Volume 13, Issue 1 , Pages 3-10, January 2012
The evaluation of creatinine clearance, estimated glomerular filtration rate and serum creatinine in predicting contrast-induced acute kidney injury among patients undergoing percutaneous coronary intervention☆☆☆
Article Outline
- Abstract
- 1. Introduction
- 2. Methods
- 3. Results
- 4. Discussion
- 5. Conclusion
- 6. Summary
- Acknowledgments
- References
- Copyright
Abstract
Purpose
The purpose of the study was to compare creatinine clearance (CrCl), estimated glomerular filtration rate (eGFR) and serum creatinine (SCr) in predicting contrast-induced acute kidney injury (CI-AKI), dialysis and death following percutaneous coronary intervention (PCI).
Methods and Materials
Data were prospectively collected on 7759 consecutive patients within the Dartmouth Dynamic Registry undergoing PCI between January 1, 2000, and December 31, 2006. Renal function was measured at baseline and within 48 h after PCI using three methods: CrCl using the Cockcroft–Gault equation, eGFR using the abbreviated Modification of Diet in Renal Disease equation and SCr. We compared CrCl, eGFR and SCr in predicting CI-AKI, post-PCI dialysis-dependent renal failure and in-hospital mortality. Areas under the receiver operating characteristic curve (ROC) were calculated using logistic regression and tested for equality.
Results
On univariable analysis, CrCl [ROC: 0.69; 95% confidence interval (CI): 0.67–0.72] predicted CI-AKI better than eGFR (ROC: 0.67; 95% CI: 0.64–0.70) (P=.013) and SCr (ROC: 0.64; 95% CI: 0.61–0.67) (P<.001). Creatinine clearance (ROC: 0.73; 95% CI: 0.69–0.77) and eGFR (ROC: 0.70; 95% CI: 0.65–0.74) outperformed SCr for predicting in-hospital mortality. On multivariable analysis, CrCl (ROC: 0.77; 95% CI: 0.75–0.80), SCr (ROC: 0.78; 95% CI: 0.76–0.80) and eGFR (ROC: 0.77; 95% CI: 0.75–0.80) predicted CI-AKI well. Creatinine clearance (ROC: 0.88; 95% CI: 0.85–0.90) and eGFR (ROC: 0.87; 95% CI: 0.85–0.90) were strong independent predictors of in-hospital mortality.
Conclusions
Creatinine clearance, eGFR and SCr predict CI-AKI equally well. Creatinine clearance and eGFR are strong independent predictors of in-hospital mortality.
Keywords: Angioplasty, Contrast media, Kidney, Contrast-induced nephropathy
1. Introduction
Contrast-induced acute kidney injury (CI-AKI) has been reported as the third most common cause of hospital acquired renal failure [1], [2]. Nash et al. [1] found that 11% of hospital-acquired renal insufficiency cases were due to contrast media, with coronary angiogram and angioplasty measuring as the most common causes. Contrast-induced AKI is a potentially serious adverse effect of percutaneous coronary intervention (PCI) and is often associated with significant morbidity and mortality [3]. Three commonly used methods of assessing renal function are serum creatinine (SCr), creatinine clearance (CrCl) as measured by the Cockcroft–Gault equation and estimated glomerular filtration rate (eGFR) as measured by the Modification of Diet in Renal Disease (MDRD) equation. The Cockcroft–Gault and MDRD equations are frequently used in clinical practice, but there is no agreement over which measurement strategy is more accurate in assessing renal function [4], [5] or which measure provides the best insight into patient-specific procedural risk. Estimating renal function accurately is important because renal dysfunction is a major risk factor for developing CI-AKI [6]. Therefore, identifying those patients at risk for developing CI-AKI post-PCI may be beneficial as these patients could receive protective therapies to potentially decrease their risk. The purpose of this investigation is to compare SCr, CrCl and eGFR in their ability to predict CI-AKI, need for post-PCI dialysis and in-hospital mortality.
2. Methods
2.1. Study population
We analyzed data from 7759 consecutive patients from the Dartmouth Dynamic registry who underwent PCI from January 1, 2000, to December 31, 2006. The Dartmouth Dynamic registry is a large prospective clinical registry of consecutive patients undergoing diagnostic or interventional cardiovascular catheterization procedures at Dartmouth-Hitchcock Medical Center [7]. The registry was approved by the Institutional Review Board (Dartmouth Center for the Protection of Human Subjects).
