Cardiovascular Risk Prediction Is Improved by Adding Asymptomatic Coronary Status to Routine Risk Assessment in Type 2 Diabetic Patients

The patients were prospectively recruited in the Diabetes Department of Jean Verdier Hospital between 1992 and 2006. This study was approved by the ethical committee of Reims, France. Each patient gave informed consent for enrollment in accordance with the European directives. Eligibility criteria included type 2 diabetes; no history of myocardial infarction or angina pectoris; normal 12-lead resting electrocardiogram (ECG); and presence of at least one of the following additional cardiovascular risk factors: dyslipidemia, hypertension, smoking, nephropathy, family history of premature CAD, and peripheral/carotid occlusive arterial disease (PCOAD) (14). Exclusion criteria included congenital heart disease or known cardiomyopathy. Diabetic retinopathy was graded according to the Early Treatment of Diabetic Retinopathy Study severity scale and defined as absent or present, and also as absent, mild/moderate (minimal and moderate nonproliferative retinopathy), and severe (severe nonproliferative or proliferative retinopathy). The diagnosis of peripheral neuropathy was based on the presence of any two or more of the following: neuropathic symptoms, decreased distal sensation, or decreased or absent ankle reflexes. Data on the treatments used during the follow-up were not available. Because antidiabetic, lipid-lowering, and antihypertensive treatments have been more intensively used since the year 2000 (15), we considered the percentage of follow-up time spent after 2000 to estimate the quality of treatment. When the follow-up duration spent after year 2000 was ≥10% of the overall follow-up duration, the patient was considered as being on current multifactorial care. Finally, the 10-year cardiovascular UKPDS (3) and the 10-year cardiac Framingham (4) risk scores were calculated.

The protocol was previously reported (10,13,14). Each patient underwent a ²⁰¹Tl myocardial scintigraphy after an ECG stress test, a pharmacological stress test (dipyridamole injection), or both. The ECG stress test was performed in patients who could exercise on a bicycle ergometer and were expected to have an interpretable exercise ECG. When the patient was unable to exercise or when the ECG stress test result was indeterminate, a pharmacological stress test using dipyridamole was carried out. SMI was defined as an abnormal ECG stress test, an abnormal myocardial scintigraphy imaging (i.e., defects in at least 3 out of 17 segmental regions), or both. A selective coronary angiography was performed in the patients with SMI within a period of 2 months after the noninvasive investigation. CAD was defined either as a ≥70% narrowing of the luminal diameter in the left anterior descending artery, the circumflex artery, a well-developed marginal vessel, or the right coronary artery or as a ≥50% narrowing of the left main coronary artery diameter.

The following measurements were recorded at the time of screening for SMI: HbA_1c (Dimension technology, Siemens Healthcare Diagnosis Inc., Newark, NJ), serum total cholesterol, HDL cholesterol and triglycerides (enzymatic colorimetry, Hitachi 912, Roche Diagnostics, Meylan, France), creatininemia (colorimetry, Kone Optima, Thermolab System, Paris La Défense, France), 24-h proteinuria, and the 24-h urinary albumin excretion rate (laser immunonephelometry, BN100, Dade-Behring, Paris, France). LDL cholesterol was calculated according to the Friedewald formula and creatinine clearance with the Cockcroft formula.

The date of the noninvasive cardiac testing was considered to be the beginning of the follow-up. The follow-up procedure included cardiovascular examination at least once a year. The patients were evaluated for cardiovascular signs and symptoms (angina, dyspnea, edema, and arrhythmia) and had a 12-lead ECG. For each cardiac event, medical records were obtained from the hospital or the primary care physician. When a patient died, the cause of death was documented with the help of the family, the general practitioner, or the cardiologist. The following cardiovascular events were considered: death of cardiac origin (sudden death, death caused by myocardial infarction, or congestive heart failure), nonfatal acute coronary syndrome, heart failure (New York Heart Association stage III or IV and need for hospitalization), secondary need for coronary revascularization, lower limb or carotid revascularization procedure, lower leg amputation, and stroke. The follow-up was stopped when the first event occurred.

