Author + information
- Received June 2, 1998
- Revision received August 24, 1998
- Accepted October 2, 1998
- Published online February 1, 1999.
- Frederick Van Lente, PhDa,* (, )
- Ellen S McErlean, MSNa,
- Sue A DeLuca, BSNa,
- W.Franklin Peacock, MDa,
- J.Sunil Rao, PhDa and
- Steven E Nissen, MD, FACCa
- ↵*Reprint requests and correspondence: Dr. Frederick Van Lente, PhD., Department of Clinical Pathology/L11, The Cleveland Clinic Foundation, 9500 Euclid Ave., Cleveland, Ohio 44195
The purpose of this study was to investigate the utility of cardiac troponin T and troponin I for predicting outcomes in patients presenting with suspected acute coronary syndromes and renal insufficiency relative to that observed in similar patients without renal disease.
Cardiac troponin T and troponin I have shown promise as tools for risk stratification of patients with acute coronary syndromes. However, there is uncertainty regarding their cardiac specificity and utility in patients with renal disease.
We measured troponin T, troponin I and creatine kinase MB in 51 patients presenting with suspected acute coronary syndromes and renal insufficiency and in 102 patients without evidence of renal disease matched for the same peak troponin T or I value, selected from a larger patient cohort. Blood samples were obtained at presentation to an emergency room 4 hours, 8 hours and 16 hours later. The ability of biochemical markers to predict adverse outcomes in both groups including infarction, recurrent ischemia, bypass surgery, heart failure, stroke, death or positive angiography/angioplasty during hospitalization and at six months was assessed by receiver-operator curve analysis. The performance of both troponins was compared between groups.
Thirty-five percent of patients in the renal group and 45% of patients in the nonrenal group experienced an adverse initial outcome; over 50% of patients in all groups had experienced an adverse outcome by 6 months, but these differences were not significant. The area under the curve (AUC) for the ROC curve for troponin T as predictor of initial outcomes was significantly lower in the renal group than in the nonrenal group: 0.56 ± 0.07 and 0.75 ± 0.07, respectively. The area under the curve was also significantly lower in the renal group compared with the nonrenal group for troponin T as predictor of six month outcomes: 0.59 ± 0.07 and 0.74 ± 0.07, respectively. The area under the curve was also significantly lower in the renal group compared to the nonrenal group for troponin I as predictor of both initial and six month outcomes: 0.54 ± 0.06 vs. 0.71 ± 0.07 and 0.53 ± 0.06 vs. 0.65 ± 0.07, respectively. The sensitivity of troponin T for both initial and six month adverse outcomes was significantly lower in the renal group than in the nonrenal group at a similar level of specificity (0.87): 0.29 vs. 0.60 and 0.45 vs. 0.56, respectively. Troponin I also exhibited similar differences in sensitivity in the renal group (0.29 vs. 0.50 and 0.33 vs. 0.40, respectively).
The ability of cardiac troponin T and troponin I to predict risk for subsequent adverse outcomes in patients presenting with suspected acute coronary syndromes is reduced in the presence of renal insufficiency.
The myocardial contractile proteins, cardiac troponin T (cTnT) and cardiac troponin I (cTnI) are becoming established as important laboratory indicators of myocardial injury (1–5). In addition to detecting acute myocardial infarction, troponin T and troponin I may identify high-risk patients with unstable cardiovascular disease that tend to suffer subsequent cardiac-related events. The relative clinical value of cTnT and cTnI in the presence of renal disease, however, has been questioned. Several studies have questioned the usefulness of cTnT, in particular, in hemodialysis patients and patients with renal insufficiency (6–9). Both small series of patients and anecdotal reports of patients on hemodialysis without evidence for cardiac disease have shown that cTnT may exhibit false positive values (6–9). However, the prognostic significance of troponin values in patients with renal disease actually presenting with suspected acute coronary syndromes has not been extensively studied.
We chose to study a cohort of patients with evidence of renal insufficiency who presented with suspected acute coronary syndromes. These cases were matched with a similar group of patients without renal insufficiency based either on their peak cTnT or cTnI values. The ability of cTnT and cTnI to predict both in-hospital and 6 month outcomes in each cohort was compared.
