The Relationship Between Cardiothoracic Ratio and Left Ventricular Ejection Fraction in Congestive Heart Failure FREE

MEASURES OF left ventricular contractile function, including ejection fraction (EF), are relevant to diagnosis,^1- 4 prognosis,^3- 11 and treatment^1- 4,7,11 in patients with congestive heart failure (CHF). The EF is inversely related to mortality,^3- 11 and is perhaps the best prognostic index in heart failure. Thus, experts recommend that left ventricular systolic function be determined in patients with CHF.^3,12- 13 Although EF can be readily measured using angiographic, radionuclide, or echocardiographic techniques, these tests are expensive and may not be readily available in all clinical settings. A simple, convenient, noninvasive, and inexpensive method of accurately estimating left ventricular EF would be of great value in such patients.

Cardiac enlargement is associated with an adverse outcome in patients with heart disease.^8,14- 15 The size of the heart can be estimated from a chest roentgenogram by expressing it as a proportion of the thoracic diameter, the cardiothoracic ratio (CTR).¹⁶ Most cases of chronic CHF due to acquired left ventricular systolic dysfunction are associated with chamber dilation.^17- 18 In this context, one would expect an association between overall heart size, or CTR, and EF. However, one study demonstrated only a weak relationship between CTR and measured EF in patients with CHF.⁸ Furthermore, earlier work has shown a weak or inconsistent relationship between radiographic indexes and left ventricular size^19- 21 or function,^19,22 although these studies may have been marred by small sample size and patient selection bias. Despite the absence of solid supportive data, clinicians continue to use roentgenographic heart size as a clue to differentiate systolic from diastolic dysfunction in patients with CHF.^1,4

The Digitalis Investigation Group (DIG) Trial was a study of the effects of digoxin use on mortality and morbidity in patients with stable CHF and normal sinus rhythm. This trial included a large number of female and nonwhite patients.²³ Relative to other CHF studies, the DIG Trial also included a large proportion of patients with a left ventricular EF greater than 45%,²³ thereby providing a unique opportunity to examine clinical variables in a broad spectrum of CHF patients. We used the DIG Trial database to test the hypothesis that the CTR measured from the chest roentgenogram can be used to estimate left ventricular EF in individuals with CHF.

The design of the DIG Trial has been previously reported.²³ In brief, patients with stable symptoms of CHF (New York Heart Association functional classes I-IV) who had normal sinus rhythm were enrolled in this study. Previous treatment with digitalis was allowed. Investigators were encouraged to prescribe angiotensin-converting enzyme inhibitors or other vasodilators prior to randomization. Exclusion criteria included age younger than 21 years and the presence of complex congenital heart disease, unstable or refractory angina, atrial fibrillation or flutter, cor pulmonale, constrictive pericarditis, acute myocarditis, hypertrophic cardiomyopathy, recent myocardial infarction, and uncorrected severe valvular heart disease. The underlying primary cause of CHF was determined locally by the DIG investigators as either ischemic, hypertensive, valvular, idiopathic, alcohol-related, or "other" using common clinical definitions. Participants in the DIG study with valvular and other causes of CHF, representing only 3.6% of the cohort, were excluded from the current analysis. In addition, DIG participants with missing or potentially erroneous data (ie, CTR, <0.16 or >0.84; EF, <6% or >79%) were also excluded from the current analysis. For the purpose of this study, clinical right ventricular dysfunction was considered present when patients had elevated jugular venous pressure and/or peripheral edema within 1 month of enrollment in the DIG study.^24- 25 The DIG Trial was approved by the institutional review boards of all the participating centers and all patients signed informed consent forms.

Prior to randomization in the DIG Trial, patients had a chest roentgenogram performed to measure CTR. Patients also underwent assessment of left ventricular EF. Both studies could be performed within 6 months of randomization. In cases in which clinical events occurred in the interim that might have affected the test results (such as myocardial infarction or revascularization), these studies were repeated prior to randomization and the most recent results recorded in the database. Vital patient data, including CTR and EF, were conveyed to the DIG Trial's data coordinating center prior to randomization. When investigators reported a CTR greater than 0.70 or less than 0.40 or an EF greater than 70% or less than 6%, they were asked by the data coordinating center to confirm the validity of these data.

