Between November 2010 and July 2012, a total of 137 consecutive STEMI patients whose baseline thyroid hormone levels had been determined were enrolled in the present study with the approval of the Chosun University Hospital Research Ethics Committee (approval no. CHOSUN 2012-09-003). Patients were included if they were over 18 years of age and underwent primary percutaneous coronary intervention (PCI) within 12 hours of symptom onset. Patients who did not consent to CE-CMR imaging or who exhibited contraindications to such imaging were excluded. Ultimately, 40 patients were included.

STEMI was defined as ST-segment elevation of at least 1 mm in two or more standard leads, or of at least 2 mm in two or more neighboring precordial leads, or a (presumed) new-onset left-bundle branch block.

All patients received dual oral antiplatelet treatment, consisting of aspirin (300 mg) and clopidogrel (600 mg) prior to intervention followed by daily aspirin (100 to 200 mg) and clopidogrel (75 mg) subsequently. Coronary angiography and stent implantations were performed using standard intervention techniques. Glycoprotein IIb/IIIa receptor antagonists were administered intravenously as judged appropriate by attending physicians.

Venous blood samples were collected into serum-separating tubes (BD Vacutainer Systems, Franklin Lakes, NJ, USA) on each index day. Thyroid-stimulating hormone (TSH), T3, and FT4 levels were measured by an immunoradiometric assay (TSH) or radioimmunoassays (RIAs; T3 and FT4), using RIA-gnost FT4, TSH, and T3 kits (CISbio International, Cedex, France) and a Cobra E 5005 gamma counter (Packard, Ramsey, MN, USA). All assays were performed within 2 hours of sample collection. The normal ranges are as follows: TSH, 0.25 to 4 mIU/L; FT4, 0.7 to 1.8 ng/dL; and T3, 60 to 190 ng/dL.

We used a 1.5-T magnetic resonance (MR) scanner (Avanto, Siemens Medical Solutions, Erlangen, Germany) f itted with a SENSE cardiac coil. The magnetic resonance imaging (MRI) protocol consisted of T2-weighted imaging, cine-imaging, first-pass perfusion, evaluation under adenosine-induced stress, rest perfusion, and delayed (5 and 15 minutes) enhancement imaging. Cine-MRI images were obtained using a fast-gradient echo sequence (steady-state free precession) after scout and localizer image acquisition. The slice thickness was 8 mm (with no gap), and short-axis left ventricular (LV) images were acquired from the apex to the base (thus including the entire LV volume). Repeated breath holds were required to acquire adequate images. The temporal resolution was 25 to 30 frames per RR interval. T2-weighted MRI was performed in the cardiac short-axis direction using a dark-blood T2-weighted inversion-recovery fast-spin echo sequence. Next, 0.2 mmol/kg gadolinium-diethylenetriaminepentaacetic acid (Magnevist, Bayer Schering Pharma, Berlin, Germany) was injected intravenously at 3 mL/sec followed 4 minutes later by a saline flush under adenosine infusion. The first pass of the contrast agent (after intravenous bolus injection) through the myocardium was recorded using a T1-weighted dynamic sequence. The slice thickness was 8 mm, the field of view 35 × 35 cm, and the image matrix 128 × 128. Images were taken on 40 occasions of four locations (the apex, the lower mid-axis, the upper mid-axis, and the basal short axis) every two heart beats. Delayed hyperenhancement and the extent of MVO were evaluated 5 and 15 minutes after contrast administration in 10 to 12 contiguous slices (each 8 mm thick; no gap) using a phase-sensitive inversion recovery-spoiled gradient-echo sequence (echo time 4 ms; repetition time 8 ms; and flip angle 30°). The field of view and image matrix were 35 × 35 cm and 256 × 256, respectively. The inversion delay time was varied within the range of 200 to 300 ms.

