Reassessing the clinical significance of electrocardiographically unrecognized myocardial infarctions: Radionuclide infarct size and its impact on long‐term prognosis

Abstract

Background

Silent or unrecognized myocardial infarction (UMI) diagnosed by surveillance electrocardiography (ECG) carries similarly poor prognosis as recognized MI (RMI) for poorly understood reasons.

Methods

This study included 5430 consecutive patients who presented to the nuclear laboratory and underwent 2‐day stress and rest Tc‐^99m sestamibi and ECG studies between March 1991 and June 1999. UMI was diagnosed if ECG showed Q‐wave MI in the absence of a history of RMI. We measured infarct size (% defect size as compared with the entire left ventricular sestamibi uptake), ejection fraction (EF, %), and summed difference score (SDS, sestamibi uptake by myocardium in stress minus sestamibi uptake in rest images as a marker of ischemia). Survival was determined by follow‐up survey (median 6 years).

Results

We identified 346 UMIs, 628 RMIs, and 4456 subjects without MI (No MI). As compared with RMI, UMI patients had lesser abnormalities on nuclear scans (p < .0001 for all), including smaller infarct size (5.7% vs. 12.2%), higher EF (58% vs. 53%), and lesser ischemia (SDS; 3.9% vs. 2.7%). UMI prognosis was as poor as that of RMI (annual mortality rate 4.7% vs. 4.8% with No MI rate of 2.9%; p < .001 for all comparisons), and this persisted after multivariate analysis. Infarct size quantification successfully risk‐stratified ECG‐UMI patients, but UMI patients continued to predict mortality even if the infarct size was 0%.

Conclusions

Although UMI patients have lesser abnormalities on nuclear scans, ECG‐based UMI continues to independently predict mortality, indicating the continuing relevance of ECG in clinical practice.

(A) Comparison of 5‐year survival curves of UMI with RMI and No MI subjects with Kaplan–Meier survival analyses, showing that UMI and RMI yield similar mortality prediction. (B) Comparison of UMI subjects with different infarct sizes measured as percent of left ventricle. Despite dividing the group into four subgroups, each subgroup had 17 or more events out of the 300+ patients. Infarct size successfully prognosticated UMI patients, with those with infarct size >5% having significantly greater mortality risk than those with infarct size less than 5% (data not shown). ECG‐UMI patients with 0% defect size still had lower survival than No MI controls (p < .05). (C) The rising burden of ischemic heart disease from No MI to UMI to RMI with progressively larger infarct size, greater systolic dysfunction, and ischemia is seen in this comparison based on nuclear scan (p < .0001). EF, ejection fraction; LV, left ventricle/left ventricular; MI, myocardial infarction; RMI, recognized MI; UMI, unrecognized MI.

1. INTRODUCTION

Unrecognized myocardial infarction (UMI) has been defined as myocardial infarction (MI) diagnosed by a surveillance electrocardiogram (ECG) without a history of clinically recognized MI (RMI) (Ammar et al., 2004; Sheifer et al., ²⁰⁰¹). Although multiple prior studies have suggested a similarly poor prognosis for RMI and UMI (Grimm et al., 1987; Kannel et al., ^{1970, 1990}; Medalie & Goldbourt, ¹⁹⁷⁶; Rosenman et al., ¹⁹⁶⁷; Sheifer et al., ²⁰⁰⁰; Sigurdsson et al., ¹⁹⁹⁵; Yano & MacLean, ¹⁹⁸⁹), the reasons behind this poor prognosis are not well understood. Specifically, the impacts of infarct size, ischemia, and ventricular function on the long‐term prognosis of UMI have not been studied (Ammar et al., 2007).

Stress technetium‐^99m sestamibi imaging is commonly performed to evaluate ECG‐based UMI and provides a unique opportunity to study the natural history of UMI. We hypothesized that UMI and RMI would manifest comparable degrees of infarct size, reversible ischemia, and ejection fraction (EF). In order to test this hypothesis, we queried a large tertiary care center's nuclear cardiology database.

2. METHODS

2.1. Study setting

After obtaining authorization from the Mayo Clinic Institutional Review Board, we performed a retrospective analysis of the Mayo Clinic Nuclear Cardiology database to identify all patients who underwent 2‐day stress single‐photon emission computed tomography (SPECT) sestamibi between March 1991 and June 1999. Medical record abstraction was carried out by trained nurse abstractors.