Patients and study protocol
The study was approved by the Institutional Review Board of the Cleveland Clinic Foundation. Seven hundred and two patients presenting to two emergency departments with suspected acute coronary syndromes during the period December 1995 through February 1997 were evaluated for entry into the study groups. Patients were excluded if they had undergone cardiopulmonary resuscitation within seven days of presentation, angiography or thrombolytic therapy within three weeks of presentation, or were given vasopressors. Patients with serum creatinine concentrations equal to, or greater than 20 mg/L were included in the renal study group. Two patients without evidence of renal insufficiency were matched with each case after enrollment was completed based either on the peak cTnT or cTnI value observed without consideration of presenting diagnosis, history or outcomes.
Diagnosis and outcome definitions
Q wave and non-Q wave myocardial infarctions were diagnosed using the WHO criteria of demonstration of at least two of the following: chest pain consistent with cardiac origin, appropriate 12-lead ECG changes or characteristic changes in CK and CK-MB (total CK ≥220 U/L and RI ≥4%). Unstable angina was diagnosed if the patient demonstrated classic angina at rest or sudden increase in episodes of previously stable angina with ST segment depression or T wave inversion in the absence of increased CK-MB. Diagnoses were confirmed by retrospective review of the medical record by a blinded investigator. ECG data was subsequently evaluated independently by a ECG core facility, blinded to other data. ECG were classified as positive if there was a new or unknown left bundle branch block, ST deviations of ±1 mm in 2 contiguous leads (any 2 of 2,3 and VF or any 2 of I, V1, V5, V6 or any 2 of V1–V4); ECG were classified as negative if there were no Q waves, no negative T waves, no ST deviations of ±1 mm, no new or unknown left bundle branch blocks and no new hemiblocks. All other ECG findings were considered moderate. Adverse outcomes were defined as death, coronary surgery, recurrent ischemia, reinfarction, congestive heart failure, and positive catheterization/intervention. Angiograms were considered positive if greater than 70% stenosis was present in at least one major coronary artery. In addition, presenting myocardial infarction was included as an in-house adverse outcome. Initial hospitalization events were documented by review of the medical record and confirmation of appropriate indicators by a blinded investigator. Six month outcomes (6.3 ± 0.9 months) were documented by phone survey by a blinded investigator.
Blood specimens were obtained at presentation, 4, 8 and 16 hours later. Total creatine kinase activity, CKMB, cTnT, cTnI were measured on each specimen. Routine sampling for CKMB in the existing clinical protocol did not include a four hour sample and CKMB was not reported on this specimen. Clinicians caring for the patients, therefore, were blinded to all troponin values and the 4 h CKMB value.
Blood samples were collected in evacuated tubes and centrifuged upon arrival in the laboratory; plasma were frozen at −70°C until analysis for cTnT and cTnI. Total CK activity was measured with a kinetic enzymatic method on a CX7Delta analyzer (Beckman Instruments, Brea, California). Creatine kinase MB mass (reference range, ≤5 ug/L) was determined by immunoassay on an IMx analyzer (Abbott Laboratories, Abbott Park, Illinois) using the manufacturer’s reagents. The lower limit of detection for this assay was 0.7 μg/L.
Cardiac troponin T (reference value ≤0.1 ug/L) was measured by second generation commercial ELISA on a ES300 analyzer (Enzymun Troponin T; Boehringer Mannheim, Indianapolis, Indiana) employing streptavidin-coated tubes and two monoclonal antibodies against cTnT. This assay is reported to exhibit no signficant cross-reactivity with skeletal muscle TnT (10). The lower limit of detection of this assay is 0.02 ug/L.
Cardiac troponin I (reference value ≤0.6 ug/L) was quantified by commercial immunoassay using two monoclonal antibodies against cTnI on a Stratus II analyzer (Dade International, Miami, Florida). The detection limit of this assay is 0.35 ug/L.
