This study derived from the first 10,000 participants from the Hamburg City Health Study (HCHS, www.​hchs.​hamburg) recruited between 2016 and 2019. The HCHS is a single-centre, prospective, long-term, population-based cohort study placed in Hamburg, Germany [9]. 8,245 subjects of the first 10,000 participants received a transthoracic echocardiogram (TTE). Exclusion criteria were insufficient image quality to perform standardised measurements for grading DD, LVEF < 50%, atrial fibrillation (AF) at the day of the examination, moderate or severe mitral and aortic valve disease, lacking clinical or laboratory variables for HFpEF classification, and implanted pacemakers. Our final cohort comprised 5913 subjects (Fig. 1).

The research protocol of the study was approved by the HCHS steering board and the local ethics committee (PV5131, Medical Association Hamburg). All participants gave written informed consent.

All measurements were conducted by medical professionals at a baseline visit at the HCHS Epidemiological Study Centre in Hamburg following the published HCHS protocol [9]. Blood samples were withdrawn at the day of examination under fasting conditions. Demographics and clinical parameters were assessed by self-reported questionnaires and standardized interviews. Diabetes mellitus was determined by a fasting glucose level of ≥ 126 mg/dl, or the use of antidiabetic drugs. Coronary artery disease (CAD) was defined as suffering from one or more of the following conditions: status post myocardial infarction, percutaneous coronary intervention (PCI) or history of coronary bypass surgery assessed by questionnaire and personal medical record. Physical activity was defined as the number of hours per week spent on any type of sport. Socioeconomic-status-index (SES-index) was calculated by a combination of education, occupational status, and income. Quality of life was assessed by the validated SF-8™ (short form) health survey [10, 11].

Transthoracic echocardiography (TTE), carotid and abdominal aorta ultrasound examinations were performed and analysed by cardiologists and professional sonographers (technicians) at the baseline visit on dedicated ultrasound machines (Siemens Acuson SC2000 Prime, Siemens Healthineers, Erlangen, Germany) according to the guidelines of the American Society of Echocardiography (ASE) and the European Society of Cardiovascular Imaging (EACVI). All TTE standard views were assessed in two-dimensional echocardiography, including a three-dimensional four-chamber view for chamber quantification. Image analysis was performed using an off-line workstation with the commercially available Siemens syngo SC2000 version 4.0 software. Pulsed-wave Doppler examination of mitral inflow as well as Doppler tissue imaging of the mitral annulus were performed in each subject. Left ventricular ejection fraction as well as left-sided volumes were calculated using the two-dimensional biplane method of disks summation (modified Simpson’s rule).[6] Left ventricular mass (two-dimensional) was calculated according to the ASE and EACVI guidelines [6]. Tricuspid annular plane systolic excursion (TAPSE) was assessed by M-mode echocardiography in the apical four-chamber view. Right ventricular fractional area change (FAC) was measured in a right ventricular focused four-chamber view. Valvular heart disease was detected by a combination of colour Doppler and continuous wave-Doppler following the current ASE and EACVI guidelines [12].

Diastolic function was assessed on the basis of three conditions: average E/eʹ ratio > 10; septal eʹ < 7 cm/s or lateral eʹ < 10 cm/s, and left atrial volume index (LAVI) > 34 ml/m². DD was considered to be present if ≥ 2 conditions were positive.

For the diagnosis of HFpEF, the 2016 ESC guidelines for the diagnosis and treatment of acute and chronic heart failure were applied [13]. HF was considered present if subjects showed the combination of symptoms/signs, laboratory data, and echocardiographic criteria. Symptoms and signs included: oedema, dyspnoea, history of heart failure or medication (loop diuretics, aldosterone antagonists). Subjects presenting with preserved LVEF (LVEF ≥ 50%), NT-proBNP levels exceeding 125 pg/ml, symptoms or signs of HF and either left ventricular hypertrophy (defined as LV mass indexed to BSA > 95 g/m² for women, > 115 g/m² for men), left atrial enlargement (defined as LAVI > 34 ml/m²) or DD were classified as HFpEF individuals. In our final analysis, we only included those subjects with HFpEF who were diagnosed with DD.

Carotid and femoral arteries were assessed using a linear array transducer (Siemens Acuson S2000 9L4, 4–9 MHz) at high-resolution B-mode. Images of the common carotid artery (CCA) were obtained, and intima-media thickness (IMT) was measured 10 mm caudal of the carotid bulb at three different points. IMT was assessed for the both sides and averaged. Pulsed-wave-Doppler-mode was used to evaluate the flow-velocity of the internal carotid artery (ICA) and the common femoral artery (CFA). The ankle-brachial index was measured using a sphygmomanometer and Doppler probe. The score for both sides was calculated, and the lower one was chosen. The abdominal aorta was examined using a convex transducer (Siemens Acuson S2000 6C1 HD, 1.5–6 MHz). It was assessed in strict orthogonal orientation until the aortic bifurcation, and the largest outer-to-outer-wall diameter was assessed manually. Pulsed-wave-Doppler-mode was used to measure the peak systolic velocity.