All measurements were performed at our MRI core laboratory. After the short-axis images were acquired at end-diastole and end-systole and the endocardial borders traced, the left ventricular end-diastolic volume (LVEDV), left ventricular end-systolic volume (LVESV), and ejection fraction (EF) were computed using Simpson's algorithm (Argus, Siemens Medical Solutions). Myocardial mass was calculated by multiplying the myocardial volume by the myocardial density (1.05 g/mL). Each infarct volume was quantified based on visual border detection using a manual drawing method with a commercial software package (INFINITT PACS, INFINITT Healthcare, Seoul, Korea). The infarct volume was calculated as the sum of the areas exhibiting delayed hyperenhancement within each segment of the short-axis images multiplied by the slice thickness over the entire left ventricle. The volume and extent of the MVO, defined as late-hypoenhanced regions within the infracted myocardium evident on delayed enhancement images, were calculated in the same manner. Endocardial and epicardial borders were delineated by planimetry to allow calculation of myocardial areas and were summed as described above to calculate LV myocardial volume. The infarct volume (i.e., infarct size) and the extent of MVO were expressed as percentages of the LV myocardial volume. T2-weighted images were used to explore whether myocardial hemorrhage was evident; this presented as a central hypointense signal core within the area of increased T2 signal intensity [26]. The myocardial AAR, defined as myocardium with a signal intensity greater than two standard deviations (SDs) above the mean signal intensity of remote normal myocardium, was quantified on T2-weighted images using a similar algorithm and expressed as a percentage of the LV myocardial volume. The ratio between infarct size and the AAR was calculated for each patient. The myocardial salvage index was computed as follows: myocardial salvage index = (AAR - infarct size) × 100/AAR [27]. To allow regional analysis, the LV was divided into 17 myocardial segments, as described previously [28]. The extent of infarct transmurality in each segment was calculated by dividing the maximal hyperenhanced thickness by the full thickness of the affected myocardium. Transmural infarction was considered to be present when the extent of infarct transmurality was > 75%.

All values are expressed as means ± SDs, medians (with interquartile ranges [IQRs]), or numbers (with percentages). The Mann-Whitney test was used to compare continuous variables and the chi-square test to compare noncontinuous variables, to establish the baseline characteristics of the various groups.

Receiver operating characteristic (ROC) analysis was used to determine the sensitivities and specificities (with 95% confidence intervals [CIs]) of thyroid hormone cutoff values. Multivariable logistic regression analysis of the factors associated with the extent of transmural infarction (> 75% of infarct transmurality) was conducted using a forward conditional stepwise model. Baseline clinical factors with p values < 0.1 in the univariable analysis were entered into this model. The independent variables were the presence of diabetes mellitus (DM), a T3 level ≥ 68.3 ng/dL, and the use of glycoprotein IIb/IIIa inhibitor(s). The relationships between thyroid hormone levels and other clinical variables were evaluated using the aid of Pearson correlation analysis. All statistical analyses were performed using SPSS version 15 (SPSS Inc., Chicago, IL, USA), and a p value < 0.05 was considered to reflect statistical significance.

The mean patient age was 57.8 years, and 82.5% of patients were male. The clinical, angiographic, and CE-CMR imaging characteristics of the entire cohort are shown in Tables 1, 2, and 3, respectively. The overall mean T3, FT4, and TSH levels were 73 ± 23.7 ng/dL (median, 68.3; IQR, 59.2 to 85.1; normal range, 60 to 190); 1.09 ± 0.28 ng/dL (median, 1.07; IQR, 0.90 to 1.25; normal range, 0.7 to 1.8); and 1.137 ± 1.14 mIU/L (median, 0.71; IQR, 0.38 to 1.63; normal range, 0.25 to 4), respectively.

Compared with patients with T3 levels < 68.3 ng/dL, the mean extent of infarct transmurality was higher in those with T3 levels ≥ 68.3 ng/dL (47% ± 27.2% vs. 67% ± 23.9%; p = 0.02). Furthermore, the group with higher T3 levels exhibited a greater extent of transmural infarction than did the lower T3 group (p = 0.003) (Fig. 1). However, no significant difference in the mean levels of infarct transmurality or the transmural extent of infarction was evident between those with higher and lower levels of TSH or FT4.

The cutoff T3 level predicting the extent of transmural infarction (> 75% of infarct transmurality) was evaluated via ROC analysis. The T3 cutoff value was 68.3 ng/mL, with 80% sensitivity (95% CI, 51.9 to 95.7) and 68% specificity (95% CI, 46.5 to 85.1) (area under the curve [AUC] = 0.691, p = 0.046) for detection of transmural infarction (Fig. 2).

The clinical and demographic characteristics of these two groups are shown in Table 1. Patients with high T3 levels were more likely to be male, showed a lower incidence of Killip class ≥ 2 disease, were more likely to have a higher initial heart rate, and exhibited a greater probability of a greater free T4 level. No significant difference in any other baseline clinical characteristic was evident between the two groups.

Table 2 shows the angiographic and procedural results. No patient with a low T3 level had disease of the left circumflex artery. No other difference in angiographic or procedural characteristics was evident between the two groups.

