Additional prognostic value of heart rate reserve over left ventricular contractile reserve and coronary flow velocity reserve in diabetic patients with negative vasodilator stress echocardiography by regional wall motion criteria

Abstract

Aims 

In diabetic patients, a blunted left ventricular contractile reserve (LVCR) and/or a reduced coronary flow velocity reserve (CFVR) identify patients at higher risk in spite of stress echocardiography (SE) negative for ischaemia. Cardiac autonomic dysfunction contributes to risk profile independently of inducible ischaemia and can be assessed with heart rate reserve (HRR). We sought to assess the added prognostic value of HRR to LVCR and CFVR in diabetic patients with non-ischaemic SE.

Methods and results 

Six-hundred and thirty-six diabetic patients (age 68 ± 9 years, 396 men, ejection fraction 58 ± 10%) with sinus rhythm on resting electrocardiogram underwent dipyridamole SE in a two-centre prospective study with assessment of wall motion, force-based LVCR (stress/rest ratio, normal value > 1.1), CFVR of the left anterior descending coronary artery (stress/rest ratio, normal value >2.0), and HRR (stress/rest ratio, normal value >1.22). All-cause death was the only considered endpoint. During a median follow-up of 39 months, 94 (15%) patients died. Independent predictors of death were abnormal CFVR [hazard ratio (HR) 1.59, 95% confidence interval (CI) 1.0–2.52, P = 0.05], reduced LVCR (HR 1.76, 95% CI 1.15–2.69, P = 0.009), and blunted HRR (HR 1.92, 95% CI 1.24–2.96, P = 0.003). Eight-year death rate was 9% for patients with triple negativity (n = 252; 40%), 18% for those with single positivity (n = 216; 34%), 36% with double positivity (n = 124; 19%), and 64% for triple positivity (n = 44; 7%) (P < 0.0001).

Conclusion 

Diabetic patients with dipyridamole SE negative for ischaemia still may have a significant risk in presence of an abnormal LVCR and/or CFVR and/or HRR, which assess the underlying myocardial, microvascular, and cardiac autonomic dysfunction.

Clinical trials

Gov Identifier NCT 030.49995.

Introduction

The prevalence of diabetes mellitus (DM) continues to rise, with an estimated >600 million individuals developing DM worldwide by 2045.¹ DM confers a two-fold excess risk of vascular outcomes independently of other risk factors, as shown by a meta-analysis of 102 prospective studies.² Risk stratification is effectively performed in patients with chronic coronary syndromes by functional testing and stress imaging,³ but the results have been much less impressive in DM. Recent 2019 European Society of Cardiology guidelines recommend functional imaging with exercise or pharmacological stress with a class of evidence IIb (‘may be considered’), level of evidence B.⁴ There is a pathophysiological explanation behind the unsatisfactory performance of functional testing in DM. DM determines multiple and complex vulnerabilities affecting outcome, but the standard approach to functional testing with stress echocardiography (SE) is based on regional wall motion abnormalities (RWMA) and underlying flow-limiting and ischaemia-producing coronary artery disease (CAD). This reductionistic approach identifies the patient risk with coronary stenosis severity and misses other targets of prognosis-limiting cardiovascular damage, independent of CAD, and particularly relevant in DM, such as abnormalities of myocardial function,⁵ coronary microcirculation,⁶ and cardiac autonomic system.⁷ Recently, SE was enriched by a more comprehensive approach allowing to expand functional testing to myocardial function (and underlying fibrosis or necrosis) with evaluation of left ventricular contractile reserve (LVCR), coronary microcirculation with coronary flow velocity reserve (CFVR) of the left anterior descending coronary artery (LAD), and cardiac autonomic dysfunction with assessment of heart rate reserve (HRR).⁸ LVCR,⁹ CFVR,^10,11 and HRR¹² refine outcome stratification provided by RWMA in DM, but the relative and possibly incremental value of the combined assessment of the four criteria remain unknown.

In this two-centre, prospective, longitudinal study, we assessed the prognostic contribution of combined evaluation of CFVR, LVCR, and HRR in diabetic patients with negative SE by wall motion criteria, with or without resting RWMA.

