1.0 mg/dl were associated with catastrophic survival rates at 1,250 days of only 37%, 31%, and 4%, respectively (Figure 2B). Finally, cumulative survival rates as a function of eGFR at 30 d showed marked differences in survival ranging from 78% in those with no changes in the eGFR to 5% for those with decreases >20 ml/min (Figure 2C).

Figure 2.

Cumulative survival rates depending on increases in creatinine at 48 h (A) and at 30 d (B), and decreases in eGFR at 30 d (C) were calculated by Cox regression models and shown in the panels. Cumulative survival was found to be significantly different between the various subgroups in each panel with pronounced differences in patients with creatinine increases >1.0 mg/dl at 48 h, creatinine increases >0.25 mg/dl at 30 d, and eGFR decreases >10 ml/min at 30 d (p values given beyond curves, log rank test). Decreases in the eGFR expressed as percentages of baseline eGFR (see Figure 1, lower right panel) were not found to be associated with significant differences in cumulative survival and are therefore not shown with a separate panel in Figure 2

Multivariable Analyses of Long‐Term Mortality

To evaluate which of the significant parameters found in univariate analyses were independent predictors of long‐term mortality, 2 multivariable Cox regression models were established: first, only different laboratory measures of renal function were compared to identify which of them provided a better prognostic value with regard to long‐term mortality (Table 3). Thus, only creatinine at 48 h (p = 0.049) and 30 d (p<0.001) turned out to be significant and independent predictors of long‐term mortality. Neither changes in eGFR at 48 h nor changes in eGFR at 30 d showed such a significant association. Second, in an additional multivariable model all these laboratory measures of renal function together with the basic clinical parameters were assessed (Table 4). Thus, only a left ventricular ejection fraction <35% and changes in creatinine at 30 d remained independent and significant predictors of long‐term mortality.

Table 3.

Multivariable Cox regression analysis of long‐term mortality by different laboratory measures of renal function

Factor	HRR	95% CI	p
Parameters in the model
Change in creatinine at 48 h	1.608	1.002–2.581	0.049
Change in creatinine at 30 d	2.685	1.598–4.511	<0.001
Parameters not included
Change in eGFR at 48 h	–	–	0.257
Change in eGFR at 30 d	–	–	0.550
Change in eGFR at 48 h (%)	–	–	0.280
Change in eGFR at 30 d (%)	–	–	0.220

The impact of different measures of renal function on long‐term mortality were assessed by multivariable Cox regression models using a forward conditional algorithm, with hazard rate ratios (HRR), 95% confidence intervals (CI) and p values

Table 4.

Multivariable Cox regression analysis of all parameters associated with long‐term mortality

Factor	HRR	95% CI	p
Parameters in the model
Ejection fraction<35%	0.294	0.168–0.517	<0.001
Change in creatinine at 30 d	2.839	1.664–4.843	<0.001
Parameters not included
Age	–	–	0.079
Diabetes	–	–	0.285
Hypertension	–	–	0.098
Number of diseased coronary vessels	–	–	0.762
Previous bypass grafting	–	–	0.252
ACE‐inhibitor usage	–	–	0.304
Change in creatinine at 48 h	–	–	0.051
Change in eGFR at 48 h	–	–	0.221
Change in eGFR at 30 d	–	–	0.971
Change in eGFR at 48 h (%)	–	–	0.241
Change in eGFR at 30 d (%)	–	–	0.270

The impact of different measures of renal function on long‐term mortality were assessed by multivariable Cox regression models using a forward conditional algorithm, with hazard rate ratios (HRR), 95% confidence intervals (CI) and p values

Discussion

CIN is a major and emerging problem in patients undergoing heart catheterization. It has an unfavorable prognosis in patients in all stages of CKD12, 13 since deterioration in renal function is a key predictor of in‐hospital and long‐term mortality.4, 5, 7, 8 However, of the large number of studies on CIN, only a few4, 5, 7 have assessed the long‐term outcome of these patients. Among these our recently published DVD‐trial10 is the largest and provides the longest follow‐up.

Consistently, all long‐term observations have reported dramatically higher mortality rates ranging from 30% to 40% after 1 year for those without dialysis‐dependent CIN,4, 5, 10 and from about 50% at 1 year4, 10 to 80% at 2 years7 for those who developed dialysis‐dependent renal failure. All of these studies have used serum creatinine to determine renal function at baseline, and to define subsequent renal failure due to CIN. However, serum creatinine is affected by many factors like muscle mass, diet, nutritional status, and drugs.14 These factors may lead to serious errors in the determination of renal function using creatinine.2 None of these previous studies4, 5, 7 used the estimated glomerular filtration rate (eGFR) for determination of CIN which is currently accepted as the best measure of kidney function and has for these purposes been implemented in several national guidelines and recommendations.2, 15, 16, 17, 18 Based on current guidelines,2, 16, 17 there is wide consensus that at present, calculation of the eGFR by the MDRD formula which takes age, gender, and race into account, is the most suitable way to determine renal function—much more accurate than creatinine.