Posteroanterior chest roentgenograms were performed using standard radiologic techniques. The CTR was determined using the Danzer method¹⁶ (Figure 1). Left ventricular EF was also measured locally using either angiographic, radionuclide, or 2-dimensional echocardiographic techniques, with the choice being left to the investigators. However, quantitative measurements, not visual estimates, were required. For those patients undergoing echocardiographic measurement, the area-length method, modified Simpsons rule, or some other standard equation was specifically required.

Data archival and statistical analyses were performed using SAS software.²⁶ Student t test and analysis of variance methods were used to compare the results of continuous variables between groups. The Pearson product moment correlation coefficient was used to evaluate the relationship between CTR and EF. After the relationship between CTR and EF was examined as continuous variables, categorical analysis was performed to test the predictive value of CTR for EF in individual patients. Specifically, a series of CTR cutoff points from 0.45 and below to 0.62 and higher were examined for their ability to differentiate patients with EF greater than 35% or 35% and lower; greater than 40% or 40% and lower; and greater than 45% or 45% and lower.

In addition to statistical analyses for the entire study group, results were stratified by the method of EF measurement, the cause of CHF, and the presence of clinical right ventricular dysfunction. Because the anatomical pattern of cardiac chamber enlargement (cavity dimension and wall thickness) might differ between patients with and without hypertensive conditions, separate analyses were performed for all patients without hypertensive heart disease combined into the all nonhypertensive cohort. In this article, results are presented as mean (±SD) unless otherwise noted.

Between February 1991 and August 1993, a total of 7788 patients were enrolled in the DIG Trial at 302 centers in the United States and Canada. Of the total population, 254 participants were excluded from this analysis on the basis of the presence of valvular or other heart disease. An additional 58 patients were excluded on the basis of missing or potentially erroneous data. Demographic and clinical characteristics of the study cohort, including the causal distribution of CHF, are shown in Table 1.

The EF was calculated using angiographic methods in 5.5% of patients, radionuclide techniques in 65.6%, and echocardiography in 28.9%, and the EF was similar among these groups (angiography, 32.2±11.2%; radionuclide, 31.7±12.3%; and echocardiography, 31.4±12.0%). The mean EF for the entire group was 31.7±12.2%. An EF greater than 35% was noted in 2463 patients (32.9%), greater than 40% in 1513 patients (20.2%), and greater than 45% in 886 patients (11.9%). The mean EF was similar among patients with clinical right ventricular dysfunction and those without. The mean EF was higher in the group with hypertension than in the others (P<.001). The mean CTR for the entire sample was 0.53±0.07 and varied little by causal subgroup. Within causal subgroups, patients with right ventricular dysfunction had a slightly higher CTR. The mean CTR was similar among the 3 EF method groups (angiography, 0.54±0.08; radionuclide, 0.53±0.07; and echocardiography, 0.54±0.08).

For the entire cohort, age was weakly related to CTR (r=0.08) and EF (r=0.107). For the entire sample, weight was not related to CTR (r=−0.033) or EF (r=0.039). Moreover, body mass index (calculated as the weight in kilograms divided by the square of the height in meters) was not related to CTR (r=−0.002) or EF (r=0.021). Similar findings were observed within each of the causal subgroups.

For the entire study group, a weak negative correlation between CTR and EF was observed (r=−0.176). Because of the large number of patients studied, this correlation coefficient was of high statistical significance (P<.001). A regression equation for the prediction of EF from CTR was derived:

with an SE of the estimate of 12.0%. Similar results were observed when the data were stratified by clinical right ventricular dysfunction (Table 2 and Table 3), cause of CHF (Table 2 and Table 3), and method of EF measurement (Table 3). Among those causal–right ventricular dysfunction–EF method subgroups with 100 or more patients, a correlation coefficient larger than −0.35 was observed in only 1: the group with idiopathic causes of CHF and right ventricular dysfunction who underwent radionuclide studies (r=−0.388).

