In-hospital outcomes stratified by renal dysfunction quartile (Q) in the development of congestive heart failure (CHF) (P<.001 for trend across Qs 1-4). To convert renal dysfunction values to milliliters per second, multiply by 0.0167. AMI indicates acute myocardial infarction; VT/VF, ventricular tachycardia/ventricular fibrillation; and ESRD, end-stage renal disease.
Thirty-day outcomes by renal dysfunction group (P = .001 for trend in the rehospitalization; P>.05 for recurrent AMI and death). To convert renal dysfunction values to milliliters per second, multiply by 0.0167. Abbreviations are explained in the legend to Figure 1.
Rates of cumulative myocardial infarction, development of heart failure, or death (composite end point) at 30 days by renal dysfunction strata (P<.001 for trend across quartiles 1-4). To convert renal dysfunction values to milliliters per second, multiply by 0.0167. Abbreviations are explained in the legend to Figure 1.
McCullough PA, Nowak RM, Foreback C, Tokarski G, Tomlanovich MC, Khoury N, Weaver WD, Sandberg KR, McCord J. Emergency Evaluation of Chest Pain in Patients With Advanced Kidney Disease. Arch Intern Med. 2002;162(21):2464–2468. doi:10.1001/archinte.162.21.2464
Increased rates of myocardial infarction, heart failure, arrhythmias, and death occur in patients with chronic kidney disease. We sought to evaluate the processes of care and outcomes in patients with chronic kidney disease presenting to an emergency department with chest discomfort.
We enrolled 817 consecutive patients who underwent evaluation for a possible acute myocardial infarction in a prospective study of cardiac biomarkers. Renal dysfunction did not exclude patients from this study, and baseline renal function and 30-day outcomes were available in 808. Patients were stratified by corrected creatinine clearance rate into quartiles, with those undergoing dialysis (n = 51) as a fifth comparison group.
Those patients with advanced renal dysfunction (corrected creatinine clearance rate, <47.0 mL/min [<0.8 mL/s] per 72 kg) or who underwent dialysis had higher rates of diabetes, hypertension, and prior coronary disease. More than 99% of all patients were admitted to a chest pain observation unit or to the hospital. Rates of stress testing were lower as renal dysfunction worsened. Rates of revascularization, however, were similar for all groups. The most frequent in-hospital complication was the development of heart failure, which occurred in 36.5% of those with a corrected creatinine clearance rate of less than 47.0 mL/min per 72 kg. At 30 days, this group had the highest rates of cumulative myocardial infarction, development of heart failure, and death (40.2%).
Chronic kidney disease is a marker for in-hospital and 30-day outcomes in patients presenting to the emergency department with chest discomfort.
OTHERS1,2 HAVE described the increased risks for complications and death in patients with advanced kidney disease who are experiencing a cardiovascular event. Once hospitalized, patients with chronic kidney disease (CKD) have higher rates of myocardial infarction (MI), arrhythmias, congestive heart failure (CHF), and cardiac death.3 We sought to evaluate this process before the hospitalization, starting with the evaluation of chest discomfort in the emergency department (ED). Our group4 has previously shown that most patients (approximately 85%) from the general population presenting to the ED with a possible acute coronary syndrome do not have an acute MI (AMI), and for those patients, the obligatory process to rule out AMI is costly and time-consuming. No previous studies, however, have examined patients with CKD with respect to diagnostic evaluation, including electrocardiography (ECG), processes of care, and outcomes, given the advanced risk for MI and its complications in these patients.
The methods of patient enrollment and data management have previously been described.5 Briefly, we used a database taken from a prospective diagnostic testing study of cardiac biomarkers. The patient population consisted of 1024 consecutive encounters of patients who underwent evaluation for possible AMI in the ED at Henry Ford Hospital in Detroit, Mich, from January 1 through May 31, 1999. Patients were included if the ED physician suspected possible AMI and a standard cardiac biomarker panel was ordered from the central laboratory (4 sets of creatine kinase [CK]-MB measurements during a 9-hour evaluation). We included patients with advanced kidney disease, including those with end-stage renal disease (ESRD) who were undergoing dialysis. We excluded all patients with ST-segment elevation AMI who were receiving thrombolytic therapy or immediate angioplasty.
