Heart failure with reduced, mid-range and preserved left ventricular ejection fraction in Chinese with type 2 diabetes: risk factors and prognosis from time of rst index hospitalization

Background The clinical predictors and prognosis of heart failure (HF) by categories of left ventricular ejection fraction (LVEF) have not been well studied in people with diabetes. In a retrospective cohort of Chinese with type 2 diabetes, we examined 1) clinical factors associated with incident decompensated HF, and 2) mortality post-HF, stratied by LVEF. We conducted a retrospective analysis of the Hong Kong Diabetes Register comprising 23,348 people with type 2 diabetes without history of HF enrolled between 1993–2015, followed for incident decompensated HF until 2017. Heart failure subtypes were dened according to LVEF on echocardiography. Multivariate Cox proportional hazards models were used to identify clinical factors associated with incident HF versus no HF, stratied by HF subtypes. All-cause mortality rates were compared by HF subtypes. ltration density-lipoprotein; mid-range ejection with preserved ejection with reduced ejection fraction; ratio; LDL, density-lipoprotein; RAAS, renin-angiotensin-aldosterone system


Abstract Background
The clinical predictors and prognosis of heart failure (HF) by categories of left ventricular ejection fraction (LVEF) have not been well studied in people with diabetes. In a retrospective cohort of Chinese with type 2 diabetes, we examined 1) clinical factors associated with incident decompensated HF, and 2) mortality post-HF, strati ed by LVEF.

Methods
We conducted a retrospective analysis of the Hong Kong Diabetes Register comprising 23,348 people with type 2 diabetes without history of HF enrolled between 1993-2015, followed for incident decompensated HF until 2017. Heart failure subtypes were de ned according to LVEF on echocardiography. Multivariate Cox proportional hazards models were used to identify clinical factors associated with incident HF versus no HF, strati ed by HF subtypes. All-cause mortality rates were compared by HF subtypes.

Results
Over median follow-up of 7.1 years from enrolment, 1,195 (5.1%) people developed decompensated HF.

Conclusions
In Chinese with type 2 diabetes, HFpEF was the predominant HF subtype. One-year mortality following decompensated HF was lowest in HFpEF group but 10-year mortality was similar between HFpEF and HFrEF.

Background
Heart failure (HF) is a serious complication of type 2 diabetes resulting in impaired health-related quality of life and functional status as well as frequent hospitalization. Type 2 diabetes independently increases the likelihood of HF by 1.7-fold in men and 2.0-fold in women (1). People with type 2 diabetes and comorbid HF have a 20-30% excess risk of all-cause mortality than their counterparts with HF without diabetes (2,3). Established risk factors for developing HF in people with type 2 diabetes include old age, obesity, poor glycemic control, impaired kidney function and pre-existing coronary heart disease (4,5).
Several glucose-lowering drug classes such as thiazolidinediones and dipeptidyl-peptidase 4 inhibitors may potentially induce or worsen HF (6, 7), whereas sodium-glucose co-transporter 2 (SGLT2) inhibitors bene t not only diabetes but also people with HF (8). Given the rising prevalence of type 2 diabetes and changing clinical pro le of people living with diabetes with many living to older ages, the burden of HF is also increasing (9,10).
Heart failure may be classi ed according to left ventricular ejection fraction (LVEF) on echocardiography as HF with reduced LVEF (HFrEF), HF with mid-range LVEF (HFmrEF), and HF with preserved LVEF (HFpEF) (11). Reduced LVEF (LVEF < 40%) is frequently agged as an adverse prognostic index and primarily denotes impaired ventricular systolic dysfunction with adverse remodelling (12,13). Heart failure with preserved LVEF (LVEF ≥ 50%) is characterised by LV hypertrophy and impaired LV lling, also referred to as diastolic dysfunction. Heart failure with mid-range LVEF (LVEF 40-49%) is a newly de ned HF subtype that remains poorly understood. Notably, LVEF is only one of the measures of cardiac function, is subjected to inter-observer variability, changes over time, and does not infer underlying aetiology or pathophysiology of HF (14,15). Despite these limitations, LVEF is used commonly for selection of participants in clinical trials and its measurement is widely available. Previous observational studies based on HF registries have shown that clinical characteristics and outcome of HF vary according to LVEF (13,(16)(17)(18). However, limited data are available in people with type 2 diabetes (19). Using a longitudinal cohort of Hong Kong Chinese with type 2 diabetes without HF at baseline, we examined 1) clinical factors associated with incident decompensated HF, and 2) mortality post-HF, strati ed by LVEF.