Continuous variables were compared with the Mann-Whitney U test due to absence of normal distributions. Categorial variables were compared with the Fisher-Exact Test or Pearson Chi-square. Receiver-operator characteristic (ROC) curves were generated and the area-under-the curve determined by the method of Hanley and McNeil (11); areas under the curve (AUC) were compared by determining the z value also according to Hanley and McNeil (12). Sensitivity and specificity were evaluated at the upper left inflection point of the ROC curve (13). Statistical significance was defined as P< 0.05.
The demographics of the study groups matched by peak cTnT values are shown in Table 1. The average age of patients was statistically higher in the renal cohort. Serum creatinine concentrations ranged from 20 mg/L to 148 mg/L (median: 29 mg/L) in the renal group. Creatinine values exhibited an average change of less than 1.0 mg/L during the sampling period. Creatinine values were not correlated cTnT or cTnI values in the renal group (r = −0.07, p = 0.62 and r = −0.08, p = 0.60, respectively). Nine percent of patients in the renal cohort were dialysis-dependent. The magnitude of change of cTnT concentration from the initial to the eight h sample was significantly higher in the nonrenal group (p = 0.02). Similar findings were observed in the nonrenal group matched by peak cTnI values. The median elapsed time between onset of symptoms and presentation to the emergency department in the overall cohort was 4.79 hours, the 25% quartile was 2.05 hours and the 75% quartile was 13.45 hours.
Both renal and nonrenal patient groups shown in Table 1presented with abundant risk factors for cardiovascular disease. The overall frequency of risk factors in the two cohorts were not statistically different. A history of hypertension is present in over 90% of renal patients and in 75% of nonrenal patients. A history of myocardial infarction was present in 63% of renal and in 36 to 37% of nonrenal patients, respectively, and this difference was significant. Initial ECG findings were not statistically different in the two groups. Diagnoses at time of initial discharge were not significantly different between the two groups although the nonrenal group trended towards a greater incidence of acute myocardial infarction.
The percentage of patients in the renal and nonrenal groups that experienced an adverse outcome during their initial presentation or hospitalization was not statistically different (Table 2). Some patients experienced multiple events so the total of specific outcomes listed in Table 2exceed the number of patients in each group. Interestingly, the rate of myocardial infarction and reinfarction was significantly higher in the cTnT matched nonrenal group but, although higher, failed to reach significance in the cTnI matched group. The number of patients identified with congestive heart failure (CHF) was significantly higher in the renal group when matched by cTnT values and this difference was somewhat less significant when matched by peak cTnI value.
Overall adverse outcomes by six months were similarly high in all groups (Table 2). However, the rate of CHF remained significantly higher in the renal group when compared to both nonrenal groups. The six month outcomes are cumulative and include those observed initially except for the incidence of presenting myocardial infarction.
Troponin detection of in-hospital adverse outcomes
Figure 1shows the ROC curves for cTnT and cTnI prediction of adverse in-hospital outcomes in renal and nonrenal groups. The area under the curve (AUC) for cTnT was significantly lower in the renal group (0.56 ± 0.07 vs. 0.75 ± 0.07, p < 0.01). This difference is further evident in the lower sensitivity (Table 3)found for cTnT at the indicated inflection point in the nonrenal renal group compared with the nonrenal group (0.29 and 0.60, respectively) at similar specificities (0.87 and 0.87, respectively). At the threshold value for cTnT, considered indicative of myocardial injury (0.1 μg/L), sensitivity and specificity are considerably lower in the renal group (Table 3). The AUC for cTnI is also significantly lower in the renal group (0.54 ± 0.06 vs. 0.71 ± 0.07, p < 0.01). At the optimal threshold, sensitivity of cTnI for in-house adverse outcomes (Table 3)was also lower in the renal group compared to the nonrenal group (0.33 and 0.50, respectively) as was specificity (0.77 and 0.93, respectively). The optimal threshold shown for cTnI for the renal group is the lower detection limit of the cTnI assay. At the more accepted cTnI threshold of 0.6 μg/L, sensitivity was reduced almost half in the renal group at a similar specificity (Table 3). The performance of cTnT in the renal and nonrenal groups as reflected in the ROC AUC was not significantly different from that of cTnI.
