Outcomes of decreasing versus increasing cardiac troponin in patients admitted with non-ST-segment elevation myocardial infarction: the Atherosclerosis Risk in Communities Surveillance Study

Abstract

Background

The fourth universal definition of myocardial infarction requires an increase or decrease in cardiac troponin for the classification of non-ST-segment elevation myocardial infarction. We sought to determine whether the characteristics, management, and outcomes of patients admitted with non-ST-segment elevation myocardial infarction differ by the initial biomarker pattern.

Methods

We identified patients in the Atherosclerosis Risk in Communities Surveillance Study admitted with chest pain and an initially elevated cardiac troponin I, who presented within 12 hours of symptom onset and were classified with non-ST-segment elevation myocardial infarction. A change in cardiac troponin I required an absolute difference of at least 0.02 ng/mL on the first day of hospitalization, prior to invasive cardiac procedures.

Results

A total of 1926 hospitalizations met the inclusion criteria, with increasing cardiac troponin I more commonly observed (78%). Patients with decreasing cardiac troponin I were more often black (45% vs. 35%) and women (54% vs. 40%), and were less likely to receive non-aspirin antiplatelets (44% vs. 63%), lipid-lowering agents (62% vs. 80%), and invasive angiography (38% vs. 64%). Inhospital mortality was 3%, irrespective of the cardiac troponin I pattern. However, patients with decreasing cardiac troponin I had twice the 28-day mortality (12% vs. 5%; P=0.01). Fatalities within 28 days were more often attributable to non-cardiovascular causes in those with decreasing versus increasing cardiac troponin I (75% vs. 38%; P=0.01).

Conclusion

Patients presenting with chest pain and an initially elevated cardiac troponin I which subsequently decreases are less often managed by evidence-based therapies and have greater mortality, primarily driven by non-cardiovascular causes. Whether associations are attributable to type 2 myocardial infarction or a subacute presentation merits further investigation.

Introduction

Cardiac troponin I (cTnI) levels exceeding the upper limit of normal (ULN) can be indicative of myocardial infarction (MI) and acute coronary syndromes (ACSs).¹ However, similar elevations may also be seen in patients with myocardial injury due to congestive heart failure, arrhythmias, sepsis, pulmonary embolism, and renal failure.² When accompanied by ischemic symptoms, a dynamic change in troponin has been proposed as one way to distinguish acute myocardial necrosis related to ACS from myocardial injury due to non-ACS causes.³ Accordingly, the fourth universal definition of MI requires evidence of a dynamic change in troponin (either an increase or decrease) for the classification of non-ST-segment elevation myocardial infarction (NSTEMI).³ Previous work suggests a significant incremental value of increasing, but not decreasing, troponin levels for the diagnosis of patients evaluated for chest pain.⁴

In addition to its utility as a diagnostic biomarker, troponin also conveys significant prognostic information, with elevations predicting adverse outcomes in patients with both ischemic and non-ischemic etiologies.^2,5,6 Whether the direction of change in cTnI relates to outcomes in patients with NSTEMI is uncertain. In this investigation from the Atherosclerosis Risk in Communities (ARIC) Surveillance Study, we compared the characteristics, clinical management, and outcomes of patients with confirmed NSTEMI who presented with either increasing or decreasing cTnI.

Methods

ARIC study community surveillance

The ARIC study’s data and materials are publicly available.⁷ Since 1987, the ARIC study has conducted ongoing surveillance of hospitalizations for acute MI in four geographically defined regions of the USA: Forsyth County, North Carolina; Washington County, Maryland; Jackson, Mississippi; and eight northwest suburbs of Minneapolis, Minnesota. All surveillance protocols were approved by local institutional review boards. Informed consent was not required because all data were anonymized by redacting personal identifiers. As previously described,^8,9 hospitalizations eligible for physician review were selected on the basis of age (35–74 years of age from 1987 to 2004, and 35–84 years of age from 2005 onwards), residence in the community, and discharge code (International Classification of Diseases (ICD), 9th revision, clinical modification codes 402, 410 to 414, 427, 428, and 518.4. Hospitalizations were randomly sampled within the strata of ICD-9 codes and demographic groups based on race and sex. Clinical data were collected from the hospital records by trained abstractors, using the physician notes, laboratory reports, patient histories, and discharge summaries. Classification of NSTEMI was determined by physician review of the abstracted medical record, as described next. At the time of writing, adjudication was complete for hospitalizations through 2014. We excluded hospitalizations prior to 1996, to coincide with national trends in cardiac troponin testing.