Study cohort
We conducted a retrospective analysis of the Hong Kong Diabetes Register (HKDR) which enrolled adults aged ≥ 18 years with physician-diagnosed diabetes who underwent structured assessment of metabolic control and diabetes complications at the Diabetes and Endocrine Centre, the Prince of Wales Hospital, Hong Kong Special Administrative Region (20). A liated with the Chinese University of Hong Kong and governed by the Hong Kong Hospital Authority (HA), the hospital serves approximately one seventh of the Hong Kong population of 7.39 million as of 2017. Referral sources included hospital-based specialist and family medicine clinics as well as community out-patient clinics. For this study, Chinese men and women with type 2 diabetes without history of HF at baseline enrolled since inception of the HKDR on 1 Jun 1993 until 30 June 2015 were included for analysis. People with type 1 diabetes, diabetes or unknown type, who were non-Chinese, or with history of HF were excluded. All participants have provided written informed consent for the collection and analysis of their clinical information for research purpose. The study has received approval by the Joint Chinese University of Hong Kong -New Territories East Cluster Clinical Research Ethics Committee.
Baseline assessment At enrolment, information on socio-demographics, past medical history and drug use was collected by a trained nurse. Anthropometric measurements (body weight, body height, waist circumference) and vital signs (blood pressure [BP], pulse) were collected. Fundus was examined for the presence or absence of diabetic retinopathy using ophthalmoscopy or fundus photography, interpreted by trained endocrinologists. Lower extremities were examined for evidence of sensory neuropathy using mono laments and 256-Hz tuning fork. Sensory neuropathy was present if a person ful lled two or more of the following criteria: reduced sensation to mono lament, reduced vibration sense on testing using tuning fork, and self-reported abnormal sensation over one or more lower limbs. After an 8-hour fast, blood samples were collected for HbA1c, plasma glucose, lipids (total cholesterol, low density-lipoprotein (LDL) cholesterol, high density-lipoprotein (HDL) cholesterol, triglyceride), and serum creatinine. Spot urine sample was collected for urine albumin-to-creatinine ratio (ACR). Estimated glomerular ltration rate (GFR) was determined using the Chronic Kidney Disease Epidemiology Collaboration Eq. (21). Chronic kidney disease was de ned as estimated GFR < 60 ml/min/1.73m 2 . Albuminuria was de ned as urine ACR ≥ 3.0 mg/mmol creatinine.

Follow-up and event ascertainment
The Hong Kong HA is a statutory organisation that governs all public hospitals and majority of outpatient clinics, and provides healthcare to over 90% of people living in Hong Kong. Clinical information including vital status, hospital admissions, investigations and drug prescription are stored in an electronic medical record system of the Hong Kong HA, and are retrievable using the unique Hong Kong Identity Card number compulsory for all local residents. Incident decompensated HF was identi ed as hospital admission with principal diagnosis of HF (ICD-9 code 428) from the date of enrolment into HKDR until 30 June 2017. Next, cases of incident HF in which rheumatic heart disease, valvular heart disease, endocarditis, congenital heart disease and other non-atherosclerotic causes of structural heart disease were recorded in secondary diagnoses were excluded (Supplementary Table 1). Finally, medical records of the remaining people with incident HF were reviewed for echocardiography performed within 24 months of the event. Left ventricular ejection fraction documented by echocardiography closest to the date of HF was used to categorise HF into HFrEF (LVEF < 40%), HFmrEF (LVEF 40-49%), and HFpEF (≥ 50%) (11). People with incident HF were followed until death or 30 June 2017, whichever came rst.
Additionally, repeated measurements of HbA1c, lipids, serum creatinine and urine ACR between the date of enrolment and 30 June 2017 were obtained.