Troponin detection of six month adverse outcomes
Figure 2shows the ROC curves for cTnT and cTnI prediction of adverse 6 month outcomes in both the renal and nonrenal group. The AUC for the renal group was again significantly lower than that for the nonrenal group (0.59 ± 0.07 vs. 0.74 ± 0.07, p < 0.01). At the upper-left inflection point of the ROC curve, cTnT exhibited a sensitivity of 0.45 and specificity of 0.72 in the renal group at a threshold of 0.10 μg/L and a corresponding sensitivity of 0.56 and a specificity of 0.87 in the nonrenal group (Table 3). In a manner similar to cTnT, cTnI demonstrates a significantly lower AUC in the renal group (0.53 ± 0.06 vs. 0.65 ± 0.06, p < 0.02). The sensitivity of cTnI in the renal group at 0.6 μg/L was markedly lower than that seen in the nonrenal group at a similar degree of specificity (Table 3). The most accurate threshold value of 0.35 μg/L for the renal group was the same observed for in-house outcomes. Because CKMB was not used clinically to assess 6 month outcomes, its ability to predict outcomes was compared using the cTnT matched nonrenal group and the ROC curves observed for CKMB (Fig. 2)differ in that the AUC is not significantly different in the renal and nonrenal groups. The AUC found for CKMB in the renal group is higher than either cTnT or cTnI, but the difference fails to achieve statistical significance.
Evaluation of the predictive value of cTnT and cTnI for ischemic endpoints at six months including reinfarction, CHF, death and ECG-confirmed ischemic episodes was also conducted. Results in the renal group (Table 4)were not significantly different from those shown in Figure 1and Table 3. In the nonrenal groups there was a trend towards higher sensitivity and lower specificity for all markers (Table 4). However, the comparison of AUC for ROC curves was virtually unchanged.
We evaluated the ability of serum cardiac troponin T and troponin I concentrations in a group of patients with renal insufficiency presenting with suspected acute coronary syndromes to predict adverse cardiac outcomes and compared this performance with that observed for both markers in a control group of patients without renal insufficiency matched for their peak cTnT or cTnI value. Using a case-matched approach strengthens the ability to assess the relative diagnostic performance of biochemical markers in relatively small patient groups. In order to evaluate the comparative performance of the biochemical markers over their full range of values, ROC curve analysis was employed. The high degree of correlation of cTnT and cTnI values in the groups in a matched study enhances the comparative ROC analysis of the diagnostic information provided by these markers (12).
Noncardiac factors that could affect cardiac troponin values in patients with renal insufficiency
The finding of detectable troponin concentrations in some sera of patients with chronic renal disease has been attributed to noncardiac causes. These patients, generally those with end-stage renal disease requiring dialysis, routinely failed to exhibit evidence for ischemic heart disease (7–9). These apparently false positive values have been attributed to lack of assay specificity or to expression of cardiac troponin in skeletal muscle (14,15). Similar conclusions had been reached previously to explain apparently false positive CK-MB results in similar patient groups (16). Unfortunately, most of these studies were performed in the absence of evidence for active coronary syndromes and the results are difficult to extrapolate to the symptomatic patient. One recent report of patients with acute chest pain identified 7 of 773 patients with positive cTnT and negative cTnI, six of which had renal disease (17). Because renal disease was present, these were considered false positive results.
We found little evidence that the false positive or false negative rate of cTnT or cTnI was related to the magnitude of renal insufficiency; there was no correlation of cTnT or cTnI values with serum creatinine in patients without adverse outcomes. It is possible that the length of time the patients had renal disease or the degree of left ventricular hypertrophy can affect the pattern of release of cTnT and cTnI in these patients. There was also a tendency towards less marked increase in troponin values after presentation in the renal group. This finding could indicate a more stable condition, cardiac or noncardiac, resulting in release of troponins into the circulation.
