Author + information
- Received March 4, 1999
- Revision received July 30, 1999
- Accepted October 5, 1999
- Published online January 1, 2000.
- Stuart E Sheifer, MD∗,* (, )
- Bernard J Gersh, MB, ChB, D.Phil, FACC†,
- N.David Yanez III, PhD‡,
- Philip A Ades, MD, FACC§,
- Gregory L Burke, MD∥ and
- Teri A Manolio, MD¶
- ↵*Reprint requests and correspondence: Stuart E. Sheifer, Fellow, Division of Cardiology, Georgetown University Medical Center, 4 Main, 3800 Reservoir Road, N.W., Washington, DC 20007
This study was designed to determine the prevalence of unrecognized myocardial infarction (UMI), as well as risk factors, and to compare prognosis after detection of previously UMI to that after recognized myocardial infarction (RMI).
Past studies revealed that a significant proportion of MIs escape recognition, and that prognosis after such events is poor, but the epidemiology of UMI has not been reassessed in the contemporary era.
The Cardiovascular Health Study (CHS) database, composed of individuals ≥65, was queried for participants who, at entry, demonstrated electrocardiographic evidence of a prior Q-wave MI, but who lacked a history of this diagnosis. The features and outcomes of this group were compared to those of individuals with prevalent RMI.
Of 5,888 participants, 901 evidenced a past MI, and 201 (22.3%) were previously unrecognized. The independent predictors of UMI were the absence of angina and the absence of congestive heart failure (CHF). Six-year mortality did not significantly differ between the two groups.
1) In the elderly, UMI continues to represent a significant proportion of all MIs; 2) associations with angina and CHF may reflect complex neurological issues, but they also may represent diagnosis bias; 3) these individuals can otherwise not be distinguished from those with recognized infarctions; and 4) mortality rates after UMI and RMI are similar. Future studies should address screening for UMI, risk stratification after detection of previously UMI, and the role of standard post-MI therapies.
Past cohort studies have demonstrated that 25% to 40% of myocardial infarctions (MIs) are clinically unrecognized. Because these infarctions are accompanied by minimal or no symptoms, they escape detection until ultimately an electrocardiogram (ECG) is performed for screening or other clinical purposes. The risk associated with unrecognized MI (UMI) has been demonstrated to be substantial, with a long-term total mortality as poor or worse than that of recognized myocardial infarction (RMI) (1–7).
Significant limitations exist to our current understanding of UMI. In each of the prior cohort studies, either all or most of the follow-up interval occurred before 1990, and prevalence and prognosis may have changed over time. These studies included few women and very few elderly. Also, it remains uncertain whether the predisposing factors for UMI are simply the traditional risk factors for coronary atherosclerosis, or whether UMI patients have unique characteristics that distinguish them from those with recognized events. In large part, this is due to a paucity of multivariate analyses designed specifically to predict infarct recognition. Finally, there is insufficient evidence to determine whether “recognition status” is independently associated with prognosis after MI, especially in older adults.
The Cardiovascular Health Study (CHS) is an ongoing population-based investigation of elderly men and women that employs multiple modalities, including serial ECGs, to identify novel markers of risk for cardiovascular disease. It therefore provides an opportunity to address several of the unanswered questions about UMI. This analysis was performed to determine the prevalence of UMI and the factors that independently distinguish individuals with UMI from those with RMI. In addition, it was designed to compare the long-term prognoses of these two types of MI.
The design of the CHS has been fully described elsewhere (8,9). In brief, this longitudinal cohort study follows 5,888 individuals aged 65 years or more for cardiovascular events. Enrollment at four field centers in the U.S. began in 1989. Participants, the majority of whom were free of overt cardiovascular (CV) disease at the time of entry, are tracked annually with questionnaires regarding medical history, medication use, and CV risk factors. In addition, they receive annual evaluations, which include physical examinations, neurologic and psychosocial assessments, laboratory studies, and noninvasive tests, including ECGs.