In 206 patients, 9 hours of biomarker testing were not clinically appropriate after the initial assessment was complete. Of these, 116 were discharged home, and 90 were admitted to the hospital. At 30-day follow-up, 1 cardiac death had occurred in the first group and 1 noncardiac death in the second group. One additional patient was excluded because of lost data. Thus, 817 patient encounters were studied. Of these, 808 had baseline serum creatinine levels and 30-day outcomes available for analysis. The corrected creatinine clearance rate (CorrCrCl) was calculated using the following formulas6:
where Cr indicates the baseline creatinine level in milligrams per deciliter and agey, the age in years.
The Institutional Review Board at Henry Ford Hospital approved the original study. Research nurses and technicians (working 24 hr/d, 7 d/wk) obtained blood samples on patient arrival and completed a case report form. Five milliliters of heparinized blood were collected for measurement of CK-MB levels and a biochemistry profile, including serum creatinine level, in the central laboratory. The CK-MB level was measured using an immunoassay analyzer (AxSYM; Abbott Laboratories, Abbott Park, Ill). All central laboratory results, including creatinine levels, were available to the treating clinicians. Disposition of patients and final discharge diagnoses from the hospital were determined by the treating physicians.
Acute myocardial infarction was defined as (1) at least 1 CK-MB value above the upper reference limit (9 ng/mL) as measured in the central laboratory during the 9-hour test period, and (2) independent agreement between 2 cardiologists (P.A.M. and J.M.) that an AMI had occurred after reviewing the pattern of change of CK-MB levels and the medical record. In case of disagreement, a third cardiologist adjudicated the case. The discharging physician determined all other final diagnoses and documented them in the medical record. Electrocardiograms were interpreted by a cardiologist masked to the clinical cases using the classification proposed by the Thrombolysis in Myocardial Infarction Study Group.7
Baseline characteristics are reported in mean ± SD with proportions and 95% confidence limits as appropriate. We performed group comparisons using the χ2 test or 1-way analysis of variance as appropriate. We performed multiple logistic regression for the cumulative combined end point of AMI, CHF, or death, with the highest-risk group (fourth quartile) vs all others, controlling for significant risk factors found in the univariate analysis. A 2-sided α level of .05 was used for determining significance.
The demographics and medical histories of the study population are given in Table 1. The overall mean age of the study population was 63.8 ± 16.0 years; 54.3% were women; 81.3% were African American; 15.7% were white; and 3.0% were of other races. We found trends for higher rates of diabetes, hypertension, and prior coronary artery disease (CAD) in patients with worsened renal dysfunction. However, we also found lower rates of smoking and family histories of premature CAD in patients with CKD. Consistent with higher rates of CAD across the renal strata, we also found higher rates of expected long-term medication use, including aspirin and β-blockers. Only 26 patients (3.2%) were in the predialysis CorrCrCl range of less than 20 mL/min (<0.3 mL/s) per 72 kg.
Table 2 highlights the initial ECG findings and processes of care by the groups studied. Only 2.0% of patients with ESRD compared with 25.4% of those with normal renal function (CorrCrCl, >99.4 mL/min [>1.7 mL/s] per 72 kg) had normal ECG findings. Those with normal ECG findings tended to be triaged to chest pain units. We found no significant differences in the rates of ischemic ST- or T-wave changes among the groups. However, we found a trend for more left ventricular hypertrophy (P = .05), right bundle-branch block (P<.001), and left bundle-branch block (P = .04) in worsening levels of CKD. Overall, some form of stress testing was performed in 29.7% of patients. Decidedly fewer regular stress tests and stress echocardiograms were ordered in patients with advanced CKD; however, we found no increase in other forms of testing, including nuclear or pharmacological stress tests. Overall, only 15.7% of patients with ESRD, compared with 39.7% of patients with normal renal function, underwent some form of a stress test (P<.001). Despite these differences in stress testing rates, the rates of coronary angiography and revascularization were similar across CKD groups. Overall, more than 99% of patients were admitted to a chest pain observation unit or to a hospital bed. Patients with worsened CKD tended to be admitted directly to a telemetry or coronary care unit, with only 13.7% of patients with ESRD observed in the chest pain unit.