2.2. Data collection
Patient, disease and procedure characteristics were collected prospectively on all patients in the study, and we report them by baseline renal function as determined by SCr, eGFR and CrCl standard thresholds (Table 1, Table 2). Patient characteristics included age, gender, body mass index (BMI), morbid obesity, hypertension, hypercholesterolemia, diabetes mellitus, peripheral vascular disease and chronic obstructive pulmonary disease. Cardiac disease characteristics included history of congestive heart failure, chest pain, stable angina, acute coronary syndrome, prior myocardial infarction, cardiogenic shock, ejection fraction (estimated and/or calculated), prior thrombolysis, prior coronary artery bypass graft (CABG) and acuity of current case (elective, urgent, emergent). Laboratory measures included baseline hematocrit and baseline SCr (mg/dl); baseline CrCl was calculated using the Cockcroft–Gault equation adjusted per 1.73 m2 (ml/min/1.73 m2): [(140−age in years)×body weight in kg×0.85 (if female)]/(72×SCr in mg/dl); baseline eGFR was calculated using the abbreviated MDRD equation (ml/min/1.73 m2): 186×(SCr in mg/dl)−1.154×(age in years)−0.203×(0.742 if female)×(1.210 if African American) [7], [8], [9]. Cardiovascular procedure characteristics recorded were type of procedure (diagnostic and interventional, staged interventional), mean dose and type of contrast agent used, coronary disease classification (single-vessel, double-vessel or triple-vessel disease; critical left main disease; significant left main disease; indeterminant) and intraaortic balloon pump use. Pre-PCI prophylactic treatment such as hydration, use of sodium bicarbonate, or use of N-acetylcysteine was not available as it was not captured in the data collection for the Dartmouth Dynamic Registry. Clinical outcomes measured in the post-PCI hospitalization period were CI-AKI, new-onset dialysis-dependent renal failure, in-hospital mortality from all causes, cardiac events (new myocardial infarction, cardiac arrest, stent thrombosis), stroke and postprocedure length of stay measured in days (Table 3). Baseline SCr was defined as the last SCr prior to PCI. We defined AKI according to the Acute Kidney Injury Network criteria as an increase in SCr by at least ≥0.3 mg/dl or ≥50% from baseline to peak SCr that was measured within 48 h post-PCI [10]. Vital status was assessed at discharge, and all-cause mortality was confirmed by comparing the data from the Dartmouth Dynamic Registry with the data from the Social Security Administration Death Master File using birth dates and social security numbers. We report in-hospital mortality in Table 3 as an outcome of the index hospitalization.
Table 1. Patient and disease characteristics by renal function
| Variables | SCr<1.5 N (%) | SCr≥1.5 N (%) | P | eGFR≥60 N (%) | eGFR<60 N (%) | P | CrCl≥60 N (%) | CrCl<60 N (%) | P |
|---|---|---|---|---|---|---|---|---|---|
| Mean±S.D. | Mean±S.D. | Mean±S.D. | Mean±S.D. | Mean±S.D. | Mean±S.D. | ||||
| N | 7013 (90.39) | 746 (9.61) | 5824 (75.06) | 1935 (24.94) | 5584 (71.97) | 2175 (28.03) | |||
| Age (years) | 63.8±12.3 | 72.1±11.3 | <.001 | 60.7±11.0 | 76.3±9.1 | <.001 | 62.1±12.0 | 71.0±11.3 | <.001 |
| Gender (female) | 2149 (30.64) | 209 (28.02) | .138 | 1359 (23.33) | 999 (51.63) | <.001 | 1390 (24.89) | 968 (44.51) | <.001 |
| BMI | 29.1±5.9 | 29.1±6.1 | .854 | 30.2±5.8 | 26.0±5.2 | <.001 | 29.2±5.9 | 28.9±6.1 | .067 |
| Morbid obesity | 561 (8.35) | 75 (10.43) | .147 | 573 (10.28) | 63 (3.39) | <.001 | 451 (8.39) | 185 (8.97) | .404 |
| Hypertension | 4317 (62.49) | 568 (76.45) | <.001 | 3480 (60.64) | 1405 (73.48) | <.001 | 3317 (60.16) | 1568 (73.37) | <.001 |