Given the large sample size, normality of continuous variables was assessed utilising normal Q–Q plots. Magnitudes of continuous variables were presented as mean ± standard deviation (SD) or median ± interquartile range (IQR), accordingly. Intra-class correlation coefficient (ICC) estimates and their 95% confident intervals (CI) were calculated based on a mean-rating, consistency, two-way mixed-effects model [14].

The unpaired t test was used to analyse differences between groups. For non-normally distributed variables, the Mann–Whitney U test was used instead. Pearson’s correlation coefficient was used to quantify the correlation between end-diastolic and mid-systolic measurements. For multiple group comparisons, overall significance levels were obtained using one-way ANOVA. For multiple pairwise comparison against the base-mean, the t test was used.

The analysed sample of 5,913 participants from the first 10,000 HCHS participants showed the characteristics of a representative middle-aged (mean age 64.4 ± 8.3 years) European population with 3032 women (51.3%) (Table 1).

DD was diagnosed in 753 subjects (12.7%). Of those, 75 (1.3%) subjects suffered from HFpEF (DDwHFpEF) and 678 (11.5%) participants were asymptomatic (ALVDD). Individuals with DDwHFpEF were older, and in contrast to individuals with ALVDD and normal diastolic function predominantly female. The prevalence of most cardiovascular risk factors gradually increased from normal diastolic function to ALVDD to DDwHFpEF, including arterial hypertension, diabetes, current smoking, coronary artery disease (CAD), and atrial fibrillation (AF). The use of heart failure medication as well as the levels of the biomarkers NT-proBNP, hsCRP and HbA1c accordingly increased between the three groups. However, the glomerular filtration rate (GFR), low-density lipoprotein (LDL), and haemoglobin showed an inverse pattern (Table 2). Electrocardiographically, the PR interval, the duration of the QRS complex and the corrected QT interval significantly increased from normal to DDwHFpEF. LVEF was similar in all groups. Nevertheless, there was a significant increase for LV mass index, LVEDV as well as markers of diastolic dysfunction such as E/e’ ratio, TR Vmax, and LASV between the three groups (Table 2).

Several demographic, functional, and morphological data, as well as biomarkers were associated with ALVDD, DDwHFpEF or both cohorts (Table 3, Fig. 2) after adjustment for age, sex, BMI, hypertension, diabetes, current smoking, and CAD. ALVDD correlated with arterial hypertension with an OR of 2.0 (95% CI 1.5–2.6, p < 0.001), HbA1c with an OR of 1.2 (95% CI 1.1–1.3, p = 0.007), renal function assessed by glomerular filtration rate with an OR of 1.2 (95% CI 1.1–1.4, p < 0.001), the PR interval with an OR of 1.1 (95% CI 1.0–1.3) and the diameter of the abdominal aorta with an OR of 1.1 (95% CI 1.0–1.2, p = 0.01). Cardiovascular risk factors that were not only associated with ALVDD but also with DDwHFpEF included BMI (OR 1.2, 95% CI 1.1–1.3, p < 0.001; OR 1.6, 95% CI 1.2–2.1, p = 0.001) and age (OR 1.7, 95% CI 1.5–1.9, p < 0.001; OR 2.7, 95% CI 1.8–4.3, p < 0.001). Furthermore, the functional and morphological echocardiographic parameters E/e’, left ventricular end-diastolic volume (LVEDV), left atrial systolic volume (LASV), and left ventricular mass indexed to BSA showed significant associations with both ALVDD and DDwHFpEF. Apart from E/eʹ, the associations were generally stronger for DDwHFpEF than for ALVDD.

However, comorbidities and demographics that were not related to ALVDD but significantly to DDwHFpEF comprised AF with an OR of 2.6 (95% CI 1.1–5.9, p = 0.024), CAD with an OR of 7.2 (95% CI 3.6–14.2, p < 0.001), COPD with an OR of 3.9 (95% CI 1.8–8, p < 0.001) and female sex with an OR of 2.5 (95% CI 1.3–4.9, p = 0.006). Individuals with DDwHFpEF demonstrated impaired physical abilities, evidenced by the EQ5D score. Functionally, DDwHFpEF in contrast to ALVDD was associated with a prolonged QRS complex (OR 1.4, 95% CI 1.1–1.7, p = 0.005).

In binary logistic regression analysis, clustering of specific risk factors increased the probability of ALVDD and DDwHFpEF (Fig. 3). While the combination of female sex, age > 65 years, diabetes, and arterial hypertension predicted a 25.7% probability for being classified in the ALVDD group, the same risk factor combination predicted DDwHFpEF by 3.1%. Adding BMI > 30 kg/m², GFR < 60 ml/min, CAD, and AF to the risk factor profiles resulted in a 68.3% probability for DDwHFpEF compared to 28.7% for ALVDD.