Statistical analysis
We compared baseline clinical characteristics across HF subtypes among people with incident HF, and between people with and without incident HF. Data were expressed as mean ± standard deviation (SD), median (interquartile range [IQR]), or percentages. χ 2 test was used for between-group comparison of categorical variables, t test for normally distributed continuous variables, and Kruskal-Wallis test for continuous variables with skewed distribution. P-values for multiple comparisons were adjusted using the Benjamini-Hochberg method. We performed multivariate Cox proportional hazards model to determine the association of clinical factors (expressed as hazard ratios [HR] with 95% con dence intervals [CI]) with incident HF, separated by HF subtypes. The variables included in multivariate Cox regression model for each HF subtype were based on demonstration of statistically signi cant association with HF subtype, versus no incident HF, on univariate analysis. High density-lipoprotein cholesterol, LDLcholesterol, estimated GFR, HbA1c, log urine ACR, history of coronary heart disease (CHD), and history of stroke were treated as time varying covariates. The proportional hazards assumption in each Cox model was veri ed by evaluating the weighted Schoenfeld residuals (22), and there was no evidence to support that the assumption was violated, with all p-values > 0.05. We estimated all-cause mortality rates and mortality rates at 1-year, 5-year and 10-year post-HF for each HF subtypes. Pairwise comparisons were conducted by using log-rank test and p-values were adjusted using the Benjamini-Hochberg method. Kaplan-Meier analysis was used to plot cumulative all-cause mortality post-HF, separated by HF subtypes. All statistical analyses were conducted using R 4.0.3 (R Foundation for Statistical Computing, Vienna, Austria). A two-tailed p-value < 0.05 was considered statistically signi cant.

Results
Baseline clinical characteristics of the study cohort Among 25,075 people enrolled into the HKDR, 809 with type 1 diabetes or diabetes of unknown type, 230 non-Chinese and 688 with history of HF at baseline were excluded ( Supplementary Fig. 1). Of the remaining 23,348 people, 1,195 (5.1%) developed de novo HF over a median follow-up of 7.1 (IQR 6.8) years. People with available echocardiography were younger and had lower frequencies of chronic kidney disease, diabetic retinopathy and stroke compared with those without echocardiography (Supplementary Compared with those without incident HF, people with HF, irrespective of HF subtypes, were older, had longer diabetes duration, higher HbA1c levels and worse control of other metabolic risk factors at baseline (Supplementary Table 3). In addition, those with HF were more likely to have microvascular complications and coronary heart disease (CHD), and had more frequent use of renin-angiotensinaldosterone system (RAAS) inhibitors, BP lowering drugs and insulin (Supplementary Table 3). Some differences were detected in baseline clinical characteristics between the HF subgroups. Compared with people with HFrEF, those with HFpEF and HFmrEF were older, had longer diabetes duration, higher body mass index (BMI), systolic BP, lower estimated GFR and higher urine ACR (Table 1). A female preponderance was observed in people with HFpEF (Table 1). Sex differences in baseline characteristics among people with incident HF are shown in Supplementary Table 4. ACR, albumin-to-creatinine; BMI, body mass index; BP, blood pressure; GFR, glomerular ltration rate; HDL, high density-lipoprotein; HF, heart failure; HFmrEF, heart failure with mid-range ejection fraction; HFpEF, heart failure with preserved ejection fraction; HFrEF, heart failure with reduced ejection fraction; LDL, low density-lipoprotein; RAAS, renin-angiotensin-aldosterone system Clinical predictors of incident HF, strati ed by HF subtypes Multivariate Cox proportional hazards models were conducted to identify clinical factors associated with incident HF (Table 2). Only variables that showed signi cant association with incident HF in univariate analysis were included (Supplementary Table 5). Old age, low estimated GFR, high urine ACR and history of CHD were independently associated with increased hazards of incident HF, irrespective of HF subtypes. In addition, female sex, BMI and history of sensory neuropathy were associated with increased hazards of incident HFpEF, and HDL-cholesterol was inversely associated with incident HFmrEF.  Table 3). Mortality rates at 5 years and 10 years post-HF remained lower in people with HFpEF than those with HFmrEF, but were similar between HFpEF and HFrEF, and between HFmrEF and HFrEF (Table 3). HF, heart failure; HFmrEF, heart failure with mid-range ejection fraction; HFpEF, heart failure with preserved ejection fraction; HFrEF, heart failure with reduced ejection fraction; IQR, interquartile range * P-values for multiple comparisons were adjusted using the Benjamini-Hochberg method.

