Predicting Heart Failure Events in Patients with Coronary Heart Disease and Impaired Glucose Tolerance: Insights from the Acarbose Cardiovascular Evaluation (ACE) Trial

Malgorzata Wamil Great Western Hospitals NHS Foundation Trust John J. V. McMurray University of Glasgow Institute of Cardiovascular and Medical Sciences Charles A.B. Scott University of Oxford Ruth L. Coleman Churchill Hospital Yihong Sun China-Japan Friendship Hospital Eberhard Standl Helmholtz Zentrum Munchen Institut fur Diabetesforschung Lars Rydén Karolinska Institutet Rury R Holman (  rury.holman@dtu.ox.ac.uk ) University of Oxford https://orcid.org/0000-0002-1256-874X


Background
It is well known that individuals with type 2 diabetes (T2D) have a higher incidence of heart failure than those without.(1) Less is known, however, about the relationship between impaired glucose tolerance (IGT) and heart failure incidence, and whether treating IGT to delay or prevent development of T2D might reduce the incidence of heart failure. China has one of the highest rates of IGT and new-onset T2D in the world. (2) In addition, the China Heart Survey showed that the prevalence of IGT is signi cantly higher in individuals with coronary heart disease (CHD), compared with those without CHD.(3) Thus, the combination of CHD and IGT potentially identi es individuals at a greatly increased risk of developing heart failure.

Page 3/21
The Acarbose Cardiovascular Evaluation (ACE) trial was a randomized, double-blind, placebo-controlled, phase 4 cardiovascular outcome trial that randomized Chinese patients with CHD and IGT to acarbose or placebo designed primarily to assess whether acarbose could reduce the frequency of cardiovascular events and, secondarily, reduce the incidence of type 2 diabetes. (4)(5)(6) Acarbose is alpha-glucosidase inhibitor that reduces postprandial hyperglycemia by delaying carbohydrate digestion and absorption after meals. (7) In a meta-analysis of seven randomized controlled trials of acarbose, the incidence of heart failure was numerically but not signi cantly lower compared with placebo. (8) Using ACE data, we aimed to identify independent baseline predictors for a composite endpoint of incident hospitalisation for heart failure (hHF), or incident hHF or cardiovascular (CV) death in this population, and to assess the impact of post-prandial glucose lowering with acarbose on this outcome, compared with placebo. We also investigated the cumulative risk of hHF, recurrent hHF (given the competing risk of death), and death following hHF, to determine the total burden of hHF in this population.

Methods
The ACE trial design The design and results of the ACE trial have been published previously. (4,6) In brief, it was a prospective, double blind, randomized, multi-centre, cardiovascular intervention trial that enrolled 6,522 patients aged 50 years or more with established CHD and IGT. CHD was de ned as: acute myocardial infarction (MI), unstable angina (UA) or stable angina. IGT was diagnosed on a single 70g OGTT, de ned as a 2-hour plasma glucose (2HPG) value ≥7.8 but <11.1 mmol/l and a fasting plasma glucose (FPG) <7.0 mmol/l within six months prior to enrolment. The primary ACE trial outcome was a ve-point composite major adverse cardiovascular event (MACE5) outcome (de ned as any of CV death, non-fatal MI, non-fatal stroke, hospitalisation for UA or hHF). Pre-speci ed secondary endpoints included a diagnosis of T2D, allcause mortality and analysis of the individual MACE5 components.

Outcomes
The four pre-speci ed outcomes for this analysis were: i) Incident hHF; ii) A composite of hHF or CV death (to account for a portion of the competing risk of death); iii) Recurrent hHF; iv) Death following hHF. Incident hHF was de ned by the ACE trial Cardiovascular Event Adjudication Committee (CVEAC) charter as an emergency/unplanned admission to the hospital setting that resulted in ful lment of at least one of the following three criteria: i) Manifestation of new or worsening heart failure (HF); ii) New or additional therapy speci cally for the treatment of HF; iii) That the CVEAC should be satis ed that HF was the primary reason for hospitalisation.

Statistical analysis
Continuous baseline variables were summarized using means and standard deviations, except where noted otherwise. Categorical variables were summarized as counts and percentages. We identi ed potential independent predictors using a Cox proportional hazards model by rst testing pre-speci ed variables for univariate associations with each of the four outcomes. Nominally statistically signi cant variables (P<0.1) were then tested in a multivariate model and required to achieve P<0.05 to be considered as independent predictors. Potential risk factors examined were: age, sex, body mass index (BMI), race, waist circumference, smoking status, plasma creatinine, systolic blood pressure (SBP), haemoglobin, FPG, 2-HPG, HbA 1c , allocation to acarbose or placebo, and presence of prior CV events, atrial brillation (AF) or hypertension, as well as use of statins, aspirin or anti-hypertensive medications.
Recurrent hHF events were analysed using the Andersen-Gill model, a generalisation of the Cox proportional hazards model. (9) This model is appropriate as correlations among events for each individual are induced by measured covariates. Thus, the dependence is captured by specifying timedependent covariates such as number of previous events. To identify predictors of death following hHF we used logistic regression analysis, with P<0.1 indicating a univariate association and P<0.05 required for retention in a multivariable model.

