Diabetes mellitus (DM) is often complicated by concomitant hypertension and associated with increased cardiovascular complications, the most common of which are coronary artery diseases and the subsequent development of heart failure (HF). However, the existence of a primary myocardial disease in diabetic patients, diabetic cardiomyopathy, has been proposed (1- 11). The existence of diabetic cardiomyopathy was first proposed by Rubler et al. (10) in 1972 on the basis of postmortem findings. Subsequently, abnormalities in both systolic and diastolic performance in diabetic subjects have been demonstrated in animal and human studies. The pathogenesis of this left ventricular (LV) dysfunction in diabetic patients is not clearly understood. Microangiopathy, increased extracellular collagen deposition, and abnormalities in calcium transport alone or in combination are considered to be associated with this dysfunction (12- 14). Furthermore, the evidence indicates that myocardial damage in diabetic patients affects diastolic function before systolic function (15). Unfortunately, there have been few population-based studies to evaluate the outcomes of pre-clinical diastolic dysfunction in diabetic patients (16).

The combination of pulsed tissue Doppler imaging velocity of the medial mitral annulus during passive filling (e′) with the early passive transmitral inflow velocity (E) has been validated as a reliable index of LV filling pressure (17). Studies have demonstrated a reduction of annular e′ in type 2 diabetes mellitus (DM) and that an increased E/e′ ratio was associated with left atrial enlargement and correlated independently with glycosylated hemoglobin (11). Thus, the E/e′ ratio may be used to detect and follow the progression of diastolic dysfunction in diabetic patients (18).

The objective of our study was to determine whether there is an association between pre-clinical diastolic dysfunction in DM patients and the subsequent development of HF. We used tissue Doppler imaging indices of diastolic dysfunction as the measure of cardiac dysfunction because tissue Doppler imaging indices correlate well with intracardiac pressure tracings of ventricular filling in patients with and without DM. We hypothesized that diastolic dysfunction in DM will be associated with the increased risk of the development of subsequent HF.

This study was conducted in the Olmsted County, Minnesota, population. Health care providers in Olmsted County include the Mayo Clinic, Olmsted Medical Center, and a handful of private practitioners. Medical records for all providers are available for review by using the comprehensive record-linkage system provided by the Rochester Epidemiology Project. The Rochester Epidemiology Project allows the indexing of all medical records of Olmsted County residents according to clinical and pathological diagnoses, surgical procedures, and billing information. Death certificates are also indexed according to cause of death. Records are quite complete because >90% of the population receives care at Mayo Clinic or Olmsted Medical Center, and residents are seen on average every 3 years at the Mayo Clinic (19). The potential of the data source was described elsewhere (19).

After approval by the Mayo Clinic and the Olmsted County Medical Center Institutional Review Boards, we retrospectively identified all diabetic patients with a tissue Doppler imaging assessment of diastolic function in the Olmsted County, Minnesota, population. To avoid inclusion of patients in whom the diagnosis of HF was triggered by the results of echocardiography, we excluded patients who were diagnosed within 30 days after the echocardiogram. Furthermore, patients were excluded if the diagnosis of HF was made before the echocardiogram or if severe mitral or aortic valve regurgitation or stenosis was present. The main outcome was the development of HF.

We identified patients with a new diagnosis of DM in Olmsted County from 1996 through 2006 using the International Classification of Disease-9th Revision code 250. This code has been validated in Olmsted County previously with an accuracy of >98% (20). Among these patients, we identified those who subsequently developed HF by using International Classification of Disease-9th Revision code 428. This code has also been validated in Olmsted County: 90% of patients with this code have a physician diagnosis of HF and 82% of patients with this code meet Framingham Criteria for HF (21). All echocardiograms were performed from September 2001 through June 2007. Ejection fraction was evaluated by a modification of the method of Quinones et al. (22). Doppler echocardiography was performed to determine the early mitral inflow velocity (E) and tissue Doppler imaging evaluation was performed of the medial mitral annulus velocity during passive filling (e′) as previously described (17). Diastolic dysfunction was defined as an E/e′ ratio >15, as previously described (17). Left ventricular size and wall thickness were measured using the 2-dimensional image if available. Otherwise, the M-mode measurement was used.

