Diagnostic and prognostic significance of transient ischemic dilation (TID) in myocardial perfusion imaging: A systematic review and meta-analysis

Background Transient ischemic dilatation (TID) of the left ventricle is a potential marker of high risk obstructive coronary artery disease on stress myocardial perfusion imaging (MPI). There is, however, interstudy variation in the diagnostic performance of TID for identification of severe and extensive coronary disease anatomy, and varied prognostic implications in the published literature. Methods We searched MEDLINE, EMBASE, and COCHRANE databases for studies where TID was compared with invasive or CT coronary angiography for evaluation of coronary artery stenosis. Two reviewers independently evaluated and abstracted data from each study. A bivariate random effects model was used to derive pooled sensitivities and specificities, in order to account for correlation between TID in MPI and anatomic disease severity. Results A total of 525 articles were reviewed, of which 51 met inclusion criteria. Thirty-one studies contributed to the analysis, representing a total of 2037 patients in the diagnostic meta-analysis and 9003 patients in the review of prognosis. The ratio above which TID was deemed present ranged from 1.13 to 1.38. Pooled sensitivity was 44% (95% CI 30%-60%) and specificity was 88% (95% CI 83%-92%) for the detection of extensive or severe anatomic coronary artery disease. Analysis of outcome data demonstrated increased cardiac event rates in patients with TID and an abnormal MPI. In otherwise normal perfusion, TID is an indicator of poor prognosis in patients with diabetes and/or a history of coronary disease. Conclusions Among patients undergoing MPI, the presence of TID is specific for the detection of extensive or severe coronary artery disease. Electronic supplementary material The online version of this article (doi:10.1007/s12350-017-1040-7) contains supplementary material, which is available to authorized users.


INTRODUCTION
Myocardial perfusion Imaging (MPI) is an established tool for the diagnosis and risk stratification of patients with coronary artery disease (CAD) for over three decades. 1 MPI has excellent diagnostic and prognostic accuracy and also provides good insight into cardiac function through the interpretation of a variety of perfusion and functional parameters. 2,3 One of these functional parameters is transient ischemic dilation (TID), which has been validated both as a marker of severe and extensive coronary artery disease and as a predictor of cardiac outcomes in independent studies. [1][2][3] To date, the pathophysiology of ischemic LV dilatation remains unclear with the theory of subendocardial ischemia gaining the widest acceptance. 4,5 Others cite data supporting ischemia induced physical LV dilation post stress. 6,7 However, several studies have demonstrated that ischemic LV dilatation may be present in patients with normal perfusion and no significant epicardial coronary disease; for example in patients with hypertrophic cardiomyopathy, 8 or in patients with hypertensive heart disease and left ventricular hypertrophy. 9 Therefore the true diagnostic accuracy of TID on MPI is debated and the optimal threshold for its definition remains undefined.
In this study, we conducted a meta-analysis of the diagnostic performance of the presence of TID, compared to anatomical coronary artery assessment. We also conducted a systematic review of the prognostic significance of TID. Both components of our study included patients who underwent either exercise or pharmacologic stress MPI.

METHODS
We employed a systematic search of the MEDLINE, EMBASE, and COCHRANE databases. We searched for English language studies, which examined the diagnostic and/ or prognostic accuracy of TID in myocardial perfusion imaging. The search words used were (transient ischemic dilation, transient ischaemic dilation, left ventricular dilation, transient dilation, SPECT, single photon tomography, CT single photon, myocardial perfusion imaging, and myocardial scintigraphy).
Two investigators (MA, HE) independently reviewed the studies and extracted the relevant data including patient demographics, the radiotracer used, the stress modality, and findings on coronary angiography. Where additional data were required to complete the meta-analysis or discrepancies existed, attempts were made to contact the original authors to obtain such information. We excluded studies where: a) there was no coronary arterial anatomic assessment (either invasive coronary angiogram or coronary CT angiography) for comparison, or b) there was no clear documentation of the method used to calculate the ratio above which TID was diagnosed.
The studies identified for inclusion by the two investigators, and data extracted, were reviewed for eligibility and accuracy by a third investigator (DSL). Methodological quality regarding the risk of bias and concerns of applicability was assessed using the Quality Assessment of Diagnostic Accuracy Studies 2 (QUADAS 2) tool. 10 QUADAS 2 is a tool used for quality assessment of diagnostic accuracy studies included in systematic reviews and meta-analyses, to assess the risk of bias and applicability for use in systematic reviews. This tool contains 4 domains: patient selection, index test, reference standard, and flow and timing.