With categorical analysis, increasing proportions of patients were shown to have low EF at incrementally higher CTR values, consistent with the negative correlation between CTR and EF. However, no CTR value was precise in the prediction of EF greater than 35% or 35% and below; greater than 40% or 40% and below; or greater than 45% or 45% and below in any patient group. For example, among the entire study population, an EF of 35% and below was present in 56% of patients with a CTR of 0.45 and lower, 59% of patients with a CTR lower than 0.50, 64% of patients with a CTR of 0.55 and lower, 65% of patients with a CTR lower than 0.60, and 75% of patients with a CTR of 0.60 and higher. Among patients with hypertensive disease, an EF of 45% and below was present in 51% of patients with a CTR of 0.45 and lower, 55% of patients with a CTR lower than 0.50, 65% of patients with a CTR of 0.55 and lower, 69% of patients with a CTR lower than 0.60, and 81% of patients with a CTR of 0.60 and higher. The sensitivity, specificity, and positive and negative predictive values of a CTR higher than 0.55 for detecting an EF of 35% and below for all patients and certain subgroups are displayed in Table 4.

Our data demonstrate a weak negative correlation between CTR and left ventricular EF among a large and diverse population with chronic, stable heart failure. Although statistically significant due to the cohort's large size, this correlation is not clinically useful because it does not allow for the accurate prediction of EF in individual patients or specific subgroups.

The cardiac silhouette on a chest roentgenogram encompasses all the contents of the pericardium. As shown in Figure 1, the transverse dimension of the cardiac silhouette, which forms the numerator of the CTR, is predominantly affected by right atrium size, the internal dimension of the left ventricle, the thickness of the left ventricular wall, pericardial thickness, and the contents of the pericardial space. In the majority of patients with chronic left-sided CHF, left ventricular systolic function decreases, filling pressure increases, and the chamber dilates.^17- 18 Dilation and/or hypertrophy of other cardiac chambers may also occur. Thus, patients with CHF generally have a larger CTR than healthy subjects.⁸ In turn, clinicians have extrapolated that the chest roentgenogram can be used to predict systolic function in patients with CHF.^1,4

In this study, care was taken to account for those clinical variables that might interfere with the relationship between CTR and EF among patients with CHF. Patients were stratified by CHF cause because the anatomical pattern of chamber enlargement (cavity dimension and increased wall thickness) might differ among groups. For instance, an increased CTR due to hypertrophy may be associated with low, normal, or high EF in patients with CHF caused by hypertensive heart disease. Patients with vaguely defined CHF causes were excluded from this analysis. In addition, patients with valvular heart disease, representing a very small portion of the DIG Trial cohort, were excluded on the assumption that they are a heterogeneous group with a varied and complex relationship between CTR and EF. For example, there are distinct differences between the classic patterns of chamber enlargement and left ventricular contractile function in primary mitral stenosis and primary mitral regurgitation. Patients with clinical right ventricular dysfunction were studied separately from those without because of the potential differences between these groups in the pattern of right atrial and right ventricular chamber size and wall thickness.^27- 28 Subgroup analysis based on modality of EF measurement was performed to control for variation in the random error of measurement among these 3 techniques.

Despite taking into consideration these factors that might impact the CTR and EF, we obtained compelling evidence that CTR cannot be used to reliably estimate left ventricular EF in individual patients with CHF due to ischemic, hypertensive, idiopathic, or alcohol-related causes. In general, a majority of our patients with stable CHF manifested high CTRs, but in a given CHF patient, the CTR did not accurately estimate EF. In other words, the CTR adds minimally to the prediction of EF in a patient known to have CHF.

Our findings are consistent with those of previous studies. Cohn et al⁸ reported a modest negative correlation between CTR and EF among 584 male patients with chronic CHF and low EF enrolled in V-HeFT I (r=−0.27) and 758 male patients enrolled in V-HeFT II (r=−0.28). Rose and Stolberg¹⁹ reported a negative correlation (r=−0.22) between CTR and angiographic EF among 256 subjects who underwent cardiac catheterization. In their study of alcohol abusers, Bertolet et al²² reported that only 8 of 29 subjects with asymptomatic left ventricular systolic dysfunction defined using echocardiography had cardiomegaly on their chest roentgenogram. Righetti et al²¹ found that serial changes in roentgenographic heart size following cardiac surgery correlated with the presence of pericardial fluid but not left ventricular diameter or function. Although our findings agree with those of previous studies, our study overcomes many of the limitations of prior investigations. We examined a larger and more diverse population, with a greater mix of race and sex. Moreover, we included a greater proportion of CHF patients with an EF greater than 45%. Finally, our statistical method allows the clinician to evaluate the predictive accuracy of CTR for EF in individual patients.