Thus, our hypothesis was that eGFR may have a better diagnostic accuracy for determination of renal failure due to CIN after exposure to contrast media. We, therefore, analyzed the prognostic impact of eGFR using the data from our prospective DVD trial. The main result was that an increase in creatinine at 48 h and at 30 d, as well as a decrease in the eGFR after 30 d, was significantly associated with a higher than 1 year mortality in univariate Cox regression models. But in multivariable Cox regression analysis, only changes in creatinine at 48 h and 30 d remained significant and independent predictors of long‐term mortality. Contrary to our hypothesis, changes in eGFR after 48 h and after 30 d did not yield an independent association to long‐term mortality despite the additional information included, like age, race, and gender. Moreover, the calculation of a relative (percentage) change in the eGFR was not predictive for outcome neither in univariate nor in multivariable Cox regression models although it includes additional information such as the degree of renal failure at baseline. Thus, the calculation of eGFR by the MDRD formula provided less significant information about long‐term outcome than simple serum creatinine in this setting. These results are in accordance to a recent report about the long‐term course of patients with chronic heart failure19 which also found limitations in the predictive value of the eGFR calculated by the MDRD formula.

These findings raise the question of why the eGFR provides no beneficial information in this setting of contrast media‐associated renal impairment although it is in general superior to creatinine. Our study was not designed to answer this question specifically. However, we observed that with each calculation step (from creatinine to eGFR; from eGFR to relative changes in eGFR expressed in percentages) the prognostic value decreased. This might be due to the fact that by every calculation step a part of the prognostic information in the parameters was lost. Moreover, the eGFR is just a mathematical method which was established in a large cohort of patients with almost stable chronic kidney disease.2, 3 In patients with acute renal deterioration as in our subset, the mathematical method might have to be refined to reflect the changes in kidney function correctly.

What Threshold of Creatinine Increase is Associated with Worse Outcome?

Most studies defined CIN as a creatinine increase of 0.5 mg/dl or more compared to baseline.5, 7, 8, 10 Moreover, some reports used an additional condition for defining CIN if creatinine worsened by 25% or more compared to the baseline value4, 8 to refer to the fact that patients with lower baseline creatinine (e.g., 1.0 mg/dl) suffer already from renal failure due to CIN if an increase (e.g., 1.3 mg/dl) occurs (what in general means a 50% loss in clearance). However, these alternative thresholds or definitions to classify CIN are historically grown and have not been derived from any prospective assessment which took distinct steps of creatinine increases and their relation on outcome into account.

Although this was not a primary target of our prospective DVD‐trial,10 post‐hoc analyses of the 412 patients included in the DVD‐trial do provide some specific data. Thus, Figure 1 showed that those 31 patients who showed a creatinine increase at 48 h of 0.5 mg/dl or more suffered from a 1 year mortality which was more than 2‐fold higher compared to all other groups. Similarly, patients who at 30 d showed an increase in creatinine levels of only 0.25 mg/dl or more compared to baseline suffered also from a 1 year mortality which was 2‐fold higher. The same accounted for changes in the eGFR at 30 d: a decrease of 10 ml/min or more was associated with a 2.5‐fold increased 1 year mortality compared to the other groups which showed a very constant mortality rate. Moreover, Cox regression models (Figure 2B,C) revealed that the cumulative survival rate at about 3 y in patients with an increase in creatinine of 1.0 mg/dl or more at 30 d was only 4%, while of the patients with a decrease in the eGFR of >20 ml/min at 30 d, 95% have deceased after 1,250 days. Nevertheless, it is much more suitable to perform controls of serum creatinine levels 48 h after contrast media administration than after 30 d, which is another argument that the current practice to classify creatinine increases of 0.5 mg/dl or more after 48 h as CIN.

In conclusion, although the eGFR may be a more accurate measure for renal function in general, this could not be confirmed in patients exposed to contrast media. In fact, determination of creatinine after 48 h and 30 d does still provide the best prognostic information for the prediction of long‐term mortality after CIN, with an increase in creatinine of 0.5 mg/dl or more at 48 h still presenting a threshold where mortality rates rose markedly.