Predictors of hHF
During median 5.0 years follow-up, 138 (2.1%) participants experienced one or more hHF events (0.46 rst events per 100 person-years follow-up). Baseline characteristics for those who did, or did not, experience hHF during the trial are listed in Table 1. Participants with, compared with those without, incident hHF were older, had lower mean haemoglobin, higher mean plasma creatinine values, and were more likely to have had a prior MI (54.4% vs. 41.3%, p=0.0021), UA (26.8% vs. 42%, p=0.0004), HF (19.6% vs. 3.4%, p<0.0001) or AF (16.7% vs. 3.6%, p<0.0001). Participants with incident hHF were also more likely at baseline to be taking ACE inhibitors/angiotensin receptor blockers, diuretics, nitrates or digitalis.
Univariate associations of variables with hHF are listed in Table 2. Of these, only higher age, higher plasma creatinine, lower haemoglobin, prior HF, prior MI and prior AF remained as statistically signi cant independent risk factors in multivariate models ( Table 3).

Predictors of hHF/CV death
The composite hHF/CV death outcome occurred in 393 (6.0%) ACE participants during median 5.0 years follow-up, triggered by 138 hHF events and 255 CV deaths. Univariate hHF/CV death predictors were: higher age, higher BMI, higher plasma creatinine, lower haemoglobin, prior HF, prior MI, prior UA, prior AF, prior stroke/TIA and hypertension ( Table 2). Of these, only age, plasma creatinine, prior HF, prior MI, prior AF and prior stroke remained as statistically signi cant independent risk factors in multivariate models ( Table 3). Measures of glycemia (FPG, 2HPG, HbA 1c ), lipids, SBP, BMI and waist circumference were not associated independently with hHF/CV death.

Recurrent hHF events
A total of 232 hHF events occurred in 138 participants during the study, of whom 40 (29.0%) experienced ≥2 events (94 recurrent events in total) with a median (IQR) of 3 (2-4) hHF events during follow-up.

Mortality following incident hHF
The proportion of participants who died following a rst hHF event was similar in those with or without prior HF (