Statistical Analysis
Data were extracted to construct 2 9 2 tables, from which the sensitivity and specificity of each study was calculated. The sensitivity and specificity estimates were pooled using a See related editorial, pp. 738-741 bivariate random effects model, as recommended by the Cochrane Diagnostic Test Accuracy Working Group. 11 The bivariate model was then used to construct a hierarchical summary receiver operator curve (ROC). A P value \.05 was considered statistically significant. We did not calculate an I 2 statistic given that it is not an accepted method of measuring heterogeneity between diagnostic studies. 11 Univariate meta-regression was used to assess the significance of key covariates that were likely to affect the diagnostic accuracy of the test. Due to the small number of studies, multivariate meta-regression was not performed as it was likely to be underpowered to detect any differences.
All statistical analyses were conducted using STATA/SE, version 12.0 (Stata Corp LP, College Station, Texas, USA).

Summary of Studies Examining TID
From the initial database search, we identified 525 citations of which 368 articles remained after removing duplicates. After reviewing the titles and abstracts of these records, 317 articles were excluded because they were not relevant to the purpose of the study.
Of the remaining 51 articles, 20 studies were excluded because: (a) there was no evaluation of coronary anatomy (invasive coronary angiogram or CT angiography) for diagnostic studies [n = 11], (b) they included only patients with left ventricular dysfunction and fixed perfusion defects for prognostic studies [n = 1], (c) there was no clear documentation of the method used to calculate the ratio above which TID was diagnosed or only visual assessment of TID was employed [n = 5], (d) incomplete data [n = 2], and (e) duplicate data [n = 1] as shown in Figure 1 (Group Z). These excluded studies are shown in Online Table A. We included 31 studies, of which 23 evaluated TID from a diagnostic perspective. Of these, 13 studies were included in the quantitative meta-analysis ( Figure 1, Group A), and 10 studies were not quantitatively synthesized because the patient-level data were only reported in aggregate, and patients could not be separated into TID positive or negative, or severe or nonsevere CAD categories (Group C). The quantitative meta-analysis encompassed a total of 2037 patients in the diagnostic evaluation and 9003 patients in the prognostic evaluation ( Figure 1). There were 8 studies examining the prognostic significance of TID, which did not report patient-level data and therefore were incorporated in a narrative synthesis (Group B).

Studies of Diagnostic Performance
Characteristics of the 13 studies included in the quantitative meta-analysis are shown in Table 1, and the 10 non-quantitatively analyzed studies are shown in Online  Table B. The definition of severe and extensive CAD and software used for TID evaluation are shown in Table 2. Technetium-99 was the most commonly utilized radiotracer (8 studies), followed by Tl-201 (2 studies), dual isotope scanning (2 studies), and Rb-82 (1 study). Coronary angiography was performed in 73% of patients. The modality of stress included exercise (4 studies), pharmacologic stress (6 studies), and both exercise and pharmacologic stress (3 studies). Methodological Quality Assessment using the QUADAS 2 tool revealed that most of the studies included in the meta-analysis demonstrated low risk for bias or concern regarding applicability ( Figure 2).
The ratio above which TID was diagnosed ranged from 1.13 to 1.38, with differences noted depending on the tracer used. Using the published TID ratios (as defined by the individual study authors), the sensitivity of TID for the detection of extensive or severe CAD ranged from 21% to 62.5% (see Table 3). One study only showed very low sensitivity (7%). 12 Specificities were higher and demonstrated less variability, ranging from 77% to 98% (Table 3). Bivariate analysis of the 13 studies, which had complete statistical data, revealed a pooled sensitivity of 44% (95% confidence interval [CI] 30%-60%) and a pooled specificity of 88% (95% CI 83%-92%) as shown in Figure 3. The pooled area under the ROC curve was 0.82 (0.78-0.85) for all studies ( Figure 4).