Continuous variables were expressed as means ± SD values and compared by one-way ANOVA or Mann-Whitney U test as adequate. The significance of differences in proportions was tested with the χ² test. Because great interindividual differences were observed in the duration of follow-up, two types of analyses were conducted.

In the first analysis, the Kaplan-Meier method was used to examine the time-dependent cumulative probabilities of cardiovascular events. Cox regression analyses were used to determine hazard ratios (HRs) for cardiovascular events in relation to the parameters that predicted cardiovascular events according to the Kaplan-Meier method. We considered the cumulative probabilities of cardiovascular events first according to the presence of SMI or CAD after adjustment on the parameters included in the Cox regression analyses and then according to subgroups, considering routine cardiovascular risk assessment and the presence of SMI or CAD.

In the second analysis, we limited the statistical analysis to a 5-year follow-up. Logistic regression was used for multivariate analyses based on models including the factors that were associated with the occurrence of a cardiovascular event during the first 5 years of follow-up with a P value ≤ 0.10 in univariate analyses. We used c statistic to determine if SMI or CAD added to the prediction of a cardiovascular event above and beyond the risk prediction based on the other parameters. Finally, we calculated the Hosmer-Lemeshow χ² statistic (HLχ²) to test the difference in expected and observed probabilities of an event in the different models.

Statistical analyses were carried out using SPSS software (SPSS, Chicago, IL). The 0.05 probability level was considered for statistical significance.

A total of 731 patients were enrolled between 1992 and 2006. Among them, 688 were followed, whereas 43 (6.0%) were lost to follow-up. The latter patients did not differ significantly from the former when considering either the main clinical and biological criteria or the SMI status (data not shown). The main baseline characteristics of the 688 patients are described in Table 1. SMI was diagnosed in 207 of them (30.1%). A coronary angiography was subsequently performed in 191 of the 207 subjects with SMI. Out of them, 76 (i.e., 11.0% of the 688 patients) had CAD, including one-vessel disease in 47 and two- and three-vessel disease in 15 and 14 patients, respectively.

Of the 76 patients with silent CAD, 22 were treated by coronary angioplasty and 6 by coronary artery bypass, whereas the remaining patients were medically treated, according to the cardiologist’s decision. These initial revascularization procedures were not counted as cardiovascular events. A total of 98 patients had a first cardiovascular event during a 5.4 ± 3.5 (range: 0.1–19.2) year period: 10 cardiac deaths, 39 acute coronary syndromes, 10 nonfatal congestive heart failures, 1 secondary coronary revascularization procedure, 21 strokes, 12 peripheral revascularization procedures, and 5 lower-leg amputations.

Kaplan-Meier survival analyses showed that SMI (adjusted log-rank test 21.2, P < 0.0001), the presence of both SMI and silent CAD (log rank 47.2, P < 0.0001), retinopathy whatever its stage (log rank 11.7, P < 0.001), severe retinopathy (log rank 5.8, P < 0.05), diabetic nephropathy (log rank 5.1, P = 0.025), macroproteinuria (log rank 16.0, P < 0.001), PCOAD (log rank 56.6, P < 0.001), a Framingham risk score ≥20% (log rank 7.6, P < 0.01), a UKPDS risk score ≥20% (log rank 8.0, P < 0.01), and current multifactorial care (protective, log rank 83.3, P < 0.0001) were significant predictors of cardiovascular events. Three models of Cox regression analyses were built (Table 2). The routine variables that predicted cardiovascular events, including Framingham or UKPDS risk score, were entered into model 1. SMI (model 2) or CAD (model 3) was then added to model 1. In model 1, macroproteinuria, current multifactorial care, and PCOAD were independently predictive of cardiovascular events (model 1 with UKPDS risk score: χ² 65.5; model 1 with Framingham risk score: χ² 61.1). SMI in model 2 was additionally and independently predictive of cardiovascular events (χ² 69.0 with UKPDS risk score and 64.5 with Framingham risk score) as CAD was in model 3 (χ² 73.5 with UKPDS risk score and 71.0 with Framingham risk score). The cumulative probabilities of cardiovascular event according to the presence of SMI or CAD after adjustment on the parameters from model 1 (with UKPDS risk score) are shown in Fig. 1A and B. This figure also shows that the cumulative probability of cardiovascular events increased with the presence of at least one of the following criteria: macroproteinuria, no current multifactorial care, and PCOAD or with SMI (Fig. 1C), CAD (Fig. 1D), or both.

