This study was performed according to the recommendations of the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) statement.12 Studies were identified by searching the PubMed electronic database (MEDLINE) and scanning reference lists of articles. The following search terms were used: (“Diastole”[MeSH] OR diastolic OR “myocardial relaxation” OR “cardiac relaxation” OR “myocardial stiffness” OR “cardiac stiffness”) AND (“Prognosis”[MeSH] OR “Heart Failure”[MeSH] OR “Mortality”[MeSH]) AND (community OR “general population”). A language filter was used to restrict the search to English, Portuguese and Spanish papers. There were no date restrictions applied to the electronic searches – all reports from 1974 until October 26 2017 (when the last search was conducted) were eligible. There were no other methodological filters.

Cohort studies that included adults from the community assessing the impact of DD on cardiovascular events (including heart failure, myocardial infarction, and hospitalization from cardiovascular cause) and/or mortality (both cardiac and all-cause) over time were eligible. Studies in specific population groups, such as end-stage renal disease on dialysis, were excluded. Only studies clearly describing how DD was defined and providing data on incident MACE and/or mortality were selected for inclusion. Studies comparing outcomes between individuals with and without DD were eligible for meta-analysis. Studies that only assessed diastolic function parameters as continuous variables, without defining a cut-off to distinguish DD from normal diastolic function, were excluded.

After studies were identified using the search query, they were screened by one investigator (M.A.) based on the title and abstract. Eligibility was assessed and article data were extracted by two reviewers independently (M.A., R.L.) using a standardized form. Any disagreement was subsequently resolved by the two authors. Information was extracted from each included report on the following: characteristics of participants (number of individuals, age, gender, ethnicity, risk factors such as hypertension, diabetes, smoking, body mass index, systolic blood pressure, diastolic blood pressure, medication, left ventricular [LV] hypertrophy, systolic dysfunction, heart failure, coronary artery disease) and diastolic echocardiographic measures (e’ velocity, E/e’ ratio, left atrial size, tricuspid regurgitation velocity, E/A ratio); imaging method used for the diagnosis of DD and diagnostic criteria; and type of outcome measure (MACE and/or mortality).

To avoid double counting of a cohort, one set of results was selected when multiple publications were available for the same cohort. Priority was given to the study with the longest follow-up.

Regarding the studies eligible for meta-analysis and not providing the required data in the full-text and supplementary material publications, the first and/or corresponding authors were contacted by email in order to provide the required information.

The methodological quality of all studies was assessed using the revised and validated version of the Methodological Index for Non-Randomized Studies (MINORS),13 as detailed in Supplementary Table 1.

Study-specific measures were pooled using random-effects model meta-analysis to provide a single summary estimate. Random-effects model meta-analysis makes allowance for between-study heterogeneity. Pooled estimates along with their 95% confidence intervals (CI) were provided. Heterogeneity between studies was assessed using Q and I2 statistics (I2 values of ≤25%, 50%, and ≥75% represent low, moderate, and high levels of heterogeneity, respectively (www.cochrane-handbook.org).

A forest plot was constructed showing the individual studies with the pooled estimates. Publication bias was assessed using the Egger test and the funnel plot analysis. All statistical analyses were performed using STATA software (version 13.1, StataCorp LP, College Station, TX, USA).

Of a total of 604 initially identified studies, only 12 matched our eligibility criteria and were included (Figure 1). Seven additional studies were included after checking the reference list of the articles, yielding 19 studies for the qualitative analysis. Ten studies were excluded from quantitative analysis (meta-analysis) because they provided data only for patients with DD without a clearly defined group with normal diastolic function (n=4), or provided insufficient data for the required calculations (as full-text and supplementary material publications) and did not respond to multiple contact attempts (n=6).

Figure 1.

Flowchart showing search strategy for published data and selection process for inclusion in the systematic review and meta-analysis (according to the PRISMA flow Diagram12).

(0.29MB).

