The use of cardiovascular imaging in prognostic stratification: Table 1

Several cardiologists investigate the prognostic value of novel measurements obtained with cardiovascular imaging. These measurements are most often related to remodelling or cardiac function. For this purpose, they need to develop an extensive database. In addition, they have to provide a precise and reliable long-term follow-up. It is mandatory to have big numbers to reliably enter multiple variables in Cox proportional-hazards regression models or logistic regressions. This turns out to be a considerable venture. It takes many years before prognostic results become available. The size of the database and the quality of the observations both matter, to provide prognostic information, which is additive to what we have known. We fortunately have several such ventures on both sides of the Atlantic. For showing that a new measurement is useful for improving the prognostic prediction of a subject or cardiac patient, the investigator most often adds this measurement to clinical parameters, laboratory measurements, and baseline echocardiographic parameters (Table 1). The investigator has to demonstrate that he is capable of improving the explained variance, increasing the area under the curve (AUC), and net classification improvement (NRI) for nested models. Cardiovascular imaging and prognosis General population Patient population Hypertension, heart failure, or myocardial infarction Clinical parameters Age, sex, race, height, weight, blood pressure and components, and heart rate NYHA class Lifestyle, co-morbidities (diabetes, frailty, …), and hospitalizations Laboratory tests Haematocrit, eGFR, sodium, potassium, fasting glucose, HbA1c, lipid fractions, albuminuria, and inflammation NT-pro-BNP or BNP Established parameters LV mass index, LV volume indexes, LA volume index, and EF Tissue Doppler parameters: s′, e′, a′ LV filling pressures evaluated with E/e′ Tricuspid regurgitation velocity Right ventricular function Novel parameters Myocardial strain and strain rate Arterial function and arterial wave reflections Ventricular–arterial coupling Other … Cardiovascular morbidity Hospitalization for myocardial infarction, heart failure, and cerebrovascular accident Cardiovascular mortality Total mortality General population Patient population Hypertension, heart failure, or myocardial infarction Clinical parameters Age, sex, race, height, weight, blood pressure and components, and heart rate NYHA class Lifestyle, co-morbidities (diabetes, frailty, …), and hospitalizations Laboratory tests Haematocrit, eGFR, sodium, potassium, fasting glucose, HbA1c, lipid fractions, albuminuria, and inflammation NT-pro-BNP or BNP Established parameters LV mass index, LV volume indexes, LA volume index, and EF Tissue Doppler parameters: s′, e′, a′ LV filling pressures evaluated with E/e′ Tricuspid regurgitation velocity Right ventricular function Novel parameters Myocardial strain and strain rate Arterial function and arterial wave reflections Ventricular–arterial coupling Other … Cardiovascular morbidity Hospitalization for myocardial infarction, heart failure, and cerebrovascular accident Cardiovascular mortality Total mortality eGFR, estimated glomerular filtration rate; BNP, brain natriuretic peptide; LV, Left ventricular; LA, left atrial; EF, LV ejection fraction. Cardiovascular imaging and prognosis General population Patient population Hypertension, heart failure, or myocardial infarction Clinical parameters Age, sex, race, height, weight, blood pressure and components, and heart rate NYHA class Lifestyle, co-morbidities (diabetes, frailty, …), and hospitalizations Laboratory tests Haematocrit, eGFR, sodium, potassium, fasting glucose, HbA1c, lipid fractions, albuminuria, and inflammation NT-pro-BNP or BNP Established parameters LV mass index, LV volume indexes, LA volume index, and EF Tissue Doppler parameters: s′, e′, a′ LV filling pressures evaluated with E/e′ Tricuspid regurgitation velocity Right ventricular function Novel parameters Myocardial strain and strain rate Arterial function and arterial wave reflections Ventricular–arterial coupling Other … Cardiovascular morbidity Hospitalization for myocardial infarction, heart failure, and cerebrovascular accident Cardiovascular mortality Total mortality General population Patient population Hypertension, heart failure, or myocardial infarction Clinical parameters Age, sex, race, height, weight, blood pressure and components, and heart rate NYHA class Lifestyle, co-morbidities (diabetes, frailty, …), and hospitalizations Laboratory tests Haematocrit, eGFR, sodium, potassium, fasting glucose, HbA1c, lipid fractions, albuminuria, and inflammation NT-pro-BNP or BNP Established parameters LV mass index, LV volume indexes, LA volume index, and EF Tissue Doppler parameters: s′, e′, a′ LV filling pressures evaluated with E/e′ Tricuspid regurgitation velocity Right ventricular function Novel parameters Myocardial strain and strain rate Arterial function and arterial wave reflections Ventricular–arterial coupling Other … Cardiovascular morbidity Hospitalization for myocardial infarction, heart failure, and cerebrovascular accident Cardiovascular mortality Total mortality eGFR, estimated glomerular filtration rate; BNP, brain natriuretic peptide; LV, Left ventricular; LA, left atrial; EF, LV ejection fraction. This commonly used and conservative approach is not devoid of limitations. Established parameters are usually not questioned. The parameter under scrutiny may have an important shared variance with an established parameter. It nevertheless may be more reliable, reproducible, and with a better prognostic yield. Such a parameter will most likely not come out as a significant improvement in the regression procedure. The most striking illustration is the prediction of cardiovascular events with the SCORE system, using age, sex, smoking, blood pressure and cholesterol as parameters.1 It has been notoriously difficult to demonstrate the added value of inflammation parameters, genetic or epi-genetic markers,2 as these novel markers affect outcome largely through the SCORE variables. It therefore could be useful in a next phase of the analysis to substitute novel imaging markers to established parameters, to show their superiority. It would be advisable to search for easily measurable, reproducible, and validated alternatives, for example alternatives for NYHA class, ejection fraction (EF), and some of the conventional echocardiographic measures. The Copenhagen City Heart Study provides a most interesting illustration of some key aspects of how to improve prediction of cardiovascular morbidity and mortality with sensitive, reproducible, and easily measurable echocardiographic measurements.3 The study population comprises >2000 subjects (mean