Table 1 shows the variables that were associated with the occurrence of a cardiovascular event during the first 5 years of follow-up. The variables predicting cardiovascular events were entered into three logistic regressions: UKPDS risk score ≥20% (age, sex, smoking, diabetes duration >20 years, and HbA_1c ≥10% were not entered into the model because they were already included in this score), retinopathy, nephropathy, peripheral neuropathy, and PCOAD in model a; plus SMI in model b; and plus CAD in model c. The results are shown in Table 3; macroproteinuria, PCOAD (model a, b, and c), and SMI (model b) or CAD (model c) were independently predictive of a cardiovascular event during the first 5 years. The area under the receiver operating characteristic (AROC) curve for model a parameters to predict a 5-year cardiovascular event was 0.705 (95% CI 0.616–0.794; P < 0.001). When the presence of SMI (model b) or CAD (model c) was added into the model, the AROC curve increased to 0.788 (0.720–0.855; P < 0.0001) or 0.779 (0.701–0.857; P < 0.001), respectively. HLχ² were 1.34, P = 0.932; 5.13, P = 0.643; and 1.77, P = 0.940 for models a, b, and c, respectively.

The same statistic was built using the Framingham risk score ≥20% (but not age, sex, or smoking, which were already included in this score), diabetes duration >20 years, HbA_1c ≥10%, retinopathy, macroproteinuria, peripheral neuropathy, and PCOAD in model a; plus SMI in model b; and plus CAD in model c. Framingham risk score and PCOAD (model a, b, and c), HbA_1c and macroproteinuria (model a), and SMI (model b) or CAD (model c) were independently predictive of 5-year events (Table 3).

The present data show that in this cohort of asymptomatic type 2 diabetic patients with at least one additional cardiovascular risk factor, the performances of UKPDS or Framingham risk scores to predict cardiovascular events are limited and may be improved by considering the presence of macroproteinuria and PCOAD. Furthermore, we show here for the first time that the presence of SMI or silent CAD is independently associated with cardiovascular events and improves cardiovascular risk prediction.

Although some studies (16) suggest that the presence of diabetes should be regarded as a risk of coronary mortality similar to established CAD, we show in the current study that type 2 diabetic patients may be further stratified by evaluating their a priori cardiovascular risk using the calculation of a specific (UKPDS) or nonspecific (Framingham) risk score. However, the association between a high risk score and the occurrence of events disappeared in multivariate analyses (Table 2). Recent studies have also shown that risk equations are likely to overestimate cardiovascular risk (5,6), partly because the current multifactorial therapy has markedly improved the cardiovascular prognosis in the diabetic population. In the current study, we considered the year 2000 as the threshold time, from when the treatment of risk factors has been intensified in accordance to the current guidelines (15). It is interesting that shorter time exposure to contemporary treatment (expressed as percentage of follow-up duration spent after 2000) was independently predictive of events.

Other parameters may be useful to evaluate the cardiovascular prognosis in the diabetic population, such as the presence of retinopathy or nephropathy. The presence of these microangiopathic complications has usually been considered as a good marker of exposure (in time and intensity) not only to hyperglycemia but also to other risk factors including hypertension. In the current study, any stage of retinopathy and severe retinopathy both predicted events, although only severe retinopathy was previously shown to be associated with a high cardiovascular risk (9). Our data also support the high risk of events associated with incipient nephropathy and the even higher risk associated with macroproteinuria (8).