2.2. Technetium‐^99m sestamibi imaging analyses for infarct size quantification, EF, and ischemia

We measured infarct size, ischemia, and EF as described in multiple prior publications from the Mayo Clinic Nuclear Laboratory (Gibbons et al., 2004; Hurrell et al., ²⁰⁰⁰; Ramakrishna et al., ²⁰⁰⁶).

Infarct size quantification was performed after the patients received an intravenous injection of technetium‐^99m sestamibi. Images were acquired using a rotating gamma camera with an all‐purpose collimator (Elsinct; or Adac Laboratories). Standard filter back projection algorithms and a ramp‐Hanning filter were used for processing and reconstruction. The entire left ventricle (LV) was divided into five representative short‐axis slices, from base to apex, and then circumferential count profiles were generated. Infarct size was quantified using a threshold value of 60% of peak counts. The defect size was expressed as a percentage of the LV.

LVEF was quantitated by the first pass technique (Ramakrishna et al., 2006) with the formula of EF = (end‐diastolic volume – end‐systolic volume)/end‐diastolic volume.

For ischemia measurement, all rest (day 1) and stress (day 2) gated studies were performed with a high dose of approximately 30 mCi Tc‐^99m sestamibi. Rest and poststress SPECT acquisition began 30–60 min after injection. The stress study was performed with the injection of 30 mCi Tc‐^99m sestamibi at peak stress. The stress modality involved exercise treadmill testing with a Bruce protocol or pharmacologic stress with adenosine (140 μg per kilogram per minute for 6 min) or dobutamine (3‐min interval‐graded dose starting at 10 μg per kilogram per minute up to 40 μg per kilogram per minute to achieve the target heart rate). Images were processed by using filtered back projection and reconstruction algorithms.

The rest and poststress static short‐axis images were analyzed for presence, location, and extent of defect reversibility by consensus of two experienced observers. Short‐axis images were divided into 16 segments, and a 5‐point scoring system was used to assess perfusion in each segment (4, normal perfusion; 3, mild hypoperfusion; 2, moderate hypoperfusion; 1, severe hypoperfusion; 0, no perfusion). Summed rest score (SRS) and summed stress score (SSS) were calculated as the summation of perfusion grades in the short‐axis segments for the rest and stress images, respectively. A normal image has a score of 64. The difference in myocardial perfusion imaging scores (SRS – SSS) is the summed difference score (SDS).

2.3. Diagnostic criteria for RMI and UMI

The standard resting 12‐lead electrocardiogram performed just prior to the stress test was analyzed in the Mayo Clinic Core ECG Laboratory. RMI was diagnosed if there was a clinical history of documented MI by asking the patients prior to the stress test whether they had ever had a heart attack. This is a standard method used in other studies, including the Framingham Heart Study. In addition, the charts were abstracted by nurse chart abstractors for prior history of MI. UMI was diagnosed if the patient denied having a prior history of heart attack but the ECG fulfilled criteria for ECG‐MI.

ECG‐MI was diagnosed by visual overread of the automated read provided by the GE Marquette ECG system by the nuclear cardiologists assigned to the laboratory. In general, it was based on the presence of significant Q waves (≥40 ms in duration and ≥1/3 of the total height of the QRS complex) in at least two contiguous leads, while ignoring Q waves that are a normal variant, such as in lead III. The status No MI was assigned to those with no clinical history of MI and no MI seen on the ECG. In order to increase the diagnostic accuracy of ECG‐UMI and to create a clinically “pure” group, subjects with conditions that may produce false‐positive Q waves and subjects who already carried a diagnosis of coronary artery disease were excluded since our purpose was to look at the prognosis of UMI subjects without any other coronary diagnosis. Therefore, the exclusions included hypertrophic cardiomyopathy, pacemaker, left bundle branch block, Wolff‐Parkinson‐White syndrome, LV hypertrophy (LVH) (if anterior wall MI diagnosis was accompanied by LVH as LVH may cause Q complexes in V₁ and V₂), complete heart block, indeterminate prior MI status due to missing data, prior coronary intervention, or prior bypass surgery.

2.4. Statistical analysis

Categorical data are summarized as a percent of the group total with corresponding 95% confidence intervals based on the normal approximation and comparison between groups based on the chi‐square test for association. Continuous variables are summarized as mean ± standard deviation, and comparisons between groups were based on analysis of variance (ANOVA) models. Post‐ANOVA comparisons of continuous variables were based on the t‐test, but no adjustments for multiple comparisons were made. In case of variables with non‐normal distribution, like infarct size, both mean and median are reported, as well as statistical tests used that do not assume normal distribution (e.g., Kruskal–Wallis test).