CE-CMR imaging was performed at a median interval of 41 days after the index event (IQR, 25 to 61 days). The interval from the procedure to CE-CMR tended to be greater in the high T3 group (49.5 days [IQR, 23 to 80.5]) than in the low T3 group (40.5 days [IQR, 25.8 to 55.5]). Fig. 3 shows representative coronary angiographic images before reperfusion and CE-CMR images of reperfused STEMI patients. The results of CE-CMR are shown in Table 3. No significant difference in LVEDV, LVESV, LV mass, left ventricular ejection fraction (LVEF), infarct size, the extent of hemorrhagic infarction, or MVO area was evident between the two groups. The AAR was significantly larger and the myocardial salvage index higher in the low T3 group than the high T3 group upon CE-CMR imaging. However, the numbers of segments exhibiting > 75% infarct transmurality as well as the maximal infarct transmurality were higher in the high T3 group. In addition, transmural infarction was detected more frequently in the high T3 than the low T3 group.

The relationships between cardiac enzyme peak levels and the extent of transmurality were evaluated using Pearson correlation analysis. Peak creatine kinase-MB (CK-MB) and peak high-sensitivity cardiac troponin T (hs-cTnT) levels were significantly (positively) associated with the extent of transmurality (r = 0.392, p = 0.012; r = 0.465, p = 0.002, respectively).

The cutoff levels of cardiac enzymes predictive of the transmural extent of infarction (> 75% of infarct transmurality) were evaluated by ROC analysis. The cutoff values were 137.15 ng/dL for CK-MB, with 86.7% sensitivity (95% CI, 59.5 to 98.3) and 52% specificity (95% CI, 31.3 to 72.2; AUC = 0.637, p = 0.150), and 4.485 ng/dL for hs-cTnT, with 86.7% sensitivity (95% CI, 59.5 to 98.3) and 60% specificity (95% CI, 38.7 to 78.9; AUC = 0.725, p = 0.018).

Univariate analysis showed that DM positivity, a T3 level ≥ 68.3 ng/dL, and use of glycoprotein IIb/IIIa inhibitors were significantly associated with the extent of transmural infarction. We included the significant univariate variables in the multivariate logistic regression analysis. The only variable that remained as an independent risk factor for transmural infarction was a high T3 level (≥ 68.3 ng/dL) (Table 4).

Univariate analysis showed that a high peak CK-MB level (≥ 137.15 ng/dL), a high peak hs-cTnT level (≥ 4.485 ng/dL), a long interval between the index day and the day of CE-CMR imaging (> 41 days; the median value), the transmural extent of infarction, and a T3 level ≥ 68.3 ng/dL were all significantly associated with a low median myocardial salvage index when the entire cohort was evaluated. We included the significant univariate variables in the multivariate logistic regression analysis. The only variable that remained as an independent risk factor for a low myocardial salvage index was a long median interval from the index day to the CE-CMR imaging day.

Univariate analysis showed that only a longer symptom to balloon time (> 189 minutes; the median value) was significantly associated with a high AAR (> 13.3%; the median value).

The relationships between the T3 level and other clinical variables were evaluated using Pearson correlation analysis. All continuous variables, including age, Killip class, systolic blood pressure at admission, initial heart rate, door to balloon time, symptom to balloon time, peak levels of cardiac enzymes (CK-MB and hs-cTnT), levels of creatinine and high-sensitivity C-reactive protein, number of diseased vessels, "baseline thrombolysis in myocardial infarction" flow grade, and LVEF, were evaluated. Of these, only the initial heart rate was significantly associated with the T3 level (r = 0.379, p = 0.016) (Fig. 4). Age tended to correlate negatively with the T3 level (r = -0.302, p = 0.058).

To determine if initial heart rate and age served as surrogate markers of T3 levels, we determined the cut-off values predictive of a high T3 level (≥ 68.3 ng/dL) for both variables using ROC analysis. The cutoff initial heart rate was 67 beats per minute, with 75% sensitivity (95% CI, 50.9 to 91.3) and 55% specificity (95% CI, 31.5 to 76.9; AUC = 0.631, p = 0.156), and the cutoff age was 62 years, with 80% sensitivity (95% CI, 56.3 to 94.3) and 50% specificity (95% CI, 27.2 to 72.8; AUC = 0.626, p = 0.172), with regard to predicting a high T3 level (≥ 68.3 ng/dL). However, neither heart rate nor age was predictive of the extent of transmural infarction (Table 4).

The present study was performed on a relatively small sample size, rendering imprecise ORs and wide CIs. Furthermore, the small sample size limited our ability to consider additional (possibly relevant) factors. In addition, the investigation was not prospective, and our conclusions are thus subject to the limitations inherent in such analyses.

Selection bias was possible because we included only STEMI patients in whom thyroid hormone levels had been measured and infarct transmurality assessed via CE-CMR imaging.

In conclusion, a high T3 level (> 68.3 ng/mL) was an independent predictor of the extent of transmural infarction in patients with STEMI. This predictive power was independent of glycoprotein IIb/IIIa inhibitor use and DM status.