Methods

Patients

The initial population comprised 788 patients with DM prospectively enrolled at two Italian cardiology institutions (Lucca and Benevento) from July 2009 to December 2018 as a part of the Echo Persantine International Cooperative-Flow Reserve study until 1 September 2016,¹³ and as a part of the International Stress echo 2020 study from 1 September 2016 onwards.⁸ All patients underwent dipyridamole SE with assessment of CFVR of LAD by transthoracic Doppler, LVCR, and HRR. Of 788 patients, 40 (5%) were excluded for atrial fibrillation at rest making the interpretation of HRR difficult. Of the remaining 748 patients, 43 (6%) were excluded for inadequate acoustic window precluding satisfactory imaging of LAD flow Doppler (for CFVR assessment), 13 (2%) for inadequate acoustic window precluding satisfactory imaging of endocardial borders (for LVCR assessment), 9 (1%) for side effects requiring premature test interruption, and 47 (6%) for test positivity by wall motion criteria. Accordingly, 636 [396 (62%) men; mean (±SD) age 68 ± 9 years] with a negative SE by wall motion criteria and complete follow-up data formed the study group. Of the 636 patients evaluated in this study, 273 (43%) were previously reported (with a substantially shorter follow-up) in a previous study on triple imaging (wall motion, CFVR, and LVCR, without HRR) in diabetics.¹³

Indication for SE was suspected CAD in 347 (55%) and risk stratification of known CAD (i.e. history of myocardial infarction, coronary revascularization, and/or angiographic evidence of ≥50% diameter coronary stenosis) in 289 (45%) subjects. Exclusion criteria were significant valvular or congenital heart disease, and prognostically relevant non-cardiac diseases (cancer, end-stage renal disease, or severe obstructive pulmonary disease). According to individual needs and physician’s choices, 247 (39%) patients were evaluated after anti-ischaemic drugs had been discontinued, and 389 (61%) patients were evaluated under anti-ischaemic therapy (Table 1). Methylxanthine-containing drugs or beverages were discontinued at least 24 h before testing. SE data were collected and analysed by stress echocardiographers not involved in patient care. Written informed consent was obtained from all patients before testing. The study protocol was reviewed and approved by the institutional ethics committees, in its last version as a part of the more comprehensive SE 2020 study (148-Comitato Etico Lazio-1, 16 July 2016; Clinical trials. Gov Identifier NCT 030.49995).

Transthoracic echocardiography

We used commercially available ultrasound machines (iE 33 Philips, Bothell, WA and GE Vingmed Vivid 7, Horten, Norway) equipped with multifrequency phased-array sector scan probe and with second harmonic technology. All patients underwent comprehensive transthoracic echocardiography at rest. All measurements were taken by certified cardiologists according to the recommendations of the American Society of Echocardiography and European Association of Cardiovascular Imaging.¹⁴ Patients underwent SE with dipyridamole (0.84 mg/kg over 6 min) according to the recommended protocols.^15,16 Electrocardiogram and blood pressure were monitored continuously. Criteria for interrupting the test were severe chest pain, diagnostic ST-segment shift, excessive blood pressure increase (systolic blood pressure ≥240 mmHg, diastolic blood pressure ≥120 mmHg), limiting dyspnoea, maximal predicted heart rate, significant arrhythmias, or limiting side effects.¹⁷ The quadruple ACDE protocol of SE was used when each laboratory had completed the upstream quality control process.¹⁸ Echocardiographic imaging was performed from parasternal long-axis view, short-axis view, and apical four-, three-, and two-chamber views, using conventional 2D echocardiography. Step A included assessment of RWMA. Wall motion score index (WMSI) was calculated in each patient at baseline and peak stress, in a four-point score ranging from 1 (normal) to 4 (dyskinetic) in a 17-segment model of the left ventricle.¹⁵ Step C of protocol included the force-based assessment of LVCR as the stress/rest ratio of force, calculated as systolic blood pressure/end-systolic volume from biplane Simpson’s method (16). Step D of protocol included pulsed-Doppler assessment of CFVR, defined as the ratio between hyperaemic peak and basal peak diastolic coronary flow velocities in mid-distal LAD.¹⁰ The procedure for acquisition between centres was standardized through a web-based learning module before starting data collection. All readers (one for each centre) underwent a quality control as previously described for RWMA, end-systolic volume, and CFVR. Imaging-independent Step E of the ACDE protocol included electrocardiogram-based assessment of HRR as peak/rest ratio of heart rate as an index of cardiac autonomic dysfunction.¹⁹

