Author + information
- Received October 3, 2005
- Revision received November 29, 2005
- Accepted January 2, 2006
- Published online May 16, 2006.
- Anja Wagner, MD⁎,
- Heiko Mahrholdt, MD†,
- Louise Thomson, MD⁎,
- Stefan Hager, MD†,
- Gabriel Meinhardt, MD†,
- Wolfgang Rehwald, PhD⁎,
- Michele Parker, MS, RN⁎,
- Dipan Shah, MD‡,
- Udo Sechtem, MD†,
- Raymond J. Kim, MD⁎,1 and
- Robert M. Judd, PhD⁎,1,⁎ ()
- ↵⁎Reprint requests and correspondence:
Dr. Robert M. Judd, Duke Cardiovascular Magnetic Resonance Center, Duke University Health System, P.O. Box 3934, Durham, North Carolina, 27710.
Objectives This study sought to investigate the influence of time, dose, and inversion time (TI) and their interactions on myocardial infarct size measurements to establish the foundation for a standardized protocol for multicenter trials.
Background There is growing interest in using magnetic resonance imaging (MRI) infarct size measurements as an end point in clinical trials. However, no standardized protocol exists, and there are limited data concerning the effects of time, contrast agent dose, and TI.
Methods First, we determined the influence of postcontrast imaging time (5 to 40 min), contrast agent dose (0.1 vs. 0.2 mmol/kg), TI, and their interactions in an animal model (n = 14). Second, we tested whether the findings of the animal study apply to patients and are generalizable. Therefore, we retested the diagnostic window in a multicenter study. A total of 48 patients with first acute myocardial infarction (AMI) from three centers were imaged twice (5 and 30 min) after injection of 0.15 mmol/kg gadolinium diethylenetriamine-pentaacetate using an adjusted TI.
Results The animal study showed that the infarct size is independent of time and dose (p = 0.9 and p = 0.16, respectively) using an adjusted TI. Using a fixed TI, however, infarct size is a function of time and dose (p = 0.0001 and p = 0.01, respectively). The multicenter study showed that MRI 1 (16.9 ± 12% of left ventricle) was not statistically different from MRI 2 (16.4 ± 12% of left ventricle, p = NS) with no difference between sites (p = NS).
Conclusions The AMI size can be measured with MRI using a contrast dose between 0.1 and 0.2 mmol/kg and a time window of 5 to 30 min after contrast administration, provided that the TI is adjusted.
It has been recently shown that the magnetic resonance imaging (MRI) delayed-enhancement technique has a high sensitivity for the detection of myocardial infarction (1–3), and there is growing interest in the use of MRI infarct size measurements as an end point in clinical trials. The accuracy of the MRI delayed-enhancement technique has been proven in animal models (4,5), and has been tested in patients for the assessment of acute infarction (2,4) as well as chronic infarction (1,5). However, these studies were single-center trials including only a small number of patients.
Currently, there is no standardized protocol for assessing myocardial infarct size. Consequently, different scanner operators use different imaging parameters such as imaging time after contrast agent administration, contrast agent dose, and time of inversion (TI) (6–8). The effects of these scan parameters as well as their interactions on the assessment of infarct size are unknown. However, the knowledge of the limits of these scan parameters is essential for establishing a standardized protocol and conducting multicenter studies with MRI infarct size as an end point. Therefore, we designed a multifactor study to determine the effects of imaging time, contrast agent dose, TI, and their interactions on infarct size measurements in an animal model with direct comparison to histology. Additionally, we examined whether the findings of the animal study apply to humans by retesting the diagnostic window established in the animal model in a second study in patients with first acute myocardial infarction (AMI). To determine whether our findings were generalizable, we included consecutive studies from a variety of patients with AMI that were acquired by different scanner operators at three different MRI centers.
Effects of time, dose, and inversion time
The effects of time after contrast agent application, dose of the contrast agent, and inversion time and their interactions were assessed in a multifactorial animal model for which true infarct size could be assessed by histology.