| Hypercholesterolemia | 4331 (62.71) | 442 (59.89) | .073 | 3662 (63.81) | 1111 (58.32) | <.001 | 3478 (63.05) | 1295 (60.86) | .065 |
| Diabetes mellitus | 1773 (25.83) | 308 (41.85) | <.001 | 1530 (26.82) | 551 (29.06) | .156 | 1324 (24.16) | 757 (35.72) | <.001 |
| PVD | 549 (8.20) | 106 (14.64) | <.001 | 386 (6.93) | 269 (14.50) | <.001 | 389 (7.25) | 266 (12.93) | <.001 |
| COPD | 581 (8.71) | 116 (16.04) | <.001 | 420 (7.57) | 277 (14.97) | <.001 | 430 (8.05) | 267 (13.00) | <.001 |
| Cardiac disease | |||||||||
 Congestive heart failure | 635 (9.41) | 196 (26.89) | <.001 | 408 (7.29) | 423 (22.52) | <.001 | 399 (7.40) | 432 (20.71) | <.001 |
| Chest pain | 4279 (65.65) | 412 (59.71) | .005 | 3623 (66.85) | 1068 (59.73) | <.001 | 3442 (66.09) | 1249 (62.45) | .013 |
| Stable angina | 1052 (17.08) | 103 (15.65) | .312 | 889 (17.34) | 266 (15.74) | .316 | 863 (17.48) | 292 (15.52) | .022 |
| Acute coronary syndromes | 2031 (32.48) | 227 (33.68) | .029 | 1678 (32.19) | 580 (33.82) | .456 | 1623 (32.45) | 635 (32.97) | .888 |
| Prior myocardial infarction | 3523 (52.57) | 444 (62.18) | <.001 | 2893 (52.02) | 1074 (57.90) | <.001 | 2736 (51.24) | 1231 (59.30) | <.001 |
| Cardiogenic shock | 84 (1.20) | 33 (4.42) | <.001 | 59 (1.01) | 58 (3.00) | <.001 | 46 (0.82) | 71 (3.26) | <.001 |
| Ejection fraction | |||||||||
| ≥70 | 383 (14.69) | 12 (14.63) | .057 | 328 (14.62) | 67 (15.02) | <.001 | 324 (14.84) | 71 (14.06) | <.001 |
| 60–69 | 988 (37.90) | 22 (26.83) | 874 (38.97) | 136 (30.49) | 845 (38.69) | 165 (32.67) | |||
| 50–59 | 669 (25.66) | 25 (30.49) | 585 (26.08) | 109 (24.44) | 571 (26.14) | 123 (24.36) | |||
| 40–49 | 333 (12.77) | 9 (10.98) | 279 (12.44) | 63 (14.13) | 272 (12.45) | 70 (13.86) | |||
| <40 | 234 (8.98) | 14 (17.07) | 177 (7.89) | 71 (15.92) | 172 (7.88) | 76 (15.05) | |||
| Prior therapy | |||||||||
| Prior thrombolysis | 528 (8.19) | 46 (6.73) | .350 | 454 (8.47) | 120 (6.80) | .049 | 398 (7.72) | 176 (8.92) | .144 |
| Prior CABG | 937 (13.79) | 179 (24.55) | <.001 | 740 (13.10) | 376 (20.04) | <.001 | 697 (12.82) | 419 (20.09) | <.001 |
| Laboratory measures | |||||||||
| Baseline hematocrit | 39.0±4.8 | 35.5±5.6 | <.001 | 39.6±4.6 | 35.9±4.9 | <.001 | 39.5±4.6 | 36.6±5.3 | <.001 |
| Baseline SCr (mg/dl) | 1.0±0.2 | 2.0±1.0 | <.001 | 1.0±0.2 | 1.4±0.7 | <.001 | 0.9±0.2 | 1.5±0.7 | <.001 |
| Baseline CrCl (ml/min/1.73 m2) | 93.6±38.6 | 42.8±17.5 | <.001 | 103.7±34.4 | 43.6±11.8 | <.001 | 101.3±36.8 | 56.5±28.1 | <.001 |
| Baseline eGFR (ml/min/1.73 m2) | 76.4±25.7 | 35.8±10.5 | <.001 | 80.3±25.6 | 49.1±17.5 | <.001 | 85.0±19.0 | 40.4±18.0 | <.001 |
Table 2. Procedural characteristics
| Total number of patients: 7759 | |||||||||
|---|---|---|---|---|---|---|---|---|---|
| Variables | SCr<1.5 N (%) | SCr≥1.5 N (%) | P | eGFR≥60 N (%) | eGFR<60 N (%) | P | CrCl≥60 N (%) | CrCl<60 N (%) | P |
| Mean±S.D. | Mean±S.D. | Mean±S.D. | Mean±S.D. | Mean±S.D. | Mean±S.D. | ||||
| 7013 (90.39) | 746 (9.61) | 5824 (75.06) | 1935 (24.94) | 5584 (71.97) | 2175 (28.03) | ||||
| Acuity of current case | |||||||||
| Elective | 4087 (58.29) | 379 (50.80) | <.001 | 3427 (58.84) | 1039 (53.75) | <.001 | 3336 (59.74) | 1130 (52.00) | <.001 |
| Urgent | 1839 (26.23) | 231 (30.97) | 1481 (25.43) | 589 (30.47) | 1484 (26.58) | 586 (26.97) | |||
| Emergent | 1085 (15.48) | 136 (18.23) | 916 (15.73) | 305 (15.78) | 764 (13.68) | 457 (21.03) | |||
| Catheterization | |||||||||
| Combined diagnostic and interventional | 6855 (97.75) | 723 (96.92) | .153 | 5702 (97.91) | 1876 (96.95) | .016 | 5461 (97.80) | 2117 (97.33) | .224 |
| Pre-resting aortic pressure | 91.7±25.8 | 89.0±23.9 | .0067 | 92.0±26.4 | 89.6±23.4 | <.001 | 92.1±26.4 | 89.8±23.5 | <.001 |