Time to death was summarized using the Kaplan–Meier estimate. Comparisons between groups were based on the log‐rank test for univariate analyses. We ran Cox proportional hazards regression unadjusted models for mortality related to UMI and then adjusted them for confounding, as well as intervening, variables. The confounding variables included standard coronary risk factors including age, sex, diabetes, hypertension, smoking, and hyperlipidemia, as an increased prevalence of these in UMI patients will increase the hazard ratio (HR) of UMI patients, although it would, in actuality, be confounded by these standard mortality predictors. The intervening variables included infarct size, ischemia, and LVEF, as these variables are the primary pathway leading to death in patients with MI, recognized or unrecognized. Prior studies have shown their importance in predicting outcome, with radionuclide EF being a powerful predictor of cardiac death and SPECT perfusion being a powerful predictor of nonfatal MI (Taqueti & Di Carli, 2015).

Additional survival analyses were carried out with the Cox proportional hazard analysis model, in which the increased mortality seen in UMI subjects was adjusted for potential intervening variables in order to measure the relative impact of each intervening variable. A decrease in HR (≥20%) for all‐cause mortality of UMI subjects would indicate a significant intervening variable. The cutoff of 20% for a meaningful change in HR has been an arbitrary cutoff quoted in textbooks of statistics, including “Active Epidemiology” by David Klinebaum, which is accessible electronically (http://web1.sph.emory.edu/activepi/).

Two‐sided p‐values <.05 were considered significant. Analyses were performed on SAS version 8 (SAS Institute).

3. RESULTS

3.1. Study participants

Among the 5430 participants, the mean age was 64 ± 11.2 years, 58% were men, mean body mass index was 31 kg/m², 12% were current smokers, 41% were past smokers, 22% had diabetes, 58% had hypertension, and 55% had hypercholesterolemia (Table 1). The prevalence of coronary risk factors, including age, male sex, diabetes, current smoking, and hypercholesterolemia, was higher in UMI than No MI subjects, but lower than in RMI subjects (p < .0001). The prevalence of cardiac symptoms was also higher in UMI subjects than RMI subjects, with RMI subjects more likely to have typical angina and UMI subjects more likely to have atypical angina, noncardiac chest pains, and dyspnea.

TABLE 1.

Baseline characteristics.

	UMI (n = 346)	RMI (n = 628)	No MI (n = 4456)	p Value
Demographics
Age (years) a , b	64.8 ± 10.69	66.0 ± 11.37	63.2 ± 11.14	<.0001 c
Male sex a	227 (65.6)	436 (69.4)	2464 (55.3)	<.0001
BMI (kg/m2) b	31.6 ± 6.00	30.9 ± 6.73	31.2 ± 6.76	<.05
Coronary risk factors
Diabetes a	93 (26.9)	184 (29.3)	932 (20.9)	<.0001
Hypertension a	222 (64.2)	390 (62.1)	2560 (57.5)	.0072 d
Current smoker	45 (13)	92 (14.6)	496 (11.1)	.0264 d
Past smoker b	157 (45.4)	299 (47.6)	1756 (39.4)	.0001 d
Pack years a , b	40.4 ± 31.01	46.6 ± 43.83	40.5 ± 39.43	.0002 c
Clinical characteristics
Symptomatic a , b	217 (62.7)	334 (53.2)	3155 (70.8)	<.0001
Chest pain a	160 (46.2)	249 (39.6)	2448 (54.9)	<.0001
Typical angina	61 (17.6)	134 (21.3)	780 (17.5)	.0634 d
Atypical angina a , b	92 (26.6)	111 (17.7)	1511 (33.9)	<.0001
Noncardiac chest pain	9 (2.6)	7 (1.1)	184 (4.1)	.0005
Dyspnea	57 (16.5)	85 (13.5)	707 (15.9)	.2916 d
Family history of CAD	99 (28.6)	213 (34)	1544 (34.7)	.07
Earliest age of onset of CAD in family (years)	54.2 ± 11.74	54.0 ± 10.35	55.6 ± 10.86	.054
Hypercholesterolemia a , b	200 (57.8)	373 (59.4)	2385 (53.5)	.01
Total cholesterol (mg/dL)	230.8 ± 45.32	218.3 ± 63.22	231.7 ± 54.71	<.0001

Note: Data presented as mean ± standard deviation or n (%).

Abbreviations: BMI, body mass index; CAD, coronary artery disease.

^a UMI vs. No MI and RMI vs. No MI p < .05.

^b UMI vs. RMI p < .05.

^c Kruskal–Wallis.

