Prognostic value of vasodilator stress perfusion cardiovascular magnetic resonance after inconclusive stress testing

Between December 2008 and January 2020, we conducted a single-centre longitudinal study with retrospective enrollment of consecutive patients with a first non-CMR inconclusive noninvasive stress test as the main indication for vasodilator stress perfusion CMR. Inconclusive stress test was defined by exercise electrocardiogram (ECG) or stress echocardiography or single photon emission computed tomography (SPECT) without positive or negative conclusion regarding the diagnosis of CAD [6, 13]. Two expert physicians reviewed the first stress test, using the definitions of positive or negative tests presented in Additional file 1, in accordance with previous studies [6]. Patients without angina or dyspnea on exertion underwent the first stress test during the work-up of known CAD, or because of relatively high CVD risk defined by the presence of at least 2 CVD risk factors (age > 50 years for men or > 60 years for women, diabetes, hypertension, smoking, dyslipidemia, family history of CAD and obesity defined by body mass index (BMI) ≥ 30 kg/m²). Exclusion criteria were: (i) contraindication to CMR (cerebral clips, metallic eye implant); (ii) contraindication to dipyridamole (severe asthma or chronic obstructive pulmonary disease, second- or third-degree atrioventricular block); (iii) known cardiomyopathy (e.g. hypertrophic, dilated, or infiltrative) and acute or chronic myocarditis; (iv) known allergy to gadolinium-based contrast medium; and (v) glomerular filtration rate < 30 ml/min/1.73 m². Clinical data were collected according to medical history and clinical examination on the day of stress CMR. All patients gave informed written consent for clinical CMR examination and enrolment in the clinical research study at baseline. The study was approved by the local ethic committee of our institution and conducted in accordance with the Declaration of Helsinki. This study followed the STROBE reporting guidelines for cohort studies.

The follow-up consisted of a clinical visit as part of usual care (72%) or by direct contact with the patient or the referring cardiologist (28%). Data collection was ended on June 2020. Cardiovascular events were checked by medical reports collected from the corresponding hospitals. The primary composite endpoint was the occurrence of at least one of the combined major adverse clinical events (MACE) defined as cardiovascular mortality or nonfatal myocardial infarction (MI). The secondary endpoints were all-cause mortality, hospitalizations for heart failure (HF), late coronary revascularizations and sustained ventricular arrhythmias. All these clinical events were defined according to standardized definitions [14, 15], and are detailed in Additional file 2. Annualized event rates were expressed as the number of patients having the event as a proportion of the number of patients at risk divided by the number of patient-years follow-up. The adjudication of the cause of death between cardiovascular and non-cardiovascular was performed by two senior cardiologists (TP and MK), with a third cardiologist (JG) to reach final consensus. For patients who underwent percutaneous coronary intervention (PCI) < 90 days after index examination, the nine peri-procedural events (seven nonfatal MI or two CVD mortality) were excluded from the analysis.

The detailed stress CMR protocol has been previously published [16, 17], and is described in Additional file 3. Briefly, CMR was performed on a 1.5T CMR scanner (Siemens Healthineers, Erlangen, Germany). Vasodilation was induced with dipyridamole injected at 0.84 mg/kg during 3 min [18]. Then, a bolus of gadolinium-based contrast agent (Dotarem^®, Guerbet LLC, France, 0.1 mmol/kg) was injected at a rate of 5.0 ml/s. Stress perfusion imaging was performed using an ECG-triggered saturation-prepared balanced steady-state free-precession sequence. A series of six slices (four short-axis views, a 2-chamber, and a 4-chamber view) were acquired every other heartbeat. No motion compensation was performed before analysis. Ten minutes after contrast injection, breath-hold contrast-enhanced 3D T1-weighted inversion-recovery gradient-echo sequence was acquired to detect late gadolinium enhancement (LGE). CMR sequence parameters are detailed in Additional file 4.

Left ventricular (LV) end-diastolic volume (LVEDV), end-systolic volume (LVEDV) and systolic function were quantified on the short-axis cine stack. Stress perfusion and LGE images were evaluated according to the American Heart Association 17-segment model [19]. The analysis of perfusion images was done visually by two experienced physicians blinded to clinical and follow-up data. Inducible ischemia was defined as a subendocardial perfusion defect that (1) occurred in at least one myocardial segment, (2) persisted for at least three phases beyond peak contrast enhancement, (3) followed a coronary distribution, (4) in the absence of co-location with LGE in the same segment [11, 12]. An MI was defined by subendocardial or transmural LGE [20]. A myocardial segment was considered viable if LGE thickness was < 50% and nonviable when LGE thickness was ≥ 50% of the myocardial wall [21]. The total number of ischemic and LGE segments was assessed for each patient.