DD is a complex syndrome originating from the pathological interplay of left ventricular pressure decline during diastole, volume load conditions, and chamber stiffness. The assessment of DD can be performed invasively with specific high-fidelity catheters, which remains the gold-standard, or non-invasively, by the use of echocardiography [3]. When assessed non-invasively, there is no single variable that reliably reflects DD, but a combination of Doppler and native measurements is applied. Until now, numerous approaches have been proposed for reliably diagnosing and classifying DD. It is pivotal to emphasise the contrasting DD definitions and the heterogeneity both in the acquisition techniques and study populations for understanding the striking differences in the reported prevalence of DD [15]. Before the introduction of the 2016 ASE/EACVI recommendations, the reported prevalence of DD ranged between 11.1% and 36% within the normal population [16, 17]. Applying the 2016 ASE/EACVI recommendations for the diagnosis of DD, the prevalence reported seemed to be significantly reduced. Two studies by Huttin et al. and Almeida et al. showed a prevalence of 1.3% and 1.4% [7, 18]. In line with this striking antagonism, in our population-based cohort, we reported, that DD was present in 12.7% and that ALVDD was present in 11.5% of subjects aged 46 and 78 while only 0.46% of all subjects were classified in the DD group applying the 2016 ASE/EACVI recommendations (Supplements). The main novelty of the 2016 ASE/EACVI recommendations compared to previous DD algorithms was the integration of TR peak velocity and a higher E/eʹ cut off, resulting in a high specificity on the cost of a rather low sensitivity highlighted by several simultaneous echocardiographic-catheterization studies [19]. Furthermore, the ASE/EACVI recommendations classified anyone with underlying myocardial disease as DD. However, the recommendations lack a definition of relevant myocardial disease. Accordingly, in line with the studies cited above, we strictly applied the echocardiographic criteria for defining DD [7, 18]. This of course led to a lower prevalence of DD in our study than expected.

Nevertheless, our study primarily screened for diastolic dysfunction in the general population as a possible biomarker in mainly asymptomatic and only a small proportion of symptomatic subjects. For this approach, in line with most population-based studies, we chose a DD definition with a reasonable sensitivity based on the most robust, feasible, and prognostically-relevant parameters e’ velocity, E/e’ and LAVI [20, 21].

ALVDD is considered a precursor of manifest heart failure [8]. Mechanisms underlying the formation of ALVDD are poorly understood. Despite the absence of symptoms, subjects with ALVDD demonstrated signs of cardiac remodelling including left ventricular hypertrophy and left atrial enlargement in our study, both established prognostic markers for adverse heart failure events [22].

Moreover, our results support the hypothesis, that not only the formation of HFpEF but also of ALVDD might be comorbidity-driven [23]. Consistently, in our study ALVDD and DDwHFpEF shared independently associated risk factors including age and BMI, which have been linked to the progression from ALVDD to DDwHFpEF before [8, 24]. The observed distribution of BMI in our study population is consistent with that of the Gutenberg Health Study, a German cohort study of similar design and age [25, 26]. Interestingly, the risk factors arterial hypertension and the biomarker HbA1c were associated with ALVDD but not with DDwHFpEF beyond the risk factors adjusted for. Possibly, arterial hypertension and elevated blood sugar levels might play crucial roles as driving factors especially for the early formation of DD [27]. In contrast, the widespread diseases AF, CAD, and COPD were strongly associated with DDwHFpEF but not with ALVDD and the presence of AF and CAD boosted the probability of HFpEF in our risk profile-based analysis. Population-based studies from Olmsted County, as well as results from the Framingham Heart Study, revealed, that new-onset heart failure in subjects with preserved ejection fraction was associated with AF, COPD, and CAD. [4, 28] Accordingly, the effects of AF, COPD, and CAD on a possible progression from ALVDD to HFpEF should be prospectively investigated in future studies. Additionally, the width of the QRS complex was associated with HFpEF but not with ALVDD. Accordingly, QRS duration could be either considered as a marker of heart failure or possibly represent dyssynchrony of the ventricles which might play a role in the transition from ALVDD to HFpEF. [29]

Our study cohort of the first 10,000 HCHS participants originates from the population of Hamburg. Hence, most study participants were of Caucasian ascend. The functional translation of our results into other populations is, therefore, limited. Therefore, our findings should be examined in other ethnic and racial groups.

Our classification of subjects as DD with and without HFpEF is based on diagnostic algorithms. There was no specific gold-standard for diagnosing DD and HFpEF, e.g., by invasive catheterization. Thus, our study did not assess the diagnostic accuracy of the different algorithms. Furthermore, dyspnoea, as the leading symptom of HFpEF, was assessed without clinical testing by a validated questionnaire.

Finally, our study setting is cross-sectional. We cannot tell which variables were the cause or the effect of ALVVD and HFpEF. Having shown multiple new associated factors of ALVDD and DDwHFpEF, further prospective studies are needed to evaluate their role in the genesis or preservation of heart failure.

Our study provides new data on the prevalence and correlates of DD both with and without symptomatic heart failure in the general population. The prevalence of DD was 12.7%. Of those 11.5% were free of symptoms (ALVDD) and 1.3% suffered from HFpEF (DDwHFpEF). We identified different patterns of cardiovascular risk factors, comorbidities, and functional parameters associated with either ALVDD or DDwHFpEF. These results warrant further research concerning the exact role of the identified correlates for the formation of DD and HFpEF.