^d Chi‐square.

3.2. Nuclear and ECG variables

UMI subjects had a significantly lower prevalence of ECG ischemia, higher exercise times, and higher Duke Treadmill Scores than RMI subjects (Table 2). The prevalence of exercise testing increased from 48% in RMI subjects to 50% in UMI subjects to 55% in No MI subjects. The rest had pharmacologic testing, predominantly adenosine stress tests (>98%). UMI subjects were significantly abnormal in terms of myocardial ischemia, infarct size, EF, systolic dysfunction, and ventricular enlargement, indicating that this ECG abnormality is identifying a high‐risk coronary group, but generally the abnormalities were less than those seen in RMI subjects. Although the mean LVEF is normal in all groups (Table 3), there is a trend of increasing mean LVEF from the RMI to UMI to No MI groups, and clinically meaningful categorization showed rising odds of mild (EF ≤50%; odds ratio of No MI, UMI, and RMI were 1, 2.4, and 4, respectively) or moderate (EF ≤40%; odds ratio of No MI, UMI, and RMI were 1, 4.5, and 8, respectively) systolic dysfunction.

TABLE 2.

Nuclear and stress test characteristics of RMI, UMI, and controls (No MI) ^a

	UMI (n = 346)	RMI (n = 628)	No MI (n = 4456)	p Value
Exercise ECG parameters
Exercise time (minutes) a , b	7.4 ± 2.28	7.0 ± 2.65	7.9 ± 2.59	<.0001 c
ECG ischemia present b	53 (15.3)	102 (16.2)	870 (19.5)	.0313 d
Nuclear imaging parameters
Cardiac enlargement a , b	61 (17.6)	205 (32.6)	363 (8.1)	<.0001
Final impression a , b				<.0001
Abnormal	237 (68.5)	554 (88.2)	2351 (52.8)
Ischemia, 1 coronary dist	80 (23.1)	96 (15.3)	1260 (28.3)
Ischemia, ≥2 coronary dist	44 (12.7)	82 (13.1)	553 (12.4)
Infarction, 1 coronary dist	21 (6.1)	126 (20.1)	149 (3.3)
Infarction, ≥2 coronary dist	4 (1.2)	33 (5.3)	17 (0.4)
Ischemia and infarction, same coronary dist	28 (8.1)	89 (14.2)	120 (2.7)
Ischemia and infarction, different coronary dist	51 (14.7)	132 (21)	168 (3.8)
Sum stress score a , b	48.7 ± 8.94	43.9 ± 9.29	52.3 ± 5.76	<.0001 c
Sum difference score a , b	3.9 ± 4.93	4.7 ± 5.29	2.7 ± 4.36	<.0001 c
Median	2.0	3.0	0.0
First pass EF (%) a , b	58.2 ± 11.65	52.8 ± 12.08	61.2 ± 9.05	<.0001 c
EF ≤50% a , b	78 (24.5)	226 (41.5)	438 (10.7)	<.0001
EF ≤40% a , b	30 (9.4)	93 (17.1)	90 (2.2)	<.0001
Infarct size (% defect size) a , b	5.7 ± 12.35	12.2 ± 15.39	1.6 ± 4.81	<.0001 c
Median	0.0	6.0	0.0
Q1, Q3	0.0, 6.0	0.0, 19.0	0.0, 0.0
Range	0.0–64.0	0.0–76.0	0.0–58.0

Note: Data presented as mean ± standard deviation or n (%), unless otherwise indicated.

Abbreviations: Dist, distribution(s); ECG, electrocardiogram; EF, ejection fraction; MI, myocardial infarction; RMI, recognized MI; UMI, unrecognized MI.

^a UMI vs. No MI and RMI vs. No MI p < .05.

^b UMI vs. RMI p < .05.

^c Kruskal–Wallis.

^d Chi‐square.

TABLE 3.

Ischemia and infarction patterns in the three groups based on SPECT imaging.