Discussion
Type 2 diabetes and HF are both age-related conditions. As life expectancy of the population increases, the prevalence of comorbid diabetes and HF is expected to rise. In our longitudinal cohort of Chinese with type 2 diabetes followed for incident decompensated HF, we report the following observations: 1) Heart failure with preserved LVEF was the predominant HF subtype accounting for 60% of acute HF in people with type 2 diabetes; 2) Clinical predictors for incident HF were similar across HF subtypes with a few salient differences. Correlation of female sex, high BMI and sensory neuropathy was detected with incident HFpEF but not with other HF subtypes, and low HDL cholesterol was associated only with HFmrEF; 3) Mortality at 1 year post-HF was the lowest in the group with HFpEF but longer term mortality rates were similar between HFpEF and HFrEF with a tendency of higher mortality in the group with HFmrEF.

Distribution of HF subtypes
The distribution of HF subtypes as reported in HF registries varies depending on the region, setting (inpatient vs out-patient), age range, and calendar period. Heart failure with reduced LVEF was the predominant HF subtype in earlier studies but more recent studies found a larger representation of HFpEF among those with acute or chronic HF, independent of diabetes status (16,18,23,24). In the U.S-based National Cardiovascular Data Registry Practice Innovation and Clinical Excellence Registry of 1,103,386 participants with chronic HF, 57% had HFpEF and 36% had HFrEF (24). Several East Asian registries also found a preponderance of HFpEF (25,26). In the present study, almost two third of people with decompensated HF and underlying type 2 diabetes had preserved LVEF and another 10% had mid-range LVEF on echocardiography. Collectively, the high prevalence of HFmrEF and HFpEF in contemporary cohorts of people with HF may re ect a shift in risk pro le of predisposed individuals including aging and rise in obesity.