Electrocardiogram analysis and diagnostic criteria for UMI
Standard 12-lead resting ECGs were obtained via the MAC PC-DT ECG recorder (Marquette Electronics, Milwaukee, Wisconsin). The results were transmitted to the CHS Electrocardiographic Reading Center for analysis using the NOVACODE measurement and classification system (10). To identify individuals with prior UMI, the CHS database was queried for subjects who evidenced a prior myocardial infarction on the entry ECG, but who answered either “no” or “don’t know” to the entry question “Has a doctor ever told you that you had a myocardial infarction or heart attack?” Electrocardiographic criteria for prior infarction included the presence of Q-waves that were of sufficient duration and amplitude to meet the lead-specific standards of the Minnesota Code (codes 1-1 through 1-2, except 1-2-8). Alternative criteria included the presence of smaller Q-waves (code 1-2-8 or 1-3), when combined with significant ST-segment or T-wave abnormalities (codes 4-1 through 4-3, or codes 5-2 through 5-3) (11).
Identification of risk factors for UMI
The SPSS for Windows Release 8.0.0 software was used for all statistical analyses in this investigation, and for all hypothesis testing, a p value of ≤0.05 was considered statistically significant (12). To evaluate factors potentially associated with UMI, characteristics of those with prevalent UMI, those with prevalent RMI, and those with no prior MI were compared (three-way analysis of variance [ANOVA] for continuous variables; three-way chi-square for categorical variables). Factors of interest included demographic characteristics, traditional risk factors for coronary artery disease, the promising novel risk marker Factor VII, past CV diagnoses and symptoms, the results of noninvasive tests (such as FEV1 [forced expiratory volume in 1 s by pulmonary function testing]), and several psychosocial variables.
Next, to identify factors that predict whether an infarct will be recognized, bivariate logistic regression analyses were performed, with each of the characteristics described above tested as an independent variable, and with infarct type (UMI vs. RMI) as the dependent variable. Finally, to identify factors that independently distinguish those with UMI from those with RMI, a stepwise logistic regression model predicting infarct recognition was generated. Factors tested in this model included those that were significant, at a p value ≤0.10, in bivariate analysis.
Annual mortality was tabulated both for subjects with prevalent UMI and for those with prior RMI. Median follow-up to date in the CHS is 5.4 years (mean: 4.8 years). Cause of death was ascertained using a standardized review process, which considered death certificate data, coroner/medical examiner reports, and interviews with next of kin (13). Cardiovascular (CV) death was defined as that due to coronary artery disease, stroke, congestive heart failure (CHF), or peripheral vascular disease. Non-CV death was defined as death from any other cause. Total, CV, and non-CV mortality of the two groups were compared, using both chi-square and Kaplan-Meier analyses. Finally, to identify independent predictors of mortality in the entire MI cohort, a Cox proportional hazards model was assembled. Factors tested in this model included age, gender, traditional coronary risk factors, CHF, and infarct recognition status (UMI vs. RMI). Each of the bivariate predictors of UMI were also tested for significance.
Definitions of selected clinical variables
Prevalent RMI was documented as present when, at entry, a participant did recall being told by a doctor that he or she had had a MI or heart attack. Angina was defined as reported symptoms at baseline that were confirmed by medication use, prior coronary events, a discharge abstract form, and/or a physician questionnaire. Congestive heart failure was documented as present if there was reported CHF at baseline and confirmation by the subject’s medication list, a discharge abstract form, or a physician questionnaire. Cerebrovascular disease was defined as subject-reported or physician-diagnosed stroke or transient ischemic attack. Claudication was based on symptoms reported at baseline and/or on baseline examination.
Hypertension (HTN) was defined as systolic blood pressure (SBP) ≥160 mm Hg, diastolic blood pressure (DBP) ≥95 mm Hg, or a reported history of HTN that was confirmed by current use of an antihypertensive medication. Diabetes mellitus was defined as a reported history of diabetes, insulin or oral hypoglycemic use, or fasting blood glucose ≥140. Family history of coronary artery disease was based on reported MI in a sibling.
