Our first unexpected result revealed that neither hospitalization for cardiovascular causes nor all cause mortality are a function of diabetes duration.
Despite this unexpected finding the prognostic value of cardiovascular biomarkers changes dynamically in relation to duration of type 2 diabetes mellitus, which implies that pathophysiology of cardiovascular events changes over time.
In patients with a diabetes duration below 7 years GDF-15 and hs-TnT are the best predictors of the composite endpoint of unplanned hospitalization for cardiovascular causes and all cause mortality.
In patients with longstanding type 2 diabetes (diabetes duration > 7 years) age and NT-proBNP are throughout the best predictors of the composite endpoint, whereas GDF-15 and hs-TnT have lost their prognostic value.
Unplanned hospitalization for cardiovascular causes and all-cause mortality were also analysed separately as two different secondary endpoints, providing additional interesting information. Regarding unplanned hospitalization for cardiovascular causes NT-proBNP was the only neurohumoral biomarker being predictive in all classes of diabetes duration. This extreme stability confirms the clinical value of this parameter.
Considering all cause mortality GDF-15 remained a stable marker for the earlier phase of diabetes until NT-proBNP gets superiority to predict mortality after 12 years of diabetes duration.
From a pathophysiological point of view NT-proBNP reflects volume and pressure overload in a failing heart. All stages from subclinical cardiovascular dysfunction to advanced stages of heart disease and heart failure can be detected with increasing blood levels depending of the severity of alterations. Tremendous increases can be seen in situations with acute volume or pressure overload of the ventricles. This mainly occurs when cardiac function worsens acutely and might explain the predictive power to detect cardiac hospitalizations better than other neurohormones. Moreover BNP and NT-proBNP are also induced by inflammation and interact directly with glucose metabolism 16. Many data suggest that NT-proBNP is a mediator of gluco-inflammatory myocardial crosstalk in diabetes, which also explains the high and independent predictive power of this peptide 17.
GDF-15 is a cytokine expressed in adipocytes physiologically decreasing food intake, body weight and adiposity. There is a clear correlation of GDF-15 with body mass index, body fat, glucose and C-reactive protein making it an attractive biomarker in diabetes. From a cardiovascular point of view, GFD-15 works as a stress responsive cytokine expressed also in cardiomyocytes and vascular smooth muscle cells. Unlike NT-proBNP, GDF-15 is primarily a marker of inflammation fibrosis, cell injury and atherosclerosis 18, 19. These malignant three are the main trigger of progression of cardiac disease. Thus high GDF-15 might reflect the activity of cardiac processes leading to myocardial or coronary dysfunction. Elevated levels of GDF-15 can already be seen in early hypertension, when hypertrophy occurs 20 but also fuels atherosclerosis by inflammatory processes 21, 22 whereas NT-proBNP reflects the severity. Heart failure with preserved ejection fraction (HFpEF) which is mainly based on fibrosis is a common cardiac disease in patients with diabetes. Even here GDF-15 appears to be a significant biomarker of the disease 23.
It can be speculated that the complementary effect of GDF-15 in early diabetes and the prognostic effect of NT-proNBP regarding mortality is based on different immanent information. GDF-15 might be a better marker of disease activity during development of cardiac disease (during the first years of diabetes onset), whereas NT-proBNP is more a marker of disease severity being a better predictor in later stages of full blown cardiac disease. This would be in accordance with a strong trend to a decrease of GDF-15 in longstanding diabetes but an increase in NT-proBNP over time.
Beside this two most prominent markers also UACR has some independent information as already known from earlier publications 24.
UACR is a marker for atherosclerosis and diabetic nephropathy 25. As shown urinary albumin to creatinine ratio performs pretty well in patients with longstanding diabetes while there is no predictive information in patients having a diabetes duration below 12 years. This is well in line with studies showing that about 40% of patients with "diabetic" nephropathy have no albumin excretion, especially at the onset of the disease26. In accordance UACR has a a poor performance at the beginning of the disease with probable early onset of kidney disease.
In a data driven cluster analysis Ahlqvist and colleagues have defined 5 different subgroups of diabetes mellitus. Moreover they impressively demonstrated differences in risk of diabetic complications in relation to the dominating pathophysiologic factor. These findings are well in line with our results also showing that cardiovascular event rates are not directly related to duration of diabetes mellitus, but might be based on different phenotypes. Based on this heterogenity in phenotypes of diabetes our data add some further information as there is an additional heterogeneity over time in the biomarker profile. We have to confess, that in this work only few, although important-markers were tested and larger datasets will give a clearer picture in the future.