Discrimination of Patients With Type 2 Myocardial Infarction

Johannes Tobias Neumann; Nils Arne Sörensen; Nicole Rübsamen; Francisco Ojeda; Thomas Renné; Vazhma Qaderi; Elena Teltrop; Solveig Kramer; Laura Quantius; Tanja Zeller; Mahir Karakas; Stefan Blankenberg; Dirk Westermann


Eur Heart J. 2017;38(47):3514-3520. 

In This Article


Baseline Characteristics

Out of 1548 patients with suspected MI, 188 were diagnosed with T1MI and 99 (34.5% of all MI) with T2MI (Table 1). The median age for the overall study population was 65 (25th and 75th percentile 51, 75) years, while patients with T2MI were significantly older than T1MI patients (69 vs. 72 years, P-value = 0.039). Patients with T2MI were more often female (48.5% vs. 28.2%, P-value < 0.001) and more likely to have atrial fibrillation (34.3% vs. 10.6%, P-value < 0.001). Not having a history of CAD (70.7% vs. 48.9%, P-value < 0.001), not presenting with typical radiating chest pain (82.7% vs. 56.1%, P-value < 0.001) and no pathological changes in echocardiography or ECG (43.8% vs. 30.3%, P-value = 0.034) were more common in T2MI patients compared to T1MI. The concentration of hs-TnI was lower in T2MI compared to T1MI patients. This effect was most pronounced 3 h after admission with a median hs-TnI concentration of 108.4 ng/L (30.3, 602.2) among T2MI patients compared to 505.3 ng/L (97.8, 2008.4, P-value < 0.001) among T1MI patients (Supplementary material online, Figure S1). Patients with T2MI underwent angiography in 38.3% (T1MI 86.7%, P-value < 0.001) and none of these patients received PCI. In T2MI patients undergoing angiography, 60.5% had only non-significant CAD or microvessel disease, while 10.5% had one-vessel disease, 7.9% two-vessel disease, and 21.1% had three-vessel disease (all non-obstructive, Supplementary material online, Table S1). The most common causes of T2MI (covering 74% of all cases) were severe hypertension, arrhythmias, and acute decompensated heart failure. The exact causes of T2MI are displayed in Supplementary material online, Table S2.

Figure S1.

Dynamic hs-TnI changes in patients with T1MI and T2MI
Presented are the median concentrations of high-sensitivity troponin I for T1MI and T2MI patients over the course of 3 hours after admission.

When comparing non-MI and T2MI patients, those with T2MI were significantly older (72 vs. 63 years, P-value < 0.001), more often female (36.5 vs. 48.5%, P-value = 0.023), had more often hypertension (65.3 vs. 77.8%, P-value = 0.015), congestive heart failure (13.5 vs. 25.3%, P-value = 0.002), atrial fibrillation (17.2 vs. 34.3%, P-value < 0.001) and worse renal function (estimated glomerular filtration rate 79.6 vs. 62.88 mL/min for 1.73 m2, P-value < 0.001) (Supplementary material online, Table S1). All serial troponin concentrations were significantly higher for T2MI compared to non-MI patients.


The events of death, incident MI, PCI, and cardiac rehospitalization were documented for non-MI, T1MI, and T2MI patients separately (Figure 1). The unadjusted 1-year mortality rate was higher for T1MI and T2MI patients compared with non-MI patients (4.1% for the latter vs. 9.4% and 13.8%, respectively), while the difference between T1MI and T2MI was not statistically significant. There were no incident MIs in T2MI patients and only one PCI within the follow-up period. In non-MI patients, the unadjusted rates of incident MI were 0.6% and 3.7% for PCI. Both events were most common in T1MI patients with an unadjusted event rate of 3.9% (MI) and 12.0% (PCI). Adjustment for age, sex, and history of CAD did not change the magnitude and direction of the reported differences in event rates (Supplementary material online, Table S3). Cardiac rehospitalization within 1 year after admission occurred in 18.8% of non-MI, 19.3% of T2MI, and 33.8% of T1MI patients (significantly higher for T1MI vs. T2MI in the unadjusted model with P = 0.025, but not significant after adjustment for age, sex and CAD).