Categorical variables were summarized as percentages and continuous variables as mean ± SD. Comparison between groups was based on 2-sample t tests for continuous variables and Pearson's chi-square test for categorical variables. The major end point was the development of HF. Other end points analyzed included death and atrial fibrillation. Kaplan-Meier analysis was performed to estimate probabilities of events and the probabilities were compared among groups using the log-rank test statistic. Univariable and multivariable associations of clinical and echocardiographic variables with each end point were assessed with Cox's proportional hazard modeling using the event of interest and time from 30 days after the echocardiogram to the date of the event or last follow-up as the outcome. In patients without an event, the date of the latest follow-up was the time of the data collection (June 2007) or date of death. Patients with missing data were excluded from the multivariable analysis. However, fitting the model with missing indicators revealed similar results. Analyses were performed using JMP version 6.0.0 (SAS Institute Inc., Cary, North Carolina).

Of the 12,014 Olmsted County residents with DM, 4,571 had echocardiographic data. From this group, we identified 2,770 diabetic patients with a tissue Doppler echocardiographic assessment of diastolic dysfunction. We excluded 845 patients with a diagnosis of HF before the echocardiogram, 130 patients with a diagnosis of HF made within 30 days after the echocardiogram, 1 patient with severe mitral or aortic valve regurgitation, and 34 patients with severe aortic or mitral stenosis. Overall, 1,760 diabetic patients were included in this study and 411 (23%) patients had pre-clinical diastolic dysfunction defined as an E/e′ >15 without a diagnosis of HF. Average time from echocardiogram to death or latest follow-up was 2.9 ± 1.8 years. Baseline characteristics are shown in (Table 1). The DM patients with diastolic dysfunction were older, were more often female, and had a higher prevalence of hypertension and coronary artery diseases compared with the DM patients without diastolic dysfunction. Echocardiographic findings of DM patients with diastolic dysfunction include greater left atrial volume and LV mass index.

Using Cox's proportional hazard modeling, we determined that the E/e′ ratio was independently associated with the subsequent development of HF after adjustment for age, sex, body mass index, hypertension, coronary disease, ejection fraction, left atrial volume, deceleration time, and LV mass index. This analysis suggests that for every 1-unit increase in the mitral E/e′ ratio, the risk of HF increases by 3% (hazard ratio [HR]: 1.03, 95% confidence interval [CI]: 1.01 to 1.06; p = 0.006).

Diastolic dysfunction defined as an E/e′ >15 was predictive of HF in Kaplan-Meier analysis (Figure 1). The cumulative probability of the development of HF for DM patients with diastolic dysfunction was 13.1% at 1 year and 36.9% at 5 years compared with 5.2% at 1 year and 16.8% at 5 years for DM patients without diastolic dysfunction (p < 0.001). In multivariable Cox's proportional hazard regression analysis, we determined that diastolic dysfunction was associated with the subsequent development of HF after adjustment for age, sex, body mass index, hypertension, coronary disease, ejection fraction, left atrial volume, deceleration time, and LV mass index (HR: 1.61, 95% CI: 1.17 to 2.20; p = 0.003) (Table 2).

Diastolic dysfunction (defined as an E/e′ ratio >15) was predictive of death in Kaplan-Meier analysis (Figure 2). The cumulative probability of death for diabetic patients with diastolic dysfunction was 6.9% at 1 year and 30.8% at 5 years compared with 3.1% at 1 year and 12.1% at 5 years for diabetic patients without diastolic dysfunction (p < 0.001). In multivariable Cox's proportional hazard regression analysis, we determined that diastolic dysfunction was associated with death after adjustment for age, sex, body mass index, hypertension, coronary disease, ejection fraction, left atrial volume, deceleration time, and LV mass index (HR: 2.01, 95% CI: 1.32 to 3.06; p = 0.001).