Fourteen 20- to 30-kg mongrel dogs were studied. The care of all animals was in accordance with the Position of the American Heart Association on Research Animal Use adopted November 15, 1984. Myocardial infarction was produced by occlusion of the left anterior descending coronary artery under sterile conditions for 90 min followed by reperfusion. The details of this experimental protocol have been described in detail elsewhere (4). All animals were scanned within two days of infarction.
The MRI scans were performed under gaseous anesthesia (isoflurane) using a 1.5-T Siemens Sonata (Erlangen, Germany) and a flexible surface coil. Typical in-plane resolution was 1.2 × 1.2 mm, and slice thickness was 6 mm, no gap.
To identify regions of contractile dysfunction, electrocardiogram gated cine magnetic resonance images encompassing the entire left ventricle (LV) were acquired during repeated breath-holds. Short-axis views were obtained from 1 cm below the level of the mitral valve insertion, then every 6 mm throughout the whole LV. Based on the cine images, one short-axis slice with significant contractile dysfunction was selected for delayed-enhancement imaging to maximize the number of time points acquired after injection of the contrast agent. Delayed-enhancement imaging of the selected slice was performed every 5 min for 40 min after intravenous injection of a contrast agent using both an adjusted TI and a fixed TI at each 5-min time point.
For the adjusted TI, inversion-recovery turboFLASH imaging was performed as previously described in detail elsewhere (8). In brief, the TI was set by the scanner operator to null signal from normal myocardium and electrocardiogram gated images were acquired during repeated breath holds in mid-diastole.
For fixed TI, the same slice was scanned with a fixed (not adjusted) TI. The scanner parameters were identical to those for the adjusted TI except that the TI was held constant at a predetermined value. To test the influence of the contrast agent concentration on the accuracy of the delayed-enhancement technique, contrast agent was given in two different concentrations, either 0.1 or 0.2 mmol/kg gadoversetamide (Mallinckrodt) (seven animals for each dose). The value of the predetermined fixed TI was 300 ms for 0.1 mmol/kg and 250 ms for 0.2 mmol/kg.
Animals were sacrificed, and the hearts were excised and then sectioned into 6-mm-thick short-axis slices starting 1 cm below the level of the mitral valve insertion to match the cine images using a commercial rotating meat slicer. The slices matching the MR short axis chosen for the delayed-enhancement technique were then stained with 2% triphenyltetrazolium chloride (TTC) and digitally photographed. The infarct extent was determined quantitatively using the image analysis program NIH Image (National Institutes of Health, Bethesda, Maryland) as previously described (4).
Images at each 5-min time point and each contrast dose were randomized and analyzed blinded to these parameters as well as to the results of the histology. Cine images were used to define the endocardium. The infarct size per slice by MRI was assessed semiautomatically by computer counting of all enhanced pixels within the short-axis image. Enhanced pixels were defined as pixels with signal intensities more than two standard deviations above the mean of image intensities in a remote myocardial region in the same image (4). Infarct size was determined as a percent of the slice by dividing the sum of enhanced pixels by the total number of pixels within the myocardium multiplied by 100%. Subendocardial hypoenhanced regions surrounded by enhancement (no reflow zone) were included as part of the infarct territory (5,9).
We retested the diagnostic window in humans in a multicenter study setting. Therefore, we imaged patients with acute MI at three different CMR sites at two different time points (5 and 30 min after contrast agent injection) using a dose of 0.15 mmol/kg gadolinium diethylenetriamine-pentaacetate and an adjusted TI. All MRI scans were performed using a 1.5-T Siemens Sonata.
Forty-eight consecutive patients with first documented AMI (infarct age 4 ± 2 days, peak creatine kinase 1,433 ± 1,285 U/l) were prospectively enrolled at three institutions (Robert Bosch Medical Center, Stuttgart, Germany [site 1], Duke Cardiovascular Magnetic Resonance Center, Durham, North Carolina [site 2], and the Nashville Cardiovascular MRI Institute, Nashville, Tennessee [site 3]). All patients had enzymatically proven myocardial necrosis (peak total creatine kinase >125 IU/l and peak creatine kinase-MB isoenzyme >9 · 0 μg/l).