| Intraaortic balloon pump | 180 (2.57) | 48 (6.43) | <.001 | 132 (2.27) | 96 (4.96) | <.001 | 114 (2.04) | 114 (5.24) | <.001 |
| Contrast media | |||||||||
| Isosmolar | 4948 (72.59) | 568 (78.89) | .004 | 4086 (71.96) | 1430 (76.96) | <.001 | 3920 (72.02) | 1596 (76.25) | .001 |
| Low osmolar | 1866 (27.38) | 152 (21.11) | 1591 (28.02) | 427 (22.98) | 1522 (27.96) | 496 (23.70) | |||
| Amount per procedure (ml) | 254.4±139.1 | 234.0±133.3 | <.001 | 257.0±139.4 | 238.7±135.4 | <.001 | 257.1±138.9 | 240.3±137.1 | <.001 |
| Coronary disease | |||||||||
| Single vessel | 2046 (29.17) | 127 (17.02) | <.001 | 1831 (31.44) | 342 (17.67) | <.001 | 1718 (30.77) | 455 (20.92) | <.001 |
| Double vessel | 1974 (28.15) | 169 (22.65) | 1665 (28.59) | 478 (24.70) | 1597 (28.60) | 546 (25.10) | |||
| Triple vessel | 1606 (22.90) | 207 (27.75) | 1288 (22.12) | 525 (27.13) | 1233 (22.08) | 580 (26.67) | |||
| Critical left main | 134 (1.91) | 34 (4.56) | 89 (1.53) | 79 (4.08) | 90 (1.61) | 78 (3.59) | |||
| Significant left main | 333 (4.75) | 55 (7.37) | 239 (4.10) | 149 (7.70) | 240 (4.30) | 148 (6.80) | |||
| Indeterminant | 844 (12.03) | 142 (19.03) | 646 (11.09) | 340 (17.57) | 641 (11.48) | 345 (15.86) | |||
Table 3. Index admission outcomes by renal function
| Total number of patients: 7759 | |||||||||
|---|---|---|---|---|---|---|---|---|---|
| Variables | SCr<1.5 N (%) | SCr≥1.5 N (%) | P | eGFR≥60 N (%) | eGFR<60 N (%) | P | CrCl≥60 N (%) | CrCl<60 N (%) | P |
| Mean±S.D. | Mean±S.D. | Mean±S.D. | Mean±S.D. | Mean±S.D. | Mean±S.D. | ||||
| N | 7013 (90.39) | 746 (9.61) | 5824 (75.06) | 1935 (24.94) | 5584 (71.97) | 2175 (28.03) | |||
| CI-AKI | 326 (4.65) | 126 (16.89) | <.001 | 202 (3.47) | 250 (12.92) | <.001 | 193 (3.46) | 259 (11.91) | <.001 |
| Dialysis | 10 (0.14) | 25 (3.35) | <.001 | 7 (0.12) | 28 (1.45) | <.001 | 5 (0.09) | 30 (1.38) | <.001 |
| Mortality | 111 (1.58) | 50 (6.70) | <.001 | 64 (1.10) | 97 (5.01) | <.001 | 64 (1.15) | 97 (4.46) | <.001 |
| Cardiac events | 150 (2.14) | 24 (3.22) | .059 | 113 (1.94) | 61 (3.15) | .002 | 114 (2.04) | 60 (2.76) | .055 |
| Stroke | 17 (0.24) | 7 (0.94) | .001 | 9 (0.15) | 15 (0.78) | <.001 | 7 (0.13) | 17 (0.78) | <.001 |
| Length of stay postprocedure (days) | 2.0±3.9 | 3.4±8.1 | <.001 | 1.9±3.8 | 2.9±6.2 | <.001 | 1.9±3.9 | 2.7±5.7 | <.001 |
2.3. Statistical analysis
Univariable and multivariable logistic regression was used to evaluate the association of baseline renal function using SCr, CrCl and eGFR with CI-AKI, new onset of dialysis-dependent renal failure and in-hospital mortality. Hosmer–Lemeshow goodness-of-fit χ2 tests were conducted to evaluate the fit of the univariable and multivariable logistic models. We calculated and plotted the area under the receiver operating curve (ROC) by testing the equality of ROC areas using the Bonferroni-adjusted significance probability for each comparison. Stata version 11.0 (College Station, TX, USA) was used to conduct these analyses. For the multivariate modeling, we used a backward stepwise logistic regression approach to identify the best risk factors for inclusion in the final parsimonious multivariate logistic model with P values<.10. The final models varied by the inclusion of risk factors. For the outcome of CI-AKI, we adjusted for age; hematocrit; history of MI or peripheral vascular disease; and current diagnosis of stable angina, cardiogenic shock, heart failure or diabetes. For the outcome of dialysis, we adjusted for hematocrit, history of prior PCI or CABG, and current diagnosis of heart failure and cardiogenic shock. For the outcome of in-hospital mortality, we adjusted for age; hematocrit; history of MI; and current diagnosis of cardiogenic shock, heart failure, hypercholesterolemia or diabetes.
