In this general population study, we demonstrate that measures of LV structure derived from an echocardiographic examination independently predict the development of DM 10 years following examination. Specifically, we demonstrate that a LV concentric geometric pattern is an independent predictor of DM in the general population, even after adjustment for established risk factors for DM. Furthermore, we show that the risk of incident DM rises continuously with increasing degree of LV concentricity as evaluated by RWT. Finally, we show that this prognostic information is incremental to established predictors of DM, indicative of a strong association between LV concentric geometry and DM.
Echocardiography, DM and diabetic cardiomyopathy
The notion of a true and distinct diabetic cardiomyopathy has been around for a long time, dating all the way back to the 1950s when Lundbæk, an internist in Denmark, observed frequent myocardial dysfunction in elderly patients with DM [23]. He suggested that DM patients may have heart disease in the absence of coronary blockage [23]. He became the first to suggest the existence of a specific diabetic form of cardiomyopathy. Now, two distinct phenotypes of diabetic cardiomyopathy have been suggested: an eccentric, dilated form with reduced systolic function (HFrEF phenotype), and a restricted, concentric form characterized by diastolic dysfunction (HFpEF phenotype) [13]. The HFpEF phenotype is the one most often encountered by clinicians [13]. Several cross-sectional studies have demonstrated impaired diastolic function and increased LV mass in DM individuals [4, 5, 7, 24]. In our study, diastolic parameters (A, DT, E/A, E/e′, e′, a′) were impaired and LVMI was increased in participants who developed DM during follow-up, supporting the findings of previous studies. However, only peak A-wave remained significant after adjustment for clinical characteristics, suggesting that the prognostic value of the other diastolic parameters was secondary to associations with clinical characteristics. In early diastolic dysfunction and with increasing age along with subtle wall hypertrophy, the A wave increases in magnitude due to increased atrial contribution to LV filling to compensate for the decreased inflow during early diastole (E wave), often referred to as “atrial kick” [25]. However, peak A-wave did not remain significant in a final model adjusting for clinical characteristics and LV concentric geometry. In this model, only LV concentric geometry and RWT were independent predictors of outcome, suggesting that the significance of A was secondary to clinical characteristics and the degree of LV concentricity. It is widely recognized that cardiac concentric remodelling can contribute to diastolic dysfunction, and thus our results support the hypothesis that diastolic parameters are associated with DM secondary to LV concentricity. A recent study by our group, examining cardiac function in a population of DM individuals, found that individuals with DM were mainly characterized by increased LV concentricity evaluated as RWT due to increased LV wall thickness and smaller LV cavities [7]. In the same study, we found that the severity of LV concentricity evaluated by RWT correlated significantly to the duration of DM [7]. Similar relationships between DM and LV concentricity have been found by other authors: In a recent cross-sectional study of type 2 DM patients, De Jong et al. [26] reported that LV concentricity and RWT were significantly increased in both obese and non-obese type 2 DM patients as compared to metabolically healthy obese patients, even in the absence of hypertension. Roberts et al. [27] compared the exercise capacity of type 2 DM patients to that of healthy age and sex-matched controls and found that increased sedentary behaviour and reduced LVIDd were significantly associated with reduced exercise capacity in patients with type 2 DM (in addition, type 2 DM patients also displayed increased RWT as compared to healthy controls). These findings concur with our results, since LV concentric geometry and RWT were the only echocardiographic predictors of incident DM in the present study, and since the risk of DM increased continuously with increasing degree of LV concentricity as evaluated by RWT, suggesting that LV concentricity is indeed correlated to abnormal glucose metabolism.
The finding that the prognostic value of diastolic parameters is secondary to indices of LV concentricity is supported by the suspected pathophysiological mechanisms underlying diabetic cardiomyopathy. It is thought, in the HFpEF phenotype, that hyperglycaemia, lipotoxicity and insulin resistance cause cardiomyocyte hypertrophy and increased resting tension, along with collagen deposition between cells [13]. This leads to LV wall thickening with a concomitant increase in stiffness and decrease in diastolic function. These pathophysiological considerations are reflected in our results, since participants with concentric LV geometry had significantly higher levels of blood glucose, blood triglycerides and cholesterol levels. They also had significantly higher body mass index and blood pressure and lower estimated glomerular filtration rates. These are all risk factors for CVD [28,29,30,31], and they contribute significantly to the increased risk of CVD seen in DM [32]. Thus, the degree of LV concentricity appears to scale with the degree of dyslipidaemia, hyperglycaemia, hypertension, and body mass index even before development of DM, supporting its prognostic value in predicting DM.