Patients were excluded if they had a prior history of myocardial infarction or contraindications to MRI. No patient was excluded from the study for technical or image quality reasons. The study was approved by the institutional review board of each enrolling institution, and all patients gave written informed consent.
Patient images were acquired in a manner identical to that of the adjusted TI animal study, i.e., using an inversion-recovery turboFLASH pulse sequence with scanner settings determined as previously described (8). All patients were scanned twice: once at 5 min and again at 30 min after administration of contrast agent (0.15 mmol/kg gadoversetamide).
Patient images from each site and each time point were randomized for blinded analysis. Infarct size by MRI was determined in a manner identical to that of the animal studies, i.e., semiautomatically by computer counting of all enhanced pixels in the myocardium on each of the six to eight short-axis images. Infarct size was determined as a percent of left ventricular mass (%LV) as the sum of enhanced pixels from each of the six to eight short-axis images divided by the total number of pixels within the LV myocardium multiplied by 100%. Mass in grams was calculated assuming a specific gravity of 1.05 g/cm3.
Continuous data are expressed as mean ± SD. The reproducibility of MRI was analyzed using the repeatability analysis method of Bland-Altman (10). The mean and 95% confidence intervals were calculated as described by Bland-Altman (10). One-way analysis of variance with repeated measures was used to determine the effects of time, dose, and their interactions on infarct size measurements; p < 0.05 was regarded as statistically significant.
Effects of time, dose, and inversion time and their interactions
We tested the effects of imaging time after contrast agent administration and contrast agent dose for a large variety of infarct sizes, ranging from 7.9% to 46.3% of the slice (mean 17.9%). Figure 1shows a typical example from the animal study arranged in chronological order. Using an adjusted TI to null normal myocardium (upper row), both the appearance of the images and the spatial extent of enhancement did not change over a time of 40 min. Conversely, using a fixed TI (lower row), both the general appearance of the images and the spatial extent of enhancement changed over time.
Figure 2compares the MR images acquired 40 min after administration of contrast agent of one animal using an adjusted TI (middle) and fixed TI (right) to the gold-standard histology (TTC). For adjusted TI, the enhanced area matches the infarct defined by histology, whereas using a fixed TI, the enhanced region acquired 40 min after contrast agent appears smaller than the true infarct size defined by TTC.
The inversion times selected by the scanner operators to null normal myocardium were higher for the 0.1 mmol/kg contrast dose than for the 0.2 mmol/kg contrast agent dose and increased over time. For all animals, the TI was increased from 265 ± 12 ms and 237 ± 5 ms for the scan performed 5 min after administration of contrast to 372 ± 13 ms and 331 ± 11 ms for the scan performed 40 min after administration of contrast, 0.1- and 0.2-mmol/kg contrast agent dose, respectively.
Figure 3shows the effects of time, dose, and TI on infarct size measurements expressed as an MRI/TTC ratio for all animals. Using an adjusted TI, the infarct size is independent of dose and time within the tested limits. Using a fixed TI, however, infarct size becomes a function of dose and time. Analysis of variance with repeated measures showed that there are no significant differences in infarct size measurements attributable to dose (p = 0.16) or time (p = 0.9) or the combination effect of time and dose interaction (p = 0.28) within the tested limits using an adjusted TI. With the use of a fixed TI, however, the measured infarct size was a function of time and dose (main TI effect p = 0.003). The effect of the interaction between fixed TI and time and dose was significant, with p = 0.0001 and p = 0.01, respectively. The combined effect of the coefficients for fixed TI, fixed TI × time, and fixed TI × dose (at 40 min) implies that the overall effect on infarct size would be 58.5% of the size measured by TTC for a dose of 0.1 mmol and 67.4% of the size measured by TTC for a dose of 0.2 mmol.
Table 1summarizes the clinical characteristics of the patient population. Of the 48 patients, 20 had anterior infarction and 46 were successfully revascularized before the magnetic resonance images were acquired.