Patients with renal disease are at risk for cardiac disease (18)and our study group with renal insufficiency exhibited a high prevalence of hypertension and history of myocardial infarction. The risk in this population is also evident in the relatively high incidence of adverse outcomes, 35% initially and 59% at 6 months. Both the number of cardiac-related risk factors and frequency of adverse outcomes was similar in patients without renal disease matched for peak troponin values. Thus, it is evident that the relative performance of cTnT and cTnI in renal patients cannot be attributed to differences in prior probability of adverse outcomes in this patient group.
The spectrum of complications was somewhat different in patients in the renal group compared with the nonrenal group even though the peak release of troponins was identical. The renal group exhibited more deaths and CHF than the nonrenal group, and the latter was statistically significant. Also, patients in the nonrenal disease group tended to have a greater incidence of myocardial infarctions.
Usefulness of cTnT and cTnI in risk stratification of patients presenting with acute coronary syndromes and renal disease
There have been several recommendations regarding the appropriate troponin concentration threshold to employ when attempting to risk stratify patients with acute coronary syndromes. Unfortunately, most investigators have taken the simplistic approach of defining a single threshold without evaluating markers over the full range of values. Katus et al. used 0.5 μg/L in assessing unstable angina patients using a first generation cTnT assay (19). Using a more sensitive assay the same group evaluated cTnT at a threshold of 0.2 μg/L for risk evaluation of similar patients and found that it exhibited almost perfect sensitivity (0.97) and specificity (0.98) for predicting subsequent myocardial infarction or death (20). More recently, a cTnT threshold of 0.1 μg/L demonstrated a 0.43 sensitivity and 0.85 specificity for adverse 30-day outcomes in patients presenting with ECG-documented acute myocardial ischemia (21). Troponin I values greater than 0.4 μg/L were shown to predict a significantly higher risk of mortality (22). As would be expected, risk has been shown to increase with increasing cTnT or cTnI concentrations (21,22). Our thresholds for both the renal and nonrenal groups were optimized for overall accuracy from ROC curve analysis and indicate that a higher cTnT threshold of 0.5 μg/L is needed in patients with renal insufficiency for predicting in-hospital outcomes compared with 0.1 μg/L for the non-renal patients. The net effect is a considerably lower sensitivity for predicting adverse outcomes at the same specificity. The same affect is seen for predicting 6 month outcomes albeit at the same optimized threshold of 0.1 μg/L. Troponin I exhibits a similar effect of decreased sensitivity in renal patients although at a somewhat decreased specificity compared with cTnT. This was observed for both in-house and six month outcomes.
Comparison of troponin T, troponin I and CK-MB
The direct comparison of cTnT and cTnI in study shows no significant differences in marker performance in renal patients with suspected coronary syndromes. Therefore, the advantage of one marker versus the other in this patient group cannot be clearly delineated. Several previous studies have concluded that cTnT is less specific than cTnI in patients with renal disease, but our data do not support this view. It is clear that both markers have a somewhat different relationship to adverse outcomes in renal patients than they do in nonrenal patients. Threshold adjustment can optimize either sensitivity or specificity.
Creatine kinase MB could only be evaluated for prognostic information for six month outcomes because CK-MB values were used as part of the active clinical protocol during the in-house phase of the clinical course. Creatine kinase MB did not demonstrate decreased prognostic value in renal patients. Neither cTnT nor cTnI was significantly better than CK-MB at predicting outcomes in patients with renal insufficiency. However, the value of CK-MB for indicating risk of adverse outcomes in nonrenal patients was clearly inferior to either cTnT or cTnI. Although there have been reports of falsely increased CK-MB in patients with renal disease with suspected acute coronary syndromes, the body of literature is small.
The nonrenal groups were matched to the renal group based on peak cTnT or cTnI values observed during the sampling protocol. This study design should maximally challenge the relative information provided by these markers. This approach also avoids the problems of matching based on relatively small numbers of certain demographic factors such as an admitting diagnosis of myocardial infarction. However, it is clear that the spectrum of cardiac disease and outcomes is not equivalent in the two groups and this could have affected comparative performance of cardiac markers. It cannot be ruled out that renal and nonrenal groups matched by admitting diagnosis or history of hypertension, two factors significantly different between groups, would demonstrate different relative utility of cTnT or cTnI. The differences in the natural history of cardiac disease in patients with renal insufficiency could lead to altered patterns of troponin leakage. Nonetheless, the study design did provide for assessment of this affect by matching patients on the basis of troponin release.
