Echocardiographic epidemiology--an emerging tool for early diagnosis, studying pathophysiology, predicting prognosis, and testing treatments

Advances in echocardiography mean that it is now possible to measure myocardial function accurately and non-invasively. Paradoxically, objective quantification of myocardial motion, contraction, and relaxation in each segment of the left ventricle actually makes it more difficult to define normality and disease in the individual patient, but the technique raises intriguing possibilities in population studies for diagnosing early heart muscle disease and testing new approaches to treatment. The catalyst to this approach was the development of tissue Doppler echocardiography to measure the velocity of myocardial motion, and then the processing of digitally stored velocity data to obtain measurements of regional myocardial deformation (strain and strain rate). Machines do not apply the Doppler equation when processing these signals, and so an alternative and more accurate term that encompasses all available methods is myocardial velocity imaging. Mogelvang et al. have reported their findings from performing echocardiography in >1000 subjects in the Copenhagen City Heart Study.1 These subjects comprised 17% of the total sample which had been drawn randomly from population registers, who attended when echocardiography with myocardial velocity imaging could be performed. They were aged between 20 and 93 years, and included substantial numbers of subjects with hypertension (>140/90 mmHg), diabetes [plasma glucose >11.1 mmol/L, or glycosylated haemoglobin (HbA1c) >7%, or on treatment], and/or ischaemic heart disease (IHD, diagnosed from history or abnormal ECG). Compared with ‘control’ subjects with none of these conditions, the subjects with hypertension, diabetes, or IHD had lower systolic velocities of long-axis shortening of the left ventricle (on average, –13%), less systolic mitral annular excursion (also –13%), and lower velocities of long-axis lengthening in early diastole (by an average of –33%). Overall, only 2% of the subjects with disease had a low ejection fraction (<50%) and only 8% had any degree of diastolic dysfunction (applying age-related cut-off points for mitral E/A ratio and mitral deceleration time). Risk factors for cardiovascular disease commonly co-exist and, unsurprisingly, their groups reflect this. Subjects included in the group with hypertension had no diabetes or IHD, whereas 80% of the diabetic group had hypertension (but no overt IHD), and 73 and 14% of the IHD group had hypertension or diabetes, respectively. Also, the controls were almost 20 years younger and they had a lower mean body mass index. Interactions between these variables were tested and adjustments were made in the statistical models. The mean differences between groups were small but they were highly significant on statistical testing. They were demonstrated most clearly by the cumulative probability distributions—which are much more useful in such studies than reporting categorical data using arbitrary cut-off points. Mogelvang and colleagues therefore provide excellent and interesting corroborative evidence for the presence of progressive, continuously declining subclinical myocardial dysfunction in subjects with increasing severity of common cardiovascular risk factors. They confirm that standard indices of global left ventricular function cannot demonstrate these clinically important relationships, while myocardial velocity imaging can do so. The velocities of longitudinal shortening and lengthening, and the displacement of the mitral annulus, can all be measured with good reproducibility. Coefficients of variation are ∼10%,2 which means that these measurements have excellent statistical power in clinical trials, even if changes of ∼25% between consecutive observations in an individual subject can occur by chance. Loss of longitudinal motion of the left ventricle is the most obvious functional change that occurs with ageing. Obviously, during each beat, the heart that has shortened in systole must lengthen by the same amount to return to its original shape and position before the start of the next beat. Mogelvang and colleagues observed that systolic velocity and motion were reduced to a similar extent in the subjects with disease, compared with the controls, but the differences in the velocity of early diastolic lengthening were greater. This does not mean that diastolic dysfunction occurs before systolic dysfunction, since both always change together.3 Instead, the greater reduction in diastolic velocity can only be explained by a simultaneous prolongation of the duration of early diastolic filling—which would be expected. It is important when studying subclinical disease to measure not just the rate of change of function (e.g. myocardial velocity, or strain rate) but also its timing, duration and amplitude (e.g. myocardial displacement, or strain). Some investigators criticize