3. Results
Patient characteristics were remarkable for mean age of 64.6±12.5 years, 69.6% of patients were male, the mean BMI was in the overweight range, 63.9% of patients had hypertension, 78.1% of patients had ejection fraction >50% and 42.4% of patients underwent urgent and emergent procedures. The eGFR and CrCl equations categorized more patients with impaired renal function (24.94%, 28.03%) than did SCr (9.61%) (Table 1). In general, patients with lower baseline renal function as determined by SCr, eGFR and CrCl were characterized by older age; greater prevalence of hypertension, diabetes, peripheral vascular disease, chronic obstructive pulmonary disease, heart failure, prior myocardial infarction, cardiogenic shock, or prior coronary bypass surgery; lower baseline mean hematocrit; higher baseline mean SCr and lower baseline renal function as determined by eGFR and CrCl (Table 1). Baseline SCr mean was 1.0±0.5 for all the patients included in the study, and 72% of patients had baseline renal function in the normal or near-normal range (>60 ml/min/1.73 m2). Patients with lower baseline renal function were more likely to undergo urgent or emergent procedures, to need intraaortic balloon pump treatment and to present with critical coronary artery disease, and generally received lower amount of contrast media during PCI (Table 2). The three outcomes measured were CI-AKI, post-PCI need for dialysis and all-cause in-hospital mortality. Of all patients in the study, 5.8% developed CI-AKI, 0.5% needed dialysis post-PCI and in-hospital all-cause mortality rate was 2.1%, with higher frequency for each outcome in patients with lower baseline renal function (Table 3). Serum creatinine, CrCl and eGFR were analyzed individually (Table 4) then compared (Fig. 1) in predicting CI-AKI, post-PCI dialysis and all-cause mortality, respectively. By univariable analysis, CrCl (ROC: 0.69; 95% CI: 0.67–0.72) was a better predictor of CI-AKI (P=.013) than eGFR (ROC: 0.67; 95% CI: 0.64–0.70). Creatinine clearance and eGFR were both superior to SCr (ROC: 0.64; 95% CI: 0.61–0.67) in predicting CI-AKI (P<.001). When compared in their prediction of post-PCI dialysis, there was no statistical difference between eGFR (ROC: 0.85; 95% CI: 0.78–0.92), CrCl (ROC: 0.83; 95% CI: 0.75–0.90) and SCr (ROC: 0.87; 95% CI: 0.80–0.94). In predicting in-hospital mortality, CrCl (ROC: 0.73; 95% CI: 0.69–0.77) and eGFR (ROC: 0.70; 95% CI: 0.65–0.74) were not significantly different, though a trend (P=.062) existed that favored CrCl. Both CrCl (P<.001) and eGFR (P=.005) were significantly superior to SCr (ROC: 0.65; 95% CI: 0.60–0.70) in predicting in-hospital mortality. Multivariable analysis revealed that CrCl (ROC: 0.77; 95% CI: 0.75–0.80), eGFR (ROC: 0.77; 95% CI: 0.75–0.80) and SCr (ROC: 0.78; 95% CI: 0.76–0.80) were good predictors of CI-AKI, with SCr and eGFR having a better model fit. Again, there was no difference in predicting post-PCI dialysis between CrCl (ROC: 0.87; 95% CI: 0.80–0.94), SCr (ROC: 0.88; 95% CI: 0.80–0.95) and eGFR (ROC: 0.89; 95% CI: 0.84–0.94). Consistent with findings of univariable analysis, CrCl (ROC: 0.88; 95% CI: 0.85–0.90) and eGFR (ROC: 0.87; 95% CI: 0.85–0.90) were superior to SCr (ROC: 0.65; 95% CI: 0.60–0.70) in predicting in-hospital mortality. Furthermore, multivariable analysis for prediction of CI-AKI by AKI stages revealed that CrCl, SCr and eGFR were good predictors of AKI by severity, with SCr and CrCl having a better model fit for predicting stage 1 AKI with no difference in predicting stage 2 and 3 AKI (Table 5).
