Background: Recent data suggest that older men with detectable cardiac troponin I (cTnI) concentrations that remain below the 99th percentile concentration cutoff are at increased risk for subsequent cardiovascular events. We designed this study to extend this observation by examining risk prediction in both men and women presenting to an emergency department with chest discomfort.
Methods: We obtained data for all-cause mortality and hospital discharges associated with either acute myocardial infarction (AMI) or congestive heart failure (CHF) for up to 8 years after the initial presentation in 448 patients who originally presented in 1996 with acute coronary syndrome (ACS). We performed retrospective analysis for cTnI (AccuTnI™; Beckman Coulter) in frozen plasma samples based on the patients’ reported time from onset of symptoms. Peak cTnI concentration was used for risk assessment.
Results: Patients with cTnI concentrations ≥0.02 μg/L (i.e., limit of detection), including those whose peak values remained below the 99th percentile (0.04 μg/L), were at greater risk for death and AMI/CHF readmissions at 2, 5, and 8 years of follow-up compared with those with peak cTnI <0.02 μg/L. All results were statistically significant (P <0.05) except for death within 2 years among patients with normal but detectable cTnI (0.02 to 0.03 μg/L), relative to the group with values <0.02 μg/L. Kaplan–Meier analyses indicated that both men and women with cTnI ≥0.02 μg/L had worse outcomes (P <0.001).
Conclusion: Both men and women who present with possible ACS with detectable cTnI concentrations that remain below the 99th percentile are at a greater risk for future adverse events.
The criterion for identifying myocardial injury is an increased cardiac troponin (cTn)1 concentration (1). Guidelines stipulate that an increase of cTn must be greater than the 99th percentile of a reference population (1)(2). In the setting of cardiac ischemia, a cTn concentration above the 99th percentile is sufficient for a diagnosis of acute myocardial infarction (AMI). The 99th percentile cutoff, however, is influenced by both the analytical precision and the detection limit of the cTn assay, as well as the healthy/reference population being assessed (3)(4)(5)(6)(7)(8). Recently, while investigating the importance of an apparent age-dependent increase in cardiac troponin I (cTnI) concentrations, Zethelius et al. (9) demonstrated that low but detectable concentrations of cTnI (≥0.021 μg/L) measured with the Beckman Coulter AccuTnI™ assay (below the 99th percentile of a healthy reference population) were associated with long-term risks for coronary heart disease and death in an older, presumably healthy, community-based male population (9). In this study, we investigated whether similar findings were evident in men and women presenting to an emergency department with chest pain.
Materials and Methods
We have previously published details on patients, study design, and methods (10)(11)(12). The study cohort was recruited in 1996: 448 unique patients (all with valid Ontario health insurance numbers for outcome tracking) who were enrolled in a Cardiac Markers Study at a community hospital. Clinical diagnoses at that time were based on WHO criteria using creatine kinase MB mass values. Patients were selected for the study if triage staff deemed their symptoms to be possibly due to cardiac ischemia. If so, blood (heparin and EDTA plasma) specimens were collected based on the patients’ reported time from onset of symptoms: study samples were collected hourly until 6 h after onset of symptoms and thereafter at 9, 12, 24, and 48 h or until the patient was discharged, declined further participation, or was removed from the study by those responsible for his or her care. All specimens were frozen from 1996 until 2003, when the heparin samples were thawed and assayed with the AccuTnI. The stability of cTnI has been well documented with respect to freeze-thaw cycles and storage [e.g., mean (SD), 11 (2) years] with the AccuTnI (9). We assigned the highest cTnI concentration for each patient as the peak value, and we recorded the earliest time point at which this measurement was first observed as the peak time for cTnI.
troponin group assignments
The reported analytical detection limit for the AccuTnI assay is 0.01 μg/L (3)(4). We measured 40 replicates of zero calibrator in 2 reagent lots and used the mean signal plus 2 SD (3) to measure the analytical detection limit during the time of analysis to be 0.006 to 0.013 μg/L (13). Other studies have indicated that the 99th percentile with AccuTnI could be as low as 0.02 to 0.03 μg/L (3)(4)(8)(9) or as high as 0.10 μg/L (5). For this study, we used the peak cTnI concentration from each patient to assign patients into the following 4 groups in keeping with the study from Venge et al. (4): not detectable (<0.02 μg/L), low (normal) but detectable (0.02 to 0.03 μg/L), intermediate (99th percentile or above; 0.04 to 0.10 μg/L), and high (>0.10 μg/L). During the study, we used data from 6 commercial quality-control samples together with data on the observed analytical detection limit of the assay to confirm that the assay did indeed achieve the stated imprecision (i.e., 0.06 μg/L with 10% CV) (11). We have previously reported short-term outcomes using cutoffs at both the 10% CV (0.06 μg/L) and the 99th percentile (0.04 μg/L), as well as for intermediate and high peak cTnI classifications (12). Briefly, the intermediate group comprised patients with a peak AccuTnI cTnI value at or above the manufacturer’s reported 99th percentile (0.04 μg/L) and at or below the highest reported 99th percentile cutoff (5)(12).