An alternative could be the identification of vascular integrators of risk (i.e., parameters that may reflect the cumulative exposure to cardiovascular risk factors and its intensity). For example, it was shown that the presence of arterial stiffness or arteriosclerotic plaques could improve the risk prediction when added to the Systemic Coronary Risk Evaluation in healthy subjects (17). In the current study, PCOAD was an independent predictor of cardiovascular events. The procedure for diagnosing PCOAD is usually easy, with ultrasound examination being performed especially in patients with clinical signs or symptoms. Screening for SMI and subsequently silent CAD is more complicated and expensive. Silent CAD in diabetic patients was shown to be associated with a higher incidence of cardiac events (10,11). With regard to diabetic patients with SMI but no CAD, we have previously reported evidence for abnormalities of coronary flow reserve and endothelium function and shown that such functional abnormalities were also associated with a worse prognosis (18). The current study confirms that both SMI and CAD are strong predictors of cardiovascular events and shows for the first time that diagnosing CAD in asymptomatic patients improves cardiovascular prediction in addition to the risk estimation based on traditional risk factors, risk equations, nephropathy, PCOAD, and current multifactorial care.

The present results are in line with some recommendations for SMI screening in diabetic patients with high cardiovascular risk (1,2,19). This proposal is, however, under debate (20,21) because of several considerations. First, such a screening cannot be performed in all diabetic patients, and the current selection criteria still need to be improved (22,23). Second, the cardiovascular prognosis has been markedly improved in diabetic patients by intensifying preventive medical treatments. However, the current article shows that the prognosis associated with SMI remains poor despite more intensive treatment as prescribed since 2000. Finally, the Detection of Silent Myocardial Ischemia in Asymptomatic Diabetic Subjects (DIAD) study has recently shown that screening for SMI was not associated with a better prognosis (12). However, very few patients with SMI underwent coronary angiography and revascularization during this study. Nevertheless, another randomized study suggested that screening for SMI and CAD may improve the prognosis if a coronary revascularization was performed in patients with coronary stenoses (24).

Our hospital-based study has some limitations. The diabetic patients who were included had at least one additional risk factor and, therefore, the results are not necessarily generalizable to the diabetic population. CAD status was unknown in the patients without SMI because they did not undergo a coronary angiography for ethical reasons. However, the present series includes the largest series ever published in the literature of coronary angiographies in patients with SMI. The number of cardiovascular events was limited. The prognosis was not adjusted on medical therapy but on the period of treatment.

In conclusion, SMI is a common condition in patients with type 2 diabetes and at least one additional cardiovascular risk factor. SMI and silent CAD are strong predictors of cardiovascular events in diabetic patients, beyond their a priori cardiovascular risk and independent of more or less intensive medical therapy. Risk prediction is improved by adding coronary status to routine prognosis assessment. However, screening for silent coronary disease is expensive and not easily available in routine assessment and, therefore, should not be performed in all diabetic patients. The selection criteria for screening still need to be improved, and the benefit of screening and subsequently treating SMI and silent CAD remains to be extensively addressed in further studies (25).

No potential conflicts of interest relevant to this article were reported.

E.C. directed research, analyzed statistics, and wrote the manuscript. M.T.N. researched data, analyzed statistics, and contributed to discussion. B.C., I.B., and S.C. contributed to discussion. C.B. and K.T. researched data. P.V. directed research, contributed to discussion, and reviewed and edited the manuscript.

Parts of this work were presented in abstract form (abstract 511-P) at the 71st Scientific Sessions of the American Diabetes Association, 24–28 June 2011, San Diego, California.

The authors thank Dr. Frédéric Paycha, Assistance Publique–Hôpitaux de Paris (AP-HP), Louis Mourier Hospital, Colombes, France, and Professor Pierre Weinmann, AP-HP, Avicenne Hospital, Bobigny, France, for isotopic explorations and Dr. Simon Cattan, Centre Hospitalier Intercommunal, Le Raincy-Monfermeil, France, for coronary angiographic explorations.