Figure 4depicts a full set of short-axis views of MRI 1 and MRI 2 acquired in three different patients. In all patients, the presence, location, and size of the enhanced regions were similar in both MRI scans using an adjusted TI.
On average, MRI 1 was performed 6 ± 2 min after administration of contrast and the average inversion time selected by the scanner operators was 283 ± 23 ms, whereas MRI 2 was performed 27 ± 3 min after administration of contrast and the inversion time was 406 ± 35 ms. Despite these differences in postcontrast imaging time, the mean infarct size for MRI 1 (16.9 ± 12% of LV, 26.0 ± 21.1 g) was not statistically different from that of MRI 2 (16.4 ± 12% LV, 25.9 ± 21.3 g, p = NS for both). The range of human infarct sizes was 1.1% to 51% of LV mass.
Figure 5shows the results of Bland-Altman analysis of the MRI data of all 48 patients expressed in %LV mass (panel A) and mass in grams (panel B). The infarct size determined by MRI using an adjusted TI remained constant between 5 and 30 min. The difference in infarct size between scan 1 and 2 (bias) was 0.003% LV or 0.07 g. The 95% confidence intervals for infarct size comparing MRI 1 and MRI 2 were ±1.6% of LV mass or 2.5 g and −2.3 g, i.e., there were no systematic differences in infarct size between the two MRI scans.
This is the first study evaluating appropriate imaging parameters to ensure the capability of MRI as a tool for conducting multicenter studies with infarct size as an end point. The main findings of this study were that acute myocardial infarct size can be measured independent of time and dose with the MRI delayed-enhancement technique using a time window of 5 to 30 min after administration of contrast and a contrast agent dose between 0.1 and 0.2 mmol/kg, provided that the TI is adjusted for image acquisition.
Effects of time, dose, and inversion time and their interactions
We found that the MRI delayed-enhancement technique accurately measures acute myocardial infarct size independent of time and dose within our tested limits using an adjusted TI.
We decided to start imaging 5 min after contrast agent injection to have sufficient time for the blood pool signal in the LV cavity to decline and allow discrimination between LV cavity and the endocardial border of the enhanced infarct. From our experience, 30 min of imaging time is more than sufficient to perform the delayed-enhancement technique in animals and patients.
The results of the current study indicate that any protocol for assessing infarct size measurements using an inversion-recovery MRI pulse sequence must use a method of adjusting the TI. By manually adjusting the TI to null signal from normal myocardial regions, the difference in image intensity between infarcted and normal myocardium is maximized (2). It has been shown that the TI needed to null signal from normal myocardium varies from subject to subject as well as with time after contrast injection, mainly because of pharmacokinetics (11). Weinmann et al. (11) studied the pharmacokinetics of gadolinium diethylenetriamine-pentaacetate in humans for doses of 0.1 and 0.25 mmol/kg, concluding that the plasma concentration decreased by a factor of approximately 2.4 between 3 and 40 min after administration. This decrease in contrast concentration will increase myocardial T1 and thus require a corresponding increase in the TI to appropriately null normal myocardium.
The inversion recovery sequence is a magnitude reconstruction. If the TI selected is too short, typically black bands will appear in the periphery of the contrast-enhanced region (8) as shown in Figures 1 and 3, resulting in an underestimation of the infarcted area. If the TI is set too long, the magnetization of normal myocardium will be above zero and will appear gray, resulting in a reduction of the difference in image intensities between infarcted and normal myocardium (8), and may reduce the accuracy of infarct size measurements. Accordingly, if the TI is not appropriately adjusted, the apparent infarct size changes over time and is erroneous (Figs. 1 and 3). However, we did not calculate regions of interest including phase cancellation areas, and one may argue that the infarct size differences would be smaller.