Table 4. Receiver operating curve and modeling statistics for prediction of CI-AKI, dialysis and in-hospital mortality
| Outcome by predictor | Univariable analysis | Multivariable analysis | ||||||
|---|---|---|---|---|---|---|---|---|
| ROC | 95% CI | GOF χ2 | P value | ROC | 95% CI | GOF χ2 | P value | |
| CI-AKIa | ||||||||
| SCr | 0.64 | 0.61–0.67 | 81.79 | <.001 | 0.78 | 0.76–0.80 | 11.18 | .192 |
| CrCl | 0.69 | 0.67–0.72 | 101.04 | <.001 | 0.77 | 0.75–0.80 | 31.69 | <.001 |
| eGFR | 0.67 | 0.64–0.70 | 52.40 | <.001 | 0.77 | 0.75–0.80 | 23.20 | .003 |
| Dialysisb | ||||||||
| SCR | 0.87 | 0.80–0.94 | 51.35 | <.001 | 0.88 | 0.80–0.95 | 7.76 | .458 |
| CrCl | 0.83 | 0.75–0.90 | 21.48 | .006 | 0.87 | 0.80–0.94 | 18.06 | .021 |
| eGFR | 0.85 | 0.78–0.92 | 10.94 | .205 | 0.89 | 0.84–0.94 | 5.88 | .661 |
| Mortalityc | ||||||||
| SCr | 0.65 | 0.60–0.70 | 64.27 | <.001 | 0.65 | 0.60–0.70 | 64.27 | <.001 |
| CrCl | 0.73 | 0.69–0.77 | 17.73 | .023 | 0.88 | 0.85–0.90 | 6.06 | .640 |
| eGFR | 0.70 | 0.65–0.74 | 19.35 | .013 | 0.87 | 0.85–0.90 | 5.22 | .734 |
aMultivariable analysis for CI-AKI adjusted for age; hematocrit; history of MI or peripheral vascular disease; current diagnosis of stable angina, cardiogenic shock, heart failure or diabetes. |
bMultivariable analysis for dialysis adjusted for hematocrit; history of prior PCI or CABG; current diagnosis of heart failure and cardiogenic shock. |
cMultivariable analysis for in-hospital mortality adjusted for age; hematocrit; history of MI; current diagnosis of cardiogenic shock, heart failure, hypercholesterolemia, diabetes. |
Fig. 1.
Comparison of the predictive value of SCr, CrCl and eGFR. Based on univariable analysis. the figure plots the ROC curves for the pair-wise comparisons of SCr and CrCl (column 1), SCr and eGFR (column 2) and CrCl and eGFR (column 3) against three outcomes: CI-AKI (row 1), dialysis (row 2) and in-hospital mortality (row 3). The ROC curves and 95% CIs are listed in Table 4. P values for each ROC comparison using the Bonferroni-adjusted significance probability.
Table 5. Receiver operating curve and modeling statisticsa for prediction of CI-AKI by severity
| Severity of CI-AKI by predictor | ROC | 95% CI | GOF χ2 | P value |
|---|---|---|---|---|
| AKI stage 1 | ||||
| SCr | 0.78 | 0.76–0.80 | 12.36 | .136 |
| CrCl | 0.77 | 0.75–0.79 | 20.03 | .009 |
| eGFR | 0.77 | 0.75–0.80 | 29.45 | <.001 |
| AKI stage 2 | ||||
| SCr | 0.76 | 0.73–0.78 | 6.92 | .545 |
| CrCl | 0.76 | 0.73–0.78 | 9.88 | .273 |
| eGFR | 0.76 | 0.73–0.78 | 14.97 | .060 |
| AKI stage 3 | ||||
| SCr | 0.76 | 0.67–0.84 | 4.57 | .803 |
| CrCl | 0.76 | 0.68–0.84 | 7.22 | .513 |
| eGFR | 0.77 | 0.69–0.85 | 5.54 | .699 |
aMultivariable logistic regression analysis for prediction of CI-AKI by AKI stages adjusted for age; hematocrit; history of MI or peripheral vascular disease; current diagnosis of stable angina, cardiogenic shock, heart failure, diabetes. |
4. Discussion
In this study, CrCl (by the Cockcroft–Gault equation) was determined by univariable analysis to be a better predictor of CI-AKI than eGFR (by the MDRD equation) and SCr. Both CrCl and eGFR were better predictors for the development of AKI following contrast administration (P<.001) as well as for in-hospital mortality (P<.001 and P=.005, respectively) than SCr alone (Fig. 1). However, when adjusted for coexisting risk factors, CrCl did not prove to be superior, and all three metrics (CrCl, SCr and eGFR) were good predictors of CI-AKI and dialysis. Creatine clearance and eGFR were identified as strong independent predictors of in-hospital mortality. To our knowledge, this is the first large report to analyze the optimal prognostic metric, based on baseline renal function assessment, in a patient population having PCI.