At baseline, 901 of the 5,888 participants (15.3%) evidenced a prior MI, and 201 (22.3%) of these were unrecognized. Table 1demonstrates the characteristics of the no-MI, UMI, and RMI subgroups. In unadjusted comparisons, there were several factors that differed significantly among these populations. Within the demographic variables, age was highest in the UMI group and male gender was most frequent in the RMI group. Conventional risk factors also differed among the three groups, including hypertension, which was most frequent in the UMI population, and family history of coronary artery disease, which was most frequent in the RMI population. Rates of several clinical syndromes also differed significantly among these subgroups. Angina, CHF, cerebrovascular disease, and claudication were all more frequent in the RMI population, while FEV1 was lowest on average in the subjects with UMI.
Factors associated with UMI
The results of bivariate modeling, comparing the UMI and RMI subgroups, are presented in Table 2. Several factors were found to be significant bivariate predictors of UMI, including female gender, increasing age and blood pressure. Also, the absence of common CV diagnoses, including angina, CHF, and claudication, predicted that an MI would be unrecognized. Additional factors that were associated with UMI included the absence of a family history of coronary artery disease, low FEV1, increasing Factor VII level, and good or excellent self-assessed health status. Former smoking was associated with recognized infarction.
Results of the stepwise logistic regression analysis are displayed in Table 3. This revealed that the sole independent predictors of UMI were the absence of angina and the absence of CHF. There was also a trend toward an independent association with low FEV1. Given the established impact of gender and height on FEV1, we controlled for these factors in the regression analysis.
Six-year total, CV, and non-CV mortality rates of the two MI subgroups, as compared by chi-square analyses, are presented in Table 4. These analyses demonstrated that the mortality rate for those with prevalent UMI was significantly higher than that for those with no history of infarction, and it was not significantly different from that of individuals with RMI. Additional comparisons of the recognized and unrecognized infarction groups showed that subjects with UMI have a significantly lower CV mortality rate, but also a trend toward more frequent non-CV death.
Kaplan-Meier analyses (Figs. 1 through 3)⇓⇓⇓generated similar results. Specifically, they showed that mean duration of survival in the UMI group (5.36 years) was significantly worse than in the population without MI (5.68 years) (p < 0.001), but not significantly different from that among subjects with prior recognized infarctions (5.18 years) (p = 0.26). Once again, the similar survival in the two MI populations appears to represent the combined effect of significantly lower CV mortality (p = 0.015) and a nonsignificant trend toward greater non-CV mortality (p = 0.22) in subjects whose infarctions were unrecognized (Figs. 2 and 3).
In Cox proportional hazards modeling (Table 5), the independent predictors of death after MI were increased age, male gender, CHF, diabetes mellitus, and fair or poor perception of one’s personal health status. Infarct recognition was not an independent predictor of mortality. However, there was a significant interaction between infarct recognition status and family history, and there was a trend toward a significant interaction between infarct recognition and hypertension.
Among the elderly, UMI represents 22.3% of all MIs. Factors that independently distinguish individuals with UMI from those with clinically detected infarction include the absence of angina and the absence of CHF. Other demographic and clinical features, including age and traditional risk factors, are not independently associated with infarct recognition. Long-term mortality after UMI is not significantly different from that after RMI.
Frequency of UMI
Although greater than one-fifth of all MIs in the CHS were clinically unrecognized, prior studies have suggested that the relative frequency of UMI is even greater. In a 1990 Framingham analysis, UMI represented 30% of all infarctions (6). Similarly, studies of the Reykjavik cohort demonstrated that 35% of infarctions in men and 33% all infarctions in women escaped clinical detection (5,14). Also, in the Bronx Aging Study, which evaluated individuals at least 75 years of age, 43.5% of MIs were clinically unrecognized (15).
These differences are probably both methodological and demographic. The ECG criteria for MI varied among studies, and differences in baseline characteristics may have contributed as well. Also, improvements over time in patient education, physician awareness, and diagnostic testing may have reduced the frequency with which infarcts escape clinical attention. Finally, whereas the Framingham analysis addressed incident infarctions, our investigation focused on prevalent MI, and thus prevalence-incidence bias may have impacted on the findings (6,16). More specifically, incident UMIs resulting in early mortality might be underrepresented in our sample, and this might have led to an underestimate of the frequency of this event.