It is possible that some of the differences in the ability of cTnT and cTnI in predicting adverse events in patients with renal insufficiency is due to assay performance. The assays differ in their lower limits of detection and, in this case, the cTnT assay is capable of detection at 0.02 μg/L while the cTnI can detect 0.35 μg/L. The second generation cTnT assay used here has been reported to have little cross-reactivity to the skeletal muscle isoform of troponin T. However, residual cross-reactivity of troponin T from noncardiac sources in patients with renal disease cannot be ruled out although the lack of significant differences between cTnT and cTnI make this unlikely. It is possible than other cTnI assays could exhibit different diagnostic performance in similar patient groups. Additionally, this study employed batch analysis of troponin concentrations using plasma stored at −70°C; any significant degradation of specimen integrity during the storage period could also have affected the relative performance of cTnI versus cTnT (24).
We matched a total of 51 patients with serum creatinine values ≥20 mg/L with a corresponding control group from a much larger cohort of patients. The discriminating power of the study would have been enhanced with larger groups. However, the incidence of patients presenting with suspected acute coronary syndromes whose serum creatinine concentrations are increased is a fraction of those with normal serum creatinine concentrations and the logistics of a larger study was not feasible. In addition, the incidence of patients who were hemodialysis-dependent was relatively small and, therefore, this study could not specifically address the performance of cTnT in end stage renal disease separately.
The assessment of 6 month outcomes was conducted by phone survey while the in-house assessment was performed by chart review and confirmation of appropriate ECG or catheterization data. Although no positive angiographies were performed after the initial outcome period, the inherent confidence interval for the 6 month data can be assumed to be slightly larger than that for the in-house outcomes. We cannot determine whether this leads to an over or underestimation of actual events. However, the relative performance of cTnT, as well as cTnI, in both outcome categories is consistent with a relatively accurate assessment of 6 month outcomes.
The ability of cardiac troponin T and cardiac troponin I to predict risk for subsequent adverse outcomes in patients presenting with suspected acute coronary syndromes is reduced in the presence of renal disease (23).
☆ This study was funded in part by Roche Boehringer-Mannheim Corporation.
- area under the curve
- cardiac troponin I
- cardiac troponin T
- congestive heart failure
- creatine kinase MB
- percutaneous transluminal coronary angioplasty
- receiver operating characteristic
- Received June 2, 1998.
- Revision received August 24, 1998.
- Accepted October 2, 1998.
- American College of Cardiology
- Adams J.E,
- Bodor G.S,
- et al.
- Rottbauer W,
- Greten T,
- Muller-Bardorff M,
- et al.
- Antman EM, Grudzien C, Sacks DB. Evaluation of a rapid bedside assay for detection of serum cardiac troponin T. JAMA 1195;273:1279–82.
- Katus H.A,
- Remppis A,
- Neumann F.J,
- et al.
- Adams J.E,
- Schechtman K.B,
- Landt Y,
- Ladenson J.H,
- Jaffe A.S
- Collinson P.O,
- Stubbs P.J,
- Rosalki S.B
- Bhayana V,
- Gougoulias T,
- Cohoe S,
- Henderson R
- Muller-Bardoff M,
- Hallermeyer K,
- Schroder A.W,
- et al.
- Zweig M.H,
- Campbell G
- Li D,
- Keffer J,
- Jialal I
- McLaurin M.D,
- Apple F.S,
- Voss E.M,
- Herzog C.A,
- Sharkey S.W
- Chan K.M,
- Ladenson J.H,
- Pierce G.F,
- Jaffe A.S
- Hochrein J
- Katus H.A,
- Remppis A,
- Neumann F.J,
- et al.
- Lindahl B,
- Venge P,
- Wallentin L
- Wu A.H.B,
- Feng Y,
- Moore R,
- et al.