A possible explanation for the weak relationship between CTR and EF lies with the variable distortion of right atrial and right ventricular morphological characteristics and function,^27- 28 which can occur in any given left ventricular EF among patients with CHF. Moreover, the poor correlation between CTR and EF may be rooted in the inherent limitations of these 2 measures. Although the CTR was designed to compensate for such factors as sex and body habitus, it does not account for respiration and cardiac cycle–specific variation in heart position and chamber volume, and is subject to variability as a function of the heart-to-film distance. Second, a time lag may occur between the onset of left ventricular systolic dysfunction and an increase in cavity dimension, or CTR. However, a time lag alone cannot account for the reduced sensitivity in our series, since the presence of right ventricular failure, which implies relative chronicity of CHF,⁶ did not improve the CTR-EF relationship. While EF is known to be dependent on cardiac loading conditions²⁹ and can be altered by pharmacologic therapy in patients with CHF,⁷ the CTR may be less responsive to these factors. Finally, EF measurement itself is subject to random variation,³⁰ although the size of our study sample should have adequately compensated for such variations in EF assessment.

A few limitations of this study warrant comment. First, failure to use a single laboratory for determination of CTR and EF may have induced significant variability in these measurements. Second, we cannot state whether our conclusions apply to those patients with CHF secondary to other causes, such as valvular or congenital heart disease, acute myocarditis, or cor pulmonale. Third, the proportion of patients in the DIG Trial with an EF greater than 40% or 45% may be smaller than that encountered in a typical primary care practice.^1,4 Thus, the positive and negative predictive values reported in our study may not apply to populations with a different prevalence of diastolic heart failure. Fourth, because the database did not specify those medications taken at the actual time of CTR and EF measurement, we cannot state the degree to which drug use interfered with the CTR-EF relationship. Fifth, we cannot discount the possibility that CTR (or other roentgenographic findings) had important diagnostic or prognostic implications that were independent of their relationship to left ventricular EF. Finally, we cannot comment on the predictive value of other radiographic findings, such as pulmonary congestion, since these data were not recorded in the database.

Congestive heart failure is a common condition^17,31 associated with a poor prognosis.^5,17,31- 33 Left ventricular systolic function has important implications for cause,^1- 4 prognosis,^3- 11 and treatment^1- 4,7,11 in this disorder. For example, in diastolic heart failure, diuretics should be used with caution or avoided, although β-adrenergic and calcium channel-blocking agents may be effective.^1,3- 4 Evidence exists that some patients with CHF may receive suboptimal medical care.^31,34 One aspect of this quality of care issue may be related to incorrect management decisions that result from inaccurate differentiation of systolic from diastolic dysfunction.² Thus, our findings support the recommendations of expert panels^3,12- 13 that direct and quantitative methods should be used to distinguish systolic from diastolic heart failure.^35- 36 Considering the morbidity and mortality associated with CHF, estimating contractile function from the chest roentgenogram should be considered poor practice and be discouraged.

Accepted for publication June 19, 1997.

This study was supported by funds from the National Heart, Lung, and Blood Institute, National Institutes of Health, Bethesda, Md (Interagency Agreement 1Y01-HC00110), and the Department of Veterans Affairs Cooperative Studies Program. The active study drug and placebo tablets were supplied by Glaxo Wellcome Company, Research Triangle Park, NC.

Reprints: Edward F. Philbin, MD, Section of Heart Failure and Heart Transplantation, Division of Cardiovascular Medicine, Henry Ford Health System, 2799 W Grand Blvd, Detroit, MI 48202 (e-mail: ephilbi1@hfhs.org).