It should be noted, however, that, due to their reliance on the ECG to document MI, probably all of the available studies have underestimated the frequency of UMI. As Q-wave criteria were employed to make this diagnosis, many previously unrecognized infarctions were probably never detected. These would include UMIs resulting in sudden cardiac death, unrecognized non-Q-wave MIs, and unrecognized Q-wave infarcts in which, over time, the Q waves resolved (7,17).
Factors associated with UMI
A key issue in identifying factors associated with UMI is the comparison group. In several prior investigations, those with UMI were compared to all other subjects in the study population, or to all those free of prior infarction (1,2). In other studies, in which the intent was to identify factors that distinguish those with UMI from those with RMI, investigators performed bivariate analyses comparing these two groups only (3,5,6). However, to determine whether a factor is independentlyassociated with infarct recognition, multivariate regression, with recognition status as the dependent variable, is necessary. To our knowledge, this investigation is the only study to include this type of model.
Results of bivariate models
Several of the bivariate findings in CHS merit closer inspection. Consistent with prior studies, increasing SBP and DBP were significant bivariate risk factors for UMI (1–3). These findings support physiologic studies suggesting that mechanisms that control blood pressure are interrelated with those that determine pain perception (18–23).
Agehas also been a consistent bivariate predictor of UMI (6). In the Reykjavik cohort, the odds ratio for UMI, per year of age, equaled 1.10 (95% confidence interval, 1.07 to 1.12) (5). These findings are consistent with the general notion that the manifestations of disease are often blunted in the elderly. Also consistent with previous analyses is the association of female genderwith UMI (6), which may relate to gender differences in clinical presentation, as well as diagnosis bias.
Finally, while diabetesis believed to induce cardiac sensory and autonomic neuropathy, and while diabetics have been shown to have a high prevalence of silent ischemia, in CHS, as in prior studies, diabetes was not associated with infarct recognition (3,5,6,24). This was true not only for diabetes in general, but also for the subgroup of diabetics on insulin or oral hypoglycemic therapy. Confounding factors that may override the effects of neuropathy include the intensity of CV screening and counseling provided to diabetics, as well as the relatively high frequency of emergent complications accompanying infarction in diabetic patients (25).
Results of multivariate analysis
In multivariate modeling, none of these traditional risk factors independently distinguished those with UMI from those with RMI. Instead, the major independent associations were with CV symptoms and diagnoses, including angina. Among men in the Framingham study, 53% of subjects with RMIs had a history of angina, but only 24% of the individuals with UMIs reported this symptom. In women, the percentages were 45 and 33, respectively, and the Reykjavik study produced similar results (5,6). In the CHS, not only were the differences in rates of angina more dramatic (63% vs. 19%), but also the absence of this symptom independently predicted that an infarct would be unrecognized.
This finding may simply be due to diagnosis bias. Individuals with a diagnosis of angina probably receive more aggressive CV follow-up than do others without this diagnosis. When they develop new or changing symptoms, even if mild or nonspecific, they are probably more likely to undergo a thorough CV evaluation to determine whether a new event has occurred (26). These patients and their families may receive more aggressive education and counseling regarding coronary artery disease, and they may consequently be better equipped to recognize the symptoms of MI and seek medical attention.
Alternatively, this association may be grounded in a generalized hyposensitivity to myocardial ischemia, otherwise known as a “defective anginal warning system” (27). Basic science investigations have identified several possible neuropsychiatric disruptions that could block normal warning mechanisms. These include insufficient myocardial receptor stimulation, cardiac neuropathy, and a host of complex supratentorial phenomena, including stoicism and denial (28–35).
The independent association with CHF and the trend toward an independent association with low FEV1 are more difficult to interpret, but they also suggest that diagnosis bias contributes to UMI. Patients with the diagnosis of CHF, like those with angina, are typically followed closely for CV symptoms and events. Conversely, patients with a low FEV1 typically carry pulmonary diagnoses, and when they develop chest symptoms, physicians may be predisposed to attribute them to respiratory problems.
The associations of CHF and FEV1 may also in part be based on neuropsychiatric phenomena. More specifically, they may relate to anatomic and physiologic gates in the afferent nervous system, at which sensory impulses may collide and abolish each other. It has been suggested that in some, when these gates are bombarded with respiratory stimuli, pain or pressure impulses from the heart may be blocked (36).