Multicenter confirmation of the diagnostic window
This is the first study reporting the diagnostic window of acute myocardial infarct size measurements assessed by the MRI delayed-enhancement technique in a multicenter setting. In a previous single-center study of chronic infarcts, Mahrholdt et al. (12) performed two MRI scans 10 ± 2 min and 27 ± 3 min after administration of contrast using an adjusted TI and a contrast dose of 0.125 mmol/kg in 20 patients and found a variance between the two scans of ±2.4% of LV mass. The value reported by Mahrholdt et al. (12) is similar to the value of acute infarcts found in the current study (±1.6% of LV mass), although the reported value of the two scans found in the current study is better most likely because we did not remove and reposition the patients between the two MRI scans as done by Marholdt et al. (12). Thus, the variances reported in the current study are not true interstudy variances.
Implications for clinical trial design
There has been a growing interest in infarct size measurements as a surrogate end point for clinical trials. Up to this date, there has been very limited experience with the use of MRI for infarct sizing in randomized clinical trials. The current report is the first study showing the capability of MRI as a tool for measuring infarct sizing in a multicenter setting.
Currently, 99mTc-sestamibi single-photon emission computed tomography (SPECT) imaging is thought to be the best available imaging technique for infarct size determination (13,14), and infarct size measured by SPECT has served as an end point in numerous studies for evaluating potential efficacy in randomized trials and in dose-ranging studies (15–17). To define the role of infarct size assessed by MRI as an end point in clinical trials, one must evaluate whether MRI offers an advantage compared with SPECT in terms of a reduction in the sample size.
In chronic infarcts, Mahrholdt et al. (12) performed both MRI and 99mTc-sestamibi SPECT in the same patients and reported that MRI could reduce the number of patients needed for a clinical trial approximately two-fold in studies in which each patient was scanned twice. The variance reported by Mahrholdt et al. (10) (2.4% of LV mass) was similar to the observed variance of the current study (1.6% of LV mass). In general, the effect of substituting MRI for SPECT on sample size for clinical trials depends on the design of the study as well as on the intrinsic variation in true infarct size within the population of the patients under study. Currently, it is unknown whether relative drop-out rates (i.e., secondary to device implantation) as well as the fact that SPECT misses small infarcts (3) has a significant impact on sample size.
Acute myocardial infarct size can be measured independent of time and dose with the MRI delayed-enhancement technique using a contrast dose between 0.1 and 0.2 mmol/kg and a time window of 5 to 30 min after administration of contrast. Magnetic resonance imaging is suitable for clinical applications such as post-infarct risk stratification and should be considered for research applications such as use as an end point for randomized trials.
↵1 Drs. Judd and Kim are inventors on a related patent owned by Northwestern University, and are consultants for Tyco-Mallinckrodt.
This work was supported in part by a grant from the Robert Bosch Foundation (Dr. Mahrholdt), NIH-NHLBI RO1-HL63268 (Dr. Judd), K02-HL04394 (Dr. Judd), and RO1-HL64726 (Dr. Kim).
- Abbreviations and Acronyms
- acute myocardial infarction
- left ventricle/ventricular
- magnetic resonance imaging
- single-photon emission computed tomographic
- inversion time
- triphenyltetrazolium chloride
- Received October 3, 2005.
- Revision received November 29, 2005.
- Accepted January 2, 2006.
- American College of Cardiology Foundation
- Kim R.J.,
- Fieno D.S.,
- Parrish T.B.,
- et al.
- Fieno D.S.,
- Kim R.J.,
- Chen E.L.,
- Lomasney J.W.,
- Klocke F.J.,
- Judd R.M.
- Oshinski J.N.,
- Yang Z.,
- Jones J.R.,
- Mata J.F.,
- French B.A.
- Mahrholdt H.,
- Wagner A.,
- Holly T.A.,
- et al.
- Gibbons R.J.,
- Miller T.D.,
- Christian T.F.
- Gibbons R.J.,
- Valeti U.S.,
- Araoz P.A.,
- Jaffe A.S.
- Mahaffey K.W.,
- Puma J.A.,
- Barbagelata N.A.,
- et al.
- Schaer G.L.,
- Spaccavento L.J.,
- Browne K.F.,
- et al.