	UMI (n = 346)	RMI (n = 628)	No MI (n = 4456)	Total (n = 5430)	p Value
Summed stress score					<.0001 a
n	346	628	4456	5430
Mean ± standard deviation	48.7 ± 8.94	43.9 ± 9.29	52.3 ± 5.76	51.1 ± 7.05
Median	52.0	45.0	55.0	54.0
Q1, Q3	45.0, 56.0	37.0, 51.5	51.0, 56.0	49.0, 56.0
Range	15.0–56.0	13.0–56.0	18.0–56.0	13.0–56.0
Summed rest score					<.0001 a
n	346	628	4456	5430
Mean ± standard deviation	52.6 ± 6.38	48.6 ± 8.11	55.0 ± 2.63	54.1 ± 4.49
Median	56.0	51.0	56.0	56.0
Q1, Q3	52.0, 56.0	44.0, 55.0	56.0, 56.0	54.0, 56.0
Range	19.0–56.0	15.0–56.0	27.0–56.0	15.0–56.0
Summed difference score					<.0001 a
n	346	628	4456	5430
Mean ± standard deviation	3.9 ± 4.93	4.7 ± 5.29	2.7 ± 4.36	3.0 ± 4.56
Median	2.0	3.0	0.0	1.0

^a Kruskal–Wallis.

Abbreviations: MI, myocardial infarction; RMI, recognized MI, SPECT, single‐photon emission computed tomography; UMI, unrecognized MI.

3.3. Mortality

UMI predicted mortality in the unadjusted analysis (HR 1.74) as well as in adjusted analyses (Figure 1a) with statistical significance (Table 4).

FIGURE 1.

(a) Comparison of 5‐year survival curves of UMI with RMI and No MI subjects with Kaplan–Meier survival analyses, showing that UMI and RMI yield similar mortality prediction. (b) Comparison of UMI subjects with different infarct sizes measured as percent of left ventricle. Despite dividing the group into four subgroups, each subgroup had 17 or more events out of the 300+ patients. Infarct size successfully prognosticated UMI patients, with those with infarct size >5% having significantly greater mortality risk than those with infarct size less than 5% (data not shown). ECG‐UMI patients with 0% defect size still had lower survival than No MI controls (p < .05). (c) The rising burden of ischemic heart disease from No MI to UMI to RMI with progressively larger infarct size, greater systolic dysfunction, and ischemia is seen in this comparison based on nuclear scan (p < .0001). EF, ejection fraction; LV, left ventricle/left ventricular; MI, myocardial infarction; RMI, recognized MI; UMI, unrecognized MI.

TABLE 4.

Cox proportional hazard analysis of all‐cause mortality as a function of ECG‐UMI with and without nuclear or stress test abnormalities compared against No MI controls.

Model	Outcome (all‐cause mortality)	No MI	UMI	RMI
			HR, 95% CI, and p Value	HR, 95% CI, and p Value
Unadjusted model	ECG‐UMI (n = 346)	1.0	1.74 (1.57, 1.94) <.0001	1.64 (1.40, 1.92) <.0001
Adjustment for confounding and intervening variables (incremental models) as done in common clinical practice
Adjusted Model 1	ECG‐UMI, RMI, standard coronary risk factors a and infarct size b	1.0	1.40 (1.13, 1.73) .002	1.13 (0.93, 1.36) .21
Adjusted Model 2	ECG‐UMI, RMI, standard coronary risk factors and reversible ischemia (sum difference score) b	1.0	1.48 (1.21, 1.82) <.001	1.26 (1.07, 1.48) .004
Adjusted Model 3	ECG‐UMI, RMI, standard coronary risk factors and ejection fraction	1.0	1.34 (1.07, 1.69) .01	1.13 (0.93, 1.37) .20

Abbreviations: CI, confidence interval; ECG, electrocardiogram; HR, hazard ratio; MI, myocardial infarction; RMI, recognized MI, UMI, unrecognized MI.

^a Standard coronary risk factors included age sex, diabetes, hypertension, smoking and hypercholesterolemia.

^b RMI not significantly associated with mortality in these models.

Further analyses were carried out to shed light on the mechanisms of increased mortality associated with UMI (Table 4). As described in the Section 2, we sought a meaningful decrease in HR in multivariate analysis, which was preset at a HR decrease of >20%, i.e., if HR decreased below 1.392 from 1.74. The meaningful decline in mortality odds does not happen with coronary risk factors alone, but as more information is added from SPECT to the prognostic model, the odds ratios progressively decline to the point of meaningful decline only in the third adjusted model (p = .04). The first unadjusted model showed an HR of ECG‐UMI of 1.74, which declined to 1.40 when adjusted for standard coronary risk factors (age, sex, diabetes, hypertension, hyperlipidemia, and smoking) in addition to infarct size (Model 1). This indicates the contribution of infarct size to the mortality prediction. In the second adjusted model (Model 2), the HR of UMI dropped from 1.74 to 1.48 when adjusted for coronary risk factors and ischemia. The final adjusted model (Model 3) showed a drop in HR from 1.74 to 1.34 when the unadjusted model was adjusted for EF in addition to coronary risk factors, indicating that, of all the nuclear variables (infarct size, ischemia, and EF), EF was the biggest mediator of increased mortality, followed by infarct size and ischemia. In addition, this provides the insight that only EF is associated with a meaningful decline in mortality rate (>20% change in HR).