Electrocardiography

Available 12-lead electrocardiograms (ECGs) were reviewed for quality, rejecting any with missing leads, muscle tremor artifact, or technical errors. The first, third, and the last ECG tracings meeting quality standards were coded electronically at the Minneapolis ECG Reading Center.¹⁰ For the purposes of this analysis, patients identified with ST-segment elevations were excluded.

Chest pain

The presence of chest pain was abstracted from the medical records, with origin determined by a review of the physician notes. Any mention of substernal pressure, tightness, or pain precipitated by exertion or excitement was considered evidence for chest pain of cardiac origin. Chest pain specified in the physician notes as ‘unknown origin’ or ‘undiagnosed’ was considered ‘unknown’. Chest pain of cardiac origin was classified as either present or absent.

Acute MI classification

Acute myocardial infarction was classified by the ARIC study, with an algorithm that is independent of a serial change in cTnI.^8,11 As previously described,^9,10 hospitalized events were reviewed and adjudicated by ARIC study physicians, and classified as definite, probable, suspect, or no MI, based on ECG evidence, the presence of chest pain, and cardiac biomarkers (which were considered ‘abnormal’ if they were 2 or more times the ULN, and ‘equivocal’ if they exceeded the ULN but were less than 2 times the ULN).⁸ Classification criteria remained constant over the study period and are described in detail on pages 29–39 of the ARIC study surveillance manual.¹² In order to qualify as definite or probable NSTEMI, one of the following conditions was required in the absence of ST-segment elevation: (a) diagnostic ECG pattern and abnormal biomarkers greater than 2 times the ULN; (b) cardiac pain and abnormal biomarkers greater than 2 times the ULN; (c) cardiac pain and equivocal biomarkers greater than the ULN but less than 2 times the ULN, with evolving ST-T pattern or diagnostic ECG pattern; or (d) abnormal biomarkers greater than 2 times the ULN with evolving ST-T pattern.¹⁰ Our study population was limited to patients with definite or probable MI without ST-segment elevation.

Biomarkers

Laboratory values for biomarkers of cardiac injury were abstracted chronologically from the medical records. In the ARIC communities, the proportion of eligible hospitalizations with a troponin measurement increased from 8% in 1996 to 98% by 2001, with cTnI more often assayed than troponin T.⁹ We considered cTnI values exceeding the laboratory-reported ULN to be ‘elevated’. Because the ARIC community surveillance abstracts cTnI from a number of hospital laboratories using a variety of cTnI platforms, we present cTnI elevation and ratios relative to the laboratory-specified ULN.¹³ The direction of change (decreasing or increasing) was determined from the first and second cTnI assessment on the first day of hospitalization. For many of the older cTnI platforms, biomarkers were undetectable unless elevated. Consequently, our analysis of serial change in cTnI is limited to NSTEMI patients with an initially elevated cTnI. Currently, there is no consensus for ‘meaningful’ change in serial cTnI; small differences may reflect inaccuracies of the measurement rather than true change. To account for this, we excluded patients within the lowest decile of cTnI change, which corresponded to an absolute change of less than 0.02 ng/mL.

Symptom onset to admission time

The elapsed time from the onset of chest pain until hospital admission was estimated by reviewing the clinical notes. Out-of-hospital symptom onset was defined by the timing of acute pain anywhere in the chest, left arm, jaw, back, shoulder, right arm, or abdomen in the 72 hours preceding hospital arrival. For patients with chronic angina, symptom onset was defined by a change in pain prompting medical attention. Because dynamic changes in cTnI are time-dependent, we limited our study population to patients experiencing out-of-hospital NSTEMI, who presented within 12 hours of their symptom onset and had an initially elevated cTnI at presentation. Presentation times were recorded as (<1 hour, 1 to <2 hours, 2 to <4 hours, 4 to <6 hours, 6 to <12 hours, 12 to <24 hours, 24 to <36 hours, and ⩾36 hours) of symptom onset.