Coceani et al. [11] compared the prognostic value of the Cockcroft–Gault and abbreviated MDRD in a cohort of 1705 patients with ischemic heart disease and found that reduced renal function as measured by the Cockcroft–Gault equation was a better predictor of 10-year survival. Our univariable analysis extends these findings by determining CrCl by the Cockcroft–Gault equation to be the optimal prediction tool compared with MDRD and SCr in predicting both AKI and in-hospital mortality in patients having PCI, despite no differences in baseline classification based on chronic kidney disease stages. However, our analysis supports the use of either SCr, CrCl or eGFR to predict CI-AKI, and the use of either CrCl or eGFR to predict in-hospital mortality in multivariable modeling.
Prior work comparing the accuracy of the Cockcroft–Gault and MDRD equations in assessing renal function has varied. The Kidney Disease Outcome and Quality Initiative advises that the MDRD equation may perform better in assessing renal function and cautions against solely using SCr for assessing renal function [12]. The MDRD Study determined that CrCl as calculated by the Cockcroft–Gault equation overestimates eGFR by as much as 16% [9]. McCullough and Soman suggested that the abbreviated MDRD equation may be more accurate than the Cockcroft–Gault equation [6]. Stevens et al. evaluated equations used to estimate renal function and concluded that CrCl overestimates glomerular filtration rate and that MDRD gives a more accurate prediction of renal function; however, the Cockcroft–Gault equation has been used more extensively in studies involving drug dosing and should be used over MDRD in drug dose adjustments [4].
In patients with acute coronary syndrome undergoing PCI in the CRUSADE registry, Melloni et al. [5] compared the MDRD and the Cockcroft–Gault equations in classifying renal function and in identifying the need for dose adjustment of antithrombotic agents. In their analysis, the median GFR was higher as calculated by the MDRD than was calculated by the Cockcroft–Gault equation, with the highest difference noted in the smaller-BMI, elderly and female subgroups; additionally, the Cockcroft–Gault equation identified a larger percentage of patients who needed drug dose adjustment when compared to the MDRD equation. Thus, Melloni and colleagues concluded that the Cockcroft–Gault equation offers a more accurate assessment of renal function and need for drug dose adjustment than the MDRD, especially in small, elderly and female subgroups [5]. In a small study of 140 patients receiving nephrotoxic or high renal clearance drugs, Roblin et al. [13] compared the Cockcroft–Gault to the abbreviated MDRD and concluded that abbreviated MDRD cannot replace the Cockcroft–Gault in determining drug dose adjustments.
It is unclear why SCr performs as well as eGFR and CrCl in predicting CI-AKI and need for dialysis post-PCI in the multivariable analysis. It is possible that this is due to having a high proportion of patients with baseline normal or near-normal renal function, as well as to the low rate of CI-AKI. In our study, SCr was measured within 48 h post-PCI, which may not have allowed adequate time to detect a peak rise in SCr and to identify all patients who developed CI-AKI, thus potentially explaining the low prevalence (5.8%) of CI-AKI in our study. As SCr can peak as late as 5 days after contrast load administration, measuring SCr at 48 h would miss 30% of patients who develop CI-AKI as suggested by Maioli et al. [14]. Additional factors that may have contributed to the low prevalence of CI-AKI in our study include the following: baseline renal function for the majority of patients was in the normal range; 57.6% of PCI cases were elective, indicating careful timing and patient selection; 28% of cases involved single-vessel PCI with minimal exposure to contrast media and patients with lower renal function received smaller amounts of contrast media as evidenced in Table 2.
Our patient cohort reflects current real-world practice of monitoring SCr at 24–48 h post-PCI as patients may be discharged as early as the following day after an uncomplicated PCI if they did not sustain ST elevation myocardial infarction prior to PCI. A study designed to measure SCr at 72 h or longer after PCI may detect later peaks in SCr, could identify more patients who develop CI-AKI and may be useful to validate our findings and to guide practice changes. Additionally, we did not study the effects of pre-PCI hydration or use of N-acetylcysteine on the predictive ability of the biomarkers used in this study, as this information was not available. Based on prior knowledge, the use of hydration prior to PCI would decrease the risk of CI-AKI, but it remains unclear whether prophylactic treatments would change the performance of biomarkers.