Subgroup Analyses for Diagnostic Performance of TID
Subgroup analysis of the technetium studies revealed a pooled sensitivity of 42% (95% CI 23%-63%) and a pooled specificity of 86% (95% CI 78%-92%), which was similar to the overall results above (see Figure 5). Subgroup analyses also demonstrated that studies using exercise as a stressor demonstrated a significantly higher pooled area under the receiver operating characteristic curve (AUC 0.92 vs 0.78, P \ .001) for detection of severe CAD compared to studies using pharmacological stressors ( Figure 5). In studies that determined the presence of TID qualitatively (instead of quantitatively) the pooled sensitivity was 46% (95% CI 38%-54%) and pooled specificity was 88% (95% CI 79%-93%).
For studies using technetium, the TID range was 1.13-1.31. In these studies, the sensitivity ranged from 34% to 56%, while the specificity ranged from 77% to 98% for detection of severe CAD. Overall sensitivity and specificity were similar for both exercise and pharmacologic technetium studies. Based on univariate meta-regression, the TID ratio cutoff used in various studies had little effect on specificity (P = .99).

Studies of Prognostic Performance
Studies evaluating TID as a marker of increased cardiac events are summarized in Table 4. Across studies, the annualized rates of cardiac death or MI ranged from approximately 0.2% to 1% in those with no TID, 2% to 5% in those with TID and normal perfusion, and 5% to 6% among those with TID and ischemia, CAD, or diabetes. De Winter reported that TID was a significant predictor of all-cause mortality even after multivariable adjustment for resting heart rate, betaadrenoreceptor antagonism, summed rest score, and resting LV ejection fraction. 13 There was heterogeneity of patients in the different prognostic studies of TID. In a study of diabetic patients with TID and ischemia by Petretta et al, the annualized rate of cardiac death or non-fatal MI was 7.2% with post-stress LVEF ≤45% and 5.4% when post-stress LVEF was greater than 45%. 14 In another study of patients with TID, rates of fatal and non-fatal MI were substantially increased in those who were not revascularized, compared to those who underwent CABG surgery or PCI. 15 One study examined patients undergoing MPI prior to non-cardiac surgery and reported high postoperative cardiac event rates: 58% in the presence of TID, 19% with reversible perfusion defects but no TID, and 2% in patients with normal scans. These events were temporally accelerated, with the majority of cardiac events occurring within 4 months postoperatively. 16

Special Consideration of TID in Patients with Otherwise Normal Perfusion Scans
The study by Abidov et al demonstrated that TID is an independent prognostic marker for cardiac events in patients with either normal or near-normal MPI. 17 Patients in the highest TID quartile (mean TID ratio of 1.35 ± 0.14) were older and diabetic. The prognostic  CTA, CT angiography impact of TID with normal myocardial perfusion was modified by the presence of CAD or diabetes, 18 with an increased risk of cardiac death or MI reported in these patients. 19 In a study of diabetes patients with normal post-stress LVEF and no ischemia, the annual event rate was 4.9% in those with TID and 0.2% in those without TID (P \ .001). 14