Coronary revascularization procedures

Angiography and revascularization by percutaneous coronary intervention or coronary artery bypass graft surgery were abstracted from the hospital records. The elapsed time from symptom onset to revascularization was derived, based on the timing of acute pain precipitating hospitalization. Revascularization times were recorded as less than 1 hour, 1 to less than 2 hours, 2 to less than 4 hours, 4 to less than 6 hours, and 6 to less than 8 hours, 8 to less than 24 hours, and 24 hours or more after symptom onset. Because invasive cardiac procedures are known to increase cTnI, we excluded patients undergoing revascularization less than 24 hours after symptom onset, to ensure serial cTnI measurements on the first day of hospitalization were not confounded. Patients undergoing revascularization 24 hours or more after symptom onset were not excluded, as these procedures would not influence cTnI assessments on the first day of hospitalization.

Medical therapies

Medications were recorded if administered during the course of hospitalization or prescribed at hospital discharge. Aspirin required routine rather than pro re nata administration for abstraction. Non-aspirin antiplatelet therapy was recorded as a single category and included P2Y₁₂ inhibitors (cangrelor, clopidogrel, prasugrel, ticagrelor, ticlopidine), glycoprotein IIb/IIIa inhibitors (abciximab, eptifibatide, tirofiban, xemilofiban), phosphodiesterase 3 inhibitors (cilostazol), phosphodiesterase 5 inhibitors (dipyridamole), and protease-activated receptor-1 antagonists (vorapaxar). Beta-blockers included β₁ adrenergic antagonists. Angiotensin-converting enzyme inhibitors and angiotensin II receptor blockers (ACEi/ARB) were recorded as a single category. Lipid-lowering agents included statins, niacin, and fibrates.

Mortality outcomes

All-cause mortality outcomes were ascertained by linking hospitalizations with the national death index. Cardiovascular death was defined by death due to ‘diseases of the circulatory system’ (ICD-9 codes 390-459 and ICD-10 codes I00-I99).

Statistical analysis

All statistical analyses were carried out using SAS 9.4 (SAS Institute, Cary, NC, USA). Statistical tests and models were weighted by the inverse of the sampling probability, because sampling fractions varied across the sampling strata.¹⁴ Continuous variables were assessed for normality and compared using the difference in least square means from weighted linear regression. Categorical variables were compared using Rao–Scott χ² tests, accounting for the sampling design. The associations of decreasing versus increasing cTnI with recommended therapies¹⁵ (aspirin, antiplatelets, beta-blockers, ACEi, and lipid-lowering medications) and invasive procedures (angiography and revascularization) were analyzed using multivariable logistic regression, with odds ratios converted into relative risks (RRs) and 95% confidence intervals (CIs).¹⁶ Models were adjusted for demographics (age, race, sex, geographical region), and year of admission. Mortality outcomes (28-day and 1-year) were assessed by multivariable Cox regression, with adjustments for demographics, time from symptom onset to admission, comorbidities and clinical course, (diabetes, acute pulmonary edema/congestive heart failure, ventricular fibrillation/cardiac arrest, cardiogenic shock), medications and revascularization procedures. Because lipid-lowering medications were only abstracted from 1998 onwards, our adjusted mortality models assume no lipid-lowering medication use in 1996 and 1997, which is consistent with the low volume of statin medication prescriptions prior to 1998.

Results

From 1996 to 2014, a total of 17,871 hospitalized patients were classified by the ARIC study as NSTEMI. Of these, 9069 had out-of-hospital onset of chest pain and presented within 12 hours of symptom onset. Of those, 1438 had serial assessments of cTnI on the first day of hospitalization, with an initially elevated cTnI at presentation. After the exclusion of 344 undergoing revascularization procedures less than 24 hours after symptom onset, 1094 patients remained. The selection flowchart is detailed in Figure 1. The distribution of absolute cTnI change was assessed, irrespective of the direction of change (decreasing or increasing). Patients in the lowest decile of cTnI change (n=86) had a less than 0.02 ng/mL difference between the first and second cTnI measurement and were excluded. This left 1008 hospitalizations, corresponding to 1926 weighted events.

The majority of patients were white (63%) and men (57%), with a mean age of 64 years. An increasing cTnI pattern was more commonly observed (78%) than a decreasing cTnI pattern. Patients with decreasing cTnI were more often black (45% vs. 35%) and women (54% vs. 40%). Medical histories and comorbidities were comparable between the two groups. Similarly, there was no difference in elapsed time between the self-reported onset of symptoms until hospital admission. (Table 1). As shown in Table 2, the highest recorded cTnI value was lower for patients with a decreasing cTnI pattern, regardless of whether presented as the absolute cTnI value, or the cTnI value relative to the laboratory-specific ULN.¹³ Similarly, patients with decreasing cTnI had a lower magnitude of change between the first and second cTnI assessments compared to patients with an increasing cTnI.