age 58 years), representative of the population at large. The participants were questioned, were examined, and underwent an echocardiographic examination at baseline from 2001 till 2003. Participants were followed till April 2013. The primary endpoint was cardiovascular death or admission for either heart failure or myocardial infarction. This primary endpoint was reached in 277 (13.4%) of the participants. Different prediction models of the primary endpoint are presented. Model 1: adjusted for age and sex; Model 2: Model 1 and BMI, heart rate, hypertension, diabetes, prior admission for ischaemic heart disease or heart failure, and plasma pro-BNP; Model 3: Model 2 and conventional echo measures. In all three models, lower s′, lower e′, and lower a′ independently predicted the primary endpoint. Interestingly, only reduced early diastolic velocity e′ remained an independent predictor of acute myocardial infarction, whereas only reduced systolic s′ and late diastolic velocities a′ remained independent predictors of heart failure admissions and cardiovascular death. The information provided by the different parameters, therefore, is distinct. In a previous study from the same database, the authors looked at the prediction of mortality till the year 2007, with some 1000 participants and 90 events.4 In that study, lower a′ (and to a lesser extent lower s′) was a significant predictor of death in Cox proportional-hazards models adjusted for clinical variables and conventional echocardiography, but e′ was not. This part of the information is confirmed and refined in the present investigation. These studies are attractive, because the results are consonant with basic physiological principles. Left ventricular function (decreased s′) is related to death and heart failure. Impaired myocardial relaxation (decreased e′) is related to coronary heart disease and incident myocardial infarction. In addition, the authors carefully analysed atrial contraction, an often overlooked phase of the cardiac cycle in contemporary diastology. Myocardial relaxation slows with age, thereby impairing rapid filling. Myocardial relaxation may also be affected by myocardial ischaemia or maladaptive hypertrophy. The physiological compensation for impaired relaxation is a forceful contribution to filling by active contraction of the left atrium. Such information should be adequately reflected by the ratio e′/a′. There, however, is more information in the data. Early (e′) and late (a′) diastolic velocities provide independent, additive prognostic information. Reduced and not increased a′ was associated with a poor prognosis even in the subgroup of persons with a normal conventional echocardiographic examination. Whereas e′ and a′ were inversely correlated in the event-free subgroup (as expected), they were not correlated in the subgroup that reached the combined endpoint. A lower a′, therefore, is an independent predictor of outcome. This is unexpected, important, and novel information coming from both manuscripts from the Copenhagen City Heart Study. But what does a lower TDI a′ actually mean? It can be the reflection of a stiffer ventricle. A stiff ventricle expands less during atrial contraction. This results in a smaller and shorter Doppler A wave and TDI a′ velocity. Another and quite different possibility is impaired active contraction of the left atrium. Impaired atrial contractility would then parallel the impaired ventricular contractility revealed by the lower s′. Differentiating between increased ventricular stiffness and atrial contractile failure is not possible with the published data. This could be achieved with simultaneous analysis of the mitral A wave and the reversed pulmonary venous flow during atrial contraction. The amplitude of the reversed flow will reflect the strength of the atrial contraction, and discrepancies between mitral and pulmonary venous flow duration will reflect increased ventricular stiffness.5 The important and independent information on the prognostic significance of atrial contraction might trigger future research. This will require measurement of the pulmonary vein flow as well, which on top will provide information on the S/D integral ratio. This ratio is another novel prognostic factor related to diastolic function.6 A limitation of the pulmonary venous flow, however, is that it cannot be recorded in a significant proportion of subjects.5 Another interesting aspect of the study is the poorer performance of ratios, compared with the performance of the components of the ratio. It is clear from Table 2 of the manuscript under scrutiny that E/e′, e′/a′, and e′/a′*s′ (eas index) perform poorer than s′, e′, and a′. There is a logical explanation for it. A ratio works only well if the disease process induces opposite effects (respectively, increase and decrease) on the two components of the ratio. An illustration is the use of E/e′ to predict filling pressures in the particular case of HFrEF, where increasing E and decreasing e′ reflect disease severity and deteriorating heart failure. If however the disease process induces changes in the same direction, the impact on the ratio will be weakened and disappear. A striking example is the effect of ageing on EF. With increasing age, stroke volume and end-diastolic volume both decrease, so that the EF (the ratio) remains unaltered.7 Maladaptive remodelling and the contractile dysfunction are therefore concealed by the use of a ratio. What novel parameters will the near future bring us? Population and patient data, including sensitive strain and strain rate-related parameters, are acquired on a wide scale worldwide. Early myocardial dysfunction was documented in diabetes8 and in hypertensive patients.9,10 It nevertheless is largely unsettled to what extent the documentation of early systolic and diastolic dysfunction may predict maladaptive remodelling,11 but still more important global outcome. This advanced evaluation of myocardial function has to be integrated in an improved assessment of vascular load (afterload) and arterial function. Contractile parameters are load dependent. In addition, the need for this additional sophistication is a consequence of well-established and opposite effects of time - varying early and late systolic load on myocardial function,12 left ventricular remodelling,13 and outcome.14,15 While higher early systolic load surprisingly is associated with improved myocardial function and outcome, late systolic load (as induced by amplified and earlier arterial wave reflections) predicts impaired myocardial function, maladaptive remodelling, and poorer outcome. Future research on the prognostic relevance of cardiac function will deal with systolic, early diastolic, and late diastolic function. It will include as well arterial function and ventricular–arterial coupling.