Discussion
This is the rst study describing incident hHF and associated mortality in a large Asian population with CHD and IGT. In Chinese patients with CHD and IGT, we show that higher age, higher plasma creatinine, lower haemoglobin, prior HF, prior MI and prior AF were all predictive of incident hHF, and that higher age, higher plasma creatinine, prior HF, prior MI, prior AF and prior stroke were predictive of hHF/CV death. Death following hHF was predicted by older age and higher plasma creatinine. Allocation to acarbose 50 mg TID, compared with placebo, in the ACE trial, (4,6) showed no risk reductions for incident hHF, recurrent hHF or for hHF/CV death despite an 18% reduction in the risk of T2D. (5) Our within trial observations are in accord with those seen in the NAVIGATOR trial, which also showed no reduction in hHF in patients with IGT and CHD despite the improved glycaemic control achieved with nateglinide, a short-acting insulin secretagogue. (10,11) Lowering post-prandial glycemia has been suggested as a way to reduce HF risk, (12,13) but neither increasing post prandial insulin concentrations with nateglinide, nor using the insulin-sparing mode of action of acarbose to do so, reduced the risk of hHF in populations with IGT and CHD ( Figure 2). Similar ndings have been reported consistently with other glucose-lowering therapies (DPP-4 inhibitors and insulin) in patients with established T2D (10,(14)(15)(16) suggesting that therapeutic strategies focused on glycaemic control in T2D have a neutral effect on incident hHF or hHF-related death. Additionally, despite the reduced risk for new-onset diabetes seen with acarbose in the ACE trial, the within-trial risk of incident and recurrent hHF was unchanged. IGT is a precursor to T2D and has been shown to be a risk factor for cardiovascular morbidity. Thus, it is plausible that metabolic alterations leading to myocardial damage may start at an earlier stage, even before the formal diagnosis of T2D. In the ACE trial, 2.1% of participants experienced ≥1 episode of hHF, equating to 0.46 cases per 100 person-years of follow-up. (5) The hHF rate was similar in the NAVIGATOR trial (0.36 cases per 100 person-years of follow-up). For comparison, in the EXAMINE and the ELIXA trials, 3.6% and 8.3% of patients with acute coronary syndrome and established T2D experienced ≥1 episode of hHF respectively, equating to 1.09 and 1.94 cases per 100 person-years of follow-up. (17,18) The higher hHF rates observed in T2D patients, compared with IGT subjects, likely re ects their more advanced CHD and myocardial damage, which could be secondary to longer duration of abnormal glycaemic exposure.
MI is a major predictor of HF, as con rmed in our study, but given no CV bene t was evident in the threẽ 5-year duration randomized controlled trials that have compared intensive with conventional glycaemic control in T2D patients (19)(20)(21), it may not be surprising that improving glycaemic control over a similar timescale in ACE subjects with CHF and IGT had little impact on incident hHF. In the longer term, the UKPDS has demonstrated a legacy effect of the bene ts of earlier improved glycaemic control with sulfonylurea or insulin in people with newly-diagnosed T2D. The UKPDS showed a 15% lower risk of MI (P = 0.01) over a median follow-up of 16.8 years, but did not report HF rates. (22) Considering the increasing burden of HF events observed in people with IGT or T2D, early screening for those factors contributing to an individual's heightened risk needs careful consideration. Several studies have shown that prior HF is a powerful independent predictor of readmission (23, 24). Our study is in accord with these observations, con rming that in a Chinese population with CHD and IGT a prior history of HF was the most signi cant predictor of hHF. In our multivariable model, older age and increased plasma creatinine were also major predictors for incident hHF and subsequent case fatality. Kidney dysfunction is common in elderly populations and strongly associated with short-term and long-term outcomes in patients with HF (25).
In the high-risk ACE trial population with CHD and IGT, previous MI but not angina was independently associated with hHF. Of note, measures of glycaemic control, weight and blood pressure showed no association with hHF. This may re ect the fact that ACE trial participants were exceptionally wellmanaged with respect to administration of evidence-based secondary CV prevention therapies, as there was a protocol-mandated 4-week CV risk factor optimisation run-in period preceding randomisation.(4) Hence, additional HF risk factors such as age and plasma creatinine may become more prominent with good control of traditional CV risk factors such as hypertension, dyslipidaemia and smoking. However, an alternative interpretation of the ACE and the NAVIGATOR trial ndings might be that in populations with CHD and IGT already established on maximum CV secondary prevention, those with kidney disease have a poorer prognosis with a higher risk of hHF and CV death. In this context it is important to note novel glucose-lowering treatments with SGLT-2 inhibitors with their promising bene cial renal pro le have been shown to be of particular value in reducing the risk of HF in people with or without T2D. (26-30) Of note, there were some differences between the major predictors for hHF reported by the NAVIGATOR and the ACE trials. NAVIGATOR, which included 9,306 participants with IGT and either CV disease or CV risk, con rmed waist circumference among three major predictors of hHF (11). In the ACE trial by contrast, neither waist circumference nor BMI were independently associated with an increased hHF rates, possibly due to phenotypic differences in those recruited. Whilst NAVIGATOR recruited subjects with a mean BMI≥30 kg/m 2 , the mean BMI of the ACE Chinese participants was 25.4 kg/m 2 . This could suggest that the mechanisms leading to the development of HF in an Asian 'lean' cardiometabolic phenotype are distinct from those in Western patients whose IGT is closely linked to obesity.
Epidemiological data strongly support the role of AF in the development of HF in people with diabetes. The pathophysiology and risk factors for AF and HF are closely aligned, with affected patients usually being elderly with multiple comorbidities (31, 32). In our study, we found prior AF to be an independent predictor of incident hHF, as did NAVIGATOR. (11) These analyses have several limitations. Firstly, of the entire spectrum of heart failure events the ACE trial only examined hHF events. Restricting HF outcomes to hospitalised cases does not re ect contemporary management of HF, with an on-going shift towards community and ambulatory care. Secondly, left ventricular ejection fraction data was not collected so we do not know whether participants with incident hHF had a reduced or preserved ejection fraction. While T2D and IGT are associated with both types of HF, it is likely that adjudicated hHF events represent ischemic cardiomyopathy as the more predominant phenotype. Hence, more subtle changes expected in HF with a preserved ejection fraction, which are frequently observed in the setting of IGT and insulin resistance(33), have not been detected in this trial.
Thirdly, any conclusions regarding the predictive role of AF for HF in the ACE and NAVIGATOR populations need to circumspect given the low number of hHF events recorded in both trials.
In conclusion, in Chinese patients with CHD and IGT established on optimal secondary CV risk prevention, we identi ed higher age, increased plasma creatinine, lower haemoglobin, prior MI and prior AF as independent predictors of hHF. Higher age and increased creatinine predicted case fatality. Allocation to acarbose 50mg TID was not associated with reductions in hHF, hHF/CV death, recurrent hHF or death following hHF. These ndings will assist health care professionals to risk stratify patients with CHD and IGT and help identify those that might bene t from early screening for signs of HF. As this study focused on a Chinese population with a lower BMI compared with similar Western cohorts studied to date, any extrapolation of our results should consider the potential in uence of phenotypic differences on HF risk.      Kaplan-Meier plot of hospitalisation for heart failure (hHF)/CV death, split by participants allocated to acarbose or placebo.

Figure 2
Therapies that lower post-prandial glucose excursions examined in the ACE and NAVIGATOR trials had no impact on incident hospitalisation for heart failure (hHF) or hHF-related death.