ICELAND MI is an epidemiologic cohort study of diabetic and nondiabetic individuals. Participants were enrolled from January 2004 to January 2007, recruited from the AGES-Reykjavik Study (n=5,764),¹³ a randomly selected population-based cohort of men and women born between 1907 and 1935 who have been followed in Iceland since 1967 by the Icelandic Heart Association. AGES–Reykjavik was approved by the National Bioethics Committee in Iceland that acts as the institutional review board for the Icelandic Heart Association (approval number VSN-00-063) and by the National Institute on Aging Intramural Institutional Review Board. Participants were eligible to participate in ICELAND MI if they provided written informed consent and were ineligible if they could not safely receive CMR scans (e.g., implanted devices) or gadolinium contrast (e.g., severe kidney disease). Participants were recruited from AGES-Reykjavik in two phases. The first phase involved random recruitment, and a second phase recruited all eligible and willing participants with diabetes.

Participants were characterized during three clinic visits.¹³ CMR studies occurred during a separate exam that included ECG. Participant surveillance has been ongoing since 1967 through the Icelandic Heart Association¹³ and provided ascertainment of recognized MI.

Participants were defined as having a recognized MI when a history of MI was supported by hospital records or surveillance records.¹³ Participants were defined as having an UMI by ECG when there was evidence of MI by ECG criteria (Minnesota codes 1.1.1–1.2.8). ¹ UMI by CMR meant there was no prior MI by hospital records or by surveillance records, and LGE involved the subendocardium in a coronary distribution. Other “atypical” patterns of LGE were specifically not designated as MI, a strategy that yields sensitivities and specificities >90% for MI detection.^14–16 CMR studies were interpreted by cardiologists blinded to clinical information.

Participants were further characterized with demographics, risk factors related to atherosclerosis, other comorbidity, biochemical measurements from blood, coronary calcium (Agatston scores), and ECG. Participants were classified as having diabetes according to standard criteria (fasting glucose ≥7 mmol/L)¹⁷ or if they were already receiving treatment for diabetes. All cause mortality was identified by review of hospital records as well as a national mortality index with authentication of all death certificates¹³ through September 1, 2011.

CMR scans were performed on a 1.5T GE scanner (Milwaukee, WI) using a four-element cardiac phased array coil. Typical cine SSFP scan parameters resulted in pixel dimensions of 1.8×2.1 mm, slice thickness 8 mm with 3 mm gap, and 30 images per cycle. Standard long axis and short axis views were obtained to evaluate global and regional function. The presence of MI was evaluated with a prospective, ECG gated, segmented, phase sensitive gradient echo inversion recovery sequence approximately 6–25 minutes after 0.1 mmol/kg intravenous gadolinium (Magnevist, Berlex).¹⁸ LGE was designated MI by consensus of cardiologists experienced in CMR.

Results are presented with 95% confidence intervals (CI). Categorical variables were compared with the Chi-square or Fisher’s test. Continuous variables were compared with the Wilcoxon rank-sum test. McNemar’s statistic tested whether CMR was more likely to detect UMI than ECG. The log rank test compared survival curve strata. Binary response variables were further analyzed by Cox regression survival analysis, and continuous variables were analyzed by linear regression. Multivariable Cox models adjusted for variation in key baseline characteristics included in prior epidemiologic studies using ECG: age, gender, diabetes, recognized MI, and finally UMI by CMR or UMI by ECG. Proportional hazards assumptions were verified by Schoenfeld residuals and time interaction terms. Absolute risk increases were calculated by measuring the survival rate difference before and after exponentiating the 7 year Kaplan-Meier survival rate in the reference group to the power of the adjusted hazard ratio (HR) in the comparison group. The integrated discrimination index (IDI) and net reclassification index (NRI) evaluated the added predictive ability of survival models with the introduction of the UMI by CMR variable.^19,20 Follow-up was enhanced by hospital record information, a national mortality index with authentication of all death certificates, a Minimum Data Set for Nursing Home patients, and Minimum Data Set for Home-Care patients.¹³ Coronary artery calcium (CAC) was analyzed on the natural logarithm scale, ln(CAC+1). Two sided p values <0.05 were considered significant. SAS software (version 9.2) analyzed the data.

For phase one, 839 individuals were invited and 702 enrolled. In phase two, 421 participants with diabetes were invited and 290 people enrolled (1005 total). Thirty-five participants declined CMR. Of those who underwent CMR (n=970), 34 participants had nondiagnostic CMR scans due to: arrhythmia or inability to breath hold (n=14); claustrophobia (n=7); inability to gate cardiac images (n=3), technical issues with reconstruction and data transfer (n=9); or artifact from spinal implants (n=1). These participants were excluded leaving a final cohort of 936 participants. Survivors were followed for a median of 6.6 years (range 4.6– 7.7 yrs).

The median age was 76 years (range 68 to 94 years), and 52% (CI 49–55%) were women (484 of 936). Baseline characteristics are summarized in Table 1. ICELAND MI participants randomly selected in phase 1 had characteristics almost identical to the AGES-Reykjavik participants (Supplementary eTable 1).

While 91 of 936 participants (9.7% CI, 8–12%) had recognized MI, the prevalence of UMI by CMR was even higher 157 of 936 (17%, CI, 14–19%; p<0.001) as shown in Table 2. Those with diabetes had a higher prevalence of UMI by CMR than those without diabetes (n=72; 21%, CI 17–26% vs. n=85; 14%, CI 11–17%, p<0.001). Examples are shown in Figure 1.