the use of myocardial velocities because a velocity can be recorded from a segment that moves but does not contract. This is pertinent when there is a regional disease such as myocardial ischaemia or infarction, but it is not important when the whole heart is affected by subclinical myocardial dysfunction and when measuring velocities is much more reproducible than measuring either strain or strain rate (which is currently true, also for speckle tracking). Velocities and excursion measured at the base of each wall from an apical echocardiographic window are determined by the function of all the segments along that wall. Averaging the results from all the walls, as done by Mogelvang and colleagues, provides an extremely useful summary of subendocardial function in all segments of the ventricle that is very sensitive to early disease. Radial function of the left ventricle may be misleading as it is increased in subjects with early diabetic4 or hypertensive5 heart disease—perhaps because the subtle increase in ventricular sphericity that occurs in early disease6 leads to increased stretch of the mid-wall myocardial fibres. What are the most important clinical implications from ‘echocardiographic epidemiology’ in this and other similar studies that have been reported and are still being conducted? First, we can learn about the pathophysiology of subclinical myocardial dysfunction and its progression to heart failure—which generates hypotheses for secondary preventive strategies that can be tested in animal models and clinical trials. Hypertension, diabetes, and IHD are all risk factors for heart failure (whether systolic or with preserved ejection fraction). Reduced longitudinal function of the left ventricle is related to increased interstitial fibrosis in the subendocardium and to increased conduit arterial stiffness and central arterial pressures.7 It is thought to relate to inefficient ventricular–arterial coupling, although this hypothesis may need to be refined given the recent observation that conduit arterial stiffness is not the major determinant of the energy of arterial waves that are reflected back to the heart.8 Secondly, it now becomes very difficult to make a precise diagnosis in the individual patient. How can one decide if a hypertensive patient also has myocardial ischaemia, or if a diabetic patient also has hypertensive heart disease, when there is so much overlap between groups, and function varies continuously? Quantitative diagnosis using single cut-off points is impossible. Instead, we need a change of paradigm—in future, precise diagnosis will be made using computerized algorithms that can adjust for risk factors as well as simple demographic variables (similar to the simple formula reported by Mogelvang and colleagues) and then give a probability for a particular measurement to predict the presence of additional disease. Some important technical caveats should be stressed. Maximal myocardial velocities measured using a pulsed technique in real time are ∼25% higher than mean velocities obtained by processing stored images. Different manufacturers process the signals in different ways, and equivalence between machines has not been established. Results for any individual patient or for patients in a trial can only be compared when they have been obtained using the same machine and methods. Thirdly, left ventricular longitudinal function may prove to be a very powerful prognostic indicator for major cardiovascular endpoints, as shown already in specific clinical contexts.9 Its greatest value is probably that it reflects the net effects of many adverse influences on the subendocardium. In future, the simple parameter of mean longitudinal shortening could be used to monitor cardiac ageing and as a surrogate functional endpoint for guiding treatment. The potential utility of this approach is illustrated by the unexpected finding in the ACCORD trial that intensified treatment of diabetes did not reduce clinical endpoints,10 but subjects with lower glycosylated haemoglobin gained more weight, and myocardial velocity imaging has demonstrated that obesity is a strong correlate of reduced longitudinal function.11 Finally, all new diagnostic tests should be considered for their impact on clinical outcomes12 and they should be applied in routine practice only if this is beneficial. Some ‘proof of principle’ clinical studies have made sequential measurements of longitudinal velocities or deformation and then demonstrated improvements in regional myocardial function, for example from antifibrotic treatment with spironolactone in chronic heart failure13 and from valsartan and other hypotensive agents in patients with diastolic dysfunction.14 These methods have not yet been applied in large endpoint trials. There is surely a need to bring diagnostic and therapeutic experts together so that new trials are designed to take advantage of the most recent diagnostic advances. Conflict of interest: none declared.