The most frequent complication in all comparison groups was CHF, which occurred at the highest rate (36.5%) in the fourth quartile (Figure 1). Thirty-day outcomes tracked somewhat differently as depicted in Figure 2. These outcomes were dominated by all-cause hospitalization, which occurred in more than 25% of those with ESRD (P<.001 for trend). The rates for combined in-hospital and 30-day outcomes (cumulative rate) for the most frequent complications of AMI, CHF, or death are shown in Figure 3. The rate was highest in the fourth quartile, with 40.2% of all patients in this group incurring a cardiovascular complication by 30 days. Multivariate analysis, controlled for age, sex, diabetes, prior CAD, and aspirin use, found that the fourth quartile incurred a 58.5% excess risk for the combined cardiovascular end point at 30 days of follow-up.
Of the 29 patients who died during the hospital admission, 19 deaths (65.5%) were due to cardiovascular causes. Of these, 11 were due to AMI, 3 were related primarily to CHF, 4 were due to ventricular tachycardia or ventricular fibrillation without AMI or CHF, and 1 was due to aortic dissection. Of the 10 noncardiac deaths, 1 patient had endocarditis; 1, stroke with brainstem herniation; 2, metastatic cancer; 4, sepsis and respiratory failure; and 1, rhabdomyolysis. The remaining patient had an unspecified cause of death.
Our study has demonstrated a very high rate of cardiovascular end points in a group (fourth quartile; CorrCrCl, <47.0 mL/min [<0.8 mL/s] per 72 kg) presenting to the ED with chest discomfort. This group had a mean age of 75 years, 36.0% rate of diabetes, and mean serum creatinine level of 2.21 mg/dL (195.4 µmol/L); 55.0% were women and 85.7% were African American. Fewer than one third of these patients underwent some form of stress testing or cardiac catheterization. Thus, after hospitalization, these patients underwent conservative medical therapy and were ultimately discharged to home to face a 40.2% composite end point at 30 days. Patients with ESRD, on the other hand, had the highest individual rates of all-cause rehospitalization (25.5%) and all-cause deaths (5.9%) at 30 days. Most deaths (65.5%) were due to cardiovascular causes, in particular, AMI.
Another study1 from our group previously proposed the following 4 pathways by which CKD may be linked to poor outcomes: (1) uncontrolled confounding by age and other established risk factors; (2) therapeutic nihilism; (3) excess toxicity of conventionally used therapies, including medication-related adverse events; and (4) the special biology of the CKD state with respect to accelerated atherosclerosis, the development of CHF, and arrhythmias. In addition, the simultaneous nature of accelerated atherogenesis, thrombosis, and chronic decreases in cardiac, renal, and cerebral perfusion works to create a state that is quite susceptible to injury due to hemodynamic instability (hypertension or hypotension). We believe we have demonstrated at least the first 2 mechanisms at play. Our univariate and multivariate analyses support the fact that these data are heavily confounded by age. There was a 27.1-year difference between quartiles 1 and 4, and hence, a considerable amount of the variation in processes of care and outcomes could be accounted for by age alone. Most of the older patients with more advanced renal dysfunction had stopped smoking and were using insulin if they were diabetic. However, despite the older age and greater comorbidities, less diagnostic testing occurred in the fourth quartile and there was a very low rate (3.9%) of attempted revascularization. Taken together, we believe these data represent some degree of therapeutic nihilism or at least a strong trend for conservative medical management in this group, which may be the most salvageable with respect to efforts to preserve cardiac and renal function. Unfortunately, the original study design did not call for collection of details on the use of in-hospital and discharge medications or their potential adverse effects. Therefore, we cannot make inferences about the thoroughness of the medical treatment in these patients. We might infer, however, that there is room for improvement in the diagnostic evaluation and management of chest pain in these patients with a CorrCrCl of less than 47.0 mL/min (<0.8 mL/s) per 72 kg (mean creatinine level, 2.21 mg/dL [195.4 µmol/L]), given a rehospitalization rate of 14.0% and composite end point of AMI, CHF, or death of 40.2% at 30 days. We acknowledge that this conclusion should be tempered by the fact that our study included a large proportion of African American patients, and may not be completely transposable to other US populations.