Figure 1.

Outcome of non-MI, T1MI, and T2MI patients. Kaplan–Meier curves for the end points death, incident non-fatal MI, PCI, and cardiac rehospitalization are displayed for a follow-up period of 2 years. Below the Kaplan–Meier curve, the number of individuals at risk is provided. The P-value compares T1M1 and T1M2. T1M1, type 1 myocardial infarction; T1M2, type 2 myocardial infarction.

The exact cause of death, the duration until death, and the initial maximum cardiac biomarkers of T2MI patients are displayed in Supplementary material online, Table S4. Importantly, there were only two cases of a cardiac origin of death (both heart failure), while most other cases were caused by cancer and respiratory failure.

Score Development

In order to distinguish T2MI from T1MI patients, a logistic regression model was developed. The optimal hs-TnI cut-off concentration, derived from an receiver operating characteristic (ROC) curve, was ≤ 40.8 ng/L for the 0-h sample, ≤ 68 ng/L for 1 h and ≤ 330.9 ng/l for 3 h (Supplementary material online, Figure S2). In the univariable logistic regression model, female sex (AUC 0.60 [CI 0.54–0.66]), hyperlipoproteinaemia (AUC 0.59 [CI 0.53–0.65]), history of MI (AUC 0.58 [CI 0.53–0.62]), atrial fibrillation (AUC 0.62 [CI 0.57–0.67]), not having typical radiating chest pain (AUC 0.63 [CI 0.58–0.68]), not having a history of CAD (AUC 0.61 [CI 0.55–0.67]), and hs-TnI concentrations at baseline, after 1 or 3 h were the strongest predictors for T2MI (Figure 2). In the multivariable logistic regression model, female sex (Beta 1.27 [CI 0.67–1.90], not having typical radiating chest pain (Beta 1.62 [CI 0.96–2.34]) and a baseline hs-TnI concentration ≤ 40.8 ng/L (Beta 1.30 [CI 0.74–1.89]) were selected as predictors for T2MI (Table 2). This model resulted in an AUC of 0.71, which was superior to each single variable alone (Figure 3, Supplementary material online, Table S5). The calibration of this model is displayed in Supplementary material online, Figure S3. These three variables (sex, angina, and troponin) of the multivariable model were transferred to a binary score with one point per variable (Figure 4). Patients with the highest possible score of 3 had a 72% probability of T2MI, while patients with a score of 0 had a 5% probability of T2MI. The diagnostic performance and the ROC curve of the binary score are displayed in Supplementary material online, Figure S4 and Supplementary material online, Table S6.

Figure 2.

AUC of the logistic regression for differentiation of T1MI vs. T2MI. Presented is the unadjusted AUC for selected variables in order to differentiate T1MI from T2MI patients. MI, myocardial infarction; T1MI, type 1 myocardial infarction; T2MI, type 2 myocardial infarction; OR, odds ratio; CI, confidence interval; SD, standard deviation; CAD, coronary artery disease; eGFR, estimated glomerular filtration rate; ECG, electrocardiogram.

Figure 3.

Receiver operating characteristic curve for the multivariable logistic regression model. The receiver operating characteristic curve to predict type 2 myocardial infarction is displayed for each of the most important predictors and the entire multivariable logistic model. These results are based on the fitted multivariable regression model.

Figure 4.

Performance and characteristics of the diagnostic model to predict the probability of T2MI. Score based on the most important predictors from the multivariable model and the related probability of T2MI. T2MI, type 2 myocardial infarction.

Figure S2.

ROC-Curve for determination of the optimal hs-TnI cutoff concentrations

Figure S3.

Calibration plot of the multivariable logistic regression model
These results are based on the fitted multivariable regression model.

Figure S4.

ROC-Curve for the developed point score