Clinical predictors of incident HF according to subtypes
The pathophysiologic processes leading to HFpEF is not well elucidated. Besides myocardial ischemia due to coronary heart disease, endothelial dysfunction in coronary microvasculature, disturbed cardiac metabolism and systemic low grade in ammation may be contributory (27). In the PRevalence Of MIcrovascular dySfunction in Heart Failure with Preserved EjectionFraction study, local microvascular dysfunction as assessed using adenosine stress Doppler echocardiography, was correlated with markers of systemic endothelial dysfunction and albuminuria (28). In the Asian Sudden Cardiac Death in Heart Failure registry, among 2,800 people with diabetes and HF, the presence of microvascular complications was associated with 1.8-fold higher odds of HFpEF versus HFrEF, independent of other clinical confounders (19). In the present study, we found that age, low GFR, albuminuria and pre-existing coronary heart disease were independent predictors of all subtypes of HF. On the other hand, female sex, BMI and sensory neuropathy were associated with incident HFpEF but not with other HF subtypes. Our results are consistent with the notion that metabolic comorbidities, as indicated by obesity, insulin resistance and microvascular disease, favor the development of HFpEF (19,29). Notably, coronary heart disease increased the hazard of HFrEF by 16-fold and that of HFpEF by 3-fold, suggesting a larger in uence on development of HFrEF versus HFpEF by atherosclerotic stenosis or occlusion, similar to ndings in people without diabetes (13,23,29,30).
The association between female sex and incident HFpEF has also been shown in other studies (23,24,30,31). Reasons underpinning the propensity for HFpEF in women are unclear but may be due to sex differences in the distribution of risk factors. In the present study, among people with incident HF, women were older, had longer diabetes duration, worse metabolic control and a higher frequency of diabetes retinopathy, whereas men were more likely to smoke and have coronary heart disease. Although glycemic and metabolic parameters have been adjusted for in the regression model, residual confounding of unmeasured metabolic factors cannot be excluded. Men have a higher risk of premature mortality and might not have lived long enough to have HFpEF, which is also driven by age (10).
Short-and long-term mortality post-HF Our observation of lower mortality rate at 1 year after HF in the group with HFpEF as compared with the other HF subtypes concur with results from other studies (16,18,25). Findings on mid-to longer-term mortality after HF are con icting in the literatures with some studies reporting similar rates (30,32) and in others, lower mortality in those with HFpEF (13,16,26). In the present cohort of people with type 2 diabetes, we detected similar mortality rates at 5 and 10 years after HFpEF and HFrEF. People with HFrEF were more likely than those with HFpEF to have pre-existing coronary heart disease and possibly more liberal use of life-saving drugs such as statins. Conventional drug therapies for HF are effective in reducing all-cause and cardiovascular mortality in people with HFrEF but not in those with HFpEF (33)(34)(35)(36)(37)(38). Mineralocorticoid receptor antagonists and angiotensin-receptor blockers may prevent HF-related hospitalization in people with HFpEF albeit with modest impact (37,38). More recently, SGLT2 inhibitors have been shown to confer survival advantage in people with HFrEF independent of diabetes status, but evidence for those with HFpEF is pending (8). Given the high frequency of HFpEF in people with diabetes whose long-term prognosis is similar to their counterparts with HFrEF, there is an urgent need to develop new therapy with the aim to improve quality of life, reduce morbidity and mortality in this group.
The higher mortality in people with HFmrEF deserves further consideration. Mortality rates at 1, 5 and 10 years were higher in people with HFmrEF than HFpEF, and numerically higher than those with HFrEF.
These ndings are consistent with a prospective follow-up study of 1,405 consecutive individuals in whom 37-40% had diabetes attending heart failure clinic that have shown the strongest independent association between HFmrEF and the combined endpoint of all-cause mortality and HF-related hospitalization, particularly in those with frailty or fragility (39). Heart failure with mid-range LVEF may represent improved HFrEF on guideline-directed medical therapy (recovered HFrEF), worsened HFpEF with adverse cardiac remodelling and disease progression, or a discrete condition (11). In our study, the group with HFmrEF resembled those with HFrEF with respective to smoking status and BMI at baseline. In the regression analysis, the strength of the association between history of coronary heart disease and incident HF was comparable between HFmrEF and HFrEF. Additionally, HDL-cholesterol, involved in reverse cholesterol transport and has anti-atherogenic properties, was inversely related to incident HFmrEF. Taken together, these ndings point to a heavier atherosclerosis burden in people with HFmrEF, as also shown in other studies (23,24). The distinct risk pro le in this group may also contribute to the higher mortality rates when compared against people with HFpEF. Current European guidelines support the management of people with HFmrEF in a manner as for those with HFrEF (10). Availability of data and materials

Limitations
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.
analysis and approved the nal version. E.F., H.W., E.S.H.L., A.Y., E.C., A.P.S.K., R.C.W.M. and J.C.N.C. contributed to conception of the article and approved the nal version. All authors have agreed both to be personally accountable for the author's own contributions and to ensure that questions related to the accuracy or integrity of any part of the work, even ones in which the author was not personally involved, are appropriate investigated, resolved, and the resolution documented in the literature. A.O.Y.L. is the guarantor of this work, has full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Figure 1
Survival curve for mortality after incident heart failure according to heart failure subtypes

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. SupplementaryTablesAL.docx