4.1. Study limitations
It is known that baseline renal dysfunction is a major risk factor for developing CI-AKI [15], [16]. As mean baseline creatinine for patients in our study was 1.0±0.5 mg/dl, our findings may only apply to patients with normal or near-normal renal function at baseline. Furthermore, we did not employ a reference method for assessing renal function, such as 24-h urine CrCl, for comparison. The Cockcroft–Gault equation was developed using a group of 96% male subjects, while the MDRD was developed using a group of 60% male subjects [8], [9]. It is possible that analyzing male and female subjects separately may change study results and show gender superiority of either equation.
5. Conclusion
Contrast-induced AKI is associated with increased mortality, prolonged hospitalization and higher medical costs [6], [17]. Our study evaluated commonly used markers in clinical practice to predict post-PCI CI-AKI. Our multivariable analysis supports that either SCr, CrCl or eGFR may be used to predict CI-AKI in patients undergoing PCI. These findings may result in simplifying the process of risk stratification and early implementation of prophylactic measures in patients at risk by using widely available markers of renal function, and further raise the question of changing current practice by using either of these three markers to risk stratify patients prior to PCI. A larger study would be needed to confirm our results of similar SCr performance in comparison to eGFR and CrCl in predicting CI-AKI and post-PCI dialysis. As a future direction, further research is needed to evaluate pre-PCI use of novel biomarkers, such as serum cystatin C, NGAL and IL 18, to determine their ability to early and accurately predict post-PCI CI-AKI.
6. Summary
This study compared CrCl, eGFR and SCr in their ability to predict CI-AKI, dialysis and death following PCI. Multivariable analysis revealed that CrCl, SCr and eGFR predicted CI-AKI equally well and identified CrCl and eGFR as strong independent predictors of in-hospital mortality.
Acknowledgments
There are no acknowledgements to be made.
References
- . Hospital-acquired renal insufficiency. Am J Kidney Dis. 2002;39:930–936
- . How to protect from contrast media-induced nephropathy?. Clin Res Cardiol. 2007;96:125–129
- . Incidence and prognostic importance of acute renal failure after percutaneous coronary intervention. Circulation. 2002;105:2259–2264
- . Assessing kidney function — measured and estimated glomerular filtration rate. N Engl J Med. 2006;354:2473–2483
- . Cockcroft–Gault versus modification of diet in renal disease: importance of glomerular filtration rate formula for classification of chronic kidney disease in patients with non-ST-segment elevation acute coronary syndromes. J Am Coll Cardiol. 2008;51:991–996
- . Contrast-induced nephropathy. Crit Care Clin. 2005;21:261–280
- . Transient and persistent renal dysfunction are predictors of survival after percutaneous coronary intervention: insights from the Dartmouth Dynamic Registry. Catheter Cardiovasc Interv. 2008;72:347–354
- . Prediction of creatinine clearance from serum creatinine. Nephron. 1976;16:31–41
- . A more accurate method to estimate glomerular filtration rate from serum creatinine: a new prediction equation. Modification of Diet in Renal Disease Study Group. Ann Intern Med. 1999;130:461–470
- . Acute Kidney Injury Network: report of an initiative to improve outcomes in acute kidney injury. Crit Care. 2007;11:R31
- . Comparative prognostic value of glomerular filtration rate estimating formulas in ischaemic heart disease. Eur J Cardiovasc Prev Rehabil. 2008;15:423–427
- . K/DOQI clinical practice guidelines for chronic kidney disease: evaluation, classification, and stratification. Am J Kidney Dis. 2002;39:S1–266
- . Estimated glomerular filtration rate for drug dose adjustment: Cockcroft and Gault or abbreviated MDRD equation?. Clin Biochem. 2009;42:111–113
- . Sodium bicarbonate versus saline for the prevention of contrast-induced nephropathy in patients with renal dysfunction undergoing coronary angiography or intervention. J Am Coll Cardiol. 2008;52:599–604
- . Contrast-induced nephropathy: review of incidence and pathophysiology. Ann Vasc Surg. 1998;12:612–620
- . CIN Consensus Working Panel. Risk prediction of contrast-induced nephropathy. Am J Cardiol. 2006;98:27K–36K
- . Contrast-induced acute kidney injury. J Am Coll Cardiol. 2008;51:1419–1428
☆ Disclosures: There are no conflicts of interest to disclose.
☆☆ Funding sources: Dr. Brown is funded by the career development grant K01HS18443 from the Agency for Healthcare Research and Quality.
PII: S1553-8389(11)00465-9
doi:10.1016/j.carrev.2011.05.006
© 2012 Elsevier Inc. All rights reserved.
Volume 13, Issue 1 , Pages 3-10, January 2012