Figure 1a shows that all‐cause mortality is similar after UMI or RMI, with a 5‐year survival rate of 82% (UMI or RMI) versus 88% in No MI controls.

Infarct size quantification by SPECT successfully risk stratified UMI subjects into different prognostic groups, as has been shown for RMI in the past (Figure 1b), with significantly increasing mortality risk not only with medium‐sized (>10%) defects but even with small‐sized (1%–10%) defects. Interestingly, the prognosis of ECG‐UMI was significantly worse than that of the No MI group (p = .008) even if the ECG‐UMI had a 0% defect size (Figure 1b).

4. DISCUSSION

4.1. Principal findings

This study is a unique assessment of the stress technetium‐^99m sestamibi imaging characteristics of ECG‐identified UMI. Our analysis characterizes the cardiac structural and functional abnormalities of UMI patients identified by traditional ECG criteria and compares it with No MI and RMI subjects. As compared with RMI subjects, UMI subjects showed smaller differences than No MI controls in terms of infarct size, reversible ischemia, and EF (Figure 1c), yet mortality after UMI continued to be equally increased as compared with RMI, indicating that other natural history modifying forces, such as medications and disease recognition, may be playing a role. Infarct size quantification by SPECT successfully risk‐stratified these patients.

4.2. UMI subjects have smaller infarct size than RMI subjects

Patients with ECG‐UMI have been shown to have a much lower prevalence of regional wall motion abnormality (13% vs. 52%; p < .0001) and systolic dysfunction (18% vs. 28%; p < .0001) than patients with RMI in a population‐based study (Ammar et al., 2004). It has been postulated that UMI may be associated with a smaller infarct size, resulting not only in the observation of a lower prevalence of regional and global systolic dysfunction, but also lesser pain, leading to the reason why they go unrecognized. This study provides the first data regarding smaller infarct size in UMI subjects in support of the smaller infarct hypothesis, which is in contrast to the commonly prevalent hypothesis of a lack of pain in UMI patients due to neuropathy, as seen in diabetic neuropathy, creating a defective anginal warning system in UMI patients as the reason behind these infarctions going unrecognized (Sheifer et al., 2001). The smaller infarct size helps us understand the occurrence of UMI even in non‐diabetics: as it is a smaller infarct size, it is likely to be associated with lesser pain, which the patient is more likely to ignore and therefore the MI goes unrecognized.

4.3. Infarct size measurement

Infarct size measured by this technique has been validated in a phantom model (O'Connor et al., ¹⁹⁹⁵), in animal models of permanent occlusion (Verani et al., 1988) and reperfusion (Sinusas et al., 1990), and in explanted human hearts at the time of cardiac transplantation (Medrano et al., 1996). The limit for detection of infarction by this technique has been shown to be 3% of the LV (Gibbons et al., 2004; O'Connor et al., ¹⁹⁹⁵). The issue of false‐positive defects (infarcts) secondary to attenuation of counts from overlying soft tissue (diaphragm and breast), which theoretically might cause defects below the 60% threshold, has been addressed in our laboratory. We evaluated a series of 100 mostly obese patients selected on the basis of normal resting ECG and no clinical history of MI (Gibbons R, Miller T. unpublished data). Only eight patients had trivial defects measuring between 1% and 3% of the LV. This observation indicated that, in such patients, tissue attenuation played a minor role, thereby providing support for the general clinical practice of calling an MI on a sestamibi scan unequivocally if there is >3%–5% defect. This is based on the assumption that the ECG is normal; therefore, it cannot be applied to UMI subjects who do not have a history of clinical infarction but have an abnormal ECG. In light of this fact, instead of categorizing UMI subjects as MI present or absent, we compared the average defect size in UMI and RMI subjects.

The accuracy of measurement of defect or infarct size for ^99mTc has been studied in a cardiac phantom model containing no defects and defects of 5%–70% of total myocardial mass, and the best correlations were seen at threshold values between 55% and 60% (r > .99), with the lowest average absolute error in estimating defect size (<2.1%) (O'Connor et al., ¹⁹⁹⁵). Furthermore, scatter correction has reduced the average absolute error to 0.8% (O'Connor et al., ¹⁹⁹⁵). Such observations support the clinical practice of greater confidence in larger infarct sizes but not in smaller infarct sizes. In the presence of ECG‐UMI, any defect size increases the mortality risk, which is an important observation that challenges the common clinical practice of judging a defect size of <5% insignificant and potentially a false positive due to diaphragm, breast, or motion. These data argue that, in the presence of UMI, any defect size should be considered further evidence of a prior MI. ECG‐UMI indicates an “electrical scar,” which, if substantiated by even a tiny defect on a nuclear scan, merits more serious consideration than is customary practice in those with a normal ECG.