Our findings support some potentially new thinking about the evaluation and management of chest pain in patients with CKD. Renal dysfunction appears to be a clinical marker or warning sign for high event rates. Patients with CKD appear to get a bad deal from the health care system. It is sufficiently unclear why this is so. Perhaps the clinical distraction of being older and sicker and the reluctance to use medical and interventional strategies with proven benefits in healthier populations are at the root of the problem. It is reasonable to think that increased triage to intensive care units, higher rates of evaluation for CAD, higher rates of cardioprotective medication use, and possibly higher rates of revascularization may lead to improved outcomes over time. These scenarios, however, have to be tempered by the well-understood risks for acute renal failure after coronary angioplasty or coronary artery bypass surgery in this group.8,9 However, for patients in this group, it seems prudent to have systems in place for close, early follow-up with additional diagnostic testing and medical management on an outpatient basis.10
The results of our study are consistent with a variety of information sources on cardiorenal risk. The recently reported Heart Outcomes Protection Trial demonstrated incrementally higher rates of AMI and death across the strata of baseline serum creatinine levels.11 Perhaps these rates of events were augmented by less vigilant evaluations for chest pain in subjects who reported to the ED with chest discomfort. Although current guidelines for the evaluation of a possible acute coronary syndrome list age greater than 70 years and diabetes as intermediate factors, they do not speak to the risk for CKD that we have observed in this study.12 In patients with ESRD, Herzog and coworkers2 have demonstrated high rates of all-cause mortality after AMI. Perhaps these high rates are due in part to reduced rates of cardioprotective medical therapy, diagnostic testing, and revascularization. Beattie and coworkers1 have recently published an analysis of 1724 patients with ST-segment elevation AMI, and have demonstrated reduced rates of thrombolysis and primary angioplasty in the patients at the highest CKD risk, including ESRD.
Similar to many retrospective analyses performed on prospectively collected data sets, our study has several limitations. We did not have important information on prior diagnostic evaluations performed. Previous stress tests and angiograms or a well-understood history of CHF might have played into the clinical decision not to order a diagnostic test during the hospitalization we studied. In addition, we did not have data on the rates of cardioprotective medications used by patients during the hospitalization and at discharge. Differential rates of use of aspirin, β-blockers, statins, and angiotensin-converting enzyme inhibitors might have accounted for differences in the composite end point. If so, it makes an even stronger case for more aggressive therapeutic measures in those who are elderly and have impaired renal function. Our division of subjects on the basis of CorrCrCl yielded small subgroups; hence there was some instability of the point estimates and lack of robustness owing to missing data. In addition, CorrCrCl is a crude population tool that overestimates renal function in those who weigh less than 72 kg, and, conversely, underestimates it in those who weigh more than 72 kg. Use of more exact calculations, including the Cockcroft-Gault prediction13 or Modification of Diet in Renal Disease equation,14 would have been desirable if weight was in the database. Also, we did not have serum albumin levels measured in all cases and cannot comment on this important risk factor for death in patients with CKD.
Despite high rates of admission and observation, we found differential rates of diagnostic testing and outcomes when patients with chest pain were stratified by baseline renal function. Renal dysfunction appears to be a clinical signal for high combined end points at 30 days, and hence these patients constitute a high-risk chest pain population.
Accepted for publication April 3, 2002.
The study data collection was supported by an unrestricted research grant from Biosite Diagnostics, San Diego, Calif.
Corresponding author: Peter A. McCullough, MD, MPH, Cardiology Section, Department of Internal Medicine, University of Missouri–Kansas City School of Medicine, Truman Medical Center, 2301 Holmes St, Kansas City, MO 64108 (e-mail: firstname.lastname@example.org).