To assess the association between diastolic dysfunction and the subsequent development of atrial fibrillation, we analyzed the 1,450 patients without a previous diagnosis of atrial fibrillation before the date of the echocardiogram. Among this subgroup, the cumulative probability of subsequent atrial fibrillation for diabetic patients with diastolic dysfunction was 6.8% at 1 year and 18.7% at 5 years compared with 3.1% at 1 year and 8.8% at 5 years for diabetic patients without diastolic dysfunction (p < 0.001). In multivariable Cox's proportional hazard regression analysis, we determined that diastolic dysfunction was not significantly associated with the subsequent development of atrial fibrillation after adjustment for age, sex, body mass index, hypertension, coronary disease, ejection fraction, left atrial volume, deceleration time, and LV mass index (HR: 1.16, 95% CI: 0.75 to 1.76; p = 0.50).

In subgroup analysis of 1,356 diabetic patients with a diagnosis of hypertension before the date of echocardiography, diastolic dysfunction was independently predictive of subsequent HF (HR: 1.67, 95% CI: 1.20 to 2.33; p = 0.003) and death (HR: 2.14, 95% CI: 1.36 to 3.36; p = 0.001) after adjustment for age, sex, body mass index, coronary disease, ejection fraction, left atrial volume deceleration time, and LV mass index. However, after adjustment for these variables, diastolic dysfunction was not independently associated with the subsequent development of atrial fibrillation (HR: 1.01, 95% CI: 0.62 to 1.63; p = 0.95).

We used the criteria outlined by Lang et al. (23) for the measurement of relative wall thickness and the classification of left ventricle geometry. In 1,616 patients, the left ventricular geometry could be defined. A total of 554 (34%) had normal geometry, 524 (32%) had concentric remodeling, 343 (21%) had concentric hypertrophy, and 195 (12%) had eccentric hypertrophy.

In multivariable analysis with adjustment for age, sex, body mass index, hypertension, coronary disease, ejection fraction, left atrial volume, deceleration time, LV mass index, and relative wall thickness (as a continuous variable), diastolic dysfunction was independently associated with the subsequent development of HF (HR: 1.63, 95% CI: 1.20 to 2.21; p = 002) and death (HR: 2.13, 95% CI: 1.43 to 3.15; p = 002).

In multivariable analysis with adjustment for age, sex, body mass index, hypertension, coronary disease, ejection fraction, left atrial volume, deceleration time, and geometric pattern of the left ventricle (as a categorical variable), diastolic dysfunction was independently associated with the subsequent development of HF (HR: 1.62, 95% CI: 1.19 to 2.19; p = 002) and death (HR: 2.18, 95% CI: 1.48 to 3.20; p < 0.001).

The present study is the first to use a community-based cohort of diabetic patients to determine whether HF is more likely to develop in those with pre-clinical diastolic dysfunction compared with those without diastolic dysfunction. Our current findings demonstrate that pre-clinical diastolic dysfunction is common in patients with DM and confirms that pre-clinical diastolic dysfunction in those with DM is associated with an increased risk of the subsequent development of HF, mortality, and atrial fibrillation.