We obtained research ethics board approval to investigate health outcomes in our study population via linkage to the Registered Persons Data Base for mortality outcomes and the Canadian Institute for Health Information Discharge Abstract Database for Ontario hospital discharges associated with AMI or congestive heart failure (CHF) (12)(14). Based on the earliest subsequent readmission for AMI or CHF and the death date, we created indicators to reflect whether an event (death or readmission) occurred within 2, 5, and 8 years after presentation. If a patient died without previous AMI or CHF readmission, follow-up was censored at the date of death. Outcomes were captured as events postpresentation (i.e., either during the index hospitalization or afterward).
We performed all statistical analyses using SAS version 9.1.3. A P value of <0.05 was considered statistically significant. Between-group comparisons of central tendency (means, medians) were based on one-way ANOVA and the Kruskal–Wallis test. We used the Pearson χ2 test statistic to compare proportions and assessed the time to an adverse event by Kaplan–Meier survival curves with differences between groups determined by the log-rank test. We used the Cox proportional hazard model to compare time to an event while adjusting for age and sex. Hazard ratios for each detectable cTnI group relative to the not-detectable group were derived by partial likelihood estimation. Significance of the association was based on the Wald χ2 statistic.
In our study population, the male population was younger than the female population (men, median age 61 years, vs women, median age 67 years; P = 0.01). There was no difference between the sexes in the number of specimens or the time from onset to either presentation or peak cTnI measurements (Table 1⇓ ). Table 2⇓ illustrates the composition of our study population in reference to the peak cTnI concentration and its subsequent group assignment (i.e., cTnI <0.02, 0.02–0.03, 0.04–0.10, and >0.10 μg/L). Peak cTnI concentration was significantly associated with age and outcome (P <0.001) but not with sex in our acute coronary syndrome (ACS) population.
Kaplan–Meier curves constructed for the population show clear differences in mortality between patients with no detectable cTnI (i.e., <0.02 μg/L) and those with detectable cTnI (i.e., ≥0.02 μg/L; P <0.001; Fig. 1A⇓ ). We also found differences in event-free survival for the probability of AMI/CHF admission after the initial event (P <0.001; Fig. 1B⇓ ). All 3 cTnI groups had significant hazard ratios for the combined endpoints of death/AMI/CHF at 2, 5, and 8 years (Table 3⇓ ). At 2 years, only the intermediate and high cTnI groups had a significant risk for death compared with the not-detectable group (i.e., cTnI <0.02 μg/L). At 5 and 8 years, the risk of death was significantly greater for all patients with cTnI concentrations ≥0.02 μg/L (Table 3⇓ ). A separate analysis, in which the 21 in-hospital events (6 deaths, 15 AMI/CHF recurrences) were censored at the time of the index event, yielded similar risks (data not shown).
We also assessed the long-term survival for each sex. As seen in Kaplan–Meier curves (Fig. 2⇓ ) for both men and women, there were differences in mortality between those with no detectable cTnI (<0.02 μg/L) and those with detectable cTnI values (≥0.02 μg/L; P <0.001 for all plots). This trend was also observed for each sex when the AMI/CHF readmission endpoint was assessed by Kaplan–Meier analysis after 8 years of follow-up (event-free survival in women: 82% for not detectable, 64% for low, 45% for intermediate, and 35% for high; event-free survival in men: 85%, 71%, 43%, and 38%, respectively; P <0.001).
cTn values above the 99th percentile are known to be strong predictors of early adverse events in those with ACS (4)(8)(15)(16)(17). This study, in concordance with the prior studies of Venge and colleagues, probed values below the 99th percentile (4)(8)(9). Using the same assay (AccuTnI), Venge et al. (4) reported that the 99th percentile for cTnI was ≥0.02 μg/L in a young, healthy population. In older individuals, the 99th percentile concentration was higher, leading to a proposed overall cutoff value of 0.04 μg/L for the assay. Later, they reported that older men with values ≥0.021 μg/L who were followed long-term (10 years) demonstrated increased mortality and more coronary heart disease (9), suggesting that increases ≥0.021 μg/L, rather than being related to normal aging, reflect instead a cardiac comorbidity.