SE positivity criteria

All positivity criteria were determined a priori. The A criterion was considered positive in presence of stress-induced RWMA (WMSI stress > rest) when at least two adjacent segments of the same vascular territory of the left ventricle showed an increment of at least one point of WMSI during SE. The C criterion was considered positive in presence of force-based LVCR ≤ 1.1.⁸ The D criterion was considered positive in presence of CFVR ≤2.0.¹³ The E criterion was considered positive in presence of HRR ≤ 1.22.¹⁹

Follow-up data

Deaths were identified from the national health service database. Non-deceased participants were contacted directly. Follow-up data were obtained from review of the patient’s hospital record, personal communication with the patient’s physician and review of the patient’s chart, a telephone interview with the patient or a patient’s close relative conducted by trained personnel, a staff physician visiting the patients at regular intervals in the outpatient clinic. Mortality was the only endpoint. In order to avoid misclassification of the cause of death,^20,21 overall mortality was considered. Coronary artery bypass surgery and angioplasty were also registered; however, follow-up was not censored at the time of revascularization.

Statistical analysis

Continuous variables are expressed as mean ± SD. Correlations between CFVR, LVCR, or HRR were estimated with Pearson’s coefficients. Event rates were estimated with Kaplan–Meier curves and compared using the log-rank test. Univariable analyses by Cox proportional hazards models were performed to assess the association between each candidate variable and outcome. Non-proportionality of hazard was assessed using the Schoenfeld test. The primary endpoint was the time-to-event analysis by a multivariable Cox proportional hazards model. Hazard ratios (HRs) with the corresponding 95% confidence interval (CI) were estimated. Selection of independent predictors was performed both for logistic and proportional hazards model with a backward approach using a P value of 0.10 as threshold for inclusion in the model. In addition, clinical variables and the sequential steps of CFVR, LVCR, and HRR were sequentially included into the model and the global χ² was calculated after each step. A significant increase after the addition of further variables indicated incremental prognostic value. All analyses were two-sided.

Statistical significance was set at P < 0.05. All statistical calculations were performed using SPSS for Windows, release 20.0 (Chicago, IL, USA).

Results

The main clinical and echocardiographic features in study patients are reported in Table 1.

By selection no patient has inducible RWMA during stress, but 169 (27%) had RWMA at rest, 214 (34%) had prior PCI, and 288 (45%) had known CAD at study entry (Table 1).

Stress echocardiography

No major complications occurred during the test. Resting echocardiographic left ventricular ejection fraction in the entire study group was 58 ± 10%. CFVR was ≤2 in 193 (30%) patients, while LVCR was ≤1.1 in 168 (26%) patients, and HRR was ≤1.22 in 235 (37%) patients. The three parameters were all abnormal in 44 (7%) patients and all normal in 252 (40%) patients. CFVR was not related to either HRR (r = 0.02; P = 0.66) or LVCR (r = 0.00; P = 1.0). HRR was weakly related to LVCR (r = 0.16, P < 0.0001).

Three-hundred and sixty-one (57%) patients were tested on and 275 (43%) patients were tested off β-blocker therapy. The former had lower resting (66 ± 10 vs. 73 ± 10 bpm; P < 0.0001), and peak heart rate (85 ± 12 vs. 93 ± 15 bpm; P < 0.0001). However, HRR was comparable between the two groups (1.29 ± 0.15 vs. 1.29 ± 0.17; P = 0.92). In addition, HRR ≤1.22 was detected in 129 (36%) subjects investigated on and in 106 (38%) subjects investigated of β-blocker therapy, with no intergroup difference (P = 0.47).