As shown in Figure 2, patients with decreasing cTnI were equally likely to receive aspirin, beta-blockers and ACEi/ARB as patients with increasing cTnI, but significantly less likely to receive lipid-lowering medications (62% vs. 80%), non-aspirin antiplatelet therapies (44% vs. 63%), and invasive angiography (38% vs. 64%). Among the patients evaluated by angiography, the percentage undergoing revascularization also trended lower for patients with decreasing versus increasing cTnI (46% vs. 55%; P=0.1). Although the sensitivity of cTnI labs has continually improved over the years, patients with decreasing cTnI were consistently less likely to receive lipid-lowering agents, non-aspirin antiplatelets, and invasive angiography when analyzed by year (Figure 3).

After adjusting for demographics and year of admission, patients with decreasing cTnI had significantly lower probabilities of receiving non-aspirin antiplatelets (RR 0.70; 95% CI 0.55–0.86), lipid-lowering medications (RR 0.79; 95% CI 0.64–0.92), and invasive angiography (RR 0.57; 95% CI 0.43–0.74). As shown in Supplementary Table 1, associations were not significantly impacted by additional adjustment for comorbidities and time to admission following symptom onset.

There was no difference in the inhospital mortality between the groups (3% for each). However, patients with a decreasing cTnI pattern had twice the risk of 28-day all-cause mortality (12% vs. 5%; P=0.01) and 1-year all-cause mortality (23% vs. 12%; P=0.002). Among the 28-day fatalities, death was more often attributable to non-cardiovascular causes in patients with decreasing versus increasing cTnI (75% vs. 38%; P=0.01), Figure 4. A similar but non-significant trend was observed among the 1-year fatalities, with a greater percentage attributable to non-cardiovascular causes in patients with decreasing cTnI (60% vs. 43%; P=0.1). After adjustment for demographics, year of admission, symptom to admission time, comorbidities, clinical course, medications and revascularization, decreasing cTnI remained associated with a higher all-cause mortality at 28 days (hazard ratio (HR) 2.45; 95% CI 1.37–4.38) and 1 year (HR 1.60; 95% CI 1.05–2.44).

Because a lower magnitude of cTnI change was observed among patients with decreasing cTnI, we conducted a sensitivity analysis examining clinical characteristics, management, and mortality among patients with lesser versus greater cTnI change, irrespective of the direction of change. The absolute change in cTnI concentration was stratified into approximate quartiles, based on the 25th, 50th, and 75th percentiles of change. As shown in Supplementary Table 2, patients with the lowest quartile of cTnI change were more often black and women, and had a more prevalent history of coronary disease. Guideline-directed NSTEMI therapies were less often administered to patients with a lower magnitude of cTnI change. However, when disregarding the direction of cTnI change, no difference in 28-day mortality was observed among the quartiles of cTnI change.

Discussion

In this population-based surveillance of NSTEMI patients admitted with chest pain and initially elevated cTnI, several observations can be made. First, patients with a decreasing cTnI pattern were more often black and women. Second, patients with a decreasing cTnI pattern had a higher all-cause mortality at 28 days and 1 year, largely driven by non-cardiovascular causes. Third, patients with a decreasing cTnI pattern were less likely to receive evidence-based NSTEMI therapies, such as non-aspirin antiplatelets, lipid-lowering agents and an invasive strategy during their hospital stay. To our knowledge, the association between decreasing cTnI and the under-utilization of ACS therapies and higher mortality has not been shown before.

In the ARIC Community Surveillance Study, patients with decreasing cTnI were more often black and women, and the differences in NSTEMI management by race and sex have been well described.^17–20 However, our models were adjusted for demographics and year of admission. While we found no considerable differences in clinical characteristics between patients with decreasing versus increasing cTnI, those with decreasing cTnI may have a greater prevalence of unmeasured comorbidities, causing providers to elect medical therapy over an invasive strategy.^21–24 In the Can Rapid risk stratification of Unstable angina patients Suppress ADverse outcomes with Early implementation of the American College of Cardiology (ACC)/American Heart Association (AHA) guidelines (CRUSADE) registry, patients with the highest-risk non-ST-segment elevation ACS were least likely to receive evidence-based care.²¹

Other factors may potentially explain the lower angiography utilization in patients with decreasing cTnI. A decreasing cTnI pattern may cause the perception that patients have either completed their infarction or have myocardial injury from non-epicardial coronary artery disease, and therefore would not benefit from intensive treatment. Another possibility may be an under-recognition of NSTEMI in patients with decreasing cTnI. During the rising phase, the serial change in cTnI is often steep, and may have a greater likelihood of being labelled ischemic myocardial injury. In contrast, decreasing cTnI often follows a flatter, more gradual change, with the potential to be classified as chronic myocardial injury rather than acute MI. Such differential classification would lead to a predilection against invasive management and underutilization of guideline-directed NSTEMI medications. In this analysis from the ARIC Community Surveillance Study, the magnitude of change in cTnI was greater in patients with an increasing pattern, possibly influencing the management strategy.