Pre-clinical diastolic dysfunction has been broadly defined as diastolic dysfunction in patients with normal systolic function and no symptoms of HF (15,24). In diabetic patients, the existence of a pre-clinical diastolic dysfunction has been well defined and estimates of prevalence vary from 20% to 60% depending on the Doppler echocardiographic criteria that was used to define diastolic dysfunction (1- 9). Several lines of evidence indicate that LV diastolic dysfunction may precede LV systolic dysfunction in diabetic patients (15,20). This early restrictive disease seen in diabetic patients is likely due to microangiopathy, interstitial fibrosis, extracellular collagen deposition, calcium transport abnormalities, and neurohormonal alterations, alone or in combination (12- 14). Regan et al. (25) conducted a cardiac catheterization study that demonstrated that normotensive diabetic patients with normal ejection fraction without coronary artery disease and without clinical evidence of HF have an increased LV end-diastolic pressure and a decreased LV end-diastolic volume. Although systolic dysfunction, LV hypertrophy, coronary disease, and hypertension have all been shown to increase the risk of the development of HF in DM patients, the prognostic impact of pre-clinical diastolic dysfunction in DM patients has not been well defined, but the association with the subsequent development of HF has been suspected (16).

We determined that the prevalence of pre-clinical LV diastolic dysfunction in our population was 23%. We also demonstrated that pre-clinical diastolic dysfunction in DM patients was associated with an increased risk of the subsequent development of HF and mortality after adjustment for multiple demographic, echocardiographic variables, hypertension, and coronary disease. Although novel, these findings were hypothesized based on the known associations between diastolic dysfunction, HF, DM, and mortality. Previously, Redfield et al. (24) demonstrated that even mild diastolic dysfunction conferred a risk of increased mortality compared with subjects with normal diastolic function in the general population. Furthermore, DM is a well-recognized risk factor for the development of HF as demonstrated by the report from the Framingham Heart Study that showed that HF occurred 2 times more often in men with DM and 5 times more often in women with DM compared with age-matched control subjects. Likewise, we previously reported that the prevalence of DM in patients with HF is approximately 20%, and DM is associated with increased mortality in patients with HF (26), whereas Tribouilloy et al. (27) reported a markedly increased mortality rate in patients with HF and preserved ejection fraction among patients with DM compared with patients without DM.

The cited studies focus on the E-to-A ratio to diagnose diastolic dysfunction. The difference between the E-to-A ratio and the E/e′ ratio is physiopathological: The E-to-A ratio is <1 when the relaxation is abnormal and LV filling pressure is still normal, but becomes >1 when LV filling pressure increases. Indeed, there are limitations of a categorical variable such as the E-to-A ratio as shown in an analysis of the Strong Heart Study by Bella et al. (28) in which high values of the E-to-A ratio predict mortality compared with low values, which do not. On the other hand, the E/e′ ratio has a continuous behavior because its increase is always an expression of increased LV filling pressure. As we have shown with our cohort of diabetic patients, an increase in the E/e′ ratio is associated with increasing risk of subsequent HF.

Despite the current medical therapy for HF and DM, our current study demonstrates that the prevalence of pre-clinical diastolic dysfunction is high in DM patients and associated with worse outcomes. Furthermore, we recently reported that there is a direct correlation between the duration of DM and diastolic dysfunction and that significant diastolic dysfunction occurs 4 years after the onset of DM independent of coronary disease or hypertension (20). Therefore, future studies should be conducted to test the hypothesis that screening and aggressive management of diabetic patients with pre-clinical diastolic dysfunction may delay the progression to HF with improved outcomes.

The study was retrospective, we did not have data on medications or diabetes subtype, and the study relied heavily on the International Classification of Disease–9th Revision coding variables to define HF. Although these codes have been validated as a diagnostic and research tool in Olmsted County, Minnesota, they do allow for potential bias. Furthermore, the patients were not recruited from the community, but were clinically referred for echocardiography by their primary physicians, which may decrease the generalizability of the results. Finally, although the diversity of the Olmsted County population is increasing, as shown by the 2000 census (29), characteristics of the Olmsted County population are similar to those of U.S. whites (19). Thus, these findings should be examined in different racial and ethnic groups.

This study confirms that pre-clinical diastolic dysfunction is prevalent in patients with DM. More importantly, we demonstrated that an increase in the E/e′ ratio in diabetic patients was associated with subsequent HF and mortality independent of hypertension, coronary disease, and other echocardiographic parameters.