The imaging time was increased by on average 3 min for CFVR (mostly for finding LAD flow at rest, it is usually easier and faster at peak stress). The imaging time considerably declined with increasing experience of the operators from about 5 min in the early experience (100 cases) to an average of 2 min in the more recent experience. No additional imaging time was required for LVCR since the same images and projections used for RWMA are used. No additional imaging time is required for imaging-independent HRR.

Analysis time (off-line) was always <1 min (including rest and stress assessment) for CFVR, 2–3 min for LVCR (requiring end-diastolic and end-systolic volume measurements also needed for ejection fraction), and of a few (<10) seconds for HRR since rest and peak stress values are displayed automatically in the echo monitor and electrocardiographic tracing.

Figure 1 shows an example of a patient with normal response for CFVR, LVCR, and HRR, and an example of a patient with abnormal response for CFVR, LVCR, and HRR.

Follow-up events

During a median follow-up of 39 months (first quartile 25, third quartile 79 months), there were 94 (15%) deaths. In addition, 77 (12%) patients underwent coronary revascularization (13 surgery and 64 angioplasty).

Outcome prediction

The 8-year mortality was 39% in patients with abnormal CFVR, 38% in patients with abnormal LVCR, and 38% in those with abnormal HRR. All-cause death was significantly lower in patients with normal CFVR (16%), normal LVCR (16%), or normal HRR (15%) (Figure 2). Eight-year death rate associated to abnormal and normal HRR was 37% and 20%, respectively (P = 0.0003) in patients investigated on β-blocker therapy, and 38% and 12% (P < 0.0001) in those investigated off β-blocker therapy.

Univariate and multivariate prognostic predictors are reported in Table 2. Multivariable indicators of all-cause death were age, previous CABG, and the following SE variables: HRR ≤1.22 (HR 1.92, 95% CI 1.24–2.96; P = 0.003), LVCR ≤1.1 (HR 1.76, 95% CI 1.15–2.69; P = 0.009), and CFVR ≤2 (HR 1.59, 95% CI 1.00–2.52; P = 0.05) (Table 2).

The 8-year mortality in the entire study population was 24% (Figure 2). It was 9% in patients with normal CFVR, LVCR, and HRR, two-fold higher in patients with 1 abnormal index only, four-fold higher in patients with 2 abnormal indices, and seven-fold higher in patients with 3 abnormal indices (Figure 3).

By using an interactive stepwise procedure, the global X² of the clinical model for mortality was 56.4 (P < 0.0001); sequential inclusion of CFVR ≤2, LVCR ≤1.1, and HRR ≤1.22 increased it, respectively, to 66.8 (18% increase; P < 0.0001), 74.9 (12% increase; P < 0.0001), and 84.1 (12% increase; P < 0.0001) (Figure 4).

Discussion

Patients with DM still have a significant risk of all-cause death even in the absence of stress-induced ischaemia, on average 24% at 8 years. If all other SE indices (CFVR, LVCR, and HRR) are normal, this value drops to 9%. If all other indices are abnormal, this value rises to 64%. The comprehensive SE protocol introduces no relevant loss of feasibility and only negligible increase in imaging time to the standard SE. It is especially useful in contemporary patients in whom the positivity rate based on inducible ischaemia was only 6% in our population, consistently with the dramatic drop-off in positivity rate observed during the last decades with SE, probably for the rise of population with low or very low pre-test probability and increased number of patients with full anti-ischaemic therapy which decreases the sensitivity of test based on ischaemia induction.^22–24 Our data are consistent with previous evidences showing that there are different and unpredictable targets of cardiovascular involvement in the diabetic heart, with selective or multiple impairment of epicardial artery stenoses (determining stress-induced RWMA), myocardial dysfunction, coronary microvascular impairment, or reduction in cardiac sympathetic reserve. A limited assessment of myocardial ischaemia misses the equally important extra-ischaemic determinants of prognosis-limiting cardiovascular involvement in diabetic patients. Furthermore, a comprehensive assessment is best suited to achieve one of the goals of functional testing beyond CAD diagnosis and outcome stratification, that is the assessment of underlying functional heterogeneity which must be uncovered to pave the way to personalized treatment.