CMR detected 157 UMI which was more than the 46 UMI by ECG (prevalence by CMR 17%, CI 14–19% vs. ECG 5%, CI 4–6%, respectively, p<0.001). There were 27 participants (3%, CI 2–4%) with UMI by ECG that exhibited no MI on CMR, and there were 138 (15%, CI 12– 17%) individuals who had UMI by CMR yet did not meet criteria for UMI by ECG (p<0.001). In the randomly sampled cohort (n=670), 61 (9%, CI 7–11%) had recognized MI and 97 (14%, CI 12–17%) had UMI by CMR whereas only 35 (5%, CI 4–7%) had UMI by ECG, significantly less than UMI by CMR (p<0.001).

Coronary artery disease risk factors were more prevalent in participants with UMI compared with those with no MI. Compared to those without MI, participants with UMI were more frequently male, were slightly older, and had more hypertension and diabetes (Table 3). Similarly, those with UMI had more atherosclerosis with significantly higher coronary calcium scores than those without MI (Table 3). Overall, coronary calcium showed a significant graded relationship to the presence of MI, where participants with UMI had coronary calcium intermediate between those without MI and those with clinically recognized MI (Table 3).

There was also a graded relationship between the likelihood of revascularization and MI status (Table 3). For 26 of 72 diabetic (36%, CI 25–47%) and 18 of 85 nondiabetic (21%, CI 12– 30%) participants with UMI had prior coronary revascularization. Excluding those with prior coronary revascularization (n=139), diabetic and nondiabetic participants still had high rates of UMI (46/273 or 17%, CI 12–21% versus 67/524 or 13%, CI 10–16%, respectively). Thus, UMI was associated with atherosclerosis risk factors, coronary calcium, and treatment for atherosclerosis. Other characteristics of those with UMI by CMR are also provided in Table 3.

Over a median follow-up of 6.4 years (interquartile range 4.9–7.0 years), 30 of 91 participants with recognized MI died (33%, CI 23–43%) and 44 of 157 with UMI by CMR died (28% CI 21–35%) which were both significantly higher rates than the 17% (CI 15–20%) with no MI that died (119 of 688). Both UMI by CMR and recognized MI had higher mortality compared to those without MI (HR 1.81, CI 1.28–2.56; absolute risk increase 13%, and HR 2.20, CI 1.48–3.29, absolute risk increase 19%, respectively). UMI by CMR improved mortality risk stratification beyond RMI (category free NRI: 0.34; CI 0.16–0.53). UMI detected by ECG was not associated with higher mortality (HR 0.95, CI 0.49–1.87, absolute risk increase; −1%). Unadjusted Kaplan-Meier survival curves for those without MI, those with UMI by CMR, and those with clinically recognized MI are shown in Figure 2. Five years after the CMR scan, the absolute mortality rates were: 12% (CI 9–14%) for those without MI, 23% (CI 16–29%) with unrecognized MI by CMR, and 23% (CI 17–30%) in those with recognized MI. This culminated in approximately a 10% difference in absolute mortality rates between those with and without MI (eTable2).

After adjusting for age, gender, diabetes, and recognized MI, UMI by CMR remained associated with mortality (HR 1.45 CI 1.02–2.06; absolute risk increase 8%), but UMI by ECG was not associated with mortality (HR 0.88 (CI 0.45–1.73; absolute risk increase −2%). Similarly, UMI by CMR significantly improved the classification of those at risk for mortality (category free NRI 0.16; CI 0.01–0.31, p=0.042) but UMI by ECG did not (NRI: −0.05; CI −0.17– 0.05). Finally, UMI by CMR significantly improved mortality risk stratification (absolute IDI 0.008, CI 0.004–0.013, p<0.001), but UMI by ECG did not improve mortality risk stratification (IDI 0.000 (CI −0.001−0.001; p=0.71).

We observed more use of aspirin, beta-blocker, and statin medications in those with UMI by CMR compared to those without MI. Yet, the use of cardiac medications was significantly less in those with UMI compared to those with recognized MI (Table 3). Roughly half of those with UMI were taking aspirin, whereas less than half were taking statins or beta-blockers.

The AGES–Reykjavik cohort provides results that are most applicable to Caucasian participants, and may not extend to other ethnicities. The sensitivity of CMR for detecting chronic MI using a 0.1 mmol/kg gadolinium contrast dose in our study may be lower compared to higher doses.¹⁶ However, if our study actually had low sensitivity, then the ‘true’ prevalence of MI would be higher. Mitigating the issue of contrast dose, the phase sensitive LGE¹⁸ method used in this study has better signal to noise ratio at low contrast doses than conventional LGE methods. In the minority of participants with both UMI and prior coronary revascularization, we could not ascertain whether UMI occurred independently or as a clinically unappreciated consequence of revascularization. Nonetheless, revascularization complications do not explain the high prevalence of UMI since the prevalence of UMI in diabetic and nondiabetic participants remained high even after excluding prior coronary revascularization. We also did not examine more subtle ECG changes that may be associated with MI. Risk adjustment was limited. This study was designed to demonstrate comparable prognosis between UMI and recognized MI; it was not powered to permit extensive risk adjustment for all baseline differences.