Our data in patients presenting with possible ischemia support and extend these observations in 2 ways. First, they extend the observations to women [who were not included in the study of Zethelius et al. (9)], a feature that is of particular importance, because it is now clear that the increased frequency of AMI detected by contemporary cutoff values for troponin includes a large number of older women, many with troponin values that are quite low (18). It may also be that this phenomenon is why women (who are usually older when they present with ACS) in the TACTICS-TIMI (Treat Angina with Aggrastat and Determine Cost of Therapy with an Invasive or Conservative Strategy–Thrombolysis in Myocardial Infarction) study were less likely than their male counterparts to have increases in troponin (19). The use of a lower cutoff value in this population might be helpful.
Our data also extend the observations to patients presenting with possible ACS. Since our study population was not clinically triaged by cTnI in 1996 (10)(11)(12), an assessment is possible of the long-term risks associated with different cTnI concentrations absent the aggressive interventional nature of modern practice. We have previously reported that higher peak cTnI concentrations are associated with a greater risk for early death/AMI within 1 year (12). The findings from this study support the notion of a concentration-dependent relationship for an increased long-term risk for subsequent death/AMI/CHF events after 2 years. This finding supports the relationship between biomarker concentrations, the severity of disease, and prognosis (9)(20)(21). The long-term mortality risks deviate from this concept in that the mortality of those patients with cTnI >0.10 μg/L is similar to that of patients with cTnI concentrations of 0.02 to 0.03 μg/L after 5 years. Of note, the high cTnI group (n = 99) contained all the Q-wave myocardial infarctions and all but 2 patients who were classified as AMI in 1996 (59 of 61), indicating a probable beneficial treatment survival effect. Our data are consistent with prior publications identifying 0.02 μg/L as the optimal cutoff value for risk stratification with the AccuTnI assay (4) and recent work suggesting that emergency department patients with detectable values even below the 99th percentile of a reference population are at increased risk for mortality (22)(23). Our study does not necessarily lead us to advocate for lower cutoff concentrations for AMI; rather, our data support the observation that there is significant future cardiac risk for patients who have detectable cTnI concentrations that are near the 99th percentile (9)(12).
The underlying assumptions associated with these findings rest on the concept that what is presently designated as the 99th percentile contains an admixture of patients, some of whom have very low values presumably within the reference interval, whereas others have minor detectable increases that lead to an increase in the calculated 99th percentile (22). Such a contention is supported by recent work reporting concentrations in healthy volunteers roughly a log unit lower than the conventional cutoff values (24). Our work supports recent findings that minor detectable increases of cTn can be due to acute or chronic abnormalities (4)(9)(25)(26). Such information reinforces the importance of identifying a rising pattern to define patients with acute disease (10)(11)(22).
Unfortunately, we cannot discern the cause of death; it would be beneficial to know not only if it was cardiovascular in nature, but the exact mechanism as well. Our study cannot provide a data-driven answer in regard to the appropriate therapeutic response, nor can we advocate the use of lower AMI cutoffs below the 99th percentile without additional validation studies. If these patients are similar to those with slightly higher values, they may benefit from a strategy including aggressive anticoagulation and early invasive intervention (27)(28)(29)(30). These data do suggest the need for much more sensitive assays so that prospective studies assessing these interventions can be done in patients with low but detectable cTn concentrations.
The fact that our patients all presented acutely suggests that most of the increases were related to acute problems. We cannot exclude the possibility that some may have been related to structural abnormalities such as abnormal renal function, heart failure, left ventricular hypertrophy, or diabetes, as is likely the case with the other studies (9)(26). However, we suggest that in the future, more sensitive and precise assays that allow for measurements of cTnI at very low concentrations (i.e., <0.02 μg/L) may identify a cohort with these low-level increases who could benefit from existing therapies.
This work was supported by a grant from the Canadian Institutes of Health Research. The AccuTnI reagent was contributed for the study by an unrestricted grant from Beckman Coulter Inc. V.L. has received financial support for lecturing on cardiac markers from Roche Diagnostics.
1 IQR, interquartile range (25th, 75th percentiles).
2 P values for males vs females.
3 Time is from onset of symptoms.
1 IQR, interquartile range.
2 By independent review of charts and creatine kinase MB mass biomarker concentrations by a cardiologist and a specialist in emergency medicine based on the WHO MONICA criteria (12).
1 Adjusted for age at presentation and sex.
↵1 Nonstandard abbreviations: cTn, cardiac troponin; AMI, acute myocardial infarction; cTnI, cardiac troponin I; CHF, congestive heart failure; ACS, acute coronary syndrome.
- © 2007 The American Association for Clinical Chemistry