DISCUSSION
Our study showed that TID is a specific but not a sensitive marker for detection of severe and extensive CAD with a pooled sensitivity of 44% and pooled specificity of 88%. In the analysis of subgroups, we found that exercise stress resulted in a trend toward higher sensitivity than pharmacologic stress, but specificity was similar. The prognostic studies demonstrated consistently elevated risk when TID was present despite somewhat different populations studied.
This risk was heightened in those with TID and post-stress LVEF ≤ 45%, with rates of cardiac death or MI exceeding 7%·year −1 . Among patients with normal perfusion scans, the presence of TID was associated with increased risk primarily when patients had a history of CAD or DM.
While there are many disparate studies examining diagnostic test performance, few have utilized metaanalytic approaches and summary receiver operating characteristic curves to evaluate the performance of a specific high risk marker such as the presence of TID. Many systematic reviews and meta-analyses were conducted to study the overall diagnostic and prognostic role of different imaging modalities like stress echocardiography, 20 cardiac PET, 21 and coronary CT angiography. 22,23 However, meta-analyses of specific components of a diagnostic test, such as TID are less commonly encountered. Despite the importance that is imparted to the presence of TID in nuclear cardiology, to our knowledge, this is the first meta-analysis that studied its quantitative diagnostic performance for detection of severe CAD and prognostic performance for prediction of cardiac outcomes.  Mechanistic studies may explain, in part, the reason for the high specificity and low sensitivity of TID. Prior studies have demonstrated that TID results from subendocardial ischemia with apparent LV dilatation due to decrease in the radiotracer uptake in the endocardium. 4,5 Others have proposed that TID is a manifestation of LV dilatation post-stress due to ventricular dysfunction. 24 Therefore, the presence of TID usually indicates the existence of severe CAD, favoring higher specificity and lower sensitivity for less-critical or less-extensive disease. The diagnostic performance of TID could also be impacted because it may occur in those with hypertensive heart disease, hypertrophic cardiomyopathy, and in some patients undergoing 2-day protocols, with concomitantly normal epicardial coronary vessels. 8,9 TID in myocardial perfusion imaging has been proposed as a diagnostic and prognostic marker for the detection of severe and extensive CAD; however, there is variability in the literature on its utility. To our knowledge, this is the first meta-analysis to: (a) examine the range of different ratios above which TID was diagnosed, (b) systematically review the pooled diagnostic performance of TID, and (c) examine the value of TID as a prognostic tool in a systematic review. Our meta-analysis confirms the usefulness of TID in myocardial perfusion imaging as a high risk marker for stress induced myocardial ischemia and its ability to predict future cardiac events. Based on our findings, we propose a modified algorithm approach 25 in the presence of TID. Clearly, if both TID and high risk MPI are present, consideration should be given to invasive coronary angiography. 25 However, since specificity of TID is high, if TID and non-high risk MPI (SSS \ 4) are present in a patient with an intermediate clinical risk (CAD, diabetes, or chronic kidney disease), further noninvasive evaluation may be beneficial. 25 Finally, the methodology of quantitative meta-analysis to evaluate other putative high risk diagnostic markers nested within imaging modalities, 26 may be useful in future cardiac imaging research.
While meta-analyses are valuable tools for synthesizing the published literature, there are always limitations to such analyses. For example, the different patient populations, techniques, and diagnostic cutoffs all contribute to the clinical heterogeneity of the published literature. While higher TID ratio did affect sensitivity, meta-regression did not identify any other clinical variable that affected the effect estimates. However, due to the relatively small number of publications with interstudy differences in stress modality and tracer employed, the impacts of these factors may have been underrepresented. The majority of studies reported the definition of angiographic severity of coronary artery disease using percent stenosis, except the study by Rischpler. 27 We included this study because there were high rates of prior cardiac history (e.g., prior MI, documented CAD, prior coronary revascularization procedures) in the majority of patients and it was the only study that utilized Rb PET. 27 However, we did do a sensitivity analysis excluding this study and it revealed that there was no significant change in the diagnostic performance of TID after exclusion of this particular study, with a pooled sensitivity of 47% (31%-63%), specificity of 88% (82%-92%), and an AUC of 0.82 (0.78-0.85).

NEW KNOWLEDGE GAINED
The presence of TID has a high pooled area under the receiver operating characteristic curve for the detection of severe and extensive CAD. While sensitivity is low, specificity of TID is high for the detection of severe and extensive CAD.
The rate of cardiac death or MI is increased in those with TID and normal perfusion, primarily amongst those with DM, CAD, or ischemia. Rates of cardiac death or MI appear to be increased further in those with reduced LVEF.

CONCLUSION
In conclusion, in this meta-analysis, we found that transient ischemic dilation during myocardial perfusion imaging is a specific diagnostic marker of severe and extensive coronary artery disease. Transient ischemic dilation is an indicator of poor prognosis, and risks were significantly elevated among those with evidence  suggestive of coronary disease or reduced stress LV ejection fraction. The presence of TID significantly worsens prognosis even among diabetes patients with normal perfusion. Therefore, TID should be considered a high risk marker that may guide clinical management in patients with suspected or known coronary artery disease.