Another possibility is that patients with a decreasing cTnI pattern more often had type 2 MI, or a mismatch in myocardial oxygen supply and demand driven by a process other than athero-thrombotic coronary obstruction.^25–27 The release of cardiac troponin into the peripheral circulation is known to be dependent on the coronary blood flow, with a slower washout observed with totally occluded coronary arteries.^28,29 Thus, cTnI release and clearance may be accelerated in the setting of infarction without occlusion.²⁹ If so, type 2 MI may be a potential explanation for decreasing cTnI within the first 12 hours of symptom onset. The benefit of guideline-directed NSTEMI therapies has not been demonstrated in patients with type 2 MI.³⁰ Patients with type 2 MI are also known to have higher mortality more often attributable to non-cardiovascular pathology.³¹ Consistent with type 2 MI, we observed a lower utilization of NSTEMI therapies and higher non-cardiovascular mortality in patients with a decreasing cTnI pattern. However, the classification of type 2 MI was not available in the ARIC Community Surveillance Study.

Our study should be evaluated in the context of its limitations. High-sensitivity troponin was not approved for diagnostic use in the USA until 2017 and is not included in our analysis. Biomarkers were drawn as part of routine clinical practice rather than on a predefined schedule, preventing standardized timing of cTnI measurements. Assessments of peak cTnI were limited to the highest recorded cTnI value in the medical records. In addition, cTnI was not analyzed by a central laboratory, and we were unable to assess the coefficient of variation among the assays. Hospital laboratories had differing ULNs, which varied geographically and over time. We attempted to account for this by normalizing cTnI values to the laboratory-reported ULN but were limited to analyzing the qualitative direction of cTnI change. In addition, we excluded a significant number of patients with initially elevated cTnI but no serial assessment. Angiography results were not abstracted by the ARIC Community Surveillance Study. Finally, causes of death were inferred from ICD codes and were not adjudicated.

Our study also has several noteworthy strengths. The ARIC Community Surveillance Study provides a large, population-based sample of unselected patients, allowing a real-world analysis of cTnI patterns in hospitalized NSTEMI patients. Clinical and laboratory values were collected by certified abstractors following standardized protocols, NSTEMI was classified and adjudicated by physician review, and mortality outcomes were verified by the national death index.

Conclusion

A decreasing pattern of cTnI in patients hospitalized with NSTEMI is more often observed in black patients and women. Those with decreasing cTnI are less likely to be managed with evidence-based NSTEMI therapies and have a greater risk of all-cause mortality, largely driven by non-cardiovascular causes. This may be attributable to a greater prevalence of type 2 MI among patients with decreasing cTnI; however, further work is needed to understand better the possible mechanisms behind these associations.

Supplemental material

Supplementary material is available at European Heart Journal: Acute Cardiovascular Care online.

Acknowledgements

The Atherosclerosis Risk in Communities Study is carried out as a collaborative study supported by National Heart, Lung, and Blood Institute contracts (HHSN268201700001I, HHSN268201700002I, HHSN268201700003I, HHSN268201700004I, HHSN268201700005I. The authors would like to thank the staff and participants of the ARIC study for their important contributions.

Funding

MAC reports funding from AstraZeneca, Cheisi, Merck, Sanofi-Aventis, Boehringer-Ingleheim, Janssen, Novo-Nordisk, Abbott Laboratories, GlaxoSmithKline, Novartis, Takeda, and Bristol-Myers Squibb. No other authors have any relevant disclosures to report. AQ is supported by the NHLBI postdoctoral training grant T32 HL007604 and the American Heart Association Strategically Focused Research Network in Vascular Disease under award numbers 18SFRN3390085 (BWH-DH SFRNCenter) and 18SFRN33960262 (BWH-DH Clinical Project).

Conflict of interest: The authors declare that there is no conflict of interest.

References