4.4. Infarct size and mortality

The effect of infarct size on post‐RMI mortality has been studied extensively. Survival was excellent, with a 2‐year mortality of 0%, for patients who had an infarct size <12%. For patients with an infarct size ≥12%, 2‐year mortality was 13% (Miller et al., 1999). This established the importance of infarct size in predicting mortality in RMI patients. In the case of UMI subjects, mortality rises with increasing infarct size (Figure 1b). ECG‐UMI indicates an electrical scar, which may not manifest as a large morphological scar on nuclear scans or echocardiograms but nevertheless is associated with increased mortality. This suggests that mechanisms other than infarct size may be operative in the poorer prognosis. Of these possible mechanisms, the one we did not evaluate was secondary coronary preventive measures, like aspirin, beta blockers, and statins, which may have been lacking in the UMI subjects as they had not been diagnosed with coronary artery disease.

4.5. UMI is associated with increased mortality independent of infarct size, LV function, and ischemia

In Table 4, we have presented the adjusted models as they are congruent with clinical reasoning in routine clinical practice. The first adjusted model addresses the scenario of a clinician who discovers an ECG‐based UMI, but, if the patient does not have coronary risk factors and a significant infarct is not detectable on sestamibi scan, is likely to call the ECG infarct a false positive. Model 1 shows that ECG‐UMI continues to predict mortality despite adjustment for infarct size; therefore, the current clinical practice is erroneous.

The second adjusted model addresses the clinical scenario in which the clinician detects ECG‐UMI but, in the absence of ischemia, decides to ignore the clinical significance of the ECG‐UMI. This model provides evidence against this current clinical paradigm as the ECG‐UMI continues to predict mortality even after adjustment for ischemia.

The third adjusted model evaluates the clinical belief that an MI translates into increased mortality by causing LV dysfunction. It shows that, indeed, the largest drop in UMI mortality occurs on adjustment for EF, and therefore EF is the most important intervening variable.

These observations in UMI were in contrast with those in RMI (HR for mortality 1.64:1; RMI:No MI; results not shown). RMI patients were not associated with statistically increased odds of mortality when adjusted for Models 1, 2, and 3 of Table 4 (coronary risk factors plus infarct size/ischemia/EF), indicating that RMI patients prognostically act differently than UMI patients. Whereas the increased risk of mortality in RMI patients is explained by infarct size, ischemia, or systolic dysfunction, UMI patients continue to have increased risk of mortality even after adjustment for the nuclear variables in Table 4. Our data do not provide insights behind this difference between UMI and RMI. One can only speculate that if RMI patients get the usual coronary preventive measures (aspirin, beta blockers, statins, angiotensin‐converting enzyme inhibitors, etc.) and UMI patients do not get the diagnosis and therefore do not get these secondary preventive measures, the result is that UMI continues to predict increased mortality.

UMI has been shown to carry a four times higher risk of mortality as compared with No MI controls, even after correcting for echocardiographic abnormalities, like regional wall motion abnormalities, that are generally considered the gold standard for diagnosing prior MI (Ammar et al., 2004). This has raised uncertainties regarding the reasons behind the increased mortality associated with UMI, as well as concerns regarding the common clinical practice of calling an ECG‐based UMI a false positive if the confirmatory echocardiogram does not show wall motion abnormalities. This study extends these findings to UMI patients undergoing nuclear scans, with the important addition of superior infarct size quantification with nuclear imaging, leading to more precise prognostication.