Comparison with previous studies

Our data are largely consistent with previous studies. In DM, the annual hard-event rate is still 3–4% (relatively high) in patients with negative SE for RWMA,^25,26 decreases to 2.5% in patients with either normal CFVR,²⁷ or normal LVCR,¹¹ and ∼1.2% when both CFVR and LVCR are normal.¹³ HRR was shown to be a predictor of death independent of inducible ischaemia (assessed with myocardial perfusion imaging or SE) in diabetic patients studied with vasodilator stress or exercise.^12,28 Of some interest, in our experience HRR was unaffected by β-blocker use and therapy did not affect the prognostic value of HRR. No previous study compared the four variables—RWMA, CFVR, LVCR, and HRR—in the same patients evaluated during a single stress in the same setting.

Clinical implications

The results of functional testing have a practical and a potential therapeutic implication. From the practical viewpoint, a comprehensive risk assessment based on multiple parameters is more effective than the classical assessment based only on RWMA which misses important vulnerability of the patient outside and beyond CAD and ischaemia. From the therapeutic viewpoint, a more integrated functional testing is the conceptual prerequisite for targeted intervention focused on the specific vulnerabilities of the patient, who can be at increased risk for myocardial (LVCR) or coronary microcirculatory (CFVR) or cardiac autonomic (HRR) dysfunction. The overall risk may indicate the level of required metabolic and glycaemic control, but the specific type of positivity may drive specific cardiovascular therapy and might actually be used by a clinician to assist with patient management.

Study limitations

HbA1c is an important part of the metabolic characterization of the diabetic patient and microalbuminuria is now an integral component of risk assessment in patients with DM.⁴ HbA1c data should have been obtained within 3 months of SE to reflect the actual diabetic control at the time of stress performance, but this important parameter was not systematically available in our population of mostly outpatients and therefore could not be entered in the data bank. We analysed all-cause death, without further separation of specific causes of death. The use of overall mortality is an objective and unbiased endpoint.^20,21 Cardiovascular mortality could be not reliably separated from all-cause mortality in our data set. Different parameters might affect differently different endpoints. In particular, HRR abnormality determines electrical instability and an increased vulnerability to sudden death.^29,30 An abnormal CFVR is more linked to ischaemic events and a reduced LVCR may portend cardiac decompensation.⁶ Altered autonomic and coronary microvascular function also contribute to onset and progression of myocardial damage in DM.^6,30,31 However, the interplay of these factors in determining outcome is complex and largely unpredictable, and many of these risk factors for cardiac death are also potentially linked to non-cardiac causes of death. CFVR is associated to higher risk not only of cardiovascular death but also of cancer and non-cardiovascular, non-cancer deaths.³² Multi-parametric functional testing is rather an assessment of global risk, and all-cause death is a methodologically solid, statistically convenient and biologically plausible endpoint. Cardiac autonomic neuropathy is a complex disease and involves more than simply an impaired HRR.³³ In addition, abnormalities in resting heart rate are a determinant of reduced HRR, which is associated with higher values of resting heart rate as previously shown. Nevertheless, we used the HRR for its striking simplicity and since it was the only parameter available in the data set of SE registries. In our population, mean resting heart rate provided no prognostic information at univariate analysis. The high rate (57%) of patients tested on β-blockers may have affect this result. The used protocol is a significant advancement over conventional SE approach based only on RWMA but still is not state-of-the-art, since the current version of SE uses the ABCDE protocol, with B step added to image extravascular lung water with the four-site simplified scan in the third intercostal space. This step adds pathophysiologic and prognostically relevant information over the other steps, but it was not available in the recruitment years of the present study (from 2009 to 2018) since it was introduced in our SE laboratories only starting September 2016.³⁴

Conclusions

Diabetic patients with dipyridamole SE negative for ischaemia still may have a significant risk in presence of an abnormal LVCR and/or CFVR and/or HRR, which assess the underlying myocardial, microvascular, and cardiac autonomic dysfunction. These three parameters can be easily assessed during a functional stress test without a significant extension of the imaging and analysis time and show independent and incremental value in stratifying outcome.

Conflict of interest: none declared.

References