Nine studies were eventually included in the meta-analysis. For the 19 studies included in the qualitative analysis, the mean follow-up period ranged from one year14 to 11 years.15 Eleven studies were conducted in the USA,11,15–24 six in Europe,14,25–29 one in Israel30 and one in Japan.31 Six of the included studies14,17,21–23,31 were retrospective. Included studies were assessed as high-quality publications (median MINORS score of 17; 25th and 75th percentile of 16 and 18, respectively).

The included studies provided data on 63 802 participants from community-based cohorts, with mean ages ranging from 50.929 to 82 years,14 with some studies focusing on the elderly population.16,20,21,30 Most of the studies had a balanced gender distribution, except for Ren et al.,19 which included 81% males. Kuznetsova et al.29 included only white participants and Brady et al.22 had 57% black participants. As depicted in Table 1, the prevalence of major cardiovascular risk factors was in agreement with what would be expected in cohorts coming from the community. However, Blomstrand et al.25 and From et al.17 focused only on subjects with diabetes. Most of the studies included individuals with systolic dysfunction and only six studies excluded this population.14,19,22–24,31 Kardys et al.28 reported as much as 39% of systolic dysfunction in their baseline population. Symptomatic heart failure was present in eight studies,11,14,18,22,24,25,30,31 and most included individuals with known coronary artery disease (only Aurigemma et al.,16 Tsang et al.,21 Brady et al.22 and Kardys et al.28 used this as an exclusion criterion).

All studies used echocardiography (pulsed-wave Doppler and/or tissue Doppler imaging) to assess diastole and to define DD, except Brady et al.,22 who used cardiac catheterization and measurement of LV pressure. The prevalence of DD in the studies ranged from 5.3%20 to 65.2%,24 with a median prevalence of 35.1%.

Significant heterogeneity was observed regarding the definition and grading of DD. Ten studies11,15,16,18,19,21,23,24,28,29 used a one-level classification tree (criteria were presented for each grade and DD was defined as fulfillment of the criteria for any of these grades), two studies26,27 used a two-level classification tree (criteria for DD were defined and, if fulfilled, subsequent grading took place with additional variables), and seven studies14,17,20,22,25,30,31 only defined the criteria for DD, without grading. As shown in Table 2, there was also significant variability in the core echocardiographic parameters for diagnosis of DD. The most used variable was E/A ratio (10 studies11,15,16,18,19,21,23,27–29), followed by E/e’ ratio (eight studies11,17,20,23,25,27,29,30), E-wave deceleration time (seven studies11,15,18,21,23,28,31), left atrial size (five studies15,20,21,27,29) and pulmonary vein flow indices (five studies11,19,23,29,31), e’ velocity (two studies20,27) and LV end-diastolic pressure (one study22).

The median incidence of the primary outcome among individuals with normal diastolic function and participants with DD was 3% (range 2.6-16.8%) and 13.1% (range 5.2-37.4%), respectively. Of the 14 studies providing an association measure, 13 showed a significant association between DD and MACE and/or mortality in the multivariate analysis. The strongest association was reported by Redfield et al.,11 with an HR of 8.31 (95% CI 3.00-23.1) for all-cause mortality in individuals with mild DD and an HR of 10.17 (95% CI 3.28-31.00) for patients with moderate or severe DD. Only Leibowitz et al.30 did not find an association (HR 1.028; 95% CI 0.98-1.084) between E/e’ ratio (as a continuous variable) and all-cause mortality.

Two studies found an association between diastolic function variables and cardiovascular events and/or mortality: Shah et al.20 found that abnormal e’, E/e’ ratio, left atrial dimension and left atrial volume index (LAVi) were significantly associated with incident death or heart failure hospitalization, while Vogel et al.23 correlated E/A ratio with incident atrial fibrillation.

