BACKGROUND The size of the heart assessed by cardiothoracic ratio on chest radiography is often used as a screening test for the presence of heart failure and for assessing its severity.
METHODS We compared cardiothoracic ratio (CTR), left ventricular ejection fraction (LVEF) from radionuclide ventriculography, and left ventricular dimensions from echocardiography in a population of 91 patients (aged 60.4 (SD 9.6) years) with a diagnosis of chronic heart failure.
RESULTS There was a weak relation between CTR and LVEF (R=0.33) and fractional shortening from echocardiography (R=0.22). LVEF and fractional shortening correlated more closely (R=0.55). No measure of left ventricular function correlated with exercise capacity as measured by peak oxygen consumption. For the group of patients with a normal fractional shortening (n=17), the left ventricle was dilated in all but two (mean end diastolic dimension 5.9 (0.7) cm). The two with normal dimensions had a low ejection fraction. For the 12 patients with a CTR in the normal range, the left ventricular end diastolic dimension was only slightly smaller than for the rest (6.2 (0.9)v 6.9 (1.2); p=0.045).
CONCLUSIONS Chest radiography is not a reliable indicator of the degree of left ventricular dysfunction. Echocardiography and radionuclide ventriculography are more appropriate investigations for assessing cardiac function.
- chronic heart failure
- chest radiography
- radionuclide ventriculography
Statistics from Altmetric.com
An accurate diagnosis is the cornerstone of good management in chronic heart failure. Clinical examination alone is notoriously unreliable for establishing heart failure,1 2and demonstrating objective evidence of left ventricular dysfunction is crucial to make the diagnosis. Methods commonly used are echocardiography (using either the absolute left ventricular dimensions or a derived index such as fractional shortening or ejection fraction) and radionuclide ventriculography (again used to determine left ventricular ejection fraction, LVEF).
Although heart failure is common, the diagnosis can be difficult. Scoring systems have been described (primarily for use in epidemiological studies),1 3-5 which use the cardiothoracic ratio from a posteroanterior chest radiograph as an index of the size of the heart. The chest radiograph is widely available and frequently performed as a screening test for left ventricular failure. Other methods for assessing cardiac function, such as left ventricular angiography and echocardiography are less widely available.
We sought to compare methods of quantifying left ventricular function in widespread use that are easily available in district general hospitals. We thus compared LVEF derived from radionuclide ventriculography with echocardiographic indices derived from M mode echocardiography in patients with clinical heart failure. In particular, we were concerned to see whether the chest radiograph could be recommended as a method for assessment of left ventricular function.
This was a retrospective study. Ninety one patients from the Royal Brompton heart failure database were identified who had had a chest radiograph, radionuclide ventriculography, and an echocardiographic study within three months of each other. Only those patients were included for whom M mode echocardiographic tracings were available.
Subjects were injected intravenously with 0.03 mg/kg of stannous fluoride red cell labelling agent (Amersham International). After allowing a period of 15–30 minutes for equilibration of the red blood cells, the patient was positioned supine on a scanning couch, with the gamma camera, a Sopha medical DS-X rectangular field of view, connected to lead V6 of an electrocardiograph to synchronise events. A bolus of 740 MBq of 99mtechnetium in 0.3–0.5 ml was injected intravenously, with the gamma camera positioned at 45º left anterior oblique with 10º of craniocaudal tilt. An equilibrium radionuclide ventriculogram was then acquired; 16 frames were acquired in each R-R interval, and summed repeatedly, until the end diastolic image contained 300 000 counts, to ensure an accurate estimation of LVEF. Ventricular ejection fraction was calculated by comparing background subtracted images at end diastole and end systole as: ejection fraction (LVEF) = (end diastolic−end systolic counts)/end diastolic counts × 100.
Left ventricular internal dimensions were acquired from standard M mode echocardiographic images in the parasternal long axis view at the mitral valve tips. Fractional shortening was calculated as left ventricular end diastolic dimension (LVEDD) minus left ventricular end systolic dimension (LVESD) divided by LVEDD.
CARDIOTHORACIC RATIO (CTR)
CTR was determined by one of us (ALC) only where a standard posteroanterior chest film from within three months of the radionuclide and echocardiographic studies was available. Only posteroanterior erect radiographs were accepted. No anteroposterior, supine, or seated films were accepted. The measurements were made blind to the results of the other investigations.