CMR-related coronary revascularization was defined as all procedures (coronary artery bypass grafting [CABG] or PCI performed within 90 days after stress CMR. All patients were treated with optimal medical therapy according to current guidelines in patients with chronic coronary syndromes [2]. Decision-making regarding initial coronary revascularization was based on the presence of myocardial ischemia in at least two contiguous segments in symptomatic patients, and the choice between PCI or CABG was made by the Heart Team of the Institutions. All clinical data, CMR parameters and CMR-related coronary revascularization were prospectively recorded into a dedicated database (Clinigrid software, Hemolia, France).

Continuous variables were expressed as mean ± standard deviation (SD), categorical variables as frequency with a percentage, and follow-up as a median and interquartile range (IQR). Patients with and without inducible ischemia were compared using the Student’s t-test or the Wilcoxon rank-sum test for the continuous variables and the Chi-square or Fisher’s exact test for the categorical variables. Cumulative incidence rates of the outcomes were estimated using the Kaplan–Meier method and compared with the log-rank test. The data of patients who were lost to follow-up were censored to the time of the last contact. Cox proportional hazards methods were used to identify the predictors of MACE among patients with and without inducible ischemia. The assumption of the proportional hazards ratio (HR) was verified. To assess the incremental prognostic value of both the inducible ischemia and CMR-related coronary revascularization, different multivariable models were used, as follows:

The discriminative capacity of each model for predicting MACE was determined according to Harrell’s C-statistic before and after the addition of inducible ischemia. The additional predictive value of the inducible ischemia was calculated using Harrell’s C-statistic increment. In addition, the global χ² statistic was calculated for models with or without stress CMR parameters and compared using the likelihood ratio (LR) test for predicting MACE.

In the competitive risk analysis, cumulative incidence functions were used to display the proportion of patients with the event of interest or the competing event (nonfatal MI or CV mortality) as time progressed, and the Fine and Gray regression model was used for the sub-distribution hazard. A two-tailed p-value < 0.05 was considered statistically significant. Statistical analysis was performed using R software, version 3.3.1 (R Project for Statistical Computing, Vienna, Austria).

Of the 35,280 patients referred for stress CMR during the inclusion period, 1584 (4.5%) patients were referred for dipyridamole vasodilator stress CMR because of a first inconclusive noninvasive stress test. Among those, 1563 (98.7%) completed the stress protocol, as detailed in the flowchart (Fig. 1). Of the 1563 patients who successfully underwent stress CMR, the diagnosis of ischemia was inconclusive in 24 patients (1.5%) due to nondiagnostic image quality, arrythmias or artifacts. Out of these 1563 patients, 61 failed to respond to dipyridamole injection as assessed by the rate-pressure product (3.9%). No patient died during or shortly after CMR, and detailed safety results are presented in Additional file 5. Overall, 1402 patients completed the clinical follow-up and constituted our study cohort. Baseline subject characteristics and baseline CMR data are shown in Table 1. Among those 1402 patients (66.7% male, 69.5 ± 11.0 years), 58.4% had dyslipidemia, 57.6% had hypertension, 32.7% had diabetes mellitus, 30.7% had obesity, 27.9% had a family history of CAD and 24.0% were smokers. Overall, 727 (51.9%) patients had known CAD. Of note, 247 (17.6%) of patients were in atrial fibrillation or supraventricular arrhythmia. Regarding the first inconclusive stress test, 702 (50.1%) patients had a prior inconclusive stress echocardiography, 612 (43.7%) a prior inconclusive SPECT (147 dipyridamole SPECT) and 88 (6.3%) a prior inconclusive exercise ECG testing. The two main reasons for an inconclusive stress test were poor image quality (68%) and sub-maximal exercise (29%).

Among the 1402 patients, 485 (34.6%) were asymptomatic without angina or dyspnea. These asymptomatic patients were older and had a higher rate of known CAD (87.0% vs 33.3%, p < 0.001), diabetes mellitus, hypertension and smoking than symptomatic patients with angina or dyspnea (Additional file 6). Consistently, asymptomatic patients presented a higher CV risk than symptomatic using the ESC SCORE 10-year risk for fatal CAD [22] (3.4 [2.2–6.9] vs 2.1 [0.7–5.5] %, p < 0.001).

The study cohort had a mean LV ejection fraction (LVEF) of 50.4 ± 12.1%. LGE was present in 556 (39.7%) and presence of inducible ischemia was detected in 414 (29.5%) patients with a mean extent of 2.3 ± 1.2 segments (Additional file 7). Of note, the rate of inducible ischemia in the overall population of 35,280 patients referred for stress CMR during the inclusion period was 12.4%.

Patients with inducible ischemia were older, more frequently males and had a higher rate of diabetes mellitus, hypertension, dyslipidemia and history of peripheral atheroma than patients without inducible ischemia (Table 1). Of 414 patients with ischemia, 381 (92.0%) had a coronary angiography. Among those, 323 (84.8%) underwent CMR-related coronary revascularization (317 [98.1%] PCI and 6 [1.9%] CABG).