Banerjee et al.14 found a significantly worse combined outcome of all-cause mortality and hospitalization in a cohort of patients with DD and increased E/e′ ratio. Okura et al.31 reported that the cumulative survival rate of DD patients, irrespective of a history of heart failure, was significantly lower than in the general population. However, Brady et al.22 showed that the mean mortality in the DD group was similar to the general population, and therefore DD with a normal LVEF, in the absence of coronary artery disease and systolic dysfunction, had an excellent prognosis over a long period (5-6 years). Overall, 17 studies showed that DD was a significant predictor of MACE and/or mortality, while two studies did not find this association.

Based on random-effects model meta-analysis, the pooled estimate for relative risk of MACE/mortality for individuals with DD across nine studies was 3.53 (95% CI 2.75-4.53; I2=85.5%) (Figure 2). Including the studies providing data on hospitalizations and/or mortality (six studies), the pooled estimate using the random-effects model was 3.98 (95% CI 2.91-5.44; I2=84.2%); in addition, including only the two studies providing all-cause mortality as the primary outcome (Halley et al.24 and Kardys et al.28), DD was associated with a 3.13-fold (95% CI 2.92-3.35; I2=0%) increased risk of death.

Figure 2.

Forest plot showing the overall estimate of the association between diastolic dysfunction and cardiovascular events and/or mortality. CI: confidence interval; RR: relative risk.

(0.11MB).

There were significant differences between individual studies in the magnitude of the association between DD and MACE, as indicated by the statistical test for heterogeneity (Q2=55.0, I2=85.5%, p <0.001). Figure 3 illustrates the funnel plot of studies that excluded small-study effects coming from publication bias (p=0.480 from the Egger test for funnel plot symmetry).

Figure 3.

Funnel plot of the studies included in the meta-analysis for assessment of potential asymmetry and risk of publication bias.

(0.06MB).

Research during the last decade has shed some light on the pathophysiological changes leading to DD and its deleterious impact on cardiac function. Common risk factors for cardiovascular disease (such as hypertension, obesity, hypercholesterolemia and diabetes) are associated with systemic inflammation, myocardial oxidative stress and coronary microvascular dysfunction, and are significant contributors to myocardial stiffening and LV DD.32 In the cardiovascular risk continuum, intermediate stages of risk such as pre-hypertension6 and non-diabetic metabolic syndrome33 are already associated with deterioration in indices of diastolic function measured by echocardiography and cardiac MRI, which suggests that there is also a continuum of myocardial structural and functional changes that impact on diastole. Indeed, the complex interplay between a systemic low-grade proinflammatory state, endothelial dysfunction and changes in myocardial extracellular space and intrinsic cardiomyocyte properties are now accepted as the new pathophysiological paradigm for HFpEF.34 A previous systematic review and meta-analysis35 showed that asymptomatic LV DD was associated with an increased risk for incident HF (relative risk 1.7; 95% CI: 1.3-2.2), including data from five studies.15–19 Therefore, even subclinical DD seems to be strongly involved in the pathophysiology of HF. That being said, both in patients without symptomatic cardiovascular disease and in patients with full-blown cardiovascular disease (such as coronary artery disease), the presence of DD may indicate a more advanced degree of a specific but complex low-grade inflammatory state and structural and functional myocardial changes that appear to be associated with increased risk of CV events and death.

In this meta-analysis significant heterogeneity was observed between studies, which may result from three key factors: different study populations, significant differences in the definition of DD, and different definitions of the primary outcome in each of the studies. All of these are potential limitations of this study and are inherent to the methodological approach adopted, especially regarding the quantitative meta-analysis.

Despite the heterogeneity between the studies included in this work as discussed above, our findings are significant for daily clinical practice and should drive a shift towards a rapid but careful assessment of diastolic function in most patients, as DD was found to be a consistent predictor of cardiovascular events and death. Therefore, we favor the inclusion of a statement concerning diastolic function in all echocardiography reports, when feasible. Notwithstanding, a universal definition of DD is still lacking and therefore most echocardiography laboratories should adopt the one they feel most confident with and use it for consistency and reproducibility.