In the current era of multimodality imaging for resolution of clinical issues, this is the first study in the literature to evaluate the clinical entity of ECG‐UMI in the light of SPECT imaging. It fills in the void of understanding regarding how to deal with ECG‐UMI patients, who in clinical practice generally undergo echocardiography and stress SPECT imaging, mainly owing to less availability of and insurance coverage for magnetic resonance imaging (MRI). There is one population‐based study that evaluated ECG‐UMI in the light of echocardiography (Ammar et al., 2005). The prevalence of regional wall motion abnormalities increased from 2% to 13% to 42% from No MI, to UMI, to RMI patients, and so did EF, systolic dysfunction, diastolic dysfunction, and LV enlargement, which suggested a smaller infarct size in UMI patients, a finding finally confirmed by the current study. The follow‐up study (Ammar et al., 2007) elucidated the cause of mortality in UMI patients to be mediated via not only regional wall motion abnormalities but multiple other parameters of cardiac remodeling, including systolic or diastolic dysfunction and LV or left atrial enlargement, resulting in cardiac symptoms that were out of proportion to any individual echocardiographic abnormality like regional wall motion abnormality. There are multiple published MRI studies on UMI (Kwong et al., 2006; Nordenskjöld et al., ²⁰¹⁶) that show as much as a seven fold increased risk of mortality (Kwong et al., 2006) in MRI‐UMI patients. These studies have small sample sizes (n < 300) (Kramer, 2010; Kwong et al., ²⁰⁰⁶; Nordenskjöld et al., ²⁰¹⁶) and use an MRI‐first approach, whereas contemporary clinical practice moves forward with an ECG‐first approach. One recent study (Schelbert et al., 2012) did enroll a large population‐based sample (n = 936) of older community‐dwelling adults (age 67–93) and showed that ECG‐UMI was not associated with worsened prognosis but MRI‐UMI was. This finding is in contradiction with the findings of prior UMI studies and the current study. Although MRI claims to be the current gold standard for infarct size quantification, mainly based on its superior spatial resolution as compared with SPECT (0.5 mm vs. 8 mm), it really does not image the same structures as SPECT. MRI‐based UMI diagnosis is based on scar imaging, determined by gadolinium uptake in between myocardial cells, suggestive of increased interstitial space. In contrast, SPECT imaging, the pre‐MRI gold standard for infarct size quantification, detects a myocardial infarct by lack of uptake by myocardial cells, which indicates dead cells, specifically dead mitochondria, since sestamibi crosses into the mitochondria. SPECT imaging, arguably, visualizes a myocardial infarct more directly than MRI, and, therefore, its insights are still relevant from a mechanistic perspective, in addition to a clinical practice perspective, as it continues to be the main workhorse.

Last but not least, Qureshi et al. (2018) recently showed in the ARIC database (n = 9243), published in January 2018, that ECG‐UMI is an intermediate entity between No MI and RMI in terms of its association with incident heart failure with an HR increasing from 1 in No MI to 1.9 in UMI to 3.5 in RMI. The current data provide confirmatory evidence to this recent study while shining further light on the infarct size characteristics of UMI patients, as it is the smaller infarct size that leads to the lesser incidence of heart failure and, likely, the lesser amount of pain, leading to the lack of recognition of MI.

4.6. Potential limitations

Referral bias is unavoidable in such a study, which excludes subjects presenting with UMI to their primary provider who may choose to ignore the ECG abnormality, perform a stress echocardiogram, or proceed straight to providing risk factor modification, depending on the clinical scenario and varying from physician to physician. The majority of our study population (>90%) were White, which limits generalizability to other races. The potential for misclassification by ECG or sestamibi scan cannot be ignored, especially as normalization of both ECG and scintigraphic MI abnormalities may occur with time. The interobserver variability of the ECG reads has not been studied, which is another limitation.

5. CONCLUSIONS

ECG‐UMIs are associated with a smaller infarct size, lesser reversible ischemia, and higher EF. UMI status predicts mortality independent of standard coronary risk factors, but its mortality risk prediction significantly and meaningfully reduces as information from SPECT imaging is accounted for, including infarct size quantification, reversible ischemia, and EF measurement. The unequivocal presence of an electrical infarct on ECG seems to be associated with increased risk for mortality, if associated with even a tiny nuclear defect, indicating the vital role of SPECT imaging and infarct size quantification in these patients.

AUTHOR CONTRIBUTIONS

KAA and RJR designed the work, performed the data extraction and statistical analyses, and wrote the article. Both authors read and approved the final manuscript.

FUNDING INFORMATION

This study was supported by grants from the Mayo Clinic Foundation.

CONFLICT OF INTEREST STATEMENT

The authors declare that they have no competing interests.

ETHICS STATEMENT

This study was approved by the Mayo Clinic Institutional Review Board. The requirement for informed consent was waived due to the retrospective nature of the study.

Ammar, K. A. , & Rodeheffer, R. J. (2023). Reassessing the clinical significance of electrocardiographically unrecognized myocardial infarctions: Radionuclide infarct size and its impact on long‐term prognosis. Annals of Noninvasive Electrocardiology, 28, e13088. 10.1111/anec.13088

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.