During a median follow-up of 6.5 (IQR 5.6–7.5) years, there were 197 (14.1%) MACE, including 141 (10.1%) CV mortality and 73 (5.2%) nonfatal MI. Furthermore, 255 all-cause mortality (18.2%), 106 HF hospitalisations (7.6%), 99 late coronary revascularizations (7.1%), and 34 sustained documented ventricular arrhythmias (2.4%). Annualized event rates were 4.4% for MACE, 2.4% for CVD mortality, and 4.6% for all-cause mortality. Patients without inducible ischemia or LGE had the lowest annualized rate of MACE (2.1%/year), whereas the annualized rate of MACE was greater for patients with inducible ischemia without or with LGE (9.0%/year and 9.3%/year respectively, both p < 0.001) (Additional file 8). The annualized rate of MACE was lower in patients without inducible ischemia compared to patients with mild, moderate, or severe ischemia (2.4%/year vs. 4.2%/year, 20.7%/year and 27.9%/year, respectively; p_trend < 0.001) (Fig. 2). In addition, the prognostic value of the presence of inducible ischemia was consistent irrespective of age (Additional file 9).

In univariable analysis, age, male gender, hypertension, diabetes mellitus, dyslipidemia, known CAD, LVEF value, LV end-diastolic volume (LVEDV) and end-systolic volume (LVESV) indexed (LVEDVI, LVESVI, respectively) and the presence and extent of both inducible ischemia and LGE were all significantly associated with MACE (Table 2). Using Kaplan–Meier analysis, the presence of inducible ischemia was associated with the occurrence of MACE (HR: 2.88, 95% CI 2.18–3.81, Fig. 3) and CV mortality (HR: 2.44 95% CI 1.75–3.40; both p < 0.001), and the same finding was observed for LGE (HR: 1.46, 95% CI 1.16–1.89; and HR: 1.38, 95% CI 1.06–1.77, both p < 0.001; respectively). In the overall population, the CMR-related coronary revascularization was associated with the occurrence of MACE and CV mortality (HR: 2.43 95% CI 1.83–3.22; and HR: 2.04 95% CI 1.46–2.86; respectively both p < 0.001). The prognostic value of inducible ischemia to predict MACE was consistent for both women and men (Additional file 10); and both asymptomatic and symptomatic patients (Additional file 11).

In the overall population, inducible ischemia was also associated with nonfatal MI (HR: 5.04, 95% CI 3.09–8.21; p < 0.001), late elective coronary revascularization (HR: 2.61, 95% CI 1.76–3.87; p < 0.001), ventricular arrythmias (HR: 2.76, 95% CI 1.41–5.42; p = 0.003), and all-cause mortality (HR: 1.73, 95% CI 1.35–2.22; p < 0.001) (Additional file 12. The prognostic value of inducible ischemia remained consistent in different subgroups of clinical interest, such as diabetics and non-diabetics, obese and non-obese, and regardless of LVEF value or the presence of LGE (Fig. 4).

In multivariable stepwise Cox regression analysis, age, male gender, the presence of inducible ischemia and the number of ischemic segments were independent predictors of a higher incidence of MACE (HR: 1.05, 95% CI 1.03–1.06, p < 0.001; HR: 1.47, 95% CI 1.05–2.06, p = 0.027; HR: 2.53, 95% CI 1.89–3.40, p < 0.001; and HR: 1.58, 95% CI 1.47–1.71, p < 0.001; respectively) (Table 3). In competitive risk analysis, the presence of inducible ischemia was independently associated with nonfatal MI and CV mortality (HR: 4.26, 95% CI 2.60–6.89, p < 0.001 and HR: 1.80, 95% CI 1.27–2.56, p < 0.001; respectively) (Fig. 5 and Additional file 13).

Using Kaplan–Meier analysis only in patients with inducible ischemia, CMR-related coronary revascularization was not associated with the occurrence of MACE (HR: 0.95, 95% CI 0.60–1.48; p = 0.81) (Fig. 6). Inducible ischemia remained independently associated with MACE in patients without or with coronary revascularization (HR: 2.88, 95% CI 1.82–4.56; and HR: 2.44, 95% CI 1.79–3.34, both p < 0.001; respectively) (Table 3).

For the prediction of MACE, the baseline C-statistic value was 0.63 (95% CI 0.60–0.68) for model 1 with stepwise variable selection. The addition of inducible ischemia or the number of ischemic segments significantly improved the C-statistic to 0.73 (95% CI 0.66–0.79; C statistic improvement for model 1: 0.10) and 0.75 (95% CI 0.69–0.81; C statistic improvement for model 1: 0.12), respectively. Furthermore, the addition of both presence of inducible ischemia and CMR-related coronary revascularization did not improve the C-statistic compared to the model with only the presence of inducible ischemia (C-statistic 0.73 for both) (Table 4).