Target organ damage and prediction of DM
Conventionally, it has been thought that a clear temporal relationship existed between cardiovascular risk factors and the development of end organ damage such as LV diastolic dysfunction, LV hypertrophy and LV systolic impairment. However, it has recently been shown that LV hypertrophy and arterial stiffness, markers of end organ damage, are predictors of incident hypertension in both normotensive and prehypertensive individuals [33, 34]. This raises questions about the temporal relationship between cardiovascular risk factors and end organ damage, indicating that end organ damage may contribute with novel prognostic value. Besides our study, one study has previously investigated the prognostic role of echocardiography with respect to development of DM in the general population. In this study, Park and Colleagues investigated the prognostic value of echocardiography in predicting incident type 2 DM in a cohort of 1817 non-diabetic individuals from Korea [9]. The age of their population was lower (mean 54 years vs mean 57.9 years) and the follow-up was shorter (6 years vs median 12.6 years) when compared to our study. Also, the prevalence of hypertension in their sample was much lower (22.1% vs 38.7%) than in our sample indicative of a younger and healthier population. During follow-up 273 individuals (15%) developed type 2 DM, which is higher than what we found in our study. This may be explained by the lower age of their study sample. Due to the higher age of participants in our study many had already developed DM at baseline and were therefore excluded. Park and Colleagues found that markers of diastolic dysfunction (e′ and E/e′) predicted incident DM. After adjustment for clinical risk factors, only e’ and the presence of diastolic dysfunction (yes/no) were independent predictors of DM. No results regarding the prognostic value of LV concentric geometry or RWT in predicting DM in the final multivariable model are shown, however, RWT was significantly higher in individuals who developed type 2 DM during follow-up (0.39 SD 0.07 vs 0.36 SD 0.06). This suggests that LV concentric geometry may have been an independent predictor of DM in their study as well, however no results regarding LV concentric geometry was reported. Nevertheless, it is possible that the prognostic value of echocardiography differs between ethnicities with regards to prediction of DM. Hence more research is needed to explain the differences between our two studies. However, altered LV concentricity seems to correspond more closely to the pathophysiological mechanisms suspected to underlie the cardiac alterations seen in the HFpEF phenotype of diabetic cardiomyopathy, and our results suggest that the predictive value of many diastolic parameters in predicting DM is secondary to LV concentric geometry. To our knowledge, ours is the second study to evaluate the prognostic value of echocardiography in predicting development of DM in the general population, and therefore, our results should be viewed as exploratory and hypothesis generating. More research, by independent groups in similar populations, is needed before any significant conclusions can be drawn. In summary, we show that end organ damage can contribute with prognostic value in predicting DM in the general population, and that this organ damage precedes the development of DM.
Limitations and future considerations
Several limitations to this study must be acknowledged. Firstly, mean age at baseline of participants included into this study was 56.6 years (SD: 16.3 years), and thus the prevalence of DM before exclusion of diabetic individuals was 10.9% (241 individuals). These were all excluded from the study. Therefore, due to the high age of this general population sample, a large proportion had already developed DM at baseline and were therefore excluded. Furthermore, we assessed the development of DM by ICD-10 codes, and therefore we do not know how rigorously or by what methods participants have been monitored for the development of DM. Due to unknown differences in the rigour of monitoring of participants, the time from echocardiographic examination to DM diagnosis may have been overestimated, since participants may not have sought out medical consultation at the onset of symptoms from DM. This could potentially alter the temporal relationship between RWT values and DM development observed in this study. Furthermore, in the Copenhagen City Heart study, prevalent DM was defined using the old HbA1c cut-off value (HbA1c ≥ 7.0%). Today a cut-off value of HBA1c ≥ 6.5% is used [35]. Therefore, our definition of prevalent DM may have underestimated the true prevalence. Also, the general population in Denmark is mainly of Caucasian ethnicity, and therefore we cannot extrapolate our results to other races—stressing this is particularly important when considering ethnicity in itself is a significant risk factor for DM, and that ethnicity modifies the contributions of many other risk factors for cardiovascular disease in diabetic individuals [11, 12]. We did not have information on the prevalence of rare cardiomyopathies such as hypertrophic cardiomyopathy, which may affect LV geometry. However, undiagnosed cases of rare cardiomyopathies are unlikely to have affected our results given their very low prevalence in the general population. Also, since rare cardiomyopathies do not cause DM, potential inclusion of such patients would only serve to weaken our results. Yet, we still found significant associations between LV concentric geometry and DM. A final limitation is related to the method of outcome assessment in the present study. In this study, DM outcomes were assessed through the Danish National Board of Health’s National Patient Registry using ICD 10 codes. However, this means that receiving a diagnosis of DM necessitates some type of hospital contact. This hospital contact can either be directly related to DM, or it can be unrelated to the presence of DM and the presence of DM may then be discovered through patient history or examination/assessment. Given the long follow-up of our study (median 12.6 years) and the high age of participants at baseline (mean 57 years) it is reasonable to assume that most patients will have at least one hospital contact not related to a potential DM diagnosis facilitating the discovery of DM if present. Yet, this does constitute a limitation of the present study and could have affected our results. In addition, it is also possible that the DM outcomes detected may represent sicker DM patients possibly requiring hospitalisation at DM debut since outcome assessment is related to hospital contact. Thus, although our study provides important hypothesis generating results, ideally, our findings should be evaluated in multiple ethnicities, in a younger study population and using a study design including closer, dedicated outcome assessment.