Exercise data were available from maximal incremental exercise tests with metabolic gas exchange data. Patients were encouraged to exercise to exhaustion using the standard Bruce protocol with the addition of a stage 0 (three minutes' exercise at one mile per hour with a 5% gradient). Subjects breathed through a one way valve allowing the collection of expired air. Expired air was mixed with an inert indicator gas in a mixing chamber, and samples taken every 10 seconds to be analysed by mass spectrometer (Amis 2000, Odennse, Denmark). Peak oxygen consumption was derived.
Results are shown as mean (SD). The relation between variables was explored using linear regression analysis.
Demographic data are shown in table 1. The average dose of frusemide (furosemide) being taken was 83 (55) mg daily, although seven were on no diuretic. Twenty one patients were not receiving angiotensin converting enzyme inhibitors. Twenty eight were receiving digoxin.
The indices of left ventricular function are shown in table 2. LVEDD and LVESD correlated closely (R=0.95; p<0.00001). The relations between LVEF from the radionuclide study and echocardiographic measurements and cardiothoracic ratio is shown in table 3 and figs 1 and2.
From fig 1, it can be seen that most patients have both abnormal LVEF and fractional shortening. For the group of patients with a normal fractional shortening (n=17), the left ventricle was dilated in all but two (mean end diastolic dimension 5.9 (0.7) cm). The two with normal dimensions had a low ejection fraction.
There was no significant relation between exercise capacity (as assessed by peak oxygen consumption) and any of the indices of cardiac function (the strongest relation was between peak oxygen consumption and LVEF; r=0.23, p=0.05).
Twelve patients had a CTR in the normal range (0.50). For this group, the LVEDD was slightly smaller than for the rest (6.2 (0.9)v 6.9 (1.2); p=0.045), but the ejection fraction was significantly higher (35.6 (17.5)v 25.7 (14.2); p=0.03).
Establishing the diagnosis of chronic heart failure accurately is essential for the correct management of the patient.8 9Clinical examination alone is unreliable.1 2 The most widely available imaging method available is echocardiography, with the dimensions of the heart and fractional shortening being the most reliable methods for quantifying function. Ejection fraction can be estimated from long axis dimensions. These methods are limited by being unable to take account of regional abnormalities.10 More sophisticated methods are very dependent on image quality, are time consuming, and are not routinely available. Echocardiography can be limited by poor image quality,11 12 although the impression of a trained operator is suitable clinically even in the absence of accurate measurements.13
Radionuclide scanning provides a measure with which clinicians are instinctively happy, LVEF. The correlation between echocardiography and radionuclide scanning has not been good in previous reports.10 14 LVEF from radionuclide ventriculography is more reproducible than echocardiography.15 16Radionuclide ventriculograms are less readily available than echocardiograms, and provide less overall information about the heart. Chest radiographs are commonly used as an initial test for the diagnosis of heart failure, particularly in general practice,17 but the apparent size of the heart on a plain film can be very misleading.18
The results reported here show that the correlations between heart size on radiography and left ventricular function measured by echocardiography and radionuclide ventriculography are too poor for the chest radiograph to be useful clinically as a diagnostic test of left ventricular function. Similar conclusions have been reported in studies postmyocardial infarction.19 20
Radionuclide scanning and echocardiography compared well with each other. There were a few patients with a low ejection fraction who were found to have normal fractional shortening, but these patients had enlarged left ventricles. There were six patients with a low fractional shortening who had a normal ejection fraction. The majority of patients were shown to have left ventricular impairment by both methods.
Chest radiography is an unsatisfactory method for diagnosing heart failure or drawing inferences as to cardiac function. The results of radionuclide ventriculography and echocardiography in a routine clinical setting agree well with each other and are each appropriate diagnostic tools.
This was a retrospective study. We have not included patients who have normal cardiac function. It is unethical to expose normal subjects to radionuclide scanning.
There is no absolute standard for the diagnosis of heart failure. The patients in this study were diagnosed on the pragmatic basis that they had symptoms compatible with heart failure in the presence of objective evidence of left ventricular dysfunction.
As this was a retrospective study, it is not possible to be certain of the standardisation for chest radiography. A poor inspiration can give a spuriously raised cardiothoracic ratio. The wide scatter of the points suggests that this has not been a systematic error.