Prognosis of UMI
This study also confirms that unrecognized infarctions have significant clinical implications. Total mortality was significantly greater in the UMI group than in those with no prior infarction. In addition, while the association between infarct recognition status and outcome varied with the presence or absence of coronary risk factors, total mortality did not significantly differ between the UMI and RMI groups. These results are similar to those in the Reykjavik Study, in which 15-year mortality in men with UMI was 55%, as compared to 52% for those with a previously recognized infarct (5).
This poor prognosis may relate to non-CV co-morbidity. In comparison to subjects with prior recognized infarctions, individuals with prevalent UMI were approximately 30% more likely to die a non-CV death over the study period. Perhaps non-CV co-morbidity may impact on individuals’ ability to sense infarction, and therefore also on their tendency to ascribe symptoms to coronary artery disease.
Subjects with UMI may have also been adversely affected by delays in diagnosis and treatment. When a CHS examination identifies a previously unrecognized infarction, the subject’s personal physician is notified. However, even if the physician chooses to prescribe standard postinfarction medications, their initiation may occur years after the MI. This may explain why UMI subjects, as compared to participants with RMI, had more adverse coronary risk profiles, including higher mean cholesterol and blood pressure, and a greater prevalence of current smoking. This investigation was not designed to address differences in treatment between the UMI and RMI populations.
This study has several limitations. First, it addressed prevalent infarctions, and not incident events. This may have had implications for data quality, and it limited our assessment of risk factors for UMI, as we could not evaluate temporal associations. However, in the CHS to date, the number of incident infarcts is insufficient to answer with adequate statistical power the relevant questions about UMI. A second important limitation of this investigation was the use of self-report in the coding of MI, as it created the potential for recall bias. In particular, using self-report to document RMI might have led to an overestimate of its prevalence. In the CHS, each reported infarction prompts a standardized review process, including evaluation of ECGs, discharge abstract forms, and physician questionnaires. This process confirmed only 471 of the 700 prevalent recognized infarctions. However, if a substantial number of the reported infarctions were erroneous, then the true relative frequency of UMI would be even greater than reported, and this would further support our conclusion that a substantial proportion of infarctions are unrecognized. Finally, given that the CHS cohort has been followed for only six years, it is possible that, with longer follow-up, significant differences in total mortality between infarct groups will emerge.
Despite these limitations, these results have important implications for further study, and for patient care. First, the impact of the absence of angina on risk for UMI needs to be explored further. Subsequent investigations should evaluate potential neuropsychiatric explanations, as they may provide insight into the pathophysiology of UMI. Alternatively though, if they are negative, more emphasis should be placed on the potential role of diagnosis bias. Appendix
While these investigations are ongoing, cost-effective screening mechanisms should be identified, as they may promote early diagnosis of prior UMI, early management, and potentially improved outcomes. More specifically, because it appears that UMI cannot be independently predicted by traditional clinical factors, the potential costs and benefits of routine screening ECGs in all individuals with coronary risk factors should be assessed. Also, because the prognosis after UMI is as poor as that after recognized infarction, it may be prudent to evaluate coronary risk in affected individuals thoroughly. Strategies for risk stratification after detection of a previously unrecognized infarct should be evaluated. It would be particularly valuable to define tests and patient characteristics that predict outcome post-UMI, so as to identify patients who may benefit from more aggressive therapy. Finally, future studies should investigate the role of standard post-MI therapies in this interesting subset of patients.
The cardiovascular health study: participating institutions and principal staff
Forsyth County, NC—Bowman Gray School of Medicine of Wake-Forest University: Gregory L. Burke, Sharon Jackson, Alan Elster, Walter H. Ettinger, Curt D. Furberg, Gerardo Heiss, Dalane Kitzman, Margie Lamb, David S. Lefkowitz, Mary F. Lyles, Cathy Nunn, Ward Riley, John Chen, Beverly Tucker; Forsyth County, NC—Bowman Gray School of Medicine-ECG Reading Center: Farida Rautaharju, Pentti Rautaharju.
Sacramento County, CA—University of California, Davis: William Bommer, Charles Bernick, Andrew Duxbury, Mary Haan, Calvin Hirsch, Lawrence Laslett, Marshall Lee, John Robbins, Richard White.
Washington County, MD—The Johns Hopkins University: M. Jan Busby-Whitehead, Joyce Chabot, George W. Comstock, Adrian Dobs, Linda P. Fried, Joel G. Hill, Steven J. Kittner, Shiriki Kumanyika, David Levine, Joao A. Lima, Neil R. Powe, Thomas R. Price, Jeff Williamson, Moyses Szklo, Melvyn Tockman.
MRI Reading Center, Washington County, MD—The Johns Hopkins University: R. Nick Bryan, Norman Beauchamp, Carolyn C. Meltzer, Naiyer Iman, Douglas Fellows, Melanie Hawkins, Patrice Holtz, Michael Kraut, Grace Lee, Larry Schertz, Cynthia Quinn, Earl P. Steinberg, Scott Wells, Linda Wilkins, Nancy C. Yue.
Allegheny County, PA—University of Pittsburgh: Diane G. Ives, Charles A. Jungreis, Laurie Knepper, Lewis H. Kuller, Elaine Meilahn, Peg Meyer, Roberta Moyer, Anne Newman, Richard Schulz, Vivienne E. Smith, Sidney K. Wolfson.
Echocardiography Reading Center (Baseline)—University of California, Irvine: Hoda Anton-Culver, Julius M. Gardin, Margaret Knoll, Tom Kurosaki, Nathan Wong. Echocardiography Reading Center (Follow-up)—Georgetown Medical Center: John Gottdiener, Eva Hausner, Stephen Kraus, Judy Gay, Sue Livengood, Mary Ann Yohe, Retha Webb; Ultrasound Reading Center—Geisinger Medical Center, Danville, Pennsylvania: Daniel H. O’Leary, Joseph F. Polak, Laurie Funk.
Central Blood Analysis Laboratory—University of Vermont: Edwin Bovill, Elaine Cornell, Mary Cushman, Russell P. Tracy; Respiratory Sciences—University of Arizona–Tuscon: Paul Enright; Coordinating Center—University of Washington, Seattle: Alice Arnold, Annette L. Fitzpatrick, Bonnie K. Lind, Richard A. Kronmal, Bruce M. Psaty, David S. Siscovick, Lynn Shemanski, Will Longstreth, Patricia W. Wahl, David Yanez, Paula Diehr, Maryann McBurnie, Chuck Spiekerman, Scott Emerson, Cathy Tangen, Priscilla Velentgas; NHLBI Project Office: Robin Boineau, Teri A. Manolio, Peter J. Savage, Patricia Smith.
☆ The research reported in this article was supported by contracts N01-HC-85079 through HC-85085, HC-95103, and HC-35129 from the National Heart, Lung, and Blood Institute.
- congestive heart failure
- Cardiovascular Health Study
- diastolic blood pressure
- forced expiratory volume in 1 s by pulmonary function testing
- myocardial infarction
- recognized myocardial infarction
- systolic blood pressure
- unrecognized myocardial infarction
- Received March 4, 1999.
- Revision received July 30, 1999.
- Accepted October 5, 1999.
- American College of Cardiology
- Blackburn H.,
- Keys A.,
- Simonson E.,
- Rautaharju P.,
- Punsar S.
- ↵SPSS for Windows Release 8.0.0. Chicago: SPSS, 1997.
- Jonsdottir L.S.,
- Sigfusson N.,
- Sigvaldason H.,
- Thorgeirsson G.
- Falcone C.,
- Auguadro C.,
- Sconocchia R.,
- Angoli L.
- Dworkin B.R.,
- Filewich R.J.,
- Miller N.E.,
- Craigmyle N.,
- Pickering T.J.
- Berger A.K.,
- Breall J.A.,
- Zanger D.R.,
- Johnson A.E.,
- Schulman K.A.,
- Gersh B.J.
- Crea F.,
- Pupita G.,
- Galassi A.R.,
- et al.
- Keele K.D.
- Falcone C.,
- Sconocchia R.,
- Guasti L.,
